Documente Academic
Documente Profesional
Documente Cultură
by
Chris Bibby
December 2011
ii
Table of Contents
Abstract.................................................................................................................................... ii
Table of Contents ................................................................................................................... iii
List of Tables .......................................................................................................................... ix
List of Figures.......................................................................................................................... x
List of Symbols and Abbreviations ................................................................................... xvii
Acknowledgements .............................................................................................................. xix
Chapter 1: Introduction ........................................................................................................ 1
1.1 Literature Review...................................................................................................... 4
1.2 Research Objectives.................................................................................................. 8
1.3 Thesis Outline ........................................................................................................... 8
Chapter 2: Optimization Metric: Open Area Ratio ......................................................... 10
2.1 Calculating Equivalent Open Areas........................................................................ 10
2.1.1 Equivalent Open Area for Sound.................................................................... 10
2.1.2 Equivalent Open Area for Flow...................................................................... 14
2.2 Summary Discussion on the Optimization Parameter ............................................ 15
2.3 Diffuse Field Theory and Ventilation Opening Characterization........................... 17
Chapter 3: Investigating the Factors in Speech Privacy .................................................. 20
3.1 Speech Intelligibility Index..................................................................................... 20
3.2 Model Description .................................................................................................. 21
3.2.1 Model Inputs ................................................................................................... 22
3.2.2 Calculating the Source Room Sound Pressure Levels.................................... 24
3.2.3 Calculating the Transmission Loss of the Partition ........................................ 24
3.2.4 Calculating the Receiver Room Sound Pressure Levels................................. 25
3.2.5 Cautionary Note .............................................................................................. 25
3.3 Results..................................................................................................................... 26
3.3.1 Vocal Effort and SII........................................................................................ 26
3.3.2 Room Size, Furnishings and SII ..................................................................... 27
3.3.3 Background Noise and SII .............................................................................. 28
3.3.4 TL and SII ....................................................................................................... 29
3.3.4.1 Partitions without Ventilation Openings..................................................... 30
iii
3.3.4.2 Partitions with Ventilation Openings.......................................................... 31
3.4 Factors Affecting Speech Privacy: Summary ......................................................... 34
Chapter 4: Describing Flow Performance of Ventilation Openings ............................... 35
4.1 Navier-Stokes Equation .......................................................................................... 35
4.2 Bernoulli’s Equation ............................................................................................... 36
4.3 Empirical Prediction of Flow Behavior .................................................................. 38
4.4 Determining Viscous Losses in Ducts .................................................................... 39
4.4.1 Discharge and Loss Coefficients .................................................................... 39
4.4.2 Dependence on the Reynolds Number............................................................ 41
Chapter 5: Measurement of Natural Ventilation Openings ............................................ 46
5.1 Ventilator Measurement Method ............................................................................ 46
5.1.1 Ventilator Transmission Loss ......................................................................... 46
5.1.1.1 Theory ......................................................................................................... 46
5.1.1.2 Measurement Method – In Field................................................................. 49
5.1.1.3 Measurement Method – In Lab................................................................... 49
5.1.2 Ventilator Flow Rate....................................................................................... 49
5.1.2.1 Theory ......................................................................................................... 50
5.1.2.2 Measurement Method – In-Situ and Laboratory......................................... 52
5.2 Design and Performance of the Lab Facility .......................................................... 52
5.2.1 Room Dimensions and Construction .............................................................. 53
5.2.2 Acoustic Performance of Lab Facility ............................................................ 54
5.2.3 Air Flow Performance of Laboratory Test Facility ........................................ 55
5.3 Report on Field Measurements of Natural Ventilation Openings .......................... 59
5.3.1 Regent College Library................................................................................... 60
5.3.1.1 Regent College L020 .................................................................................. 60
5.3.1.1.1 L020 Results: Grille On ........................................................................ 61
5.3.1.1.2 L020 Results: Grille Removed.............................................................. 61
5.3.2 Fred Kaiser Building....................................................................................... 63
5.3.2.1 Fred Kaiser 4036......................................................................................... 63
5.3.2.1.1 4036 Results.......................................................................................... 63
5.3.3 Liu Institute..................................................................................................... 65
iv
5.3.3.1 Liu 216C ..................................................................................................... 65
5.3.3.1.1 216C Results ......................................................................................... 66
5.3.3.2 Liu 308 ........................................................................................................ 66
5.3.3.2.1 308 Results............................................................................................ 66
5.3.3.3 Liu 313 ........................................................................................................ 67
5.3.3.3.1 313 Results: Grilles On......................................................................... 67
5.3.3.3.2 313 Results: Grilles Off ........................................................................ 67
5.3.4 C. K. Choi Building ........................................................................................ 69
5.3.4.1 C. K. Choi 167 ............................................................................................ 70
5.3.4.1.1 167 Results............................................................................................ 70
5.3.4.2 C.K. Choi 321 ............................................................................................. 70
5.3.4.2.1 321 Results............................................................................................ 71
5.3.4.3 C.K. Choi 326 ............................................................................................. 71
5.3.4.3.1 326 Results............................................................................................ 72
5.3.4.4 C.K. Choi 327 ............................................................................................. 72
5.3.4.4.1 327 Results............................................................................................ 73
5.3.5 Langara College Library ................................................................................. 74
5.3.5.1 Langara L104 .............................................................................................. 75
5.3.5.1.1 L104 Results: Grille On ........................................................................ 75
5.3.5.1.2 L104 Results: Grille Off ....................................................................... 76
5.3.5.2 Langara L112 .............................................................................................. 76
5.3.5.2.1 L112 Results ......................................................................................... 76
5.3.5.3 Langara L208 .............................................................................................. 77
5.3.5.3.1 L208 Results: Grille On ........................................................................ 77
5.3.5.3.2 L208 Results: Grille Off ....................................................................... 77
5.3.6 Operational Ventilator Flow Conditions......................................................... 80
5.3.7 In-Situ Results Summary ................................................................................ 84
5.4 Analysis of BRE Ventilator Measurements............................................................ 86
5.4.1 Ventilator Configurations ............................................................................... 86
5.4.1.1 Straight Ventilator....................................................................................... 86
5.4.1.2 L-shaped Ventilator .................................................................................... 87
v
5.4.2 Results............................................................................................................. 87
5.4.2.1 Effect of Acoustic Foam ............................................................................. 87
5.4.2.2 Effect of Elbow (L-shaped silencer) ........................................................... 88
5.4.2.3 Effect of Grilles........................................................................................... 88
5.4.2.4 Effect of PVC Lining .................................................................................. 88
5.4.2.5 Effect of Angled Baffles ............................................................................. 89
5.4.2.6 Effect of Wave-Shaped Absorptive Liner................................................... 89
5.4.3 Summary of BRE Ventilator Test Results ...................................................... 89
5.5 Laboratory Measurements of Ventilation Openings and Silencers ........................ 99
5.5.1 Slot Ventilation Opening ................................................................................ 99
5.5.2 Acoustical Louver......................................................................................... 103
5.5.3 Acoustical Baffle .......................................................................................... 108
5.5.4 Acoustical Air Filter?.................................................................................... 111
5.5.5 Cross Talk Silencer ....................................................................................... 114
5.5.6 Lab Measurement Results Summary ............................................................ 119
Chapter 6: Ventilation Silencer Performance Prediction .............................................. 122
6.1 Fundamental Mode Attenuation ........................................................................... 122
6.1.1 General Cartesian Solution for Sound in a Duct........................................... 122
6.1.2 Solution for a Rigid Walled Duct ................................................................. 124
6.1.3 Non-rigid Walled Duct ................................................................................. 125
6.1.3.1 Defining the Surface Impedance............................................................... 128
6.1.3.1.1 Transfer Function Method .................................................................. 128
6.1.3.1.2 Characterizing Porous Absorptive Materials ...................................... 130
6.1.4 Attenuation of the Fundamental Mode Results ............................................ 131
6.1.4.1 Duct Liner Properties................................................................................ 131
6.1.4.2 Cross Sectional Dimensions ..................................................................... 134
6.1.4.2.1 Flow Path Height ................................................................................ 134
6.1.4.2.2 Flow Path Aspect Ratio ...................................................................... 136
6.1.5 Attenuation of the Fundamental Mode Summary......................................... 139
6.2 Acoustical FEM Predictions of Ventilation Opening Transmission Loss ............ 139
6.2.1 Acoustical FEM Design................................................................................ 140
vi
6.2.1.1 Domains and Boundary Conditions .......................................................... 141
6.2.1.2 Calculation of Model Results.................................................................... 144
6.2.1.2.1 Frequency Averaging.......................................................................... 145
6.2.1.2.2 Lp (x, f)................................................................................................ 146
6.2.1.2.3 Lp (f).................................................................................................... 146
6.2.1.2.4 Lp (x)................................................................................................... 146
6.2.1.2.5 σLp ....................................................................................................... 147
6.2.1.2.6 Ii(x, f)................................................................................................... 147
6.2.1.2.7 Win(f), Wout(f)....................................................................................... 148
6.2.1.2.8 WLp(f).................................................................................................. 148
6.2.1.2.9 TL_Win(f) ............................................................................................ 149
6.2.1.2.10 TL_WLp(f).......................................................................................... 150
6.2.1.2.11 TL_Win and TL_WLp ......................................................................... 150
6.2.1.3 Designing for Convergence of Transmission Loss FEM.......................... 150
6.2.1.3.1 Mesh Convergence.............................................................................. 151
6.2.1.3.2 Frequency Convergence...................................................................... 152
6.2.1.3.3 Diffuse Field Convergence ................................................................. 155
6.2.2 Transmission Loss FEM Validation ............................................................. 166
6.2.2.1 Analytical Validation ................................................................................ 166
6.2.2.1.1 Plane Wave Attenuation ..................................................................... 167
6.2.2.1.2 Calculating the Inlet Power through Sound Pressure Level
or Intensity ............................................................................................................ 170
6.2.2.2 Comparing Measured Results to FEM Prediction .................................... 173
6.2.2.2.1 Slot Ventilation Opening .................................................................... 173
6.2.2.2.2 CT Silencers........................................................................................ 174
6.2.3 Acoustic FEM Design Summary .................................................................. 177
6.2.4 Acoustic FEM CT Silencer Results .............................................................. 179
6.2.4.1 Frequency Dependence of CT Silencer Transmission Loss ..................... 181
6.3 CFD Predictions of Ventilation Opening Discharge Coefficients ........................ 187
6.3.1 Flow Domain Geometry ............................................................................... 187
6.3.2 Flow-Equation Model ................................................................................... 190
vii
6.3.2.1 Reynolds Averaging (RANS) ................................................................... 190
6.3.3 Boundary Conditions .................................................................................... 191
6.3.4 Mesh.............................................................................................................. 191
6.3.5 Processing Flow Model Results.................................................................... 192
6.3.6 Model Validation .......................................................................................... 192
6.3.6.1 Validation by Comparison to Published Results ...................................... 192
6.3.6.2 Validation by Comparison to Measured Results ...................................... 194
6.3.7 CFD Prediction Results for CT Silencers ..................................................... 194
6.3.7.1 Effect of Length on the Straight Silencer’s Cd.......................................... 195
6.3.7.2 Effect of CT-Silencer Type on Cd ............................................................. 196
6.3.7.3 Effect of Reynolds Number on Cd ............................................................ 197
6.4 CT Silencer Modeling Summary .......................................................................... 202
Chapter 7: Conclusions ..................................................................................................... 205
7.1 Summary ............................................................................................................... 205
7.2 Future Work .......................................................................................................... 211
7.2.1 Set Optimization Objectives ......................................................................... 211
7.2.2 Improve Physical and Numerical Performance Prediction Methods............ 211
7.2.3 Optimize Natural Ventilation Openings ....................................................... 212
References ............................................................................................................................ 214
Appendices........................................................................................................................... 221
Appendix A: Manufacturer Data Sheets ........................................................................... 222
A.1 CertainTeed OEM Acoustic Fiberglass Absorber ............................................ 222
A.2 Kinetics Noise Control Acoustical Louver ....................................................... 224
A.3 Filter Data Sheets.............................................................................................. 226
Appendix B: MATLAB Code........................................................................................... 229
B.1 Diffuse Field Sound Transmission Model ........................................................ 229
B.2 Attenuation in a Lined Duct.............................................................................. 236
B.3 Processing COMSOL Results........................................................................... 241
Appendix C: Numerical Prediction Appendices............................................................... 244
C.1 Predicted CT Silencer Transmission Loss ........................................................ 245
C.2 Velocity- and Pressure-Field Figures for CT Silencers at High-Re ................. 246
viii
List of Tables
Table 1: Background noise level L90 [dB]. ............................................................................. 22
Table 2: Speech power levels [dB] [23]. ................................................................................ 23
Table 3: Absorption coefficients of room surfaces [41]. ........................................................ 23
Table 4: Room surface configuration types. ........................................................................... 23
Table 5: Room volumes. ......................................................................................................... 23
Table 6: Transmission loss and STC of various partitions [36].............................................. 23
Table 7: Equivalent open area of the partitions. ..................................................................... 32
Table 8: Flow-rate coefficients. .............................................................................................. 50
Table 9: Performance measures for Regent College, L020. ................................................... 62
Table 10: Performance measures for the ventilator in Fred Kaiser 4036. .............................. 65
Table 11: Performance measures for ventilation openings in the Liu Institute. ..................... 67
Table 12: Performance measures for ventilation openings in the C. K. Choi building. ......... 73
Table 13: Performance measures for ventilation openings in the Langara College Library. . 78
Table 14: Selected minimum required air exchange rates [59]. ............................................. 81
Table 15: Assumed ventilation opening operating conditions................................................ 83
Table 16: In-situ measurement results summary. ................................................................... 85
Table 17: BRE ventilator measurement results summary, (xxx) – estimated values. ............ 90
Table 18: Performance measures for the slot ventilation opening........................................ 102
Table 19: Performance measures for the acoustical louver. ................................................. 105
Table 20: Performance measures for the acoustical baffle. .................................................. 109
Table 21: Performance measures for the acoustical air filters. ............................................. 112
Table 22: Performance measures for the 0.3 m CT silencers. .............................................. 118
Table 23: Laboratory measurements results summary. ........................................................ 121
Table 24: Modes in each third-octave band.......................................................................... 157
Table 25: Number of modes in each third octave band, 2D geometry (1 x 1.26 m). ........... 160
Table 26: CT silencer dimensions......................................................................................... 180
Table 27: CT silencer performance....................................................................................... 180
Table 28: Measured and predicted silencer discharge coefficients – high Re...................... 194
Table 29: Predicted flow performance of CT silencers at high Re....................................... 197
Table 30: CT silencer performance prediction summary. .................................................... 202
ix
List of Figures
Figure 1: Typical natural ventilation configuration. ................................................................. 2
Figure 2: Acoustic characterization of ventilation openings. ................................................... 4
Figure 3: Air flow characterization of ventilation openings. .................................................... 4
Figure 4: Variation of SII vocal effort for various room sizes. .............................................. 27
Figure 5: Variation of SII with surface absorption in various room sizes .............................. 28
Figure 6: Variation of SII with increasing surface absorption for various BGN.................... 29
Figure 7: Variation of SII with STC for various vocal efforts (Large Office, Surface
Configuration 4)...................................................................................................................... 30
Figure 8: Variation in SII with EOAs for different wall types................................................ 32
Figure 9: Variation in the normalized SII as a function of EOAs for different wall types. .... 33
Figure 10: Variation in the normalized SII as a function of EOAA_V /EOAA_W, for different
wall types. ............................................................................................................................... 33
Figure 11: Flow meter types ................................................................................................... 43
Figure 12: Flow meter error resulting from using turbulent discharge coefficient................. 44
Figure 13: Discharge coefficient of orifice plates with various geometries ........................... 44
Figure 14: Measured reverberation time with 95% confidence intervals for source and
receiver rooms......................................................................................................................... 56
Figure 15: Measured receiver room total absorption and the maximum permissible value
according to ASTM E90. ........................................................................................................ 57
Figure 16: Uncertainty in measured Lp - 95% confidence values.......................................... 57
Figure 17: Measured transmission loss of laboratory partition - 95% confidence values. ..... 58
Figure 18: Measured equivalent open area of high pressure room (leakage). ........................ 58
Figure 19: Measured SEOAf of 0.6 x 0.6 m opening.............................................................. 59
Figure 20: Regent College L020............................................................................................. 62
Figure 21: Regent College L020 grille. .................................................................................. 62
Figure 22: Fred Kaiser 4036. .................................................................................................. 64
Figure 23: Fred Kaiser 4036 ventilator cross section ............................................................. 64
Figure 24: Liu 216C (showing ventilation openings blocked). .............................................. 68
Figure 25: Liu 308. ................................................................................................................. 68
Figure 26: Liu Institute 308 ventilator cross section .............................................................. 68
x
Figure 27: Liu 313 (showing ventilation openings blocked). ................................................. 69
Figure 28: Liu 313 grille. ........................................................................................................ 69
Figure 29: C. K. Choi 167....................................................................................................... 73
Figure 30: C. K. Choi 321....................................................................................................... 73
Figure 31: C. K. Choi 326....................................................................................................... 74
Figure 32: C. K. Choi 327....................................................................................................... 74
Figure 33: Langara L104. ....................................................................................................... 78
Figure 34: Langara L208. ....................................................................................................... 78
Figure 35: Langara L104 ventilator cross section................................................................... 79
Figure 36: Langara L104 and L208 ventilator grilles............................................................. 79
Figure 37: Langara L208 ventilator cross Section.................................................................. 79
Figure 38: Langara L112 (ventilation opening at top left). .................................................... 80
Figure 39: BRE ventilator Configuration 1. ........................................................................... 91
Figure 40: BRE ventilator Configuration 5. ........................................................................... 92
Figure 41: BRE ventilator Configuration 2. ........................................................................... 93
Figure 42: BRE ventilator Configuration 3. ........................................................................... 94
Figure 43: BRE ventilator Configuration 4. ........................................................................... 95
Figure 44: BRE ventilator Configuration 8. ........................................................................... 96
Figure 45: BRE ventilator Configuration 6. ........................................................................... 97
Figure 46: BRE ventilator Configuration 7. ........................................................................... 98
Figure 47: Slot ventilation opening, no fiberglass. ............................................................... 100
Figure 48: Slot ventilation opening, 1 m x 1 m x 25 mm fiberglass..................................... 100
Figure 49: Slot ventilation opening, 1 m x 1 m x 50 mm fiberglass..................................... 100
Figure 50: Slot ventilation opening, away from floor/ceiling............................................... 100
Figure 51: Measured transmission loss of slot ventilation openings. ................................... 102
Figure 52: Measured transmitted speech spectrum of slot ventilation openings. ................. 103
Figure 53: Louver configuration 1, view from receiver room (perforations not visible). .... 104
Figure 54: Louver configuration 2, view from receiver room (perforations not visible). .... 104
Figure 55: Louver configuration 3, view from receiver room (perforations visible). .......... 104
Figure 56: Louver configuration 4, view from receiver room (perforations visible) ........... 104
Figure 57: Louver configuration 5, view from receiver room (perforations not visible) ..... 104
xi
Figure 58: Measured transmission loss of acoustical louver. ............................................... 106
Figure 59: Measured transmitted speech spectrum of acoustical louver. ............................. 106
Figure 60: Measured (diffuse field receiver) and published (free field receiver) louver
insertion loss. ........................................................................................................................ 107
Figure 61: Measured and published SEOAf of louver. ......................................................... 108
Figure 62: Acoustical baffle.................................................................................................. 109
Figure 63: Measured acoustical baffle transmission loss. .................................................... 110
Figure 64: Measured acoustical baffle transmitted speech spectrum ................................... 110
Figure 65: Acoustic fiberglass .............................................................................................. 111
Figure 66: Pink air filter........................................................................................................ 111
Figure 67: White air filter ..................................................................................................... 111
Figure 68: Measured acoustical air filter transmission loss.................................................. 112
Figure 69: Measured acoustical air filter transmitted speech spectrum................................ 113
Figure 70: Measured acoustical air filter insertion loss. ....................................................... 113
Figure 71: 0.3 m Straight CT silencer diagram..................................................................... 114
Figure 72: 0.3 m Straight CT silencer................................................................................... 115
Figure 73: 0.3 m L-shaped CT silencer diagram .................................................................. 115
Figure 74: 0.3 m L-shaped CT silencer. ............................................................................... 115
Figure 75: 0.3 m Z-shaped CT silencer diagram .................................................................. 116
Figure 76: 0.3 m Z-shaped CT silencer. ............................................................................... 116
Figure 77: 0.3 m Straight CT silencer diagram – fiberglass removed. ................................. 116
Figure 78: 0.3 m Straight CT silencer – fiberglass removed. ............................................... 117
Figure 79: Measured transmission loss of 0.3 m CT silencers. ............................................ 118
Figure 80: Measured transmitted speech spectrum of 0.3 m CT silencers. .......................... 119
Figure 81: Absorption coefficient of OEM fiberglass measured and predicted by the Delaney-
Bazley model for different flow resistivities. σ – [MKS Rayl/m] ....................................... 133
Figure 82: Variation of normal incidence absorption coefficient for various liner thicknesses
as predicted by Delaney-Bazley, σ = 60,000 MKS Rayl/m. ................................................ 133
Figure 83: Silencer dimensions............................................................................................. 134
Figure 84: Predicted transmission loss for various duct heights and liner thicknesses ........ 135
xii
Figure 85: Predicted transmission loss for various duct heights and liner thicknesses with
transmission loss plotted on a log scale ................................................................................ 136
Figure 86: Predicted transmission loss for various aspect ratios: dy=dx=25 mm................. 137
Figure 87: Predicted transmission loss for various aspect ratios: dy=dx=50 mm................. 138
Figure 88: Predicted transmission loss for various aspect ratios: dy=dx=100 mm............... 138
Figure 89: 3D acoustic domain. ............................................................................................ 143
Figure 90: Source volume sub-domain. ................................................................................ 143
Figure 91: Source volume sampling sub-domain. ................................................................ 143
Figure 92: Ventilator flow path sub-domain......................................................................... 143
Figure 93: Ventilator liner sub-domain................................................................................. 144
Figure 94: Receiver volume sub-domain.............................................................................. 144
Figure 95: PML sub-domain................................................................................................. 144
Figure 96: Point sound power source.................................................................................... 144
Figure 97: Variation of the predicted sound power with mesh resolution............................ 152
Figure 98: Error in average Lp as a function of frequency sampling resolution for various
source volume dimensions and reflection coefficients. ........................................................ 154
Figure 99: Lp Standard deviation as a function of frequency sampling resolution for various
source volume dimensions and reflection coefficients. ........................................................ 155
Figure 100: Standard deviation of Lp vs minimum order of axial mode.............................. 159
Figure 101: External view of source volume with diffusing surfaces. ................................. 159
Figure 102: Standard deviation of sound pressure in an empty office sized room, with and
without diffusers ................................................................................................................... 160
Figure 103: Third-octave source volume surface Lp distribution. N=6, color scale in dB... 162
Figure 104: Sample area on source volume surface area...................................................... 162
Figure 105: Standard deviation of sample Lp averages vs. to sample side-length to
wavelength ratio for various source volume dimensions. Square sample surface, away from
source volume corner. ........................................................................................................... 164
Figure 106: Standard deviation of sample Lp averages vs. sample width to wavelength ratio
for various sample height to wavelength ratios. N=6, sample surface away from source
volume corner. ...................................................................................................................... 164
xiii
Figure 107: Standard deviation of average sample Lp vs. sample width to wavelength ratio
(w/l) for various sample height to wavelength ratios (h/l). Sample surface at source volume
edge. ...................................................................................................................................... 166
Figure 108: 3 m lined duct geometry for analytical validation with 503 Hz Lp solution..... 167
Figure 109: Transmission loss in center section of duct – discrete data points.................... 169
Figure 110: Analytical solution and 2D FEM solution with locally reacting liner – discrete
data points. ............................................................................................................................ 169
Figure 111: Predicted transmission loss calculated using source sound pressure level and inlet
intensity................................................................................................................................. 172
Figure 112: Difference in transmission loss result when calculated using source sound
pressure level and inlet intensity. 95% confidence intervals provided................................. 172
Figure 113: Measured and predicted TL of a 50 mm slot. 95% confidence intervals.......... 174
Figure 114: Measured and predicted TL of 0.3 m straight CT silencer. 95% confidence
intervals................................................................................................................................. 175
Figure 115: Measured and predicted TL of 0.3 m L-shaped CT silencer. 95% confidence
intervals................................................................................................................................. 176
Figure 116: Measured and predicted TL of 0.3 m Z-shaped CT silencer. 95% confidence
intervals................................................................................................................................. 176
Figure 117: Shapes and dimensions of CT silencers. Clockwise from top left: Straight, L, Z,
and U..................................................................................................................................... 180
Figure 118: Predicted transmission loss of 1 m CT silencers. 95% confidence intervals
shown. ................................................................................................................................... 183
Figure 119: Predicted transmitted speech spectrum, 1 m CT silencers. 95% confidence
intervals given....................................................................................................................... 183
Figure 120: Predicted Lp (left) and sound pressure (right) at 500 Hz, 1 m Straight silencer.
............................................................................................................................................... 184
Figure 121: Predicted Lp (left) and sound pressure (right) at 500 Hz, 1 m L-shaped silencer.
............................................................................................................................................... 184
Figure 122: Predicted Lp (left) and sound pressure (right) at 2500 Hz, 1 m Straight silencer.
............................................................................................................................................... 185
xiv
Figure 123: Predicted Lp (left) and sound pressure (right) at 2500 Hz, 1 m L-shaped silencer.
............................................................................................................................................... 185
Figure 124: Predicted Lp (left) and sound pressure (right) at 8000 Hz, 1 m Straight silencer.
............................................................................................................................................... 186
Figure 125: Predicted Lp (left) and sound pressure (right) at 8000 Hz, 1 m L-shaped silencer.
............................................................................................................................................... 186
Figure 126: Flow domain of ventilator model (sample velocity contours shown). .............. 189
Figure 127: Published experimental results for the orifice plate discharge coefficient........ 193
Figure 128: Predicted results for the orifice plate discharge coefficient. ............................. 193
Figure 129: Predicted Straight CT silencer discharge coefficients as a function of Re. ...... 198
Figure 130: Predicted flow velocity in 1 m Straight CT silencer – Re = 27,000. ................ 198
Figure 131: Predicted flow pressure in 1 m Straight CT silencer – Re = 27,000. ................ 198
Figure 132: Predicted flow velocity in 0.05 m Straight CT silencer – Re = 22,000. ........... 199
Figure 133: Predicted flow pressure in 0.05 m Straight CT silencer – Re = 22,000. ........... 199
Figure 134: Predicted discharge coefficients of 0.3 m CT silencers as a function of Re. .... 199
Figure 135: Predicted discharge coefficients of 0.5 m CT silencers as a function of Re. .... 200
Figure 136: Predicted discharge coefficients of 1 m CT silencers as a function of Re. ....... 200
Figure 137: Predicted discharge coefficients as a function of pressure loss for all silencer
configurations. ...................................................................................................................... 201
Figure 138: Predicted open area ratio as a function of pressure loss for all silencer
configurations. ...................................................................................................................... 203
Figure 139: Predicted open area ratio as a function of Reynolds number for all silencer
configurations. ...................................................................................................................... 204
Figure 140: Predicted transmission loss of 0.3 m CT silencers. 95% confidence intervals
shown. ................................................................................................................................... 245
Figure 141: Predicted transmission loss of 0.5 m CT silencers. 95% confidence intervals
shown. ................................................................................................................................... 245
Figure 142: Predicted transmission loss of 1 m CT silencers. 95% confidence intervals
shown. ................................................................................................................................... 246
Figure 143: Predicted flow velocity in 0.05 m Straight CT silencer – Re = 22,000. ........... 246
Figure 144: Predicted flow velocity in 0.3 m Straight CT silencer – Re = 25,000. ............. 246
xv
Figure 145: Predicted flow velocity in 0.5 m Straight CT silencer – Re = 25,000. ............. 247
Figure 146: Predicted flow velocity in 1 m Straight CT silencer – Re = 27,000. ................ 247
Figure 147: Predicted flow pressure in 0.05 m Straight CT silencer – Re = 22,000. ........... 247
Figure 148: Predicted flow pressure in 0.3 m Straight CT silencer – Re = 25,000. ............. 247
Figure 149: Predicted flow pressure in 0.5 m Straight CT silencer – Re = 25,000. ............. 248
Figure 150: Predicted flow pressure in 0.5 m Straight CT silencer – Re = 27,000. ............. 248
Figure 151: Predicted flow velocity in 0.3 m L-shaped CT silencer – Re = 15,000. ........... 249
Figure 152: Predicted flow velocity in 0.5 m L-shaped CT silencer – Re = 15,000. ........... 249
Figure 153: Predicted flow velocity in 1 m L-shaped CT silencer – Re = 18,000. .............. 249
Figure 154: Predicted flow pressure in 0.3 m L-shaped CT silencer – Re = 15,000............ 250
Figure 155: Predicted flow pressure in 0.5 m L-shaped CT silencer – Re = 15,000............ 250
Figure 156: Predicted flow pressure in 1 m L-shaped CT silencer – Re = 18,000............... 250
Figure 157: Predicted flow velocity in 0.5 m U-shaped CT silencer – Re = 20,000............ 251
Figure 158: Predicted flow velocity in 1 m U-shaped CT silencer – Re = 21,000............... 251
Figure 159: Predicted flow pressure in 0.5 m U-shaped CT silencer – Re = 20,000. .......... 252
Figure 160: Predicted flow pressure in 1 m U-shaped CT silencer – Re = 21,000. ............. 252
Figure 161: Predicted flow velocity in 0.3 m Z-shaped CT silencer – Re = 15,000. ........... 253
Figure 162: Predicted flow velocity in 0.5 m Z-shaped CT silencer – Re = 12,000. ........... 253
Figure 163: Predicted flow velocity in 1 m Z-shaped CT silencer – Re = 14,000. .............. 253
Figure 164: Predicted flow pressure in 0.3 m Z-shaped CT silencer – Re = 15,000............ 254
Figure 165: Predicted flow pressure in 0.5 m Z-shaped CT silencer – Re = 12,000............ 254
Figure 166: Predicted flow velocity in 1 m Z-shaped CT silencer – Re = 14,000. .............. 254
xvi
List of Symbols and Abbreviations
ε Porosity
f Frequency [Hz]
k Wavenumber, or flow-pressure loss coefficient
µ Dynamic viscosity
ν Kinematic viscosity
p Pressure (acoustic, or flow)
ρ Density of air
ρ0 Density of air at STP
A Surface area
Aα Total absorptive area
AR Aspect ratio
Cd Discharge coefficient
Dh Hydraulic diameter
EOAf Equivalent open area for air flow
EOAs Equivalent open area for sound
I Sound intensity
IL Insertion Loss
Leq Equivalent sound pressure level
Lp Sound pressure level
xvii
Lw Sound power level
NR Noise reduction
OAR Open area ratio
Q Air flow rate
R Room constant
Re Reynolds number
Retr Transitional Reynolds number – the Reynolds number above which the
discharge coefficient is constant with respect to Reynolds number. This may
not correspond to transition to turbulence.
SEOAf Specific equivalent open area for air flow
SEOAs Specific equivalent open area for sound
SII Speech intelligibility index
STC Sound transmission class
T60 Reverberation time
TL Transmission loss
U Average flow velocity
V Room volume
W Sound power
Z Propagation impedance
Zs Surface impedance
xviii
Acknowledgements
There are many people to whom I owe my gratitude for their contribution to this work.
Thank you to my supervisor, Dr. Murray Hodgson, who has helped guide this research,
provided technical support, and assisted with measurements throughout all stages of this
project and my M.A.Sc. program. I am very grateful to have had such a knowledgeable,
caring, and engaged supervisor.
Thank you to the Natural Sciences and Engineering Research Council of Canada for the
financial support by way of a NSERC CGS-M scholarship and a NSERC CREATE grant.
Thank you to Fitsum Tariku, Steve Rogak, Sheldon Green, Max Richter, Andrea Frisque,
Greg Johnson, Albert Bicol, Alireza Khaleghi, and many other people for your technical
guidance and advice. Additionally, thank you to Fitsum Tariku and BCIT for loaning us their
blower door which was essential for all of our airflow measurements, and to Sheldon Green
for providing us with ANSYS Fluent which was essential for airflow modeling.
Thank you to Richard Anthony and Mehrzad Salkhordeh of Kinetics Noise Control Inc. for
your interest and support toward this project. It is unfortunate that we were not able to
collaborate further.
Thank you to the many building managers who spent time to show and discuss their
buildings, and provided us with assess to complete our measurements.
Thank you to my wife, family and friends for the financial and emotional support, as well as
the late nights helping with my measurements.
xix
Chapter 1: Introduction
In developed nations buildings are responsible for 20-40% of all energy consumption; in
Canada this number is 30% [1]. In order to reduce our national energy consumption, building
scientists and engineers must develop methods to reduce the energy used by buildings. The
use of natural ventilation, as opposed to mechanical ventilation, is one such method. In
addition to a reduction in energy consumption, natural ventilation can provide a host of other
occupant health and satisfaction benefits, including reduced mechanical noise and sick
building syndrome, if implemented properly [2].
Natural ventilation works by using wind or buoyancy (stack effect) induced pressure
differentials to drive the ventilation air through a building [3]. Typically these pressures are
small compared to those available in a mechanically ventilated building, often not exceeding
10 Pa [3]. In order for the small pressure to drive a sufficient volume of air it is necessary to
have low resistance to air flow throughout the building [4, 5]; to achieve this, large openings
are often created in partitions, allowing air to flow from one room to the adjacent space. The
large openings prove detrimental to the noise isolation between the spaces [3-8]. Figure 1
shows a typical configuration for a naturally ventilated building. In this example, the natural
ventilation openings between the offices and corridors could result in reduced privacy of
conversations inside an otherwise private office, as well as increased annoyance for the
occupants of the office due to noise generated in the corridor.
1
Corridor
Office
The UBC Acoustics and Noise Research Group at UBC has had previous involvement with
ventilation openings in naturally ventilated buildings, which came about due to occupant
complaints in then-new buildings on UBC campus; the Liu Institute for Global Studies [5],
the C.K. Choi building, the AERL building[9], and Regent College. The complaints
surrounded a lack of noise isolation, causing either a lack of privacy or increased distraction,
between the offices and the corridor. In the Liu Institute for Global Studies project the UBC
Noise and Research Group, in combination with Stantec Architecture Inc., designed silencing
devices to increase the noise isolation to an acceptable level without overly effecting airflow;
the devices proved successful and the building occupants claim to be satisfied with the
resulting noise isolation.
It was clear that there is a great need for further understanding of ventilation openings in
naturally ventilated buildings to provide engineers and architects with design techniques
which allow sufficient air flow while providing adequate noise isolation. Under the direction
and supervision of Murray Hodgson, I have continued this work.
2
In order to provide a vocabulary for discussing ventilation openings, the general descriptors
of the acoustic and airflow performance of a ventilation opening are provided here. Further
discussion about this characterization method appears in many areas throughout this thesis.
The transmission loss of a ventilation opening affects the difference in sound levels between
the two rooms that it separates – the noise reduction (NR). Assuming the sound fields in both
rooms are diffuse, a concept which will be discussed in detail later (sections 2.3, 3.2, 5.1.1,
and 6.2.1.3), the difference in sound levels is found to be:
A
NR = Lp s − Lp r = TL − 10 log v (2)
Aα , r
With reference to Figure 2, Lps is the sound pressure level in the source room containing the
sound source, Lpr is the sound pressure level in the receiver room, Aα, r is the total receiver
room acoustic absorption, Av is the cross sectional area of the ventilator, and TL is the
transmission loss of the ventilator. Sound generated in the source room enters, propagates
through, and is emitted from, the ventilation opening resulting in a reverberant sound field
with some sound pressure level in the receiver room. Here it is assumed that the ventilation
opening is the only path for sound to propagate from the source to the receiver room. Eq. (2)
shows that the ventilator’s transmission loss is a key factor in the noise isolation between the
two rooms, but not the only factor. The reverberant conditions in the receiver room and other
noise propagation paths are examples of other factors.
3
Source Room Receiver Room
Lps Lpr, Aα, r
Ventilator
Av, TL
2(Ps − Pr )
0.5
Q = Av C d (3)
ρ
The form of Eq. (3) is a result of its empirical derivation from conservation of energy in a
flow along a streamline. It states that the flow rate through the ventilator (Q) is directly
proportional to the ventilator area (Av) and the pressure differential squared, and inversely
proportional to the fluid density (ρ). Ps and Pr are the static pressures in the source and
receiver rooms. The discharge coefficient is determined experimentally and is constant over
suitable flows; for a thin orifice it takes a value of around 0.61. There will be discussion later
in the thesis about the suitability and implications of using a constant value for Cd (sections
4.4.2, 5.3.6, and 6.3.7.3).
Ventilator
A v, C d , Q
4
provide a theoretical discussion of the physics of acoustics and airflow through openings –
that will come later.
Oldham and de Salis [3, 4] provide a discussion of concerns associated with placing natural
ventilation openings in a façade, based on the theoretical framework implied by Eq. (2) and
Eq. (3). This discussion is also relevant to interior partitions. They discuss the effect of an
aperture on the transmission loss of a partition, by finding the effective transmission loss of a
partition from the area-weighted average of the wall and aperture transmission coefficients:
τ ⋅ A + τ a ⋅ Aa
TL w+ a = 10 log w w
Aw + Aa
Subscripts w and a indicate the wall and aperture. Their theoretical analysis concluded that
the creation of a ventilation aperture in a typical façade that is large enough to provide
effective ventilation rates would be detrimental to the noise isolation provided by the façade.
Conclusions are made that natural ventilation openings must be treated acoustically; that is to
say, τ a , the transmission coefficient of the ventilation opening must be reduced. Oldham and
de Salis go on to discuss the applicability of various noise control solutions in the context of
traffic noise and apertures in a façade [3, 4]. It is suggested that, to acoustically silence
ventilation apertures:
• active noise control can be effective at very low frequencies; however, it has poor
performance at higher frequencies and for non-steady sound
• resonator absorber devices can be effective at mid frequencies
• porous, absorptive linings can be effective at higher frequencies
• hybrid solutions may be most effective
5
width was much larger than the height. Results concluded that the insertion loss increased
with aperture length and decreased with aperture height; aperture width had nearly no effect.
The effective free area of the apertures was very nearly equal to the actual aperture area in all
cases, provided the height remained above 40 mm; for heights less than 40 mm the free area
is reduced – no explanations are given. Very few details were given on the FEM and CFD
modeling techniques used in this work.
Previous work completed by the UBC Acoustics and Noise Research Group, in collaboration
with Stantec Architecture Inc., [5] investigated complaints of poor noise isolation in a
naturally ventilated building, identified the transmission issues, designed and installed
silencing devices, and completed follow-up measurements. The noise isolation problems
were determined to be associated with natural ventilation openings; one type was in the
partition between offices and a corridor; the other type comprised a vertical shaft between
floors, with inlets on each floor. A ray-tracing acoustical model was implemented to design a
silencer for the vertical shafts consisting of either a lining, or vertical splitter-type baffles. It
is not known what methods were used to specify the design criteria for flow performance;
however, guidelines were given for minimum flow path dimensions. A Z-shaped cross-talk
silencer was installed in a partition between an office and the corridor. Post-installation
measurements confirmed that the vertical shaft silencer increased the noise isolation of the
spaces to an acceptable level. The addition of the Z-shaped cross-talk silencer to the partition
between the office and corridor did not sufficiently increase noise isolation.
6
Nunes of MACH Acoustics has designed a unique silencer that is essentially a porous
absorber material in the shape of an extruded lattice [13, 14]. It is claimed to have excellent
acoustic performance; unfortunately no design details are provided. A methodology is also
proposed for measuring ventilation openings in situ, recognizing the problems associated
with differentiating the acoustic transmission through the ventilation opening and the
transmission through the remainder of the partition [14]. The approach taken is to
characterize the acoustic performance of the partition when the ventilator is in its normal
state, and when it is blocked acoustically. The difference between the two states can be used
to determine the transmission through the ventilation opening alone1. Nunes stated that CFD
simulations should be used to determine the flow performance of the ventilation silencers;
however, no methodology or results are given [13, 14].
A great amount of literature exists on the airflow performance of natural ventilation systems;
however, the bulk of it focuses on either buoyancy or wind driven flows through external
building surfaces. Research that considers interior cross-ventilation flows commonly
describes the flow, according to Eq. (3), by stating a discharge coefficient and opening area,
or equivalents such as the effective free area [3, 4, 10-12, 15-17]. In practice, as well as in
research, ventilation openings in naturally ventilated buildings are specified using either the
discharge coefficient and area, or effective free area [18-20].
Very little work, from either the natural ventilation research or design communities, focuses
on designing interior ventilation openings for a large discharge coefficient (i.e. low pressure
drop given a fixed opening dimension). It is the author’s opinion that this lack of attention is
because, when designing an entire natural ventilation system, interior ventilation openings
are relatively well understood and predictably behaved. Ventilation system design factors
such as wind and buoyancy based driving mechanisms are much more variable, less well-
understood and more challenging to research. As a result they are generally a larger concern
in the design process. To address the noise transmission through ventilation openings,
1
This work was published after the research presented in this thesis was conducted; however, a similar
approach was used for the in-situ measurements presented here.
7
however, it is very important to have a strong understanding of, and optimize for, airflow
behavior through these openings because the design of an optimal ventilation silencer will
require compromise between acoustic and airflow performance.
Recognizing the inability to provide a single optimized design, this work is based on the
following objectives:
1. Propose an optimization metric for the performance of an interior natural ventilation
opening that considers both acoustics and airflow.
2. Provide an understanding of the factors that affect speech privacy in a naturally
ventilated building.
3. Develop methods for measuring natural ventilation opening performance.
4. Develop methods for predicting natural ventilation opening performance.
5. Provide performance results for natural ventilation opening silencers.
6. Provide best practice guidelines for designing successful interior natural ventilation
openings.
8
Chapter 3 investigates how relevant factors, including the natural ventilation opening, affect
speech privacy between two spaces separated by a partition containing a natural ventilation
opening.
Chapter 4 examines discusses how air flow is analyzed and develops known flow
performance metrics. A discussion of the validity of using a single number discharge
coefficient to specify flow performance, as described by Eq. 3, is provided.
Chapter 5 presents a preliminary study of ventilation silencers, both in-situ, and in a custom-
built lab facility. The intent is to gain an understanding of the general performance
characteristics of various types of ventilation silencers. These measurements allow us to
establish which of the current silencer designs have the best performance, which of the
current silencer designs have the worst performance, as well as the common constraints or
motivating factors that would lead to the final selection of a silencer design.
9
Chapter 2: Optimization Metric: Open Area Ratio
In order to optimize the performance of a silencer a metric must exist to optimize over.
Typically a silencer’s acoustic performance would be characterized by its transmission loss, a
frequency dependent value, and its flow performance characterized by the flow rate through
it as a function of pressure. Here it is proposed to use “equivalent open areas” for the acoustic
transmission and air flow to characterize the performance of the ventilator at a specific
operating condition. The ratio of the equivalent open areas – the open area ratio (OAR) –
provides a single number metric to optimize over by maximizing:
EOA f
OAR = (4)
EOA s
EOAf is the equivalent flow area and EOAs is the equivalent acoustic area.
OAR is independent of the size of the opening to the extent that the sound energy
transmission and ventilation flow rate are directly related to the size of the opening.
The reader and/or user should keep in mind that this is a performance optimization. Other
factors such as cost, dimensions, and aesthetics will be crucial in the final design and
selection of a silencer, and are not considered here. Once the OAR is defined, this section
provides a discussion of the implications of using diffuse field room-acoustic theory to
describe ventilation opening performance.
10
2
p
Is = s (5)
4ρ 0 c
The sound power transmitted through a ventilation opening in the wall of a room ( Wt ) is
equivalent to the sound intensity travelling in one direction through a plane in the room,
multiplied by the area of the ventilation opening ( Av ):
Wt = I s Av (6)
If the sound energy transmitted through the opening is radiated into a receiver room with a
diffuse sound field, and the steady-state sound pressure in the receiver room is measured
along with the reverberation time in the room, it is possible to calculate the sound power
being transmitted into it ( Wt ):
2
A p
Wt = r r (7)
4 ρ 0c
Ar , the total acoustic absorption in the receiver room can be determined from the measured
reverberation time ( T60 ) and receiver room volume ( Vr ):
0.161Vr
Ar = (8)
T60
From the sound pressure level in the source and receiver rooms, and the receiver room
reverberation time, work backwards and deduce how large the opening must be.
Wt
Av =
I1
Now, assume that there is a treatment applied to the opening such that some of the sound
energy transmitted into the opening is dissipated prior to being emitted into the receiver
room. Based on the difference in sound pressure levels between the source and receiver
rooms, and the reverberation time, again calculate how big the opening between the rooms
must be; however, this time the area will be smaller than the actual size of the opening. The
area that is calculated is the size of an untreated opening that would transmit an equivalent
amount of sound energy. This is called the equivalent open area for sound ( EOA s ).
11
The equivalent open area of the ventilator is simply the product of the ventilator’s power
transmission coefficient ( τ v ) and the corresponding ventilator area (Av). This can be shown
through an equivalence of power transmission:
W2 = I 1τ v Av = I 1EOA s
EOA s = τ v Av (9)
In reality, acoustic fields in rooms are not generally diffuse and, certainly, sound through
small openings cannot be predicted accurately using diffuse field theory (see section 2.3).
The equivalent acoustic open area is the equivalent open area based on a diffuse field sound
transmission model and therefore will not necessarily be physically meaningful. That being
said, it is a useful measure, can be easily visualized, and can be contrasted to equivalent flow
area.
through frequency averaging the weighted EOA s at each frequency band. Weighting is done
by considering the spectral shape of the source sound power and the spectral shape of the
receiver sensitivity. In essence, the weighting makes the EOA s particularly sensitive to
frequencies that are likely to be produced at a high level by the source as well as frequencies
that receiver is sensitive to.
To choose a source spectrum weighting scheme, it is first assumed that the silencer will be
mounted in an office partition to another interior space. Noises such as computer fans are
constant and contain no information, so they are not considered as important in terms of
privacy (sounds exiting the room) or annoyance (sounds entering the room). The most
significant source of annoying noises to be controlled to acceptable levels is speech;
12
therefore speech levels have been used to generate the source spectra. Supporting this
conclusion, previous work has shown that irrelevant speech is more annoying than broadband
noise [22]. Third octave band speech pressure level spectrum, combined with the directivity
of a human speaker, at a normal speaking level were averaged to create the source power
level spectrum. Speech levels are generated by converting the ‘Normal’ Speech Spectrum
Level given by ANSI S3.5 into third octave band values using the band width ∆f [23]:
Speech Level = Speech Spectrum Level + 10 log(∆f )
This conversion corrects for the increasing band width, and therefore increasing band power,
with increasing frequency. The receiver spectrum is most simply the common A-weighting
which approximates the auditory sensitivity of an average adult with normal hearing.
However, A-weighting is only one approach. If it were desirable to find a metric specifically
to correlate with a lack of speech privacy, then a weighting scheme based on the importance
of each band for understanding speech would need to be determined and used (SII/STI
methods incorporate such a weighting but it is not readily applicable). If it were desired to
produce a metric related to distraction, which is not necessarily related to intelligibility, then
a different frequency weighting scheme would need to be developed. Research has shown
that distraction may be related to temporal variability, making it hard to correlate it with a
frequency spectrum alone [24]. A listener’s auditory sensitivity is related to both
intelligibility and distraction; therefore, the A-weighting was used as a good general
weighting scheme.
A-weighting is applied to the speech source levels (decibel addition) to give the A-weighted
speech source levels. Prior to using the A-weighted speech source levels as a weighting
spectrum they must first be normalized so that its sum is equal to zero dB. Subtracting the
ventilators transmission loss spectrum from the, normalized, A-weighted speech levels
produces a spectral shape that represents the levels that a human would hear on the receiver
side of the partition, due to transmission of speech sounds through the ventilator, reverberant
field not included. The normalized A-weighted speech levels minus the silencer transmission
loss will be referred to in this work as simply the Transmitted Speech Spectrum
( − TL A_Speech ). Peaks in this transmitted speech spectrum represent the most audible sounds,
13
and thus they are the frequencies that should be targeted for silencer improvement. Summing
all of the levels produces an effective, single value, transmission coefficient for an A-
weighted speech source. It is now possible to convert the effective transmission loss into a
single number equivalent open area to characterize the silencer:
− TL A _ speech
In North America, the most common method for assessing the capability of a partition to
provide speech privacy is the Sound Transmission Class (STC) [25]. Research has shown
the STC not to correlate well with intelligibility [26]; therefore, there was little motivation to
use it in this method. Additionally, the STC is not, by definition, the decibel representation of
a transmission coefficient; therefore, it can’t be used to directly calculate an equivalent open
area.
Conservation of energy for lossless, horizontal, streamline flow through a constriction can be
written, according to Bernoulli’s equation (conservation of energy), as:
1 1
p1 + ρ u1 2 = p 2 + ρ u 2 2
2 2
If the velocity in the constriction is much larger than the velocity upstream of it ( u 2 >> u1 ),
then we can deduce:
0.5
2∆p
Q = A (11)
ρ
The orifice area is A, and ∆p is the difference in pressure. Combining Eq. 3 and 11,
assuming identical flow rates and pressures, an effective flow area can be solved for. This is
identical to the ‘effective leakage area’ (Af) defined by ASTM E779-10 [27].
14
ρ
A f (∆p ) = C ∆p n −0.5 (12)
2
The effective leakage area can be normalized by the discharge coefficient of an opening in a
flat plate (Cd = 0.61) [12], giving the equivalent open area in a flat plate required to produce
the same flow rate as the ventilation opening in question:
A f (∆p )
EOA f (∆p ) = (13)
0.61
Provided n = 0.5, there is a simple relationship between the equivalent open area for flow and
the discharge coefficient.
Cd
EOA f = Av (14)
0.61
Note that, for n ≠ 0.5 , the equivalent open area is dependent on the applied pressure. For this
reason, the equivalent open area should be measured and calculated at a pressure that is
typical of the ventilator’s operating conditions. Unfortunately, typical operating pressures for
interior natural ventilation openings are small compared to the pressures during testing as
described by ASTM E779-10 [27]. Additionally, it is not possible to accurately extrapolate
below the measured data points because the data measured at high flow rates is not
necessarily an accurate predictor of the ventilator’s performance at low flow rates. Section
4.4.2 of this thesis discusses the implication of flow conditions on ventilator performance
characterization; section 5.3.6 discusses what the actual flow conditions are in natural
ventilation openings. In this work the lowest measurement pressure was used as the reference
pressure in calculating EOAf.
15
It is useful to introduce specific equivalent open areas for sound and flow to get non-
dimensional performance metrics that are nominally independent of the ventilator size. The
specific equivalent open area for sound is akin to the transmission coefficient, and the
specific equivalent open area for flow is akin to the discharge coefficient.
EOA s
SEOA s = 15
Av
EOA f
SEOA f = 16
Av
When using these performance metrics care needs to be taken regarding the assumptions
being made, the most significant of which are that the sound fields are diffuse, and the
equivalent open area for flow is independent of flow rate.
The OAR optimization parameter recognizes that the airflow and acoustic performance can
not be optimized independently; however, it also assumes that their physics are not coupled.
In fact, the acoustic and airflow behavior must be coupled as they are both the response to
pressure and particle velocity fluctuations in air. The two ways that acoustic-flow coupling
can be significant in ducts are flow-generated noise, and changes in the speed of sound;
however, it can be argued that in natural ventilation openings these effects are insignificant.
Flow-generated noise is caused by turbulent flow interacting with duct surfaces [28]. It has
been demonstrated analytically and experimentally that below the cut-off frequency the
sound power level generated by flow in a duct is proportional to the flow velocity to the
fourth power, and above the cut-off frequency the sound power level generated is
proportional to the flow velocity to the sixth power [28]. Because the flow velocities in
naturally ventilated buildings are much lower than in a mechanically ventilated building, the
flow noise will be much less. Even if the ventilation opening did generate flow noise, it is not
necessarily a problem. Many of the speech privacy related problems in naturally ventilated
buildings are understood to be a result of low background masking noise due to the lack of a
mechanical ventilation system [29, 30]; this is exemplified in section 3.3.3.
16
Variations in sound speed are typically due to temperature gradients or flow; in a duct
temperature can be assumed uniform. Mean flow changes the speed of sound relative to the
duct, and flow gradients in the duct cross section cause refraction; mean flow and refraction
can result in either higher or lower attenuation depending on the direction of sound
propagation with respect to flow. In outdoor sound propagation, where sound travels great
distances and slight variations in speed result in a large displacement of the wave front,
refraction and mean flow effects are significant, even for low Mach number (M) flows. In
ducts however, because the distances travelled are much smaller, refraction effects are small
provided M<0.05 [21, 31].
A diffuse sound field is one that is uniform in magnitude and direction at all locations;
additionally, the temporal and spatial relationships of the pressure modes in the room must be
random [32]. When rooms are very large, very small, have uneven absorption distributions,
large amounts of acoustically absorbing material, or large interior dimension aspect ratios,
the sound field will not be diffuse. Measurements must be made in all types of spaces, so
acousticians measure the standard deviation of sound pressure levels in the sound field
which, because a highly diffuse sound field will have a low standard deviation, is used to
imply a level of confidence in the result. Sound pressure levels are unable to give directional
information about the sound field; therefore the sound pressure level standard deviation is
only an indicator of the diffuseness, it does not tell the complete story. Diffuse field theory is
used exclusively in practice; experience is required to judge the diffuseness of a field and its
significance to the calculated result.
17
In the case of natural ventilation openings, even if the room supports an excellent
approximation of a diffuse field, diffuse field theory is incomplete and wave effects become
important. To determine the sound power entering a ventilation opening, two wave effects
that warrant discussion here are constructive interference and diffraction.
In a diffuse field, the phases of various modes are random; therefore, the time-average
pressure at any location can be found through energy addition of the modes. At room
boundaries, due to the boundary condition of zero particle velocity normal to the wall, the
spatial phase relationship of the modes is not random; they all have pressure maximums at
the surface and will interfere constructively [33]. At a wall the sound pressure level is 2.2 dB
higher than in the diffuse field; in double and triple corners the pressure level is further
increased [34]. The sound pressure level is within 1 dB of the room average value if
measured more than a quarter of the acoustic wavelength from the wall. This shows that the
sound field at the wall is, by definition, never diffuse. The implication is that the sound
pressure level that the ventilation opening is exposed to is higher if it is on a room boundary,
especially if it is at a double or triple corner.
Diffraction will cause additional energy to enter the duct. The surface adjacent to the duct is
typically hard and will therefore be the location of a pressure maximum (anti-node). At the
duct inlet surface there is, to a first approximation, no impedance change and no pressure
mode. Air adjacent to the duct will follow the pressure gradient and progress toward the duct.
The effect of diffraction will increase as the ratio of the wavelength to ventilator inlet
dimension increases.
Even if the room containing a ventilator has a highly diffuse sound field, it should not be
expected that diffuse field theory will accurately predict the sound energy entering the
ventilation opening. If the ventilation opening is near a room surface, or corner, it will be
exposed to an elevated sound level caused by the constructive sound field. If the ventilation
opening is not large compared to the wavelength, significantly more energy than predicted
will diffract into the duct inlet. As a result, even in a room with a diffuse sound field, the
location of the ventilation opening will affect the sound transmitted through it (and thus its
18
transmission loss). Additionally, because more energy may enter the ventilation opening than
predicted by diffuse field theory, we must allow measures of TL and STC to be negative, and
the SEOAs to take a value greater than one.
19
Chapter 3: Investigating the Factors in Speech Privacy
In order to understand how ventilation openings affect speech privacy between two interior
spaces a simple model has been created. This model considers two adjacent rooms with a
separating wall containing a ventilation opening. One of the rooms, the source room, contains
a talker, and the privacy is assessed for a listener in the other room, the receiver room.
Speech privacy is determined by the lack of speech intelligibility, as quantified by the speech
intelligibility index (SII). Using this model the effects that ventilation openings, partition
transmission loss, room size, room furnishings, background noise levels, and speech power
levels have on privacy can be predicted.
20
3.2 Model Description
In order to determine the SII at the receiver it is necessary to know the sound levels, as well
as the reverberation time and background noise levels, at the receiver.
Speech levels at the receiver are calculated starting in the source room with the speech source
sound power level. Knowing the source room’s dimensions and surface materials, the sound
pressure level in the source room can be determined. The transmission loss of the composite
partition can be calculated and used to determine the sound power transmitted into the
receiver room based on the source pressure level if the area and transmission loss of both the
wall and the ventilation opening are known. Using the sound power transmitted into the
receiver room, and also knowing its size and surface materials, the sound pressure level at the
receiver can finally be determined.
The reverberation time in the receiver room can be determined based on its dimensions and
surface materials.
When calculating the SII, typically one would want to use the background level most
representative of a typical office scenario; as such it is sensible to use mean, time averaged,
background levels. Here, however, the SII is being calculated for a non-typical application;
that is, to measure levels of distraction or speech privacy (the presence of unwanted
intelligibility). In an environment where one is talking on the phone, or even typing with
confidence, the background noise levels will be at an elevated level as compared to when
involved in a quieter activity such as thinking. If mean sound pressure levels were
determined in a room where the noise levels were fluctuating, the dB scale naturally biases
the durations of high levels, and the durations of low background noise levels are concealed.
By concealing the durations of low noise levels, the most likely periods of distraction or
unwanted speech intelligibility are also concealed. In order to provide a background level
that is representative of the quieter periods in the receiving room, statistical pressure levels
can be used. L90 is the level of sound that is exceeded 90% of the time, making it more
appropriate than Leq as a background level for SII measurements intended to infer privacy or
distraction. In this work, the L90 has been measured in a typical office scenario.
21
3.2.1 Model Inputs
To investigate how room and partition design effect the SII, appropriate speech levels,
background noise levels, partition types, room surface materials and room dimensions must
be specified.
The background noise level, L90 in a quiet office, is given in Table 1; the speech power levels
in dB for several vocal efforts are given in Table 2 [23]. Unless otherwise noted, the
background noise levels as stated in Table 1 will be used; however, elevated background
noise levels are created by adding a constant decibel value to each third octave band pressure
level shown in Table 2. The levels in Table 2 correspond to NC20; however, the room’s
actual NC, as calculated from Leq instead of L90, would be higher.
Table 3 shows the surface materials used and their random incidence absorption coefficients
[41]. These materials are combined to create five different room configurations, as shown in
Table 4, named 1 through 5 and increasing in average absorption.
Dimensions representative of a typical small office, large office, classroom and small
gymnasium are given in Table 5.
The transmission loss and associated STC of four different partition types are shown in Table
6 [36]. The partition types are:
• 10 mm glass
• Two layers of drywall separated by a wooden stud
• Two double layers of drywall separated by a wooden stud
• Two double layers of drywall on resilient hangars separated by a wooden stud
22
Table 2: Speech power levels [dB] [23].
Frequency [Hz]
Speech effort 125 250 500 1000 2000 4000 8000
Casual 50 59 61 55 50 45 40
Normal 58 67 69 61 56 50 44
Raised 62 71 75 70 64 57 48
Loud 65 74 79 79 73 66 54
Shout 76 75 84 88 83 75 63
23
3.2.2 Calculating the Source Room Sound Pressure Levels
The source room sound pressure level ( Lp s ) can be calculated, using Sabine diffuse field
theory, from the source power level ( Lws ) and the room’s dimensions and materials [21]:
4
Lp s = Lws + 10 log (17)
Rs
Rs , the room constant, is calculated from the room’s area-weighted average absorption
α s Ss
Rs = (18)
1−αs
Knowing the source room volume ( Vs ), the reverberation time of the source room can also be
estimated:
0.16Vs
T60 s = (19)
α sSs
transmission loss is the combined effect of the transmission loss of the wall and ventilation
opening. It is calculated, based on average energy transmission, by finding the area-weighted
average transmission coefficient (τ p ) . Here subscripts w and v represent the original wall and
τ w Aw + τ v Av
τp = (20)
Aw + Av
TL w
−
τ w = 10 10
24
TL v
−
τ v = 10 10
S p is the total partition area, and α r and S r are the area-weighted average absorption
coefficient and the total receiver room surface area. As in the source room, the reverberation
time can be calculated in the receiver room:
0.16Vr
T60 r = (22)
α s Ar
The speech levels in the receiver room have been influenced by the reverberation in the
source room and the receiver room. As such, in reality, the associated reverberation time will
be longer and have a double-sloped decay. The SII method assumes that the reverberation
time in the receiver room is the only reverberation and will therefore underestimate the
masking due to reverberation. Especially if the source room has a longer reverberation time
than the receiver room, the model will under-estimate masking due to late reverberation and
over-estimate the intelligibility.
This work will not consider the reality that a talker will change their vocal effort based on the
presence of speech masking noise or the dimensions of the room. Masking noise, from either
25
background noise or late reverberation, can increase vocal effort as explained by the
Lombard effect [42]. Room dimensions are related to the source-receiver separation distance;
vocal effort is also known to increase with this separation distance [43].
Additionally, note that the background noise levels in the source room are not considered. If
the background levels in the source room are not small compared to the speech pressure
levels, the background noise may provide some masking that is neglected in this model.
Finally, and most importantly, this model is only accurate if the sound fields are diffuse. In
very large, very small or very absorptive rooms, sound fields are not diffuse. If the aspect
ratio of the rooms is large, the sound field is not diffuse. If the speech source, or the receiver,
is near the partition, the direct sound level will not be small compared to the reverberant
sound level and diffuse field sound transmission will not be applicable.
3.3 Results
The model, as described above, was implemented in MATLAB. A copy of the code is
provided in Appendix B.1. Prior to observing the effect of transmission loss on SII, the effect
of other factors such as vocal effort, room size, room furnishings and background noise is
investigated. The effect of ventilation opening size and wall performance on SII is then
investigated. Ventilation opening size is discussed in terms of equivalent open area making
the results valid for any silencer with the specified equivalent open area regardless of the
silencer’s actual dimensions.
26
the gymnasium provided the vocal effort is not greater than Normal. In the gymnasium the
vocal effort must be increased to Shout before speech privacy is deemed Moderate (SII<0.2).
0.8
Small Office
0.7 Large Office
Classroom
0.6 Gymnasium
0.5
SII
0.4
0.3
0.2
0.1
0
Casual Normal Raised Loud Shout
Vocal Effort
27
are required to bring the signal out of the background noise. As observed in Figure 4, for a
fixed vocal effort, SII is highly dependent on room size.
0.5
Small Office
0.45 Large Office
Classroom
0.4
Gymnasium
0.35
0.3
SII
0.25
0.2
0.15
0.1
0.05
0
1 2 3 4 5
Surface Configuration
Figure 5: Variation of SII with surface absorption in various room sizes – successive surface
configurations have increasing absorption.
As in Figure 5, it can be observed that there is a worst case scenario for room absorption. As
background noise levels increase, the more reverberant rooms have higher intelligibility,
likely because the early reverberant energy is required to bring the signal out of the
background noise. For the smaller rooms, the speech levels are much higher and high levels
28
of reverberation can cause increased masking of the signal. Previous work by Hodgson and
Nosal at the UBC Acoustics and Noise Research Group has concluded, similarly, that
reverberation time is beneficial for intelligibility if background noise is present [44].
Hodgson and Nosal also note that these dependencies are not so simple. The reverberant field
due to the noise source, and thereby the noise levels for a fixed noise-source power level, will
increase with reverberation time. Additionally, the proximity of the noise source to the
receiver can be a critical factor [44].
0.35
BGN0+0dB
0.3 BGN0+10dB
BGN0+20dB
0.25 BGN0+30dB
0.2
SII
0.15
0.1
0.05
0
1 2 3 4 5
Suface Configuration
Figure 6: Variation of SII with increasing surface absorption for various BGN.
29
3.3.4.1 Partitions without Ventilation Openings
Figure 7 shows the relationship between SII and the partition type, plotted with respect to
their STC value. Here the vocal effort is Normal and the partition is separating two large
offices with the surfaces in Configuration 4. As expected, the SII is highly correlated with the
partition’s STC. For the lowest STC partition – a 10mm glass sheet (STC 16) – the speech
intelligibility is high, implying no privacy for any vocal effort. Improving the partition to a
typical stud separating two layers of drywall (STC 32) provides Moderate privacy at Casual
speech levels, but not at higher ones. A wall with two double layers of drywall separated by a
stud and connected with resilient hangars (STC 52) provides a High level of privacy for
Casual and Normal vocal efforts, Moderate privacy for Raised and Loud vocal efforts, but
insignificant privacy if the vocal effort is Shout. Pleasingly, the ANSI Standard S12.60
Acoustical Performance Criteria, Design Requirements, and Guidelines for Schools, requires
that the minimum STC of a partitions between offices is 45 dB [45]. As can be seen in Figure
7, 45 dB this agrees quite well with the 0.05 SII criterion, at a Normal vocal effort, for High
speech privacy.
0.8
Casual
0.7 Normal
Raised
0.6 Loud
Shout
0.5
SII
0.4
0.3
0.2
0.1
0
15 20 25 30 35 40 45 50 55
STC [dB]
Figure 7: Variation of SII with STC for various vocal efforts (Large Office, Surface Configuration 4).
30
3.3.4.2 Partitions with Ventilation Openings
When ventilation openings exist, without a silencing device, in a partition, building
occupants often complain about poor speech privacy (see Introduction). To examine the
effect of ventilation openings in partitions, they will be included and considered in terms of
their equivalent open area. Figure 8 shows the results, with the ventilator’s EOAs expressed
as a percentage of the total partition area (Aw =15 m2). Considered here are two Large Offices
with Surface Configuration 4; the vocal effort is a Normal level. Partition 1, even without an
opening, does not provide privacy; further reductions in transmission loss do not make the
problem much worse, as it is already nearly as bad as it can be. Partitions 2, 3 and 4, which
do provide privacy as complete partitions, have essentially lost all ability to provide privacy
if the ventilator’s EOAs reaches 0.1% of the partition area (0.12 x 0.12 m opening). With 1%
EOAs (0.39 x 0.39 m opening) they provide similar attenuation to the glass partition
(Partition 1). This figure clearly shows that, if a ventilation opening is included in a partition
with out being silenced, it is of no benefit to provide a high performance wall for the
remainder of the partition.
Figure 9 is similar to Figure 8; however, the SII for each partition has been normalized to the
SII with a complete partition. This allows examination of the relative effect of the ventilation
opening on privacy. Once again, for Partition 1, the intelligibility is not increased by an
EOAs up 1% the size of the partition. Partition 2, however, will allow the SII to double if the
ventilator’s EOAs is 0.1% of the total wall area (0.12 x 0.12 m). Partitions 3 and 4 are even
more restrictive, allowing the SII to double with the ventilator’s EOAs only 0.01% and
0.001% (4x4cm, and 1.2x1.2cm) respectively. This is sufficient to conclude that un-treated
ventilation openings can not be included in higher performance partitions without greatly
reducing the privacy that the partition would otherwise provide.
Figure 10 shows another approach to describing the effect of ventilation openings on the
privacy that a partition provides. As in Figure 9, in Figure 10 the SII is normalized to its
value with a complete partition so that the relative effect is observed. The ventilator’s EOAs,
instead of being expressed as a percentage of the wall’s area, is normalized to the equivalent
open area of the wall (EOAs_w). For reference, the EOAs_w values of the partitions are given
31
in Table 7. For Partition 1, it is observed that the privacy cannot be reduced when there is
none to begin with. For Partitions 2, 3 and 4, however, if the EOAs_v (subscript v refers to the
ventilator) is less than one tenth of the EOAs_w the reduction in privacy is small. Moreover, if
the EOAs_v reaches or exceeds 50-100% of the EOAs_w, then there is a large reduction in the
privacy provided. As a rule of thumb, it can be concluded that the EOA of the ventilator
should optimally be less than 10% of the original partition’s EOA, and it should never exceed
50%, to avoid a major loss in performance.
0.9 Partition 1
Partition 2
0.8 Partition 3
Partition 4
0.7
0.6
SII
0.5
0.4
0.3
0.2
0.1
0
0.0001 0.001 0.01 0.1 1 5
EOA s /Aw [%]
Figure 8: Variation in SII with EOAs for different wall types.
32
5
4.5 Partition 1
Partition 2
4 Partition 3
Partition 4
3.5
SII/SII0
2.5
1.5
1
0.001 0.01 0.1 1
EOA s /Aw [%]
Figure 9: Variation in the normalized SII as a function of EOAs for different wall types.
Partition 1
4.5
Partition 2
Partition 3
4
Partition 4
3.5
SII/SII0
2.5
1.5
1
0.01 0.1 1 10
EOA s v /EOA s w
Figure 10: Variation in the normalized SII as a function of EOAA_V /EOAA_W, for different wall types.
33
3.4 Factors Affecting Speech Privacy: Summary
As observed in Figure 7, SII is highly dependent on the transmission loss of the partition.
Unfortunately, as shown in Figures 4, 5 and 6, it can also be greatly affected by changes in
vocal effort, room size, room furnishings and background noise. As such, when designing a
partition to achieve a required level of privacy, it is insufficient to consider the partition’s
transmission loss alone.
The transmission loss of a partition, and the privacy it provides, can be greatly reduced if
ventilation openings without adequate silencing are included. As shown in Figure 8 for a
typical office, no significant level of privacy can be obtained, regardless of the wall
construction type, if an untreated opening one tenth of a percent of the total wall area exists
(0.12 x 0.12 m opening in a 5 x 3 m wall). It is doubtful that a successful ventilation system
can be designed with an opening restricted to these dimensions. As a result, to achieve
significant privacy, natural ventilation openings must be treated acoustically. Figure 10
shows that, in a typical office, the SII is significantly increased (privacy reduced) if the
ventilator’s EOAs exceeds 10% of the wall’s EOAs. Additionally, the SII approximately
doubles (for low values of initial SII) if the ventilator’s EOAs exceeds 50% of the wall’s
EOAs.
34
Chapter 4: Describing Flow Performance of Ventilation Openings
In this section a derivation is provided for the discharge coefficient, as well as other similar
metrics, in order to assess the applicability of these metrics for describing the flow
performance of interior natural ventilation openings. A number of texts were used to develop
this discussion [46-48].
The general solution for fluid flow is represented by the Navier-Stokes equation. It can be
developed through a general consideration of the forces and accelerations on a fluid element.
The Navier-Stokes equation can be simplified to form a more user-friendly solution known
as Bernoulli’s equation. Bernoulli’s equation is the mathematical basis for the design of duct
systems. Empirically derived energy losses must be added to Bernoulli’s equation in order to
sufficiently describe the flow; these losses are what concern us in the design of ventilators.
v
Variables are the density ( ρ ), particle velocity vector( u ), time (t), pressure (p), dynamic
v
viscosity ( µ ), and the body force vector ( g , often gravity). Insight into the behavior of the
flow can be gained by introducing non-dimensional parameters [46]. Each variable is
replaced by the product of a characteristic value and a dimensionless quantity identified by a
caret:
v v v v 1 ˆ 1 ˆ2
u = Uuˆ , t = τ tˆ , p = Ppˆ , g = Fgˆ , ∇ = ∇ , ∇2 = 2 ∇
L L
U is the characteristic velocity, τ is a characteristic time period, P is a characteristic pressure,
F is a characteristic body force and L is a characteristic dimension. Inserting these into the
Navier Stokes equation and rearranging can produce:
ρL2 ∂uˆ ˆ 2 uˆ − ρL F gˆ = − PL ∇
2
+ Re uˆ ⋅ ∇uˆ − ∇
ˆ pˆ
µτ ∂t
ˆ µU µU
35
in which Re is the Reynolds number, the well-known non-dimensional number that describes
the ratio of inertial forces to viscous forces:
ρUL
Re =
µ
For incompressible (constant density) flows, in the absence of elevation change, gravity can
be neglected. If the flow is steady, derivatives with respect to time are zero – using these
assumptions the Navier-Stokes equation can be written as:
v v ˆ 2 vˆ PL ˆ
Re uˆ ⋅ ∇uˆ − ∇ u = − ∇pˆ
µU
The pressure’s dependence on velocity is clearly dependent on the Reynolds number; general
conclusions can be drawn:
1. If Re is large, indicating relatively large inertial forces, then the forces due to
convective acceleration will dominate the shear forces; as a result, the pressure
fluctuations will be proportional to fluctuations in the velocity squared.
2. If Re is small then the viscous forces dominate the inertial forces and the pressure
fluctuations will be proportional to the velocity fluctuations.
Understanding what constitutes ‘large’ and ‘small’ Reynolds numbers is not trivial; however,
it becomes very important when using empirical, engineering analysis methods for
determining flow behavior. As shown below, the value of Re often describes the limits of the
method’s validity.
36
v
∂u v v ∇p v
+ u ⋅ ∇u = − +g
∂t ρ
v
v v u2 v v
Next, substitute the vector identity u ⋅ ∇u = ∇ − u × (∇ × u ) to obtain:
2
v v
∂u u 2 ∇p v v v
+ ∇ + − g = u × (∇ × u ) (24)
∂t 2 ρ
If we integrate the second term of Eq. 24 between points 1 and 2 along a streamline s:
v v v
2
u2 u2 u2
∫1 ∇ 2 ds = 2 − 2
2 1
Integrating the third term of Eq. (24), assuming constant fluid density:
2
∇p p 2 − p1
∫
1
ρ
ds =
ρ
Integrating the fourth term of Eq. (24):
2
v
∫ g ds = g (h
1
2 − h1 )
v
Because we are integrating along a streamline, u × ds = 0 ; thus the right hand side of Eq.
(24) becomes zero:
v v
u × (∇ × u ) = 0
Combining all of these terms gives a spatially discrete form of Eq. (24):
ρ
∂u ρ 2
∂t 2
( 2
)
+ u 2 − u1 + p 2 − p1 + ρg (h2 − h1 ) = 0 (25)
For steady flow, the change in velocity with respect to time is zero. Also, we can take the
conditions at 1 and call them constant ‘c’; i.e. a reference total pressure (Eq. (26)). These
considerations give us a common form of Bernoulli’s Equation, concisely and eloquently
describing conservation of energy in a flow along a streamline:
ρ
u 2 + p + ρgh = c (26)
2
37
The first term of Eq. (26) is the dynamic pressure (kinetic energy), the second term is the
static pressure (potential energy). If there is no elevation change, the total pressure is the sum
of the dynamic and static pressure.
ρ 1 u1 2 ρ 2u2 2
+ p s1 + ρ1 gh1 = + p s 2 + ρ 2 gh2 + ∆p t ,1− 2 (27)
2 2
∆pt ,1− 2 is a total pressure loss between 1 and 2. This is a loss that was not available in the
original lossless representation of Bernoulli’s equation, but has been added to allow changes
in energy to the system. ps1 and p s 2 are the static pressures. Eq. (26) can be simplified to:
∆p t ,1− 2 = ∆pt + ∆p se ;
where,
ρ 1 u1 2 ρ 2u2 2
∆p t = + p s1 − + p s 2
2 2
∆p se = g (ρ a − ρ )(h1 − h2 )
∆pt is the total pressure change in the fluid, and ∆pse is the change in pressure due to the
thermal-gravity effect (buoyancy). Changes in total pressure are conventionally separated
38
into friction losses over lengths of duct, friction losses at fittings, and buoyancy. The change
in total pressure over section i is expressed as [47]:
m n
∆p ti = ∆p f i + ∑ ∆p ij − ∑ ∆p seik
j =1 k =1
where ∆p f i = loss due to friction in section i, ∆pij = loss due to jth fitting in section i, and
∆pseik = stack effect due to kth stack in section i. The friction and fitting losses, commonly
known as major and minor losses, respectively, are both viscous energy dissipations. The
concept of adding the losses of each distinct section of duct together is applied successfully
in designing conventional HVAC ducting systems, where each fitting or section is separated
from the previous one by a section of straight duct. This way, the flow at the entrance of each
section is fully developed (not changing along the length of the duct) and independent of the
upstream geometry. In the case of short silencer sections, where the flow at the entrance of
one section of the silencer is not fully developed and affected by the upstream section, losses
at each section cannot be considered independent of each other. There is, however, a strong
motivation to use this formulation of discrete and independent pressure losses, as it is very
common, well understood, and because results exist for many geometries.
39
Here, p is the pressure, u is the velocity, and the subscripts refer to locations 1 and 2 on a
streamline. Assuming incompressible flow, conservation of mass results in:
A2
u1 = u2 (29)
A1
where A is the cross sectional area. Eq. (28) and Eq. (29) can be combined to produce:
A2
2
2∆p
2
u 2 1 − 2 =
A1 ρ
If the orifice is small compared to the inlet area ( A2 << A1 ), we can approximate:
2∆p
u2 =
2
ρ (30)
What follows is an expression for the flow rate through a constriction, dependent on only the
opening area, and the pressure differential. It assumes that no energy is lost, and the area
ratios are large:
0.5
2∆p
Q = A2
ρ
In reality, to account for phenomena such as the vena-contracta at the orifice, as well as
losses, an experimentally determined factor called the discharge coefficient ( Cd ) is
introduced:
0.5
2∆p
Q = AC d (31)
ρ
The discharge coefficient is useful for calculating the flow rate given a geometry and
pressure; however, another coefficient, the loss coefficient (k), is introduced into Eq. (30) to
define the pressure loss given a flow rate.
40
1
∆p = k ρU 2 (33)
2
∆p
k=
1
ρU 2 (34)
2
The loss coefficient can succinctly be defined as the ratio of the loss in total pressure to the
dynamic pressure. An important distinction is that the pressure difference used to calculate
the discharge coefficient is the difference in static pressures between two locations; the
pressure difference used to calculate the loss coefficient is the difference in total pressures
between two locations. For ventilation openings, because the ventilation opening area is very
small compared to the upstream and downstream areas, the static pressure will dominate the
total pressures up and down stream; therefore, the pressures used to calculate the discharge
coefficient and the loss coefficient are effectively identical. Both the discharge coefficient
( Cd ) and the loss coefficient (k) are used regularly in engineering application. Typically the
discharge coefficient is used for flow meters, and the loss coefficient is used for determining
the pressure head required in pipe and duct systems to achieve a desired flow. One must be
careful when selecting coefficients from the literature, the use of symbols Cd and k for the
discharge and loss coefficients are not universal.
41
It follows that, at low Reynolds numbers, the discharge coefficient is directly related to the
square root of the Reynolds number:
C d (Low Re ) ∝ Re
In the case of flow through ducts, the characteristic length is the hydraulic diameter (Dh),
which can be found from the cross section al area (A) and the perimeter length (P).
4A
Dh =
P
The Reynolds number can be conveniently written as:
UDh
Re =
ν
At low Reynolds numbers the discharge coefficient decreases with decreasing flow rate. If a
high Re discharge coefficient is used to predict pressure loss at low Re, the pressure loss will
be under-predicted. The system will be under-designed. It is therefore necessary to
understand what constitutes low and high Reynolds numbers, as it will define the lower
bound for the region of validity for the discharge or loss coefficient of a flow element. As
flow in conventional HVAC systems is, for engineering purposes, assumed turbulent, little
information exists on the low Reynolds number behavior of duct elements. In hydraulic flow
applications, however, because the viscosity is much higher and characteristic dimensions are
much smaller than in air flow applications, low Reynolds number behavior is much better
understood. As the same assumptions made for ventilation air flow are made for hydraulic
flow (incompressible and Newtonian), the same non-dimensional analysis will apply;
discharge coefficients are interchangeable.
From hydraulic flow metering, a good understanding of the discharge coefficient of various
constrictions has been developed; this information is very useful for understanding natural
ventilation openings, as they are typically both short constrictions in the flow path.
Numerical simulations have produced the discharge coefficient results for venturi, orifice, v-
cone and wedge-type flow meters [49]. Flow meter types can be identified by the cross
sections shown in Figure 11. The V-cone is more of an obstruction than a constriction so it
will not be discussed. The percentage errors in the calculated flow rate based on the turbulent
discharge coefficient are shown in Figure 12.
42
Here the transitional Reynolds number (Retr) is defined as the Reynolds number below which
errors in the calculated flow rate greater then 10% will result from using the high Re
characterization. The transitional Reynolds number (Retr) for flow through the venturi and
wedge constrictions are around 1000 and 100, respectively. Interestingly, flow through the
orifice is under-predicted near the transitional regime. From other published results this
effect can be partially attributed to the small beta value, the ratio of the orifice diameter to the
pipe diameter [50]. Figure 13 shows the discharge coefficient of various orifice plates for
different diameter ratios. As the orifice diameter becomes relatively small the discharge
coefficient over-prediction near transitional Reynolds numbers is reduced. It is safe to
conclude that the flow calculated from the turbulent discharge coefficient of an orifice
constriction is accurate for Re>100 provided the orifice diameter is small relative to the
upstream and downstream diameters. Johansen is also able to conclude that the regime where
the pressure relates linearly with flow rate is only for Re<10 [50].
43
Figure 12: Flow meter error resulting from using turbulent discharge coefficient [49].
Figure 13: Discharge coefficient of orifice plates with various geometries [50].
Another issue is that often grilles are added to the ventilation openings. The grille will have
little effect on the average flow velocity; however, the hydraulic diameter will be greatly
reduced. As a result, grilles will have a much lower Re than the duct cross section, and may
44
result in a “low” Re condition when the Reynolds number based on the ventilation opening
cross section is “high”.
Work in hydraulic flow has proposed unifying the high and low Reynolds number regions by
creating a single discharge coefficient model that describes the pressure loss in both high and
low Re regions [51-53]. This idea may be quite useful for designing composite systems
where each element is understood beforehand; however, it is less useful for the design of
individual elements, as it is required to know the transient Reynolds number of the element
(Retr) beforehand.
In summary, the loss in a flow element, be it an orifice or uniform length of duct, can only be
described by a single-number coefficient if the Reynolds number is high. Unfortunately,
there is no single guideline for what constitutes a high Reynolds number; it is dependent on
the geometry. For sudden constrictions in the flow, similar to the orifice or wedge shown in
Figure 11, and where the flow element’s length is less than one half of its diameter, a single
discharge coefficient may be valid for all Re > 100. If the constriction is gradual, as it is with
a venture flow meter (Figure 11), or if the constriction has some length, a single discharge
coefficient may only be valid for all Re > 4000.
In the end we are compelled to use a high Re characterization of the pressure loss because it
is both convenient and the industry standard. Unfortunately, in natural ventilation openings,
flow velocities are often very small and the typical HVAC assumption of high Re may no
longer be a safe one. In certain instances ventilation openings are best described as thin
orifices and a “high” Re may be as low as 100; however, this is not always be the case. In
some designs, the flow rate may be small, and the ventilation opening may be designed in a
way that the resultant Re is not large enough to be considered “high”. Regardless of this,
previous work on flow in naturally ventilated buildings uses the high Re characterization of
losses for interior ventilation openings [3, 4, 10-12, 15-20]. Reynolds number dependent loss
is mentioned at times, but not treated seriously [54]. In the following sections of this report
the Reynolds number will be discussed on a case-by-case basis to judge the accuracy of the
loss or flow rate characterization.
45
Chapter 5: Measurement of Natural Ventilation Openings
The current state of natural ventilation openings has been assessed through measurement and
analysis; in-situ measurement and laboratory measurements have been taken, and results
from a previous published work have been analyzed. A new measurement method was
developed for the in-situ measurements, and a facility was constructed and commissioned for
the lab measurements. This section will outline the measurement methods and lab
performance, and then present the results from in situ measurements, laboratory
measurements, and the analysis of results published by the UK Building Research
Establishment (BRE) [12].
5.1.1.1 Theory
In section 2.1.1, the background theory to the measurement of transmission loss was
provided. Eq. 5 and Eq. 6 show that the sound power entering the ventilator can be
determined by calculating the average sound pressure level in the source room:
46
2
p
Win = Av s
4ρ 0 c
Additionally, Eq. 7 and Eq. 8 show that the sound power exiting the ventilator can be
determined from the average sound pressure level and reverberation time in the receiver
room:
2
Ar p r 0.161Vr
Wout = ; Ar =
4 ρ0c T60
The transmission coefficient of the ventilator is equal to the ratio of the power entering to the
power exiting the ventilator:
2
A p
τv = v s2
Ar p r
The transmission performance is typically stated in decibel form as the transmission loss:
A
TL v = Leq s − Leq r + 10 log v
Ar
The standard deviation of the transmission loss ( σ TL v ) is calculated from the standard
deviation of the source and receiver sound pressure levels ( σ Leq s and σ Leq r ), and the
The ventilator’s transmission loss cannot generally be measured directly because the sound
transmitted through the ventilator cannot be measured independently of the sound transmitted
through the rest of the partition (unless the rest of the partition has a very high transmission
loss). To estimate the contribution of the transmission due to the ventilator alone, the
transmission will be measured once with the ventilator open and once with it blocked off.
Blocking the ventilator gives an approximation of the partition’s transmission loss without
the ventilator. As stated in section 3.2.3, the total transmission coefficient of the composite
partition ( τ p ) is equal to the area-weighted sum of the transmission through the ventilator
47
( τ v ) and the transmission through the rest of the wall ( τ w ). The ventilator area, unless
otherwise noted, is taken as the area of partition that it displaces:
τ wA w +τ vA v
τp =
Ap
If the transmission loss of the wall (partition without the ventilation opening) is much higher
than the transmission loss of the partition with the ventilation opening, then the transmission
through the wall is negligible compared to that through the ventilation opening and need not
be considered in the calculation. If, however, the wall has a low TL, or the ventilator a high
TL, there will be very little measureable effect due to blocking the ventilation opening. To
control error, the transmission loss of the ventilation opening will be limited to the
transmission loss of the wall. As a result, the calculated transmission loss of the ventilation
opening will represent a lower bound for its actual performance.
The difference in sound levels across the partition is called the noise reduction
( NR = Leq s − Leq r ). If the NR increases by 10 dB due to blocking the ventilator, the
ventilator TL result is independent of the wall TL. If there is not at least a 3 dB increase in
NR, the TL measurement should be treated with a large degree of uncertainty.
All of these calculations rely on the assumption that the sound field is diffuse; often, even
generally, this is not the case. In order for a room to promote a diffuse sound field it needs to
be large (but not too large), have small (but not identical or rational) aspect ratios, and
reflective surfaces. Additionally measurements must be made more than one half wavelength
from any reflective surface because, near surfaces, the sound field is never diffuse. Large
uncertainties exist in these measurements and we attempt to describe the uncertainties
through the spatial variation of the measurements; however, spatial variation alone is not
sufficient to completely recognize all uncertainties. Whenever conditions exist that are
suspected to introduce significant uncertainty they will be noted.
48
5.1.1.2 Measurement Method – In Field
To calculate the transmission loss of the ventilation openings it is necessary to have
measurements of the source room sound pressure level, receiver room sound pressure level,
and receiver room reverberation time, both with and without the ventilator blocked off. The
method provided by ASTM E336 for measuring apparent transmission loss (ATL) is
followed as closely as possible for these measurements. Corresponding to ASTM E336, the
average sound pressure levels are measured using a 1 min average, manual scan. The
reverberation time, calculated from the room impulse response, is the spatial average of six
measurements. The ventilation openings are blocked off using one half-inch thick sheet of
plywood attached to either end of the opening.
49
5.1.2.1 Theory
The volume flow rate being driven into the room, and the corresponding pressure differential
across the partition, are measured with the ventilation opening sealed and open. The
difference between the two flow rates, at a common pressure, gives the flow rate through the
ventilation opening at that pressure.
Flow rate is measured using a blower door, a calibrated fan unit designed for testing air-
tightness of buildings. It provides a method of calculating the flow rate based on the
difference between the ambient pressure and the pressure at a tap in the fan ( ∆p tap ), as well
as the pressure differential across the fan ( ∆p fan ). The equation and associated table shown
Q= m3 s
2118.882
Table 8: Flow-rate coefficients.
Range N K K1 K3
22 0.5214 486.99 -0.07 -0.12
A 0.503 259.04 -0.075 0
B 0.50075 159.84 -0.035 0
C8 0.5025 77.418 -0.03 0.004
C4 0.495 42.448 0.045 0.011
C2 0.4655 24.864 0.4 0.008
C1 0.473 13.349 0.32 0.004
The “Range” values in Table 8 correspond to the fan setting, which can be adjusted to give
different flow rate ranges. 22 is the range that provides the largest flow rate and C1 provides
the smallest. After selecting an appropriate range, Table 8 is used to select the coefficients N,
K, K1, and K3 to determine the flow rate.
The flow rate is measured at a number of different pressures, from which a log linear
regression is used to define the flow rate as a function of pressure, plus provide confidence
intervals for the curve fit. The desired expression is:
Q = C∆p n m3 s
50
C and n are the flow coefficient and flow exponent. Letting x = ln(∆p ) and y = ln (Q ) the
linear regression can be used to determine C and n from the variance and covariance of x and
y [27]:
y = ln(C ) + n ⋅ x
1 N 1 N
2
Sx = ∑ ( x i − x )2 ; Sy =
2
∑ ( y i − y )2
N − 1 i =1 N − 1 i =1
1 N
S xy = ∑ (xi − x )( yi − y )
N − 1 i =1
S xy
n= 2
Sx
C = e y − nx
It is not possible to directly subtract the flow with the ventilator sealed from the flow with the
ventilator open because the flow rates cannot be readily measured at identical pressures. To
allow the subtraction, a linear regression for the flow with the ventilator sealed is first found.
Using the linear regression, the leakage rates at the pressures corresponding to the pressures
of the open ventilator measurement are calculated and subtracted from the open ventilator
results. The corrected open ventilator measurement results are then used in a linear regression
to find an equation that describes the flow through the ventilator.
1 S y − n ⋅ S xy 2
2
Sn =
S x N − 2
1
N 2 2
∑ xi
S ln (C ) = S n i =1
N
51
5.1.2.2 Measurement Method – In-Situ and Laboratory
In order to calculate the flow characteristics of the ventilator the pressure differential across
the ventilator, and the corresponding flow rate through the room, must be measured with the
ventilator sealed and open.
To improve the accuracy of the results, any obvious flow paths exiting the room, besides the
ventilator, are sealed off. The fan unit is fitted into the door, and pressure taps are inserted in
the rooms on either side of the fan, and on either side of the ventilator2. The fan pressure is
measured as the difference between the fan tap pressure and the ambient pressure in the room
on the inlet (low pressure) side of the fan. The room to be pressurized (i.e. downstream of the
fan) is chosen as the one that offers the least obstruction to the flow of high velocity air
exiting the fan. A similar work has concluded that the flow characteristics of the ventilator
are not significantly dependent on the flow direction [12].
The ventilator is sealed off for the first set of measurements to allow calculation of the flow
rate exiting the room through paths other than the ventilator. Measurements are then repeated
with the ventilator open to allow calculation of the airflow through all flow paths. If the flow
rate with the ventilator sealed off is not small compared to that when it is open, additional
uncertainties will exist. Additionally, the pressure differential across the ventilator should be
measured with the fan off, to confirm a zero reading. A non-zero pressure would typically
indicate that the ventilation system is influencing the measurement.
2
For the in-situ measurements, the fan and ventilator were generally installed in the same partition;
therefore, the pressure differential across the fan and ventilator were the same.
52
were motivated by ASTM E90 [56], due to facility size limitations, no attempt was made to
meet its requirements.
To create two rooms, a 115 mm partition was built to nominally divide the space in two,
leaving ‘source’ and ‘receiver’ rooms with floor dimensions of 2.33 x 2.62 and 2.42 x 2.62 m
respectively. The partition was created using two sheets of ½” (12.7 mm) gypsum boards
separated by ‘2x4’ (38.1 x 88.9 mm) studs on 520 mm centers. The cavity was filled with
fiberglass batt to improve the noise isolation. All joints in the partition were caulked. This
partition is not one of high acoustical performance; however, creating a partition with a high
performance was not necessary, as the transmission loss was limited by flanking through the
walls and ceiling. To achieve a high degree of isolation, a room within a room would need to
be constructed; this was not deemed necessary.
Openings for a ventilator, and a “door” to access the source room were included in the
partition. The section for ventilators is a removable portion of the partition, bordering the
floor, 2m wide and 1m high. If necessary, it was partially filled in, with construction similar
to the rest of the partition, to accept various ventilators. To test the partition’s nominal
transmission loss, the ventilator opening was completely filled in. A door to the source room
was created by making a ‘2x4’ framed plug. The plug was covered with gypsum on one side
and 1” MDF on the other; it fits tightly into the framed opening created in the partition. The
MDF is larger than the plug, to provide an exposed lip to seal against the partition frame. The
door-partition mating surfaces were treated with gasket foam.
A second ceiling, constructed of ½” (12.7 mm) gypsum board, suspended by ‘2x4’ (38.1 x
88.9 mm) studs, was added just below the acoustic tile of the source room to reduce
53
absorption and increase transmission loss; the final source room height is 2.58 m. A 1.5 m2
trap door, constructed of one ½” ply layer and one ½” gypsum board layer, was added to the
ceiling; it can be opened to allow air to exit the room through the ceiling and into the
adjacent corridor, and closed to provide acoustic isolation. In the receiver room, the ceiling
was modified by adding a layer of gypsum to the acoustic tile such that the gypsum faced
toward the room and the acoustic tile faced toward the ceiling. The ventilation supply was
disconnected and allowed to discharge into the ceiling. A fan was placed at the door of the
room for ventilation during use.
A diffuse sound field requires a long reverberation time to create a strong reverberant field.
The reverberation time, calculated from an average of nine measurements at different
locations, is shown for the source and receiver rooms in Figure 14. At low frequencies the
reverberation time is low due to membrane absorption from the gypsum board; at high
frequencies it is low due to air absorption. In the mid-frequency range, especially considering
the small volume of the spaces, the reverberation time is fairly high, around 1.5 s. To
maintain a reverberant field, instead of specifying a minimum reverberation time, ASTM E90
specifies a maximum total absorption area (A):
V23 2000
A≤ = 2.21 m2 for ≤ f ≤ 2000 Hz
3 V13
The author suspects that the low frequency limit, which equals 776 Hz for the receiver room,
is specified because below this frequency there is not a high enough modal density for the
field to be diffuse. Increasing absorption can reduce the modal behavior of the room response
and result in a more uniform pressure field. The high frequency limit exists because, at
frequencies above 2000 Hz, the reverberation is controlled by air absorption and not by the
surfaces of the room. The total absorption area and the ASTM absorption area maximum, in
its region of validity, are shown in Figure 15. The facility meets the requirements of ASTM
E90 for maximum total absorption.
54
The uniformity of the sound field can be assessed by the standard deviation of the sound
pressure level over the room’s volume. Measurement positions in the volumes are chosen
randomly while keeping the microphone at least 1 m from any surface as required by ASTM
E90; near the surfaces the various modes are spatially in phase with each other, causing a
non-diffuse field and an elevated sound pressure level. The uncertainties in sound pressure
level associated with a 95% confidence in the mean are shown in Figure 16 for the source
and receiver rooms. The uncertainty at 125 Hz is high, around 4 dB, at 250 it drops to around
1.5 dB, and above the 500 Hz band the uncertainty is below 1 dB. Uncertainties in the
average sound pressure levels dominate the uncertainties in reverberation time when
calculating the transmission loss uncertainty.
Figure 17 shows the partition transmission loss with 95% confidence error bars. The
transmission loss is 25 dB at 125 Hz, increasing to over 35 dB between 500 and 2000 Hz.
Above 2000Hz there is a reduction in transmission loss due to coincidence; it then increases
with frequency from 31 dB at 2500 Hz to 47 dB at 10 kHz.
The equivalent open area of the high pressure (upstream) room is shown in Figure 18. The
equivalent open area increases slightly with applied pressure, possibly because gaps are
being forced open by increasing force; however, the equivalent open area never exceeds
0.005 m2. If the EOAf of any ventilator is not significantly greater that this, the results should
be used with caution.
55
Measurement results, shown in Figure 19, for a thin, square opening with a 0.6 m edge
length, give an average SEOAf of 1.024 with 95% confidence interval of 0.026. In this work,
uncertainties and errors on the order of 3% are quite small. Measurement tests were
completed at Reynolds numbers above 200k; the flow was highly turbulent.
1.4
1.2
T30 [s]
0.8
0.6
0.4
0.2
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 14: Measured reverberation time with 95% confidence intervals for source and receiver rooms.
56
7
Receiver Room
6
Max Permissible (ASTM E90)
4
A [m2]
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 15: Measured receiver room total absorption and the maximum permissible value according to
ASTM E90.
3.5
Source Room
Receiver Room
Lp Uncertainty (95% confidence)
2.5
1.5
0.5
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 16: Uncertainty in measured Lp - 95% confidence values.
57
50
45
40
35
TL [dB]
30
25
20
15
10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 17: Measured transmission loss of laboratory partition - 95% confidence values.
-3
x 10
5
4.5
3.5
3
EOAf [m2]
2.5
1.5
0.5
0
0.5 1 2 5 10 20 50
Pressure [Pa]
Figure 18: Measured equivalent open area of high pressure room (leakage).
58
1
0.8
SEOAf
0.6
0.4
0.2
0
Re=200k 10 15 20 25 30 35 40 45 50 Re=800k
Pressure [Pa]
Figure 19: Measured SEOAf of 0.6 x 0.6 m opening.
59
compensate for the reverberant field in the receiver room. The specific equivalent open area
for flow (SEOAf) is provided to evaluate the flow performance of the ventilation opening.
The change in equivalent open area for flow (%∆EOAf), is used to indicate the reduction in
flow caused by blocking the ventilator. A large difference indicates a high confidence
measure of the ventilator flow rate. Finally, the open area ratio (OAR) is given to rate the
total combined performance of the NVO.
In this work, due to limitations of the measurement equipment, the lowest flow rate was
measured at a pressure differential between rooms of 5 Pa. As such, the EOAf will be
calculated at 5 Pa (see section 2.1.2).
60
Transmission loss tests were completed using the main library as the source room and L020
as the receiver. Airflow test were completed by pressurizing L020 with the operator outside.
Measurements were completed with the grille on, and again with the grille off.
61
Figure 20: Regent College L020.
62
5.3.2 Fred Kaiser Building
Offices and meeting rooms on the third and fourth floor of the Fred Kaiser building are
naturally ventilated. Air is let in through vents in the window and is drawn out of the room
through ventilators at the top of a partition common to an atrium. The ventilators are
acoustically treated using a Z-shaped crosstalk silencer built into the partition.
Transmission loss tests were completed using the atrium as the source room and 4036 as the
receiver. Airflow test were completed by pressurizing 4036 with the operator outside.
63
Figure 22: Fred Kaiser 4036.
130mm
570mm
90mm 250mm
120mm
Figure 23: Fred Kaiser 4036 ventilator cross section– 25mm fiberglass shown as crosshatched.
64
Table 10: Performance measures for the ventilator in Fred Kaiser 4036.
∆NIC [dB] 1
ASTC [dB] 16
SEOAs 0.046
%∆EOAf -78
SEOAf 0.11
OAR 2.48
Transmission loss tests were completed using 216 as the source room and 216C as the
receiver room. Airflow test were completed by pressurizing 216C, with the operator in 216.
Error may have been induced in the flow measurement, as the fan was partially obstructed by
a desk 1.5 m away. This could lead to pressure loss not attributable to the ventilator, thus a
lower flow rate measurement.
65
5.3.3.1.1 216C Results (see Table 11)
A 13 dB change in NIC is observed between open and closed vent conditions providing a
reliable ASTC. EOAf is reduced 95% by blocking the ventilator, indicating that the air flow
results are accurate. A -1 dB ASTC and 1.04 SEOAs indicate that the ventilation opening
provides insignificant attenuation. A SEOAf of 0.74 indicates that the opening is moderately
more restrictive to airflow than a thin aperture; however, the flow performance is likely
under-estimated due to measurement conditions. This ventilator is very large, causing high
flow rates and large velocities in the rooms. The high velocities away from the ventilator
would cause increased dissipation of energy and thereby increased flow restriction. The
resultant OAR of 0.71 shows that this opening is slightly more restrictive to airflow than it is
to noise.
Transmission loss tests were completed using the corridor as the source room and 308 as the
receiver. Airflow tests were completed by pressurizing 308 with the operator in the corridor.
66
5.3.3.3 Liu 313
Room 313 (Figure 27) has a volume of 46 m3 and shares a 7.7 m2, double drywall
construction partition with the corridor. One solid wood door exists in the partition. The
partition has two 250 x 350 mm ventilation openings, each with two grilles – one on either
side. The grilles (Figure 28) are made of 1 mm thick steel fins on 20 mm centers.
Transmission loss tests were completed using the corridor as the source room and 313 as the
receiver. Airflow test were completed by pressurizing 313 with the operator in the corridor.
Table 11: Performance measures for ventilation openings in the Liu Institute.
313 313
Vent Condition 216C 308 Grille On Grille Off
∆NIC [dB] -1 0 3 4
ASTC [dB] 1.04 16 0 -1
SEOAs -95 0.030 0.85 0.88
%∆EOAf 0.74 -75 -68 -81
SEOAf 0.71 0.10 0.65 1.22
OAR -1 3.42 0.76 1.38
67
Figure 24: Liu 216C (showing ventilation openings blocked).
130mm
480mm
130mm
130mm
68
Figure 27: Liu 313 (showing ventilation openings blocked).
69
been acoustically treated by covering the steel deck with acoustic tile (rooms 321 and 326).
One additional room was modified for high acoustical privacy. The ventilation opening was
replaced with a transfer duct silencer in which a fan is mounted to assist the natural
ventilation airflow mechanisms (room 327).
Transmission loss tests were completed using the corridor as the source room and 167 as the
receiver. Due to the ventilation opening geometry, at the time of testing it was not possible to
block it acoustically and measure the partition’s transmission loss. A conservative
assumption is made that the energy transmitted through the partition is insignificant
compared to that transmitted through the ventilator. Thus, the true acoustic performance of
the ventilator will be higher than that measured. Airflow tests were completed by
depressurizing 167 with the operator in the corridor. The ventilator flow was blocked using
poly sheeting.
70
surface of the ventilation opening is acoustic tile. The bottom surface is a structural wooden
beam 0.1 m wide; as a result, the ventilation opening has a non-zero length; the ventilators
length is comparable to its height.
Transmission loss tests were completed using the corridor as the source room and 321 as the
receiver. Due to the ventilation opening geometry, at the time of testing it was not possible to
block it acoustically and measure the partition’s transmission loss. A conservative
assumption is made that the energy transmitted through the partition is insignificant
compared to that transmitted through the ventilator. The true acoustic performance of the
ventilator will be higher than that measured. Airflow test were completed by pressurizing 321
with the operator in the corridor. The ventilator flow was blocked using poly sheeting.
Transmission loss tests were completed using the corridor as the source room and 326 as the
receiver. Due to the ventilation opening geometry, at the time of testing it was not possible to
block it acoustically and measure the partition’s transmission loss. A conservative
assumption is made that the energy transmitted through the partition is insignificant
compared to that transmitted through the ventilator. The true acoustic performance of the
ventilator will be higher than it is measured to be. Airflow test were completed by
71
pressurizing 326 with the operator in the corridor. In this case the ventilation opening area is
not very small compared to the corridor or room 326 cross-sectional areas. As a result the air
velocities in this room remained high, and a fully recovered static pressure was not measured.
In addition, due to high air velocities in the room, significant pressure losses occurred away
from the ventilator. Due to the size of the opening, and fan limitations, the maximum room
pressure reached was 15 Pa.
Transmission loss tests were completed using the corridor as the source room and 327 as the
receiver. Airflow test were completed by pressurizing 327 with the operator in the corridor.
The ventilator flow area was significantly impeded by the fan.
72
5.3.4.4.1 327 Results (see Table 12)
As there is only a 1 dB change in NIC due to closing the ventilator, the transmission loss
results for the ventilator only provide a lower bound on its performance, and a large degree
of uncertainty must be acknowledged. In addition, blocking the ventilator caused only a 6%
reduction in leakage area, causing great uncertainties in the flow result. Keeping the
uncertainties in mind, the ventilator is shown to greatly reduce acoustic transmission (18 dB
ASTC, 0.0073 SEOAs); however, the SEOAf is very low at 0.0033. A OAR of 0.87 results;
however, this should be considered a lower bound because the ventilator’s measured TL was
limited by that of the rest of the partition. Additionally, as the flow was obstructed by a
stationary blower, the SEOAf is likely much lower than it would be with the duct alone.
Table 12: Performance measures for ventilation openings in the C. K. Choi building.
Vent Condition 167 321 326 327
∆NIC [dB] N/A N/A N/A 1
ASTC [dB] 1 4 6 18
SEOAs 0.92 0.51 0.22 0.0073
%∆EOAf -77 -76 -94 -6
SEOAf 0.98 1.46 0.83 0.0033
OAR 1.06 3.42 3.82 0.87
73
Figure 31: C. K. Choi 326. Figure 32: C. K. Choi 327
While the majority of the library is open-plan, there were three interior partitions with unique
ventilation openings suitable for this study. The spaces are a computer lab (L104), a printer
room (L112), and a classroom (L208).
74
5.3.5.1 Langara L104
Room L104 (Figure 33) has a volume of 360 m3 and shares a 55 m2, 0.3 m thick, concrete
partition with a large atrium. Besides the ventilator, the door is the only other apparent sound
transmission path. The ventilator (Figure 35) is an L-shaped crosstalk silencer with a mean
length of 1.8 m and has both inlet and outlet covered with grilles (Figure 36). The outlet
grille has a chevron cross section, is 20 mm thick, with 1 mm thick fins and 4 mm spacing
between the fins. The inlet ventilator is a 10 mm grid of 1 mm thick, 10 mm long fins aligned
with the flow. From a visual inspection, the outlet grille appears as if it would offer a high
restriction to airflow. Tests were completed with the grilles installed and removed.
Transmission loss tests were completed using the L104 as the source room and the atrium as
the receiver. L104 could not be used as the receiver because it was not possible to turn off the
computers in the room which produced high background noise levels.
Airflow tests were completed by pressurizing L104 with the operator in the atrium. The
mechanical ventilation system was running; however, the flow during testing was eliminated
by covering the inlet with sheet plastic and tape.
The EOAf was reduced 62% by blocking the ventilator; the airflow results are therefore
significant. The SEOAf was calculated to be 0.25.
75
5.3.5.1.2 L104 Results: Grille Off (See Table 13)
As was the case in L104 with the grille on, no significant results can be given for ASTC,
EOAs, or OAR. Blocking the ventilator reduced the EOAf by 78%, allowing accurate flow
calculation for the ventilator. The SEOAf was calculated to be 0.48, corresponding to nearly
twice the value measured with the grille on.
Transmission loss tests were completed using L112 as the source room and the atrium as the
receiver. L112 could not be used as the receiver as there was a fan in operation, resulting in
high background noise levels. The fan could not be accessed; however, it is suspected that it
was providing cooling for the printer. Airflow test were completed by pressurizing L112 with
the operator in the atrium.
The equivalent leakage area of the room was reduced 95% by blocking the ventilator,
allowing accurate ventilator airflow calculation. The equivalent open area for flow was
measured to be 160% of the actual ventilator area.
76
5.3.5.3 Langara L208
Room L208 (Figure 34) has a volume of 285 m3 and shares a 26 m2, 0.3 m thick, concrete
partition with a corridor. The ventilator (Figure 37) is an L-shaped crosstalk silencer with a
mean length of 1.2 m and has both inlet and outlet covered with grilles. Grilles are identical
to those installed in L102 (Figure 36). Tests were completed with the grilles installed and
removed.
Transmission loss tests were completed using the corridor as the source and L208 as the
receiver. Airflow test were completed by pressurizing L208 with the operator in the atrium.
The mechanical ventilation system was running; however, the flow was eliminated during
testing by covering the ventilation supply with plastic sheet and tape.
Blocking the ventilator reduced the flow rate by 50% allowing the ventilator flow to be
calculated with moderate accuracy. The SEOAf was measured to be 0.72, corresponding to
over twice the value measured with the grille on. The result is an OAR of 9.89.
77
Table 13: Performance measures for ventilation openings in the Langara College Library.
L104 L104 L208 L208
Vent Condition Grille On Grille Off L112 Grille On Grille Off
∆NIC [dB] -1 -1 2 1 1
ASTC [dB] - - - 14 14
SEOAs - - - 0.02 0.02
%∆EOAf -62 -78 95 -30 -50
SEOAf 0.25 0.48 1.61 0.32 0.72
OAR - - - 4.47 9.89
78
Top View Side View
870 mm
250 mm
500 mm
450 mm
870 mm
Figure 35: Langara L104 ventilator cross section– 50 mm fiberglass shown as crosshatched.
460mm
380mm
350mm
520mm
Figure 37: Langara L208 ventilator cross Section– 50 mm fiberglass shown as crosshatched.
79
Figure 38: Langara L112 (ventilation opening at top left).
3
Assumes the “Zone Air Distribution Effectiveness” is 1.0 {{708 Standard, A. 2007}}.
80
Table 14: Selected minimum required air exchange rates [59].
Space type Air exchange rate Air exchange rate
[L/s/m2] [L/s/person]
Office 0.3 5
Meeting room 0.3 5
Class room 0.6 5
Computer room 0.6 5
Standard ventilation rate information will be used, in combination with the measurements
taken, to estimate the pressure drop across the ventilator and the Reynolds number of the
flow. The Reynolds number is calculated as:
QDh
Re Dh =
νA
Q is the standard flow rate. At room temperature (300K) the kinematic viscosity (ν ) of air is
15.7x10-6 m2/s. The pressure required to achieve the standard air flow rate can be calculated
as:
2
ρ Q
∆p =
2 0.61EOA f
Note that, because EOAf is measured and calculated at 5 Pa, error is introduced if the
calculated pressure is very different than 5 Pa. At any rate, this analysis provides a good first
approximation of the operational conditions.
The assumed ventilator operating conditions are shown in Table 15. Langara L112, and C.K.
Choi 327 ventilator results are not shown, as the ventilators were not designed for natural
ventilation. Table 15 provides the ventilator’s EOAf, the flow rate (calculated based on Table
14), average flow velocity (Q/A), hydraulic diameter, Reynolds number, and pressure drop.
The hydraulic diameter of a ventilator with a grille is given as the hydraulic diameter of the
path between the grille fins.
The most striking result in Table 15 may be the complete lack of consistency in operational
conditions of the various designs. For adequate ventilation, the ventilators in C.K Choi 326
and Liu 216C, both very large rectangular openings with approximately 1 m2 equivalent open
area, would only need average flow velocities on the order of 0.01 m/s. This low flow rate
81
corresponds to a sub-micro Pascal pressure drop, four orders of magnitude less than the total
pressure available in a typical natural ventilation system. The ventilators in Langara are at the
other end of the spectrum. While having an EOAf that is an order of magnitude less than the
C.K Choi 326 and Liu 216C ventilators, the required flow rate is at least an order of
magnitude more. As a result, the velocities in the Langara ventilators are over 1 m/s, and the
pressure loss is tens of Pascals, which exceeds the 10 Pa value (see Chapter 1) commonly
assumed available to an entire natural ventilation system. The rest of the ventilators have
operating conditions in between these two extremes.
The operational Reynolds number of the ventilator flow also varies greatly; however, it can
be firmly concluded that the flow in these ventilation openings is not generally dominated by
inertial forces alone (Re>4000), and is certainly not dominated by viscous forces (Re<10).
Addition of grilles, besides reducing the equivalent open area, greatly reduces the hydraulic
diameter, thereby reducing Re. Unfortunately, because there is little known information on
the Re dependence of openings geometrically similar to the natural ventilation openings, it is
not possible to conclude how much error is induced by using a single high-Re
characterization for the design of these ventilators. Further study in this field, possibly
through scale modeling or CFD, is warranted.
While the operational Re may not be high for some of these ventilators, the errors associated
with using a high Re characterization may be not be of great practical consequence.
Excluding ventilation openings with grilles, any ventilator that operates at Re < 4000 has an
associated pressure loss that two to four orders of magnitude smaller than the total pressure
available to a typical naturally ventilated building. The pressure loss across such a ventilator
will be insignificant compared to pressure losses elsewhere in the system. Grilles, however,
are an exception as their very small hydraulic diameter leads to a small Re at flow rates that
result in significant pressure loss. This issue is revisited with CFD predictions in section 6.3.
82
Table 15: Assumed ventilation opening operating conditions.
Ventilator Location EOAf Q [m3/s] U (m/s) Dh Re ∆p
Regent L020 – Grille On 0.26 0.048 0.12 0.026 190 0.053
Regent L020 – Grille Off 0.49 0.048 0.11 0.60 4500 0.016
Kaiser 4036 0.19 0.044 0.16 0.17 1800 0.086
Liu 216C 0.78 0.013 0.012 0.69 550 0.00045
Liu 308 0.18 0.014 0.037 0.25 590 0.0098
Liu 313 – Grille On 0.027 0.0075 0.18 0.028 330 0.13
Liu 313 – Grille Off 0.050 0.0075 0.18 0.20 2300 0.037
C.K. Choi 167 0.54 0.013 0.024 0.37 560 0.00094
C.K. Choi 321 0.46 0.013 0.041 0.25 650 0.0013
C.K. Choi 326 1.07 0.014 0.011 0.86 600 0.00028
Langara L104 – Grille On 0.05 0.295 1.37 0.008 690 47
Langara L104– Grille Off 0.10 0.295 1.36 0.39 34000 13
Langara L208 – Grille On 0.068 0.247 1.17 0.008 590 21
Langara L208– Grille Off 0.15 0.247 1.17 0.46 34000 4.2
83
5.3.7 In-Situ Results Summary (see Table 16)
Based on the results of the in situ measurements, a number of conclusions can be made about
the effect of grilles, acoustic absorption on the ceiling above openings, and cross-talk
silencers.
Rectangular ventilation openings (Regent L020 Grille Off, and Liu 313 Grille Off) have very
little acoustic attenuation, with their SEOAs being around 0.9. Additionally, the SEOAf, is
nearly 1.2 in both cases. When grilles are installed there is almost no change in EOAs;
however, EOAf approximately halves. These results indicate that the OAR of a plain
rectangular opening will typically halve if it is covered with a non-acoustical grille. The
practical implication is that, if one intends to cover an opening with a grille, the opening will
have to be twice as large to not induce additional pressure loss at the same flow rate;
moreover, as a result of doubling the ventilator size, roughly twice the sound power will be
transmitted. The effect of grilles on airflow was also noted in Langara library’s L104 and
L208 where the EOAf was halved by installing grilles on the L-shaped cross talk silencers.
Ventilation openings that are essentially rectangular openings but have a significant length
compared to their cross sectional dimension, such as in Langara L112 and C.K. Choi 321,
have an interesting behavior. Their EOAf is around 50% greater than the result expected for a
thin orifice of the same cross section. This suggests it is possible to reduce pressure losses by
creating a short length of duct instead of a thin orifice, and effect which is also observed in
the laboratory measurements (section 5.5.5), and CFD predictions (section 6.3.7).
Measurements were made on slot openings between the top of a partition and an acoustically
reflective ceiling (Liu 216C and C.K. Choi 167), as well as an acoustically absorptive ceiling
(C.K. Choi 321 and 326). The presence of absorptive material appears to reduce the SEOAs
from around 1 down to 0.5 or less, while causing no notable reduction in SEOAf. Adding
acoustically absorptive material to the ceiling above a slot ventilation opening doubles the
OAR of the ventilation opening.
84
Z-shaped cross talk silencers (Kaiser 4036 and Liu 308) provide a large reduction in sound
transmission – around 16 dB over that of a rectangular opening; the SEOAs is around 3-5%.
This is also associated with a small SEOAf of around 10%. In this case, the acoustic benefits
outweigh the airflow disadvantages, resulting in an OAR of 2.5-3.5. Note that the acoustic
transmission may be smaller than the stated result as its performance in both cases was
largely limited by that of the partition in which it was installed.
By observing the change in noise isolation due to blocking the ventilation opening one is able
to determine if a significant portion of the sound energy passing through the partition can be
contributed to energy passing through the ventilation opening. For every CT silencer
measured (Kaiser 4036, Liu 308 and Langara L208), blocking the ventilation opening
resulted in no more than a 1 dB NIC increase. This indicates that, in all case where a designer
has made a serious attempt to silence the ventilation opening, the silencer is effective in that
the ventilation opening does not detrimentally affect the acoustic performance of the
partition.
85
5.4 Analysis of BRE Ventilator Measurements
In an attempt to produce ventilators for naturally ventilated schools, Hopkins of the BRE
proposed and tested the transmission loss and air flow characteristics of a number of
crosstalk silencer type ventilation openings. The purpose of the ventilators is to allow cross
ventilation between classrooms and corridors or cafeterias. Results and testing information
were published [12].
The acoustics and airflow performance were tested using a method effectively identical to the
laboratory experimental procedure implemented in this report. Ventilators were inserted into
the common wall of a sound transmission suite. Transmission loss was measured by standard
techniques, and airflow was measured by driving air at a known flow rate from one room,
through the ventilator, and into the other room, while measuring the pressure difference
between the two rooms. All measured values required to calculate the OAR were published,
allowing discussion of these results in the same terms as the original work presented in this
thesis.
86
5.4.1.2 L-shaped Ventilator
The L-shaped ventilator was composed of the straight ventilator, with a 0.782 m long elbow
added to the end. It was tested with no lining (Figure 41), with 50 mm thick acoustic foam on
the bottom surface (Figure 42), with 50 mm thick acoustic foam on both top and bottom
surfaces (Figure 43), with additional foam baffles inserted at an angle to the flow (Figures 44
and 45), and with smaller waves of acoustic foam (Figure 46).
5.4.2 Results
STC, SEOAs, SEOAf, and OAR were calculated for each of these configurations from the
published measurement data. Results are shown in Table 17. These results can be used to
determine the effect of the absorptive lining, the elbow, grilles, PVC film lining, angled
baffles, and wave-shaped absorptive liner.
Adding 50 mm thick acoustic foam to both top and bottom surfaces of the straight and L-
shaped silencers creates Configurations 5 and 4. The open area for flow in both cases is
reduced to half of what it was without the foam; however, this is expected as the duct height
was reduced from 20 to 10 cm by adding the foam. Equivalent acoustic open area is reduced
greatly, by a factor of 1/310 in the straight duct, and by 1/970 the L-shaped duct. As the
acoustic open area was reduced much more than the flow open area, the resulting OAR is
177 for the straight duct and 885 for the L-shaped duct.
Configuration 3 is the L-shaped silencer with foam on the bottom side only. The SEOAs is 18
times the value of the same silencer with foam on both surfaces, but less than that of the
87
unlined duct by a factor of 1/55. Airflow results are not given; however, we can assume a
value of 0.75 by interpolating between Configurations 2 and 4. The result is an OAR of 63.
As expected, these results show that an acoustic lining is a critical component of a passive
silencer. Much improved performance was observed by lining both the top and bottom
surfaces.
The L-shaped silencer has very slightly higher restriction to airflow than the straight silencer;
however, it offers much higher noise reduction. As a result the OAR of the L-shaped silencer
is 5 times that of the straight silencer. Unfortunately, as the L-shaped silencer is 1.5 times the
length of the straight silencer, this is not a fair comparison.
88
the general conclusion that the performance is not greatly affected by the addition of a PVC
film validating its use to facilitate cleaning. The reduction in airflow is surprising, in that it is
suspected to reduce the surface roughness. A possible explanation is that, without a lining in
place, some of the air passes through the foam, increasing the total flow rate.
Configuration 8, which includes baffles without a liner on the top or bottom surfaces, reduces
the EOAs to about half that of the un-lined configuration. Adding the baffles to the lined L-
shaped ventilator, as in Configuration 6, reduces the EOAs by around 30%. It is likely that, if
the airflow was measured, the EOAf would be reduced by at least 30%, showing that the
angled baffles are not beneficial to the ventilator’s performance.
89
3. Provided the SEOAf is greater than 0.5, adding a grille significantly reduces the
airflow. If the SEOAf is nearly 1, adding a grille may reduce the airflow by 50%.
Grilles do not increase transmission loss.
4. Covering the acoustic foam liner with a PVC film to facilitate cleaning did not have
strong adverse effects on the silencer’s performance (sound transmission or airflow).
5. Initial attempts at creative acoustic foam placement (angled baffles and wave-shaped
liner) did not have superior performance to the basic lined duct.
Table 17: BRE ventilator measurement results summary, (xxx) – estimated values.
Configuration STC SEOAs SEOAf OAR
1 0 1.19 1.12 0.95
1g 0 1.24 0.68 0.55
2 4 0.66 0.95 1.44
2g 4 0.65 0.62 0.96
3 24 0.012 (0.75) (63)
3g 25 0.012 - -
4 36 0.00068 0.60 885
4g 37 0.00058 0.50 860
4 PVC 31 0.00046 0.55 1180
4g PVC 31 0.00045 0.51 1140
5 28 0.0038 0.68 177
5g 29 0.0034 0.57 166
5g PVC 17 0.0048 0.55 113
6 38 0.00048 - -
6g 38 0.00050 - -
7 20 0.026 - -
7g 20 0.025 0.62 24.8
8 8 0.32 - -
8g 9 0.30 - -
90
Figure 39: BRE ventilator Configuration 1.
91
Figure 40: BRE ventilator Configuration 5.
92
Figure 41: BRE ventilator Configuration 2.
93
Figure 42: BRE ventilator Configuration 3.
94
Figure 43: BRE ventilator Configuration 4.
95
Figure 44: BRE ventilator Configuration 8.
96
Figure 45: BRE ventilator Configuration 6.
97
Figure 46: BRE ventilator Configuration 7.
98
5.5 Laboratory Measurements of Ventilation Openings and Silencers
Measurements were carried out in the purpose-built ventilator test suite to measure the
performance of a number of distinct ventilation opening and silencer types. The intent of
these measurements was three-fold. First, they were made to satisfy our curiosity about the
relative performance of some of the very different ventilator types that are currently being
used in practice. Second, they were intended to assess the performance of novel ventilators.
Finally, the results provide insight into which ventilator types should be prototyped for
further optimization and which should not be pursued.
Five different types of ventilation openings were tested, each in a variety of configurations.
These ventilation opening types are the slot, acoustical louver, baffle, air filter, and crosstalk
(CT) silencer.
The performance characteristics of the ventilation openings are stated in terms of their
SEOAs, SEOAf and OAR. Additionally, the STC rating is given, as it is a standard metric and
can be used to compare the performance of these silencers to other published acoustical data.
Plots of the third octave band transmission loss are given to show the frequency dependence.
Also the transmitted speech spectrum, defined as the transmission loss subtracted from a
normalized A-weighted speech spectrum, are given as they show the relative importance of
each frequency band on the resultant SEOAs.
The effect of acoustic absorption on the ceiling above the slot was measured. This is of
interest because, as observed in the C.K. Choi building (see section 5.3.4) ceilings are often
constructed of acoustic tiles. In order to maintain a constant ventilation opening geometry,
the absorptive tiles, when removed, were replaced with gypsum board of the same geometry.
99
Various configurations of the absorption were tested to see the effect of absorber dimensions.
With the absorber width equal to the width of the slot (1 m), and a 1 m length (the absorber
extends 0.5 m on either side of the partition), absorber thicknesses of 0, 25 and 50 mm were
tested (see Figures 47 through 50). An additional test was made to test the effect of absorber
length by having the 50 mm thick absorber only 0.5 m long, extending 0.25 m either side of
the partition (no image shown). Manufacturer data for the fiberglass absorber is provided in
Appendix A.1
It was also of interest to see the effect of the ventilation opening being located away from, as
opposed to adjacent to, the floor (or ceiling). A 1 m x 50 mm opening was created 1m away
from the corner as shown in Figure 50.
Figure 47: Slot ventilation opening, no Figure 49: Slot ventilation opening, 1 m x 1 m x
fiberglass. 50 mm fiberglass.
Figure 48: Slot ventilation opening, 1 m x 1 m x Figure 50: Slot ventilation opening, away from
25 mm fiberglass. floor/ceiling.
A summary of the slot ventilation opening measurement results are given in Table 18.
Frequency dependent transmission loss and transmitted speech spectrum are given in Figures
100
51 and 52. All of the slot openings were geometrically similar; therefore, no significant
differences in the SEOAf exist; the fiberglass does not affect airflow. A SEOAf of 1.09
indicates that, as expected, the flow performance of the slot ventilation opening is slightly
better than that of an open hole.
The slot opening without any fiberglass has a SEOAS of 2.38 resulting in an OAR of 0.46.
This is quite poor performance. Figure 51 shows that the negative transmission loss
(equivalent to SEOAS greater than one) is dominant at low frequencies; two physical
explanations for this exist, and have been previously discussed in section 2.3. Firstly, if the
wavelength is greater than the opening dimension, as it is below about 6 kHz for the 50 mm
slot, then diffraction will be significant and more energy than is predicted by diffuse field
theory will enter the ventilation opening. Secondly, diffuse field theory does not apply at
walls or corners because the phase relationship of different modes is not random. In a
reverberant environment with a diffuse sound field, the sound pressure level at a wall is 2.2
dB higher than the mean value in the room’s volume. The effect is significant within one
quarter of the wavelength of the wall; therefore, at frequencies below 2000 Hz, coherence
will occur over the majority of the 50 mm high opening, and its effect should be observed. In
double and triple corners the sound field coherence is further increased.
Adding a 25 mm thick piece of fiberglass greatly increases the transmission loss above the
400 Hz third octave band, and increases it by up to 8 dB at and above 2 kHz. The SEOAs is
reduced to 1.43. The benefits of the great increases in TL at high frequencies are not realized
in the SEOAs; as is shown in Figure 52, the SEOAs is governed by transmission in the 250
and 500 Hz octave bands.
To further reduce the SEOAs, more absorption is required in the 250 and 500 Hz bands.
Increasing the fiberglass thickness to 50 mm, as shown in Figure 51, increases the
transmission loss in these bands. As a result, the SEOAS is further reduced to 0.94, providing
an OAR of 1.15.
101
Reducing the length of the absorber from 1 m to 0.5 m has only a small negative effect on the
sound transmission, which reduces the OAR from 1.15 to 1.06.
As explained above, moving the ventilator away from the corner was expected to reduce the
sound transmission at 2000 Hz and below as the ventilation opening is exposed to a lower
sound field (see Figure 51). Indeed, the transmission loss is reduced by 1-2dB in this range.
Above 2000 Hz coherence becomes insignificant for the opening in the corner;
correspondingly, the measured TL above 2000Hz is identical for the opening near and away
from the corner. In effect, the SEOAf was reduced by 15%, and the STC was increased by
1.3dB. Surprisingly, moving the slot away from the corner increased the SEOAf by 10%,
resulting in a 54% increase in OAR.
10
No FG
8 1m x 1m x 2.5cm FG
1m x 1m x 5cm FG
6
1m x 10.5m x 5cm FG
4 Away from corner
Transmisson Loss [dB]
-2
-4
-6
-8
-10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 51: Measured transmission loss of slot ventilation openings.
102
0
-5
Transmitted Speech Spectrum [dB]
-10
-15
-20
-25
-30
No FG
-35 1m x 1m x 2.5cm FG
1m x 1m x 5cm FG
-40 1m x 0.5m x 5cm FG
Away from corner
-45
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 52: Measured transmitted speech spectrum of slot ventilation openings.
103
5. As configuration 1, but with a 50 mm thick, 1 m x 1 m piece of fiberglass on the floor
at base of louver (Figure 57)
Figure 53: Louver configuration 1, view from Figure 55: Louver configuration 3, view from
receiver room (perforations not visible). receiver room (perforations visible).
Figure 54: Louver configuration 2, view from Figure 56: Louver configuration 4, view from
receiver room (perforations not visible). receiver room (perforations visible)
Figure 57: Louver configuration 5, view from receiver room (perforations not visible)
104
Table 19 shows the results summary for all acoustical louver configurations. All
configurations have similar airflow performance and, with the exception of configuration 5,
similar transmission loss performance. Intuitively, one might expect the transmission loss to
be greater with the perforations facing the source; however, according to the theory of
reciprocity for linear systems, switching the location of source and receiver should produce
identical results [60]. In this case, as the source and receiver rooms are effectively identical,
flipping the louver front to back is the same as switching source and receiver; reciprocity
predicts that identical results are measured. The louver has a SEOAs of 0.12 to 0.14 and a
STC of 12 dB in all orientations.
Airflow is not a linear system that should obey reciprocity; however, the airflow performance
is nearly identical for any orientation. The SEOAf is 0.21-0.22 in all configurations.
The OAR of the acoustical louver varied from 1.5 to 1.8 depending on orientation, and
increased to 2.9 by adding fiberglass adjacent to the inlet.
105
25
1
20 2
3
4
Transmission Loss [dB]
15 5
10
-5
125 250 500 1,000 2,000 4,000 8,000
Frequency [dB]
Figure 58: Measured transmission loss of acoustical louver.
-10
-15
Transmitted Speech Spectrum [dB]
-20
-25
-30
-35
1
-40 2
3
-45 4
5
-50
-55
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 59: Measured transmitted speech spectrum of acoustical louver.
106
Performance data for this louver has been published by Kinetics Noise Control, allowing
comparison. Figure 60 shows the insertion loss from the measured and published data; Figure
61 shows the SEOAf results.
The measured insertion loss was calculated by subtracting the transmission loss of an empty
opening the size of the louver from the transmission loss of the louver. Ideally the
transmission loss of an empty hole is 0 dB; in reality this is not necessarily the case.
Subtracting the transmission loss of the empty opening from the transmission loss of the
louver makes the result less dependent on the room. The published insertion loss was
measured with a free field receiver. Substantial disagreement between the results exists.
Excellent agreement exists between measured and published air flow results.
16
12
Insertion Loss [dB]
10
0
125 250 500 1,000 2,000 4,000 8,000
Frequency [dB]
Figure 60: Measured (diffuse field receiver) and published (free field receiver) louver insertion loss.
107
0.25
0.2
0.15
SEOA f
0.1
0.05
Measured
Published
0
5 10 20 50
Pressure [Pa]
Figure 61: Measured and published SEOAf of louver.
108
1m
0.3 m
0.65 m
0.6 m
The performances from these two configurations of acoustical baffle are encouraging. With
the 20 mm offset the SEOAs is 0.054, and is moderately worse (0.090) with the 40 mm
offset; however, the 20 mm offset is more restrictive to airflow, with an SEOAf of 0.20 as
opposed to 0.49 with a 40 mm offset. As a result, the 40 mm offset is superior, with the OAR
being 5.43 as compared to 3.70 in the case of the 20 mm offset.
The transmission loss plots of these two configurations, shown in Figure 63, indicate that the
baffle with a 20 mm offset has a higher transmission loss than the 40 mm offset below 1000
Hz; however, interestingly, above 1000 Hz it is lower. From Figure 64 we see that it is the
lower frequencies of 250 to 500 Hz that limit the acoustic performance of this silencer, which
results in the baffle with a 20 mm offset having superior acoustic performance.
Many variables, such as baffle size, absorber thickness and location, and aperture size remain
un-explored. It may be possible to optimize this design to obtain much higher performance.
109
35
20 mm offset
30
40 mm offset
25
Transmission Loss [dB]
20
15
10
-5
125 250 500 1000 2000 4000 8000
Frequency [Hz]
-15
-20
Transmitted Speech Spectrum [dB]
-25
-30
-35
-40
-45
-50
-55 20 mm offset
4 mm offset
-60
-65
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 64: Measured acoustical baffle transmitted speech spectrum
110
5.5.4 Acoustical Air Filter?
In naturally ventilated buildings, air filtration poses a problem in the very same way as
ventilation silencing: conventional methods (fibrous filters) create flow restriction, thereby
greatly reducing the ventilation flow rates. An idea was proposed to make a device that acted
as both a silencer and a filter to reduce the total load on the system. Combining the two
systems seems possible because fibrous filters appear visually similar to fibrous absorbers,
and electrostatic precipitator filters appear similar to splitter silencers. This work represents a
preliminary investigation into combining fibrous filters and absorbers. The acoustic
transmission and air flow of one type of fiberglass acoustic absorber (Figure 65) and two
types of air filters (Figures 66 and 67) have been measured. Manufacturer data for the filters
is provided in Appendix A.3.
Figure 65: Acoustic fiberglass (see Figure 66: Pink air filter (see Appendix A.3).
Appendix A.1).
111
The acoustical fiberglass is able to reduce the SEOAs to 0.14; however, it reduces the SEOAf
to 0.005, resulting in a very poor OAR of 0.014. The air filters do not attenuate sound
effectively, with SEOAs of 0.65 and 0.73; however, are less (but still significantly) restrictive
to airflow with SEOAf of 0.20 and 0.29. The highest OAR of the filters, 0.39, is the filter that
causes the least restriction to air flow – the white air filter.
In order to see the effect of the filter alone on acoustic transmission, it is better to look at the
insertion loss – the difference in transmission loss between the empty opening and the
opening with the filter installed. Insertion loss results are shown in Figure 70. At low
frequencies the two air filters have no effect. Their insertion loss increases with frequency,
but never exceeds 5 dB at frequencies below 10 kHz. The pink air filter has slightly higher
transmission loss over mid and high frequencies. As expected, the acoustic fiberglass
provides better transmission loss; however, it is less than 5 dB at all but high frequencies.
20
Acoustic Fiberglass
Pink Filter
15
White Filter
Transmission Loss [dB]
10
-5
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 68: Measured acoustical air filter transmission loss.
112
-5
-10
Transmitted Speech Spectrum [dB]
-15
-20
-25
-30
-35
Acioustic Fiberglass
-40
Pink Air Filter
-45 White Air Filter
-50
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 69: Measured acoustical air filter transmitted speech spectrum.
16
14 Acoustic Fiberglass
Pink Air Filter
12 White Air Filter
10
Insertion Loss [dB]
-2
-4
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 70: Measured acoustical air filter insertion loss.
113
5.5.5 Cross Talk Silencer
Cross talk (CT) silencers are short lengths of lined duct. Z-shaped CT silencers were
measured in the Fred Kaiser building, and in the Liu Institute for Global Studies. L-shaped
CT silencers were measured in Langara College library. Here the performance of straight, L-
shaped, and Z-shaped CT silencers are measured. Additionally, to observe the acoustic effect
of the fiberglass liner, the straight silencer was also tested with the fiberglass removed.
To allow a comparison of the different shapes, all three of the silencers were constructed so
that the length of the flow path through the center of the silencer was 0.3 m. Unfortunately it
was not possible to accurately test CT silencers with longer lengths due to transmission loss
limitations of the lab facility. Highly accurate transmission loss measurements were required
so that they could be used to validate the numerical model presented in section 6.2.2.
Diagrams and photographs of the straight, L and Z-shaped CT silencers, as tested, are shown
in Figures 71 through 76. Figures 77 and 78 show the straight silencer with the fiberglass
removed. All of the CT silencers were constructed out of 25 mm thick plywood, and lined
with 50 mm of fiberglass on either side (fiberglass details in Appendix A.1). The height of
the flow path was 0.1 m, and the width 2 m. The fiberglass surfaces exposed to the source
and receiver rooms have been covered by a layer of 12 mm gypsum board, indicated in the
drawings by a heavy black line. All of the silencers were sealed to the floor using plasticine.
The brick section shown in the figure represents the mounting with respect to the partition. In
all cases the acoustic transmission was tested with the flow travelling from right to left, and
the acoustic transmission from left to right.
100 mm
300 mm
Figure 71: 0.3 m Straight CT silencer diagram – 50 mm fiberglass shown as crosshatched.
114
Figure 72: 0.3 m Straight CT silencer.
200 mm
100 mm
200 mm
Figure 73: 0.3 m L-shaped CT silencer diagram– 50 mm fiberglass shown as crosshatched.
115
200 mm
100 mm
150 mm
Figure 75: 0.3 m Z-shaped CT silencer diagram – 50 mm fiberglass shown as crosshatched.
50 mm
100 mm
300 mm
Figure 77: 0.3 m Straight CT silencer diagram – fiberglass removed.
116
Figure 78: 0.3 m Straight CT silencer – fiberglass removed.
A summary of the 0.3m CT silencer measurement results is given in Table 22. Frequency
dependent transmission loss and transmitted speech spectrum are given in Figures 79 and 80.
The acoustic performance of the Straight, L and Z-shaped silencers are quite similar, with
STC 9 (Straight and L) or 10 (Z) and a SEOAs near 0.3. The Z-shaped silencer has slightly
superior performance. Without fiberglass the CT silencer looses essentially all acoustical
value. For airflow, the straight CT silencer has a SEOAf greater than unity, at 1.29. This
effect of enhanced airflow through, in essence, elongated orifices has also been observed and
discussed in the in situ and numerical modeling sections of this thesis (sections 5.3 and 6.3
respectively). The L-shaped CT silencer is more restrictive than an orifice (SEOAf of 0.8);
the Z-shape (SEOAf of 0.68) is more restrictive than the L-shape CT Silencer.
As shown in Figure 79, except for local variations, the transmission losses of the three
silencers are very similar below 4 kHz, increasing from -5 dB at low frequencies up to 15 dB
at 4 kHz. Above 4 kHz the performance of the L-shaped silencer is superior. This can be
explained by the geometry of the silencers; the L-shaped silencer is the only one of the three
that had no unobstructed line-of sight path through it. High frequencies are directional will
“beam” through any unobstructed path. As is characteristic of absorptive liners, the
attenuation is very small at low frequencies and increases with frequency. Regardless of high
frequency variation in the performance, the SEOAs is essentially identical for all three
silencers because, as shown in Figure 80, the critical bands are at low and mid frequencies.
Negative transmission loss at very low frequencies can be explained by wave effects (see
section 2.3).
117
Comparing the straight silencer with and without an absorptive liner we see that they have
essentially identical performance below 250 Hz, at which point the liner becomes effective.
While the duct alone does have a small effect, above 500 Hz the vast majority of attenuation
is due to the presence of the liner.
25
Straight
20 L
Z
15 Straight w/o liner
Transmission Loss [dB]
10
-5
-10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 79: Measured transmission loss of 0.3 m CT silencers.
118
0
-5
-10
Transmitted Speech Spectrum [dB]
-15
-20
-25
-30
-35 Straight
L
-40
Z
-45 Straight w/o liner
-50
-55
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 80: Measured transmitted speech spectrum of 0.3 m CT silencers.
The silencer with the best performance in terms of OAR is the baffle with a 40 mm offset. It
caused some restriction in airflow, with a SEOAf of 0.49; however, the SEOAs is reduced to
0.09, resulting in an OAR of 5.43.
Following the baffle in performance are the CT silencers with OARs of 4.07, 2.44, and 2.32
for the Straight, Z, and L-shapes. The CT silencers all have similar acoustic performance,
with SEOAs around 0.3 which is limited by low frequency transmission. CT silencers,
especially the straight silencer, have good performance due to low airflow resistance. The
SEOAf of the straight silencer is 1.29, and reduces to 0.8 and 0.68 for the more obstructive L
and Z-shaped silencers.
119
Next in performance is the Louver. Its SEOAs of 0.12 is nearly as low as the baffle; however,
its SEOAf is much worse at 0.22, resulting in a maximum OAR of 1.79. Adding fiberglass to
the floor below the louver reduced the SEOAs to 0.075 resulting in an improved OAR of
2.79.
The slot opening, lined with 50 mm of fiberglass follows the louver in terms of performance.
It has great airflow properties, with a SEOAf of 1.1; however, it does little to attenuate noise,
with a SEOAs of 0.94. As a result the OAR is only 1.15. It is important to realize that, while
the SEOAf is near unity, the fiberglass lining has a positive effect. Because of diffraction and
sound field coherence in the corner (see section 2.3) the acoustic performance of the slot
opening without any lining is much worse, with the SEOAs increasing to 2.38.
Using acoustical fiberglass or fibrous air filters as a combined silencer and filter did not
provide good performance. The acoustical fiberglass attenuates noise, but is detrimental to
the airflow; the fibrous filters allow better airflow, but provide very little attenuation of
sound. The best performance was from the white air filter, providing an OAR of 0.39.
An interesting observation from all ventilation opening types is that the acoustic performance
in terms of the SEOAs is limited by the 500 Hz octave band. Any improvements on these
silencers must therefore seek to improve the transmission loss around this frequency.
120
Table 23: Laboratory measurements results summary.
Class Configuration STC SEOAs SEOAf OAR
Straight 9 0.32 1.29 4.07
L 9 0.34 0.80 2.32
CT Silencer
Z 10 0.28 0.68 2.44
Straight w/o fiberglass 2 1.12 - -
No FG -3 2.38 1.09 0.46
1m x 1m x 2.5cm FG 3 1.43 1.09 0.76
Slot 1m x 1m x 5cm FG 4 0.94 1.09 1.15
1m x 0.5m x 5cm FG 3 1.03 1.09 1.06
Away from corner -2 2.03 1.27 0.71
1 12 0.13 0.22 1.68
2 12 0.14 0.21 1.51
Louver 3 12 0.12 0.22 1.79
4 12 0.12 0.21 1.71
5 15 0.076 0.21 2.79
2 cm 15 0.054 0.20 3.70
Baffle
4 cm 14 0.090 0.49 5.43
Acoustic Fiberglass 6 0.14 0.0051 0.014
Filter Pink Air Filter 2 0.65 0.20 0.31
White Air Filter 1 0.73 0.29 0.39
121
Chapter 6: Ventilation Silencer Performance Prediction
Numerical methods exist, and are commonly used, for solving problems involving both
sound propagation and air flow. Using these methods for performance prediction can be
highly useful in two ways:
1. Time, facilities, and cost required to do optimization and validation through
physical experiments can be greatly reduced
2. Numerical methods can provide the entire pressure or flow field solution,
which allows valuable insight into factors that govern the ventilation
openings’ performance.
This chapter will present methods for modeling the attenuation of the fundamental mode in a
lined duct, diffuse field transmission loss using COMSOL FEM, and airflow using ANSYS
Fluent CFD.
122
Using separation of variables to find solutions, the pressure can be solved as the product of
three location-dependent functions and a time dependent function.
p(r , t ) = Px (x )Py ( y )Py ( z )T (t )
By inserting this assumption into the wave equation, spatially dependent variables can be
separated from the time dependent variable, creating multiple ordinary differential equations
from the single partial differential equation.
″ p x (x )
p x (x ) − s x c 2 = 0
″ py (y)
p y (t ) − s y =0
ODEs : c2
″ p z (z )
p z (z ) − s z =0
c2
p ″ (t ) − sp (t ) = 0
t t
where s x + s y + s z = s . Differential equations of this form can hold the following solutions:
sx
x
s
− x2 x
A1e c + A2 e c if s x > 0
2
p x , y or z ( x, y or z ) = A1 + A2 x if s x = 0
−sx −sx
j x −j x
A
1 e c2
+ A2 e c2
if s x < 0
A1e st + A2 e − st if s t > 0
pt (t ) = A1 + A2 t if st = 0
j − st
A1e + A2 e − j − st if s t < 0
Our interest is in the harmonic solution where sx, t<0. It is convenient and informative to
introduce the wave number, k, and angular frequency, ω at this point. The wave number is
the number of radians per unit distance in the direction of the associated coordinate; the
frequency is the number of radians per unit time at one location. Letting − s x = k x c 2 and
2
123
p = p x ( x ) p y ( y ) p z ( z ) pt (t )
p x ( x ) = A1e x + A2 e x
jk x − jk x
p y ( y ) = A3 e
jk y y − jk y
+ A4 e y
where : p z ( z ) = A5 e jk z z + A6 e − jk z z
pt (t ) = A7 e + A8 e
jωt − jωt
2 ω2
k x + k y + k z = k = 2
2 2 2
c
This solution form represents waves, with some amplitude and wave number, propagating in
the positive and negative directions on each axis, and propagating in both directions with
respect to time.
c2
lπ
k l = L l = 0,1, 2, 3K
x
where
k m = mπ m = 0,1, 2, 3K
Ly
124
By letting kl + k m = klm and solving for the wave number in z some properties of the
2 2 2
ω2
kz = − k lm
2
2 35
c
For relatively high frequencies or ducts of large cross section with respect to a given mode,
the pressure fluctuates sinusoidally with z:
ω2
− klm > 0
2
2
c
ω2
−j − klm 2 z
p = A cos(kl x ) cos(k m y )e e jωt c2
When the frequency becomes low, or the duct is small with respect to a given mode, the
wave number becomes complex resulting in a pressure that decays exponentially with
increasing z. This is known as the cut-off frequency for a mode in a duct. The only mode that
does not have a cut-off frequency is the plane wave mode (l=0, m=0).
ω2
− k lm < 0
2
2
c
ω2
− k lm 2 − z
p = A cos(k l x ) cos(k m y )e jωt e c2
125
Assuming that ux and uy have solutions that vary sinusoidally with time, it follows that:
− 1 dp x − kx
ux =
jρ 0ω dx
=
ρ0k c
(
A1e jk x x − A2 e − jk x x )
− ky
uy =
− 1 dp y
jρ 0ω dy
=
ρ0k c
jk y − jk y
A3 e y − A4 e y ( )
Solving for the impedance at the duct walls (hx, -hx, hy, -hy) gives:
A1e jk x hx + A2 e − jhx
Z s , x (hx ) =
− kx
ρ0k c
(
A1e jhx − A2 e − jhx )
A1e − jk x hx + A2 e jhx
Z s , x (− hx ) =
− kx
ρ0k c
(
A1e − jhx − A2 e jhx )
jk y h y − jk y h y
+ A4 e
Z s , y (h y ) =
A3 e
− ky
ρ kc
(A e 3
jk y h y
− A4 e
− jk y h y
)
0
− jk y h y jk y h y
+ A4 e
Z s , y (− h y ) =
A3 e
− ky
ρ kc
(A e 3
− jk y h y
− A4 e
jk y h y
)
0
If the impedances of opposite walls are equal, the simplifying assumption can be made that
the propagating modes will be either symmetric or antisymmetric [31]. For symmetric mode
propagation:
A1 = A2
A3 = A4
Z s , x (hx ) − Z s , x (− hx ) −k
= = cot (k x hx )
Z0 Z0 jk x
Z s , y (h y ) − Z s , y (− h y ) −k
= = cot (k y h y )
Z0 Z0 jk y
126
Z s , x (hx ) − Z s , x (− hx ) k
= = tan (k x hx )
Z0 Z0 jk x
Z s , y (h y ) − Z s , y (− h y )
tan (k y h y )
k
= =
Z0 Z0 jk y
Re-written, the system of equations for a duct in which opposite walls have equal impedance
is:
jkZ 0
k x tan (k x hx ) =
Z s,x
jkZ 0
k y tan (k y h y ) = Symmetric
Z s, y
k x + k y + k z = k 2
2 2 2
36
− jkZ 0
k x cot (k x h x ) =
Z s,x
− jkZ 0
k y cot (k y h y ) = Antisymmetric
Z s, y
k x + k y + k z = k 2
2 2 2
One important observation from this analysis is that, if the wall impedance is not infinite, the
wavenumbers will be complex. If the wavenumber in z is complex, p3, the pressure variation
with respect to the ducts length, can be written:
p3 ( z ) = A5 e j Re (k z )z e − Im (k z ) z
This result shows that the modal pressure decays exponentially along the length of the duct.
The attenuation can be conveniently expressed in dB as:
Attenuation = 20 Im(k z )z log(e ) = 8.686 Im(k z )z dB
127
6.1.3.1 Defining the Surface Impedance
A solution for the plane wave attenuation in a lined duct has been presented; however, it is
required to know the surface impedance of the absorptive liner. The transfer function method
is presented here as a simple method for converting an absorptive material’s propagation
impedance and wave number into a surface impedance. A brief background will also be
given on absorptive materials to describe how the propagation impedance and wavenumber
are determined.
( A1 − A2 )
u x (0 ) =
ρω
p x (d x ) = A1e − jk x d x + A2 e jk x d x
u x (d x ) =
(
k x A1e − jk x d x − A2 e jk x d x )
ρω
Subscript x indicates the component of the variable in the x direction. Combining these
equations gives:
ρω
p x (0 ) = p (d x ) cos(k x d x ) + ju (d x ) sin (k x d x )
kx
jk x p (d x )
u x (0 ) = sin (k x d x ) + u (d x ) cos(k x d x )
ρω
which can be equivalently expressed in matrix form as:
p (0 ) p (d x )
u (0 ) = [T ] u (d )
x
128
ρω
cos(k x d x ) + j k sin (k x d x )
[T ] = jk x
x
sin (k x d x ) + cos(k x d x )
ρω
[T] is the transfer matrix for a finite thickness layer. Transfer matrices can be defined for
many different simple geometries, and multiplied together to find the total transfer function
for compound layers and geometries. Here, we see that if we let the surface impedance at x =
d be Z s , x (d x ) , we can solve for the surface impedance at x = 0, Zs(0):
k
Z s , x (d x ) cot(k x d x ) + jZ 0
kx
Z s , x ( 0) =
Z s , x(d x )k x
j + cot(k x d x )
Z0k
If the layer is backed by a rigid surface then Z s , x (d x ) is effectively infinite and the surface
This result can be used, in combination with Eq. 36, to define the surface impedance of a
duct, provided the propagation impedance and wavenumber, Z 0 and k respectively, are
known for the porous absorber. For clarity, from here on the impedance and wavenumber of
the porous absorber will be identified as Z w and kw. The symmetric equations are:
kw jkZ 0
− jZ w cot(k w, x d x ) = j cot (k x hx )
k w, x kx
jkZ 0
cot(k w, y d y ) = j cot (k y h y )
kw
− jZ w Symmetric 38
k w, y ky
kx + k y + kz = k 2
2 2 2
This formulation allows for arbitrary incidence angle; however, kw,x must be found using
Snell’s law, as refraction occurs due to the difference in wave speed in air and a porous
absorber. With ψ and φ being the incident and transmitted angles, kw,x is [41]:
k x , w = k w 1 − sin (φ ) = k w − k 2 sin (ψ )
2
129
In practice, the wave speed in many porous materials is much smaller than it is in air; thus
the waves propagate nearly normal to the surface [63]. Considering this effect, k w, x ≈ k w .
Materials in which sound will only propagate normal to the surface are referred to as ‘locally
reacting’. The surface impedance of a rigidly backed, locally reacting absorber is:
Z s , x = − jZ w cot(k w d x )
The local reaction assumption can be expected to produce accurate results provided R<4 [63],
where R is the normalized flow resistance given by:
σd
R=
ρ 0c
where σ is the flow resistivity in MKS Rayl/m.
Porous absorbers are, as shown above, most usefully described in terms of their acoustic
propagation impedance and wavenumber. Many methods have been developed, both
empirical and analytical, to determine the acoustic impedance based on material properties
[41, 63]. Analytical methods based on models of the microscopic fluid domain have proven
successful; however, they are quite complicated as compared to the empirical methods.
Empirical methods, such as the long-popular and well-known Delaney-Bazley model [41],
provide a simple method of calculating the impedance from easily measured properties.
The Delaney-Bazley model is based on a data curve-fit of many samples of fibrous acoustic
absorbers with different flow resistivities; therefore, it should not be expected to give
accurate results for non-fibrous absorbers such as open-cell foams. The acoustic impedance
and wavenumber of a fibrous porous absorber are given as [41]:
130
Zw
= 1 + 0.0571X −0.754 − j 0.087 X − 0.732
Z0
ω
kw =
c0
(1 + 0.0978 X − 0.700
− j 0.189 X −0..595 )
X is a function of the flow resistivity ( σ ) and frequency:
ρ0 f
X =
σ
The Delaney-Bazley model is a single log-linear curve fit of the impedance and
wavenumber’s real and imaginary components to represent all fibrous absorbers. Its validity
is held when [41]:
• ε (porosity) ≈ 1
• 0.01 < X < 1.0
• 1000 < σ < 50,000 MKS Rayl/m
Z s − ρ0c
r=
Z s + ρ0c
131
The normal incidence absorption coefficient of a 1” thick fiberglass sample has been
measured using an impedance tube and a standardized measurement procedure [64]; the
fiberglass is the same material that was used in the CT silencers in section 5.5.5. A
comparison between the measured fiberglass material (called OEM, see Appendix A.1) and
the Delaney-Bazley prediction for different flow resistivities is shown in Figure 81.
Absorption data above 2000 Hz does not exist due to impedance tube limitations. Above 500
Hz the predicted absorption agrees best with the measurement when the flow resistivity is
60,000 MKS Rayl/m; however, below 500 Hz Delaney-Bazley under-predicts the measured
absorption. Using a flow resistivity higher than 60k would slightly increase the low
frequency absorption; however, it would step outside of the range of validity of the local
reaction assumption (see section 6.1.3.1). Direct measurements of the OEM fiberglass by
UBC Mechanical Engineering student Michael Gosselin showed the flow resistivity to be
46,000 MKS Rayl/m [65]. In conclusion, reasonable normal incidence absorption agreement
exists for σ = 60k MKS Rayl/m.
Absorption is also strongly dependent on the liner thickness. Using a material with a flow
resistivity of 60,000 MKS Rayl/m, the Delaney-Bazley model was used to calculate the
absorption coefficient of a layer of fiberglass with varying thickness. Results are shown in
Figure 82. All liner thicknesses generally increase in absorption with increasing frequency.
Above 1 kHz all three liners have high absorption. Decreasing thickness results in decreased
absorption at low frequency. The 25 mm liner is effectively incapable of absorbing in the 125
Hz octave band; only modest absorption is achieved in the 125 Hz band with a 100 mm liner.
132
1
0.5
0.4
0.3
0.2
0.1
0
125 250 500 1000 2000
Frequency [Hz]
Figure 81: Absorption coefficient of OEM fiberglass measured and predicted by the Delaney-Bazley
model for different flow resistivities. σ – [MKS Rayl/m]
0.9
Normal Incidence Absorption Coefficient
0.8
dy = 25 mm
0.7
dy = 50 mm
0.6
dy = 100 mm
0.5
0.4
0.3
0.2
0.1
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 82: Variation of normal incidence absorption coefficient for various liner thicknesses as predicted
by Delaney-Bazley, σ = 60,000 MKS Rayl/m.
133
6.1.4.2 Cross Sectional Dimensions
To optimize the performance of a straight section of lined duct one must consider the effect
of the silencer flow path dimensions, lining thickness, and the acoustic properties of the liner.
The height and width of the flow cavity in the silencer will have a great effect on both the
acoustic attenuation; the cross section geometry was examined by looking into the effect of
flow path height and aspect ratio, and how the behaviour is dependent on liner thickness. As
required for Eq. 38, the silencer height is equal to 2h, and the liner thickness is d (Figure 83).
The plane wave attenuation is determined by solving Eq. 36 and Eq. 37 using the Newton-
Raphson numerical iteration scheme.
2h
It is apparent that at low frequencies the absorption is governed by the absorber thickness.
Below 1000 Hz the performance of the silencer with a 25 mm liner falls off relative to that of
the 50 and 100 mm liners. Likewise, below 250 Hz the attenuation with 50 mm liner falls off
with respect to the 100 mm thick liner. This result corresponds to the normal incidence
absorption coefficient results shown in Figure 82.
134
Above 250 Hz for the 50 mm liner, and 1000 Hz for the 25 mm liner, the transmission loss is
not governed by the thickness of the liner (although it may yet be affected by the flow
resistivity). In this region the transmission loss is limited by the rate at which energy in the
fundamental mode can diffract into the absorptive material. In all cases, the frequency at
which the rate of attenuation is maximized is very close to the frequency at which the
wavelength is equal to the duct height (2h).
200
180
160
Transmission Loss [dB/m]
140
120
100
80
60
40
20
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 84: Predicted transmission loss for various duct heights and liner thicknesses: Blue – d=25 mm,
green – d=50 mm, red – d=100 mm, solid – h=20 mm, dash – h=50 mm, dot – h=100 mm.
135
100
Transmission Loss [dB/m]
10
0.1
136
thicknesses; however, as before, the transmission losses are more pronounced at lower
frequencies for thicker liners.
The increase in the absorption of a lined duct with a small aspect ratio should be expected.
With a 2D duct the wave front will form a 2D arc as it diffracts into the liner. Because the
length of an arc increases proportionally to the arc radius, the maximum energy attenuation
rate is inversely proportional to the radius. In a 3D duct with AR = 1, the wavefront will
approximate the spherical end of a 3D cone as it diffracts into the liner. The area of a sphere
increases proportionally to the radius squared; therefore the maximum attenuation rate is
inversely proportional to the radius squared. As transmission loss is energy attenuation on a
log scale, this suggest that the transmission in a duct with an aspect ratio of 1 loss will be
twice as large as in a 2D duct (or equivalently AR>10) with the same height. Figures 86, 87,
and 88 suggest that the transmission loss with AR = 1 is indeed nearly twice the 2D value for
any duct configuration and frequency.
140
2D
120
AR=10
AR=5
Transmission Loss [dB/m]
100 AR=2
AR=1
80
60
40
20
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 86: Predicted transmission loss for various aspect ratios: hy=50 mm, hx=hy*AR, dy=dx=25 mm.
137
140
2D
120
AR=10
AR=5
Transmission Loss [dB/m]
100
AR=2
80 AR=1
60
40
20
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 87: Predicted transmission loss for various aspect ratios: hy=50 mm, hx=hy*AR, dy=dx=50 mm.
140
120 2D
AR=10
AR=5
Transmission Loss [dB/m]
100
AR=2
AR=1
80
60
40
20
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 88: Predicted transmission loss for various aspect ratios: hy=50 mm, hx=hy*AR, dy=dx=100 mm.
138
6.1.5 Attenuation of the Fundamental Mode Summary
Through comparison of the absorption coefficients, it was determined that a fibrous material
with a flow resistivity of 60,000 MKS Rayl/m, as defined by the Delaney-Bazley model, has
similar acoustical performance to the fiberglass liner used in the laboratory measurements of
this thesis. Using this material with an analytical solution for plane wave attenuation in a
lined duct, the effect of varying the duct’s cross sectional dimensions has been analyzed,
providing information about how the liner thickness, duct height, and duct aspect ratio affect
attenuation.
Duct liner thickness does not affect high-frequency performance; however, it limits low-
frequency performance. The performance of a 25 mm liner falls off below 1000 Hz, and a 50
mm liner falls off below 250 Hz. From the ventilation opening laboratory measurements
(section 5.5) it was observed that the ventilation opening silencer’s performance is often
limited by the 500 Hz frequency band; as a result, the 25 mm liner is likely not thick enough
to be effective; however, the 100 mm liner may be excessive. Increasing the duct height will
reduce attenuation at all frequencies; however, if the frequency is high enough, or the duct is
large enough, so that the wavelength is shorter than the duct height, the attenuation will
decrease rapidly. In order to provide effective attenuation through the 4000 Hz band, the duct
height should not exceed 100 mm.
If the aspect ratio of a duct is greater than 10, the attenuation of its fundamental mode is in
effect identical to that of a 2D duct. Provided the duct liner and dimensions are such that the
2D silencer is effective at absorbing sound at a given frequency, reducing the aspect ratio
will result in large attenuation gains. The TL of a lined duct with AR=1 is approximately
twice the transmission loss of a 2D lined duct.
139
analytical and experimental results. The performance of various CT silencers is predicted and
discussed.
To reduce the FEM computational expense at high frequencies, a new source volume was
defined for each third-octave band, maintaining a diffuse sound field but not solving an
excessively large domain. One must be aware that, in using this approach, the dimension-to-
wavelength ratios are identical for each third octave band. This will cause the error due to
imperfections in the diffuse field to be correlated between each third octave band, thereby
reducing the apparent variation between successive third octave band predictions.
4
Geometric models that model wave effects such as interference, diffraction, and scattering [87] may have
potential in this area. Further study is required.
140
Reducing the source volume with increasing frequency allows calculation of the transmission
loss until one of two high frequency conditions is reached. The first high frequency limit that
may be reached is when the silencer’s volume alone is too large for the computer to
discretize and store in memory. Secondly, if the source volume is continually reduced with
increasing frequency, at some limiting frequency, the silencer will no longer fit on its
surface. To extend the solution to higher frequencies it may be possible, depending on the
geometry of the silencer, to use a 2D model without excessive loss in accuracy.
Previous work has proven FEM to be effective at predicting the steady-state sound field in
rooms at low frequencies [67-69], and the transmission loss in ventilation duct silencers and
mufflers [70-73]. The present work has made the additional step of combine the FEM for
rooms and silencers. COMSOL 4.1 is used for the simulations. COMSOL and MATLAB
were used for post processing. A computer with an Intel i7 processor and 24 GB of memory
was purpose-built for these simulations. Run times are typically under 24 hours; however,
they are highly dependent on geometry, mesh, and frequency resolution.
141
Source Volume Sampling Sub-domain
A sub-domain within the source volume, shown in Figure 91, is created to define the region
that is sampled in order to determine the average sound pressure level in the source volume.
This is the sub-volume in which the sound field is diffuse.
142
an anechoic termination. All surfaces of the PLM, besides the one adjacent to the receiver
volume, are hard.
Figure 89: 3D acoustic domain. Figure 91: Source volume sampling sub-domain.
Figure 90: Source volume sub-domain. Figure 92: Ventilator flow path sub-domain.
143
Figure 93: Ventilator liner sub-domain. Figure 95: PML sub-domain.
Figure 94: Receiver volume sub-domain. Figure 96: Point sound power source.
144
Ii(x, f) Acoustic intensity in direction i as a function of frequency and position
Win(f) Sound power through inlet as a function of frequency
Wout(f) Sound power through outlet as a function of frequency
WLp(f) Apparent sound power through inlet as estimated from Lp(f) as a
function of frequency
TL_Win(f) Transmission loss calculated from the inlet and outlet sound powers as
a function of frequency
TL_WLp(f) Transmission loss calculated from WLp(f) and the outlet sound power
as a function of frequency
TL_Win Band average transmission loss calculated from TL_Win(f)
TL_WLp Band average transmission loss calculated from TL_WLp(f)
If P, a set of n data points, is linearly distributed over the frequency range f i → f i +1 , then the
logarithmic average of the data over that frequency band can be found as:
1 n
f i +1
Plog =
f
∑ P log i
log n
i =1 fi 39
f1
145
vary rapidly with frequency, Eq. 39 is accurate; however, a more accurate representation may
be:
1 n
P(i ) + P(i + 1) f i +1
Plog =
f
∑ 2
log
log n
i =1 fi
f1
The average given by Eq. 39 is used in this work. From experience, the correction in using a
logarithmic average results in a small, but significant, difference of up to 0.5 dB in third-
octave bands. This correction in octave band averages, or for averages of spectrum that vary
more rapidly with frequency, will result in larger differences.
6.2.1.2.2 Lp (x, f)
The sound pressure level is calculated in the conventional way as:
p rms 2
Lp ( x, f ) = 10 * log10
2e −5 Pa
( )2
6.2.1.2.3 Lp (f)
To average the sound pressure level over a volume, the solution points defined by the mesh
are connected by a 4th-order fit, and an integral is calculated to determine the average. Before
integrating, the log scale is removed so that the result is the average squared sound pressure,
corresponding to sound power, instead of the average sound pressure level.
Lp ( x,f )
1
Lp ( f ) = 10* log10 ∫∫∫10 10
∂x∂y∂z
V V
An analogous calculation is made to average the sound pressure level over a surface or line.
6.2.1.2.4 Lp (x)
Band average sound pressure level as a function of position is calculated by finding the
average of the squared pressures, corresponding to an energy average. With respect to Eq. 39,
the energy average is found by:
Lp ( x,f i )
Pi = 10 10
146
Energy averaging is appropriate because it is representative of a random phase relationship
between the various frequencies [34]. It is not physically possible to have a fixed phase
relationship at different frequencies, as would be implied by summing sound pressures
instead of energy.
6.2.1.2.5 σLp
The sound pressure level standard deviation, in this work and context, is calculated from the
band average sound pressure level to describe the variation of a band average sound pressure
level over the volume, surface or line:
σ Lp =
1 n
[
∑ Lp ( xi ) − Lp
n − 1 i =1
]
2
1 n
Lp = ∑ Lp( xi )
n i =1
Note that Lp is a decibel average, not an energy average. Additionally, the standard
deviation is calculated from discrete points instead of from a continuous integral, as done for
Lp(f) because the data needs to be exported from COMSOL into MATLAB to calculate band
average values. Apparently it is possible to purchase a ‘livelink’ between MATLAB and
COMSOL, allowing the user to drive COMSOL from MATLAB and increase its post
processing computational flexibility. For now, because the mesh is so spatially uniform,
using a discrete representation without area weighting should give a very good
approximation to the continuous representation.
6.2.1.2.6 Ii(x, f)
The acoustic intensity is the acoustic power per unit area. From a frequency domain solution
it can be calculated from the product of the acoustic pressure and the complex conjugate of
the particle velocity in direction i:
I i ( x, f ) = pu i
*
147
∂p ∂u
= ρ0 i
∂i ∂t
Because the velocity varies harmonically with time,
∂u i
= − jkc u i
∂t
Therefore the particle velocity is:
j ∂p
ui =
kρ 0 c ∂i
6.2.1.2.8 WLp(f)
In order to calculate the predicted transmission loss for comparison to measured results, it is
necessary to calculate the predicted sound power entering the ventilation opening in the same
way as in experiment. The standard method is to measure the average sound pressure level in
the source room, assume that the sound field is diffuse, and use the sound pressure level to
determine the intensity on the duct inlet.
Calculating the inlet sound power in a 2D model can be achieved with the standard method
used in the 3D experiment method; however, an adjustment needs to be made. The inlet
sound power is calculated by assuming the sound field is composed of plane waves travelling
in all directions with equal intensity. The intensity of a plane wave is related to its pressure
by:
p2
I=
ρ 0c
The intensity passing through a plane, in one direction, is equal to the integration of the
intensity, normal to the plane, over all angles of incidence (hemisphere):
148
π
1 p2 1 2
2π cos(φ )sin (φ )dφ
2 ρ 0 c 2π ∫0
I n ,3 D =
Here the 1 2π term is the area of a hemisphere with unity radius, and φ is the azimuthal
angle. The p 2 term is multiplied by one half because only the contribution of the pressure
waves travelling in the positive direction through the plane is desired. A 90 degree azimuthal
angle corresponds to normal incidence:
π
1 p 2 2 sin (2φ )
2 ρ 0 c ∫0 2
I n ,3 D = dφ
p2
I n ,3 D =
4ρ 0 c
In the 2D model the angles of incidence, and thus the integration, take place over a semicircle
instead of a hemisphere:
π
1 p2 1
sin (φ )dφ
2 ρ 0 c π ∫0
I n,2 D =
1 p2
I n,2 D =
π ρ0c
Note that the intensity through the plane with respect to the pressure is higher in the 2D
model. This makes sense as, in the 3D case, the area weighting for high azimuthal angles is
small compared to low azimuthal angles, approaching zero at 90 degrees (normal incidence).
In the 2D case, all azimuthal angles have the same weight. This illuminates a significant
difference in the models. In a 2D model, a greater portion of the energy will strike the
ventilation opening (and any other surface) at large angles of incidence.
6.2.1.2.9 TL_Win(f)
The transmission loss calculated directly from the sound power entering the ventilator is
calculated as:
W ( f )
TL _ Win ( f ) = −10 * log10 out
Win ( f )
149
6.2.1.2.10 TL_WLp(f)
The transmission loss calculated indirectly from the sound power entering the ventilator is
estimated from the source volume sound pressure level and the actual sound power exiting
the ventilator:
W ( f )
TL _ WLp ( f ) = −10 * log10 out
W (f )
Lp
Pi = 10 10
Pi = 10 10
1. Mesh convergence
For a finite element model to be physically correct it is necessary to use small
enough elements so that the spatially discrete model creates a sufficiently
accurate reproduction of the spatially continuous model.
150
2. Frequency Convergence
The frequency domain must also be sampled with a sufficient resolution so
that, when averaged over third octave bands, the result is representative of the
result for continuous frequency sampling.
151
4
3.5 W in
W out
3
W Lp
Sound Power [dB]
2.5
1.5
0.5
0
0 5 10 15
Elements per wavelength
Figure 97: Variation of the predicted sound power with mesh resolution.
There is an additional complication, in that we know the speed of sound is slower in a porous
absorber; therefore, the wavelength in the porous absorber will be relatively short. A separate
size condition was set for the absorber domain so that the condition of at least 5 elements per
wavelength is respected. Based on the Delaney-Bazley model (section 6.1.3.1) the absorber
wavelength ( λw ) can be written as a function of its wavenumber ( k w ):
2π
λw =
kw
ω
kw =
c0
(1 + 0.0978 X − 0.700
− j 0.189 X −0..595 )
152
more modal overlap, possibly leading to a smaller standard deviation of the sound field with
respect to frequency. More absorption will provide more damping, thus lower quality factor
of the modes, and again more modal overlap. Excessive absorption will lead to a greater rate
of decrease of levels away from the source, thereby increasing the standard deviation of
levels in the sound field. Examined here is the number of equally-spaced frequency samples
required to accurately represent a continuous third-octave band.
Figures 98 and 99 show, respectively, the error in the average source volume sound pressure
level, and the spatial standard deviation in sound pressure level, as a function of the number
of frequency samples in one third-octave band. The reference sound pressure level for Figure
98 is taken from the result for 200 frequency samples. Curves are shown for different values
of r, the source volume surface reflection coefficient (0.98, 0.95, 0.9), and different values of
N, twice the ratio of the minimum source volume dimension to the maximum wavelength in
the third octave band (4 and 6). The source volume size is described in terms of N for
convenience, and is explained below. Source volume absorption is implemented by defining
the source volume surface boundary condition as ‘finite impedance’. The impedance is
written as [41]:
1+ r
Z = ρ0c
1− r
Figure 98 shows that the calculated average source pressure level has converged to within 0.1
dB of the true value when averaged over 10 or more frequencies in the third octave band, in
all cases.
As shown in Figure 99, the sound field standard deviation is strongly dependent on the
frequency resolution below a sampling rate of 10 frequencies per third octave. Above this
sampling rate, the standard deviation is nearly constant, indicating that 10 samples are
sufficient to calculate the sound field standard deviation. For low absorption surfaces,
increasing the source volume dimensions from N=4 to N=6 makes little difference; however,
for high absorption (r = 0.9) the standard deviation decreases approximately 0.5 dB.
Decreasing the absorption from r = 0.90 to r = 0.98 is responsible for approximately 1 dB
decreases in standard deviation. Erratic, highly varying results are obtained if the source
153
volume is entirely, reflective (r =1). As a result, the source volume should be highly, but not
completely, reflective. If a ventilator opening represents a large absorptive area, the sound
field standard deviation may increase.
Sampling the frequency domain 20 times in each third octave band is sufficient to provide an
accurate prediction of the sound pressure level’s average and standard deviation values. This
result holds for the source volume; however, a ventilator domain may have a more
frequency-dependent response. If the behaviour of a ventilator appears to be highly
frequency dependent it may be of interest to increase the frequency sampling resolution and
check if the third octave solution has changed.
Due to its energy-absorptive behaviour, the presence of a ventilator may increase the
standard deviation of sound pressure level with respect to frequency. As a result, the standard
deviation corresponding to each calculated average should be determined in order to assess
the significance of the result.
0.15
r=0.98, N=4
0.1 r=0.98, N=6
r=0.95, N=4
r=0.95, N=6
Average Lp error [dB]
-0.05
-0.1
154
4
r=0.98, N=4
3.5 r=0.98, N=6
r=0.95, N=4
Lp Standard Deviation [dB]
r=0.95, N=6
3
r=0.90, N=4
r=0.90, N=6
2.5
1.5
1
5 10 20 50 100
Frequencies per third octave band
Figure 99: Lp Standard deviation as a function of frequency sampling resolution for various source
volume dimensions and reflection coefficients.
155
the ratio 1:21/3:41/3 [74, 75], a well-known and often-implemented ratio for even modal
spacing, was used.
A high modal density is required so that the sound field is uniform with respect to frequency
and location. Modes in a room with three sets of parallel walls, all at right angles to each
other, occur at frequencies described by [60]:
2 2 2
c nx n y nz
f = + +
2
Lx L y Lz
for all integer values of nx , n y and n z . The room dimensions are Lx , L y and Lz . This
equation shows that the frequency separation of modes becomes larger at lower frequencies.
In addition, because results are averaged over frequency bands (octave or third octave) that
decrease in width with decreasing frequency, lower frequency results will include much
smaller numbers of room modes. Sufficient modal density for a diffuse sound field will
therefore be limited at low frequencies. The low frequency limit for a diffuse field has been
proposed by Schroeder [76]:
T60
f > 2000
V
This relationship provides a criterion for which the response at a single frequency is
composed of at least three modes. In other words, there are at least three modes within the
width of the frequency response of any single mode.
Schroeder’s limit has been shown to be unnecessarily stringent when results are taken from
band-averaged values instead of single frequencies [77]. Another approach that is more
commonly taken for band averaged values is to specify a minimum number of modes in each
band; 20 modes is common [77-79]. For a cubic volume, 20 modes per octave is achieved
when the volume is equal to the wavelength cubed [77], and 20 modes per third octave is
achieved when the volume is equal to four times the wavelength cubed [78, 79]. The
corresponding frequency limits can be written as:
c
f (octave ) > 3
V
156
c
f (third octave ) >
3 V 4
Here N, which is related to the minimum order of axial mode observed, is introduced as a
method of specifying the source-volume dimensions relative to the wavelength. N is
calculated as twice the ratio of the smallest dimension to the largest wavelength. In equation
form:
Lmin
N =2
λ max
N can take any value; however, the smallest order of axial mode, if one exists in the band,
will be N rounded up to an integer value.
The relationship between the number of modes and parameter N has been investigated for
rooms of the chosen aspect ratio. Given a room in which the smallest dimension is 1 m and
the aspect ratio is 1:21/3:41/3, the numbers of modes in each third-octave band are shown in
Table 24.
Table 24: Modes in each third-octave band.
Band 100 125 160 200 250 315 400 500 630 800 1000 1250
Modes 1 1 2 3 4 8 14 28 47 97 180 342
Over 20 modes are observed at and above the 500 Hz third-octave band. The ratio of the
room’s smallest dimension (1 m) to the largest wavelength in the 500 Hz third-octave band
(0.77 m at 445.5 Hz) is 1.3. This can be scaled to imply that, for any third-octave band, the
room’s smallest dimension should not be less than 1.3 times the smallest wavelength; thus, N
should be greater than 2.6, and there should not be an axial mode of order less than 3.
It is of interest to test this “diffuse field convergence” with the finite element model.
‘Diffuseness’ of the source volume sound field is assessed by the standard deviation of the
sound pressure level in the source volume. The source volume near the walls is not sampled,
as it cannot be diffuse. Near the walls it is not possible to have a diffuse sound field because
all modes have a pressure maximum at the wall, a result of the zero particle velocity
boundary condition normal to the wall [33]. If, at one frequency, there is equal energy
incident from all directions, then the sound field at the wall will be 2.2 dB higher than its
157
constant value away from the wall. This error is further increased in double and triple
corners. In order to avoid errors of more than 1 dB, the source volume sound field must not
be sampled less than 0.25 λ from the walls and 0.7 λ from the corners [34].
Figure 100 shows the standard deviation of the sound pressure level in the source volume as
it varies with source volume size. At N=4 the standard deviation falls below 2 dB. Above
N=4 there are diminishing returns on increased dimensions; at N=8 the standard deviation is
just under 1.5 dB. Quadratic residue diffusers, tuned to the bottom of the frequency band
being predicted, were added to the surfaces in an attempt to reduce the sound field variation.
The diffuser is a single 2D sequence with 17 elements in each direction [41, 80]. Its well
width is less than one quarter wavelength and it has no periodicity; therefore it should be an
excellent diffuser [80, 81]. An external view of the source volume with diffusely reflecting
surfaces is shown in Figure 101. As shown in Figure 100, very slight reduction in the
standard deviation of Lp was observed by implementing the diffusers, regardless of source
volume size. This result agrees with previous measurement results by the author in an empty
office sized room, shown in Figure 102, where the standard deviation in sound pressure was
not affected by the addition of a diffusely reflecting surface [82]. The diffuser used in this
measurement was known to scatter sound well above 500 Hz; however, no reduction in
sound pressure standard deviation was measured in this range. Due to the increased model
complexity, the addition of diffusers is not deemed to be beneficial; the increased
computational cost to model the diffusers can be more effectively spent on increased room
dimensions. It has been demonstrated, however, that if the source position is changed a room
without parallel walls the sound pressure level standard deviation is less likely to increase
[69]. This may imply that the diffuseness of rooms is more robust to source location if the
surfaces are diffusely reflecting.
158
3.5
Flat surfaces
Scattering surfaces
3
Lp Standard Deviation [dB]
2.5
1.5
1
1 2 3 4 5 6 7 8
N
Figure 100: Standard deviation of Lp vs minimum order of axial mode. Blue – flat surfaces, Red –
diffuser surfaces.
159
0.016
0.012
0.01
0.008
0.006
0.004
0.002
0
63 125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 102: Standard deviation of sound pressure in an empty office sized room, with and without
diffusers [82].
A 2D source volume will have many less modes than a similar geometry in 3D. As such, its
dimensions must be increased relative to the wavelength in order to maintain the modal
density. A 2D source volume with dimension ratios 1:21/3 and a minimum dimension of 1m,
has third octave band modes as shown in Table 25. This can be compared to the results for
the 3D source volume in
Table 24. In order to achieve 200 modes in the source volume for each third octave band the
smallest dimension must be approximately 10 times the largest wavelength (i.e. N=20).
Table 25: Number of modes in each third octave band, 2D geometry (1 x 1.26 m).
Band 100 125 160 200 250 315 400 500 630
Modes 0 1 1 1 1 2 2 4 6
Band 800 1000 1250 1600 2000 2500 3150 4000 5000
Modes 12 17 28 41 64 104 165 253 406
160
Source Surface Sample Area (Ventilation Opening Size and Location)
The ventilator, located on a section of the source-volume surface, must also be exposed to a
diffuse sound field. If we acknowledge that the sound pressure level must be uniform for a
sound field to be diffuse, Figure 100 shows that it is not possible to create a perfectly diffuse
sound field; some variation will exist. Figure 103 shows the predicted third-octave Lp
distribution on the source-volume surface (this is the pressure distribution on the actual
source-volume surface (see Figure 90), not the source-volume-sampling surface (see Figure
91)). Solutions from 60 frequencies have been averaged to produce the third octave result.
The ventilation opening will, however, occupy a non-zero surface area. As a result, the
average sound pressure level on the ventilator will have less variability than the sound
pressure level at any single point on the source-volume surface. This section investigates the
relationship between the ventilation opening dimensions and the uncertainty in the average
ventilation opening sound pressure level. The investigation has been completed, using a
Monte Carlo simulation, by calculating the average sound pressure level on randomly
sampled sub-sections of the source volume surface to determine the standard deviation of the
sample averages. The standard deviation of the sample averages can be used to indicate the
uncertainty in measured transmission loss due to the chosen ventilator location. With
reference to Figure 104, the Monte Carlo process is as follows:
1. Define sample area size (corresponds to the ventilator inlet area)
2. Position the sample area randomly on the source volume surface
3. Knowing the source volume surface sound pressure distribution, calculate the average
sound pressure level on the sample area
4. Repeat steps 2 and 3 many times until an accurate measure of the standard deviation
of the samples’ average sound pressure levels is found.
The standard deviation of the sound pressure level for different sample area sizes provides a
measure of the uncertainty due to the location of the ventilator on the source volume surface
given the dimensions of the ventilator inlet. Initially, the source volume surface used for
sampling was chosen as the largest surface on the source volume; however, the edges of the
surface were not included, in order to avoid the increased sound pressure level at those
locations due to constructive interference (spatial coherence) of the room modes in the
161
double corners. Elevated sound levels near the edge of the surface are clearly shown in
Figure 103. Subsequently the sound field near the edge of the source volume surface is
investigated. As ventilation openings are typically installed adjacent to the ceiling in the
partition, it was of interest to predict their performance in a corner.
Note that, because the sound field was predicted without any ventilation opening, this
analysis assumes that the sound pressure level on the source volume surface is either not
affected by the presence of the ventilator, or equally affected at each location sampled, so
that the relative comparison remains valid.
2.5
4
2 2
y location [m]
1.5 0
-2
1
-4
0.5
-6
0
0 0.5 1 1.5 2 2.5 3 3.5
x location [m]
Figure 103: Third-octave source volume surface Lp distribution. N=6, color scale in dB.
Sample area
h
Source volume
w
surface area
162
Figure 105 shows the standard deviation of the sample sound pressure level averages. It is
tested for square sample areas for which the sample’s side length to wavelength ratio is
varied from 0 to 2, for source volume sizes N=4, N=6 and N=8. The standard deviation is
well explained by the sample side length and not greatly dependent on the source volume
size (in the range tested). For very small samples the standard deviation is limited to around
1.5 dB; this is very similar to the standard deviation observed in the center of the source
volume (Figure 100). Increasing the edge length of a square sample above one half of the
longest wavelength reduces the standard deviation to less than 0.5 dB; however, many
ventilation openings are rectangular instead of square. Figure 106 shows the standard
deviation results for a rectangular samples; it is clear that both dimensions of the ventilator
do not need to be large. If either the width or the height of the sample area (ventilation
opening) is greater than one half of the longest wavelength the standard deviation will be less
than 0.5 dB. From this it can be concluded that, in order to be 95% confident (2 standard
deviations) that the sample average is within 1 dB of the total source volume surface average,
the dimensions of a square ventilator should not be less than 0.5 times the longest wavelength
of sound. This result also has implications for silencers in real rooms – if they are small
compared to the longest wavelengths of sound, the noise isolation they are able to provide
will be dependent on where they are located on the partition.
163
1.5
N=4
N=6
N=8
1
σ Lp [dB]
0.5
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
w/ λ
Figure 105: Standard deviation of sample Lp averages vs. to sample side-length to wavelength ratio for
various source volume dimensions. Square sample surface, away from source volume corner.
1.5
h/ λ = 0.1
h/ λ = 0.5
h/ λ = 1
Lp Standard Deviation [dB]
h/ λ = 2
1
0.5
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
w/ λ
Figure 106: Standard deviation of sample Lp averages vs. sample width to wavelength ratio for various
sample height to wavelength ratios. N=6, sample surface away from source volume corner.
164
If a ventilation opening’s dimensions are smaller than one half of the largest acoustic
wavelength it is still possible to obtain an accurate diffuse-field transmission loss prediction
by modeling it numerous times in different locations until an estimate of the mean
performance can be obtained. The numerous ventilation opening models could be obtained
through either multiple predictions with a ventilator in different locations, or a single
prediction with multiple ventilators positioned on the source volume surface.
Ventilation openings are often located in the edge formed by a partition and the ceiling;
therefore, it is of practical interest to model them at a corner of the source volume, knowing
the sound field there is not diffuse, because it is a better approximation of how the silencer is
used in the field. The Monte Carlo sampling process was used again to determine the
uncertainty in the average ventilation opening sound pressure level, based on its cross
sectional dimensions, given that its edge is coincident with the source volume surface.
Results are shown in Figure 107. It appears that the sound field variation is higher in the
corner than it is in the central area of the surface; however, provided the ventilator dimension
parallel to the edge (length) is greater than one wavelength, the standard deviation is less than
0.5 dB. If the ventilator width is small compared to the wavelength, the standard deviation
will be greater than 0.5 dB, even if the ventilator’s height is large.
165
1.5
h/ λ = 0.1
h/ λ = 0.5
h/ λ = 1
h/ λ = 2
1
σ Lp [dB]
0.5
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
w/ λ
Figure 107: Standard deviation of average sample Lp vs. sample width to wavelength ratio (w/l) for
various sample height to wavelength ratios (h/l). Sample surface at source volume edge.
The 3D model is solved for the 100 to 2500 Hz third octave bands, after which the duct mesh
becomes too fine for the computer to hold in memory. A 2D model is constructed and solved
for the 1600 to 10000 Hz third-octave bands.
166
Figure 108: 3 m lined duct geometry for analytical validation with 503 Hz Lp solution.
Two results from this solution are examined. First, in order to see that the attenuation and
liner performance agree with theory, the transmission loss between the duct cross sections at
1 m, and at 2 m was compared to the analytical solution for plane wave attenuation (section
6.1). The center section (1-2 m) was used because near the entrance there will be higher order
modes (which are rapidly attenuated), and near the outlet the sound field will be affected by
reflections caused by the outlet impedance change (radiation impedance). Secondly, the
transmission loss calculated directly from the sound intensity was compared to the
transmission loss determined from the source volume sound pressure level.
167
agreement, with the 3D case only slightly over-predicting the 2D TL solution. Above 5000
Hz the analytical and 2D solutions are in good agreement. Disagreement exists between 2000
and 5000 Hz, where the 3D and 2D FEM results under-predict the analytical result by up to
12 dB/m. One way in which the analytical model is dissimilar to the FEM is that the porous
layer is modeled as a surface impedance rather than a dissipative acoustic domain. Simply
put, the analytical model assumes local reaction [62] and the FEM does not.
To investigate the effect of assuming local reaction, the 2D FEM model was re-created with
the liner implemented as a complex finite impedance boundary condition instead of as an
acoustic domain. This is, in effect, identical to the boundary condition treatment in the
analytical solution (section 6.1). Figure 110 shows the result – perfect agreement between the
analytical and FEM solutions. This confirms that the disagreement between the analytical
result and the FEM results in Figure 109 can be attributed to the local reaction assumption
made in the analytical solution. The FEM solution, through modeling the absorber as a
dissipative acoustic domain (bulk absorption), should be considered the more accurate result.
This is in agreement with previous research on numerical modeling of silencers, which has
concluded that boundary element methods are inferior to FEM due to their inability to model
dissipative domains [71, 72].
168
70
3D FEM
60 2D FEM
Analytical
Transmission Loss [dB/m]
50
40
30
20
10
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 109: Transmission loss in center section of duct – discrete data points.
70
Analytical
60
2D FEM, Locally Reacting Liner
Transmission Loss [dB/m]
50
40
30
20
10
0
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 110: Analytical solution and 2D FEM solution with locally reacting liner – discrete data points.
169
6.2.2.1.2 Calculating the Inlet Power through Sound Pressure Level or Intensity
It is interesting to compare the transmission loss results based on calculating the inlet power
indirectly from sound pressure levels to that calculated directly from sound field intensity.
Figure 111 shows the third-octave band transmission losses calculated through both measures
of inlet sound power. At high frequencies there is good agreement between the two results;
however, at low frequencies the solutions diverge by up to 11 dB.
Two reasons exist for why the transmission loss solutions diverge at low frequencies, both
having to do with the sound pressure level, and diffuse field acoustic theory, not being a good
predictors of inlet intensity. These two reasons also explain why it is inadequate to study
insertion loss instead of transmission loss, as has been done in previous work [11].
1. As the ventilation opening is situated in the corner, the opening will see around 2 dB
higher sound levels, provided it is less than one quarter wavelength from the corner
[34] (see section 2.3). In the present case, where the duct is 0.1 m tall and in the
corner, elevated sound pressure levels should be expected until around 1000 Hz.
Above 1000 Hz only a fraction of the ventilation opening will be exposed to the
elevated sound field and the effect will be diminished.
2. Diffraction will also cause the sound pressure level to provide an erroneous measure
of intensity. If the wavelength is not much smaller than the dimensions of the
opening, increased sound energy will diffract into the duct due to the high impedance
adjacent walls [60, 62]. Diffraction will continue to increase with increasing
wavelength. The wavelength is smaller than the 0.1 m duct, and diffraction should be
reduced, above 3500 Hz.
The difference in the transmission loss results when the inlet power is calculated directly, as
compared to indirectly from the sound pressure level, is shown in Figure 112; 95%
confidence intervals are provided. As the outlet power is calculated in the same way in both
cases, the difference is entirely due to the inlet sound power calculation method. At 125 Hz
the inlet power is under-predicted by over 11 dB using the sound pressure level. Above 1000
Hz, and moving towards 3500 Hz, the sound pressure level is able to effectively predict the
inlet intensity.
170
There are two reasons why it is appropriate to use the sound pressure level for calculating
inlet power, even though it is not an accurate measure of actual inlet power. First, the
standardized transmission loss measurements are done this way [55, 56]; the design
community is used to interpreting and designing with them. Secondly, the transmission loss,
calculated based on the ratio of squared source sound pressure level to receiver sound
pressure level, instead of its actual definition of inlet sound power to outlet sound power, is a
more appropriate measure of its ability to isolate spaces.
Figure 111 also shows that the 2D model tends to under-predict the 3D transmission loss by
around 5 dB in the range of solution overlap. A likely cause is that the 2D model does not
include the higher order duct modes that exist in three dimensions. As higher order modes
decay more rapidly, the 2D model would under-predict the transmission loss. In addition,
because the angles of incidence are on a hemisphere in the 3D model, but on a semicircle in
the 2D case, lower angles of incidence are more heavily represented in the 3D model. Lower
angles of incidence will more easily couple with higher order and more rapidly attenuated
modes. The 3D solution will be considered as the more accurate result, and the 2D solution
will be used with caution, especially if the silencer’s performance is limited by high
frequency transmission.
171
180
160
3D TLWin
140 3D TLLp
120 2D TLWin
2D TLLp
100
TL [dB]
80
60
40
20
-20
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 111: Predicted transmission loss calculated using source sound pressure level and inlet intensity.
14
3D Prediction
12
2D Prediction
10
8
TLLp - TLWin [dB]
-2
-4
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 112: Difference in transmission loss result when calculated using source sound pressure level and
inlet intensity. 95% confidence intervals provided.
172
6.2.2.2 Comparing Measured Results to FEM Prediction
FEM predictions were compared to measurements made in a lab facility (ref section) for
validation purposes. Comparisons were made for a simple unlined slot opening and 0.3 m CT
silencers. 95% confidence intervals were plotted with the transmission loss results; for the
measured performance the uncertainty is associated with spatial variation – for the modeled
performance the uncertainty is associated with spectral variation. FEM predictions, as
plotted, are composed of the 3D solution up to and including the 2500 Hz third octave band,
and the 2D solution at and above the 3150 Hz octave band.
An important difference to note between the measurements and the predictions is that the
prediction results determine the transmission loss at each frequency, and then average those
values, reducing the effect of non-flat sound-source and source-volume frequency responses.
Physical measurements find the frequency-band sound pressure levels, and then calculate the
transmission loss from them. As such, the frequency-band transmission loss will be affected
by the frequency response of the source and source-volume within the frequency band. The
transmission loss will be more representative of any specific frequency which was relatively
loud within the frequency band.
173
6
0
Transmission Loss [dB]
-2
-4
-6
-8
-10 Measured
Predicted
-12
-14
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 113: Measured and predicted TL of a 50 mm slot. 95% confidence intervals.
6.2.2.2.2 CT Silencers
FEM predictions were made for comparison to experimental results for the CT silencers
measured in the lab facility (section 5.5.5). Below are the measured and predicted results for
the 0.3 m straight CT silencer (Figure 114), 0.3 m L-shaped CT silencer (Figure 115), and the
0.3 m Z-shaped CT silencer (Figure 116). Below the 250 Hz third octave band and above the
1000 Hz third octave band the agreement between measured and predicted results appears
quite good; note however that disagreement cannot always be explained by the random
sampling error. Between 250 and 1000 Hz the predicted results are consistently around 5 dB
lower than the measured results for all three CT silencer types. In an attempt to identify the
cause of the discrepancy, predictions have been made for these silencers using a
characteristic impedance ( ρ 0 c ) free field termination in place of the PML, and a different
point source location. Additionally, the lab measurements were repeated for the 0.3 m
straight CT silencer. All of these attempts produced effectively identical results to those
shown below. A statistically significant difference between the measurement and prediction
results exists. The difference was consistent over all CT silencers but not over the entire
174
frequency spectrum, suggesting that calculation error is not the problem, but a frequency-
dependent dissimilarity between the physical and modeled acoustic domain may be. This
dissimilarity may be due to:
• non-diffuse laboratory sound field
• reactive (complex impedance) boundary condition due to stud-mounted gypsum wall
board in the lab
• inaccurate modeling of fiberglass liner absorption.
A number of variations have been modeled in an attempt to determine the source of this
measurement-prediction discrepancy, all of which had no significant effect on the predicted
result. These variations were:
• reverberant receiver volume in-place of anechoic termination
• characteristic impedance anechoic termination in-place of the perfectly matched layer
• sound source located away from corner to change source volume sound field
• increased resolution of mesh near ventilator.
20
Measured
15 Predicted
Transmission Loss [dB]
10
-5
-10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 114: Measured and predicted TL of 0.3 m straight CT silencer. 95% confidence intervals.
175
25
Measured
20
Predicted
15
Transmission Loss [dB]
10
-5
-10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 115: Measured and predicted TL of 0.3 m L-shaped CT silencer. 95% confidence intervals.
20
Measured
15
Predicted
10
Transmission Loss [dB]
-5
-10
-15
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 116: Measured and predicted TL of 0.3 m Z-shaped CT silencer. 95% confidence intervals.
176
6.2.3 Acoustic FEM Design Summary
A FEM modeling process has been developed to predict the diffuse-field transmission loss
performance of a ventilation opening. The diffuse field is created on the inlet side of the
ventilation opening by modeling a source volume containing a point source. The ventilator
outlet is a free field termination created using a perfectly matched layer. Sound power
through the inlet of the ventilation opening can be calculated directly from the pressure field
solution, or inferred using diffuse field theory from the average source volume sound
pressure level. Outlet sound power is calculated directly from the pressure field solution.
In order to provide an accurate estimation of the source volume sound pressure level, average
and standard deviation twenty frequencies should be sampled in each third-octave band. The
frequency samples are energy-averaged on a logarithmic scale. Ventilation openings may
have very different frequency dependences; the uncertainty in band-average values should be
determined and the number of averages increased as necessary.
In order to provide a diffuse sound field, as characterized by a low sound pressure level
standard deviation, the minimum source volume dimension must be three times the longest
wavelength (N=6). It is be possible to use a smaller source volume with a modest increase in
uncertainty, or a larger source volume for a decrease in uncertainty, but at a computational
cost. Source volume aspect ratios are chosen to be 1:21/3:41/3.
In order to be 95% confident that the sound pressure level at the ventilator inlet is
independent of its position on the source volume surface within a 1 dB uncertainty, one of
the following two conditions should be met:
177
1. One dimension of the ventilator cross section is greater than one half of the longest
wavelength, and no part of the ventilator is within one quarter wavelength of a
corner.
2. The ventilator is adjacent to a corner and its dimension parallel to the corner is
greater than the longest wavelength.
A 2D model is described in the same way as the 3D model, with the exception of its shortest
dimension needing to be ten times the longest wavelength to provide equivalent modal
density (N=20).
FEM predictions agree very well with an analytical model for dissipation of the fundamental
mode in a lined duct. Any disagreement was shown to be due to the assumption in the
analytical model of a locally reacting liner. FEM predictions were also compared to lab
measurements. Good agreement was observed between predictions and measurement for a
slot-type ventilation opening. Disagreements of around 5 dB existed between predicted and
measured results for 0.3 m lined CT silencers between 250 and 1000 Hz. Comparing
predictions to measured results indicates how much uncertainty is associated with diffuse
field modeling and how difficult this uncertainty is to control. Further predictions should be
compared to carefully-conducted measurements. Physical scale modeling could be beneficial
to provide a better diffuse-field approximation in the measurements. Additionally,
comparisons should be made with silencers measured away from the corners to maximize the
sound field diffuseness.
This modeling and performance prediction method provides an excellent new tool for
developing and optimizing silencers for diffuse field application. It is a vast improvement
over the previously-used insertion loss modeling because it accounts for random incidence,
diffraction into the ventilation opening, and radiation impedance at the outlet.
178
6.2.4 Acoustic FEM CT Silencer Results
The Acoustic FEM model has been used to predict the transmission loss performance of four
types of CT silencers, each with three different total lengths (the path length along the center
of the flow path). The CT silencer types were Straight, L-shaped, U-shaped, and Z-shaped as
shown in Figure 117; the total lengths were 0.3, 0.5, and 1 m. The dimensions labeled in
Figure 117 correspond to the values shown in Table 26. Note that there is no 0.3 m U shape.
Based on results from the analytical study of plane wave attenuation in a lined duct (Section
6.1.5), the duct liner was chosen to be 50 mm thick with a flow resistivity of 60,000 Rayl/m,
and the flow path was chosen to be 0.1 m high.
Performance results for silencer predictions are summarized in terms of the STC and SEOAs
in Table 27. A complete set of frequency-dependent transmission loss plots is given in
Appendix C.1. Transmission loss performance increases with increasing length for each
silencer type. Of the silencers modeled, silencers with the same total length have very similar
predicted transmission loss performance:
• 0.3 m CT silencers have a STC of 5 to 6 dB and a SEOAs of 0.9 to 1.1
• 0.5 m CT silencers have a STC of 8 to 9 dB and a SEOAs of 0.5 to 0.6
• 0.5 m CT silencers have a STC of 8 to 9 dB and a SEOAs of 0.15 to 0.16
In the FEM validation, predicted mid frequency transmission loss was less than the measured
transmission loss by up to 5 dB. The cause of the disagreement could not be identified; as a
result, the actual transmission loss of these silencers might be up to 5 dB higher than these
predictions.
179
L1
100 mm
100 mm
L1
S1
U2
Z1
100 mm
U1 100 mm
150 mm
Figure 117: Shapes and dimensions of CT silencers. Clockwise from top left: Straight, L, Z, and U. 50
mm fiberglass shown as crosshatched. Heavy black lines are sound-hard boundaries.
180
6.2.4.1 Frequency Dependence of CT Silencer Transmission Loss
Table 27 suggests that, in the range of geometries tested, silencers of similar length have
nearly identical performance regardless of the silencer type (i.e. shape). Observing the
frequency-dependent transmission loss of the 1 m silencers in Figure 118, it is clear that the
transmission loss at high frequencies actually varies greatly between silencer types. The
reason the SEOAs is identical for each shape, regardless of high frequency variation, is
shown in Figure 119 by the transmitted speech spectrum, as introduced in 2.1.1. The
performance of all silencers is limited by low frequency transmission loss; therefore, any
variability in the high frequency transmission loss will have negligible effect on the SEOAs.
STC ratings are similarly limited by low and mid frequency transmission.
It is not necessarily the case that all silencers will be limited by low frequency transmission
loss; therefore, it is beneficial to understand the cause of the variation in high frequency
transmission loss between the various silencer shapes. The frequency dependent transmission
loss can be discussed in terms of attenuation due to the liner, and attenuation at the elbows.
Figures 120 and 121 show the sound pressure level and sound pressure at 500 Hz, a relatively
low frequency at which all silencer shapes have very similar performance. The rate of
attenuation is low throughout the silencers due to the relatively ineffective liner, and there is
no notable attenuation at the elbow. The 0.69 m wavelength at 500 Hz is long with respect to
the 0.1 m duct width, so the fundamental mode in the duct section before the elbow couples
well with the fundamental mode in the section past the elbow.
At the high frequency of 8000 Hz, the 0.04 m wavelength is much smaller than the duct
width. As concluded in section 6.1.4, if the wavelength is smaller than the duct width, the
181
fundamental mode will propagate with little attenuation, as the rate of diffraction into the
liner is small. This conclusion is confirmed visually in Figure 124. At these high frequencies,
however, because the wavelength is small compared to the duct width, the fundamental mode
upstream of the elbow couples very weakly with the fundamental mode downstream of the
elbow. Figure 125 shows that the elbow causes a large attenuation, and the attenuation at the
elbow dominates the attenuation in the straight sections.
182
90
80
1m Straight
70 1m L
1m U
60
1m Z
Transmission Loss [dB]
50
40
30
20
10
-10
-20
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 118: Predicted transmission loss of 1 m CT silencers. 95% confidence intervals shown.
1m Straight
-20 1m L
Transmitted Speech Spectrum [dB]
1m U
1m Z
-40
-60
-80
-100
-120
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 119: Predicted transmitted speech spectrum, 1 m CT silencers. 95% confidence intervals given.
183
Figure 120: Predicted Lp (left) and sound pressure (right) at 500 Hz, 1 m Straight silencer. Lp color scale
0 to 100 dB. Sound pressure color scale -0.5 to 0.5 Pa. Axis dimensions in meters.
Figure 121: Predicted Lp (left) and sound pressure (right) at 500 Hz, 1 m L-shaped silencer. Lp color
scale 0 to 100 dB. Sound pressure color scale -0.5 to 0.5 Pa. Axis dimensions in meters.
184
Figure 122: Predicted Lp (left) and sound pressure (right) at 2500 Hz, 1 m Straight silencer. Lp color
scale 0 to 100 dB. Sound pressure color scale -0.5 to 0.5 Pa. Axis dimensions in meters.
Figure 123: Predicted Lp (left) and sound pressure (right) at 2500 Hz, 1 m L-shaped silencer. Lp color
scale 0 to 100 dB. Sound pressure color scale -0.5 to 0.5 Pa. Axis dimensions in meters.
185
Figure 124: Predicted Lp (left) and sound pressure (right) at 8000 Hz, 1 m Straight silencer. Lp color
scale 0 to 100 dB. Sound pressure color scale -0.5 to 0.5 Pa. Axis dimensions in meters.
Figure 125: Predicted Lp (left) and sound pressure (right) at 8000 Hz, 1 m L-shaped silencer. Lp color
scale 0 to 100 dB. Sound pressure color scale -0.5 to 0.5 Pa. Axis dimensions in meters.
186
6.3 CFD Predictions of Ventilation Opening Discharge Coefficients
Duct fitting loss coefficients have been investigated numerically many times and the results
have shown that, using CFD, it is possible to reliably obtain discharge coefficient results for
the duct fitting within 15% of the experimental result [83]. While similar in many ways,
ventilation openings are unique from duct fittings in that the dynamic pressures upstream and
downstream from the opening are small. Only one study has predicted the performance of a
ventilation opening, and very few details are provided as to how the modeling was carried
out [11]. The following section will present the methods and results from the flow
performance predictions carried out in the present work. These results are intended to be used
in combination with the acoustic FEA results to provide a complete performance prediction
method for ventilation openings.
In this next section, the ventilation opening flow model used in this work is described in
terms of flow domain geometry, flow-equation model, boundary conditions, and mesh. The
flow domain solution provides the flow rate through, and pressure loss across, the ventilation
opening. Finite volume based CFD software ANSYS FLUENT is used for all simulations.
The model was validated by comparison with published results for an orifice, and by
comparison with measured results for CT silencers. Results are predicted for the same
silencer geometries that have been modeled using the acoustic FEM technique described in
section 6.2.4.
The discharge coefficient, in the case of natural ventilation openings, can be interpreted as a
measure of how efficiently energy is converted from potential energy (pressure on the
upstream side of the opening) to kinetic energy (velocity through the ventilation opening),
187
and back to potential energy (pressure on the downstream side). The conversion from kinetic
to potential is also known as pressure recovery [84]. In order to observe this process in the
model the flow domain must include a relatively large volume on either side of the
ventilation opening. The upstream volume must be large enough that the dynamic pressure
associated with the inlet velocity is negligible as compared to the dynamic pressure
associated with the ventilator velocity. The downstream volume must be large enough to
allow the high velocity flow in the opening to slow down and recover or dissipate its
dynamic pressure; i.e. the downstream volume must have low air velocity so that the
dynamic pressure is negligible.
The flow in the upstream volume will be relatively uniform normal to the flow direction;
therefore, its entire leading surface will be set as the inlet. For the downstream volume this
will not be sufficient as the domain would need to be very long to allow for significant
dissipation or recovery of the kinetic energy. Instead, an outlet is created in one corner of the
downstream volume so that the flow is encouraged to convert its kinetic energy in a smaller
region.
It is possible to determine the required inlet area analytically. The flow at the inlet is
expected to be nearly uniform; therefore, the velocity, at a given flow rate, is inversely
proportional to the area of the inlet. Likewise, the ratio of the average velocity in the inlet to
the average velocity in the ventilator is directly related to the ratio of the ventilator cross
sectional area to the inlet area:
U inlet A
= vent
U vent Ainlet
To assure that the dynamic pressure in the inlet is negligible, the ratio of the dynamic
pressures in the inlet and ventilator should be 1 to 100:
ρ 2
U inlet
1
= 2 U vent = 10 U inlet
100 ρ 2
U vent
2
It follows that,
Ainlet = 10 Avent
188
Provided the inlet area is ten times the ventilator area, the dynamic pressure in the inlet will
be negligible in the inlet compared to the dynamic pressure in the ventilator.
The above analysis is actually conservative, as the flow will generally not be uniform in the
ventilator. Using average velocity to calculate dynamic pressures will under-estimate the
actual value in the ventilator.
1
A ∫A
U= u dA
1 2
A ∫A
U2 ≤ u dA
This same result, unfortunately, also precludes the ability to use area ratios to determine the
dynamic pressure and domain size required in the volume downstream of the ventilation
opening. Average velocity will greatly under-predict the dynamic pressure downstream of the
ventilator.
1
1: Upstream Volume
2: Downstream Volume
4 3: Ventilator
2 4: Inlet
5: Outlet
Figure 126: Flow domain of ventilator model (sample velocity contours shown).
189
6.3.2 Flow-Equation Model
Unless otherwise noted, the information and theory presented in this section were taken from
the ANSYS Fluent 12 documentation [85].
As described in section 4.1 of this thesis, the Navier-Stokes equation (Eq. (23)), derived from
conservation of momentum and mass, provides a solution for Newtonian flow. If constant
density is not assumed, an additional equation that relates the change in density to mass flux
must be included for conservation of mass:
∂ρ ∂ρu i
Conservation of mass + =0
∂t ∂xi
If the flow is turbulent, fluctuations in the flow field variables will occur over very small
distances in space and time. In order to accurately model these properties, spatial and
temporal discretization must be extremely small. For most applications the implied
computational requirements to solve the flow field in this way, known as direct numerical
simulation (DNS), are not feasible. In order to reduce the required spatial and temporal
resolution so that the computation is possible, methods have been devised to filter out, or
average out, small-scale fluctuations. One of these methods of averaging is called Reynolds
averaging, with the class of resultant models being called RANS (Reynolds average of
Navier Stokes).
φ = φ + φ′
With the variables represented in this way, the Navier-Stokes equation for steady
incompressible flow can be presented as [48]:
∂pδ ij ∂u i ∂u j ∂
ρu j
∂u i
∂x j
=−
∂x j
+µ
∂
∂x j
∂x
+ +
∂x
(
− ρu i′u ′j )
j ∂xi i
The result is a set of equations containing only average results, save one additional term,
named the Reynolds stresses, which represents the turbulent fluctuations:
190
− ρui′u ′j
The Reynolds stresses must be modeled, and many methods exist to do so. A very popular
method is the k − ε model, where k is the turbulent kinetic energy and ε is the dissipation
rate of turbulent kinetic energy. The turbulent viscosity ( µt ) is defined to relate the turbulent
kinetic energy to its dissipation rate. The standard k − ε model assumes turbulent flow and is
therefore only valid for fully turbulent flows. Variations of the k − ε model have been
developed to make it more applicable to special classes of flow. One of those variations is the
Renormalized Group theory (RNG k − ε ). It is similar to the standard k − ε ; however, it
includes a number of improvements. One of the improvements is that the turbulent viscosity
is calculated using a differential equation, providing an improvement in accuracy for low
Reynolds number flows. The RNG k − ε RANS method has been identified as the most
appropriate model for indoor airflow [86]. For this work, the flow has been modeled using
both laminar and RNG k − ε RANS methods. Following the method used in a previous work
modeling hydraulic flow meters [49], the laminar model was used to predict the behaviour of
flows below Re = 2000, and the RNG k − ε model was used for flows above Re = 2000.
6.3.4 Mesh
The mapped mesh feature in ANSYS is used for mesh generation. Resolution of the mesh is
set by specifying the grid spacing at boundaries. Intuition and iteration were used to specify a
fine mesh near regions with high gradients in the flow. Mesh-independence of the flow
solution was confirmed by doubling the mesh resolution and confirming that the discharge
coefficient result did not change by more than 1%. Typically, at Re ≈ 25000 , convergence
was achieved with 0.1 mm mesh spacing at the ventilator wall. The maximum mesh spacing
anywhere in the flow domain was 50 mm. In a number of the predictions in this work,
doubling the mesh resolution resulted in as much as a 3% change in Cd; however, further
increases in mesh resolution would cause the numerical solver to become unstable and the
191
solution would diverge. Future work should use a structured mesh in order to produce better
solver-stability and mesh-independence. Unstructured meshes were used in this work due to
time constraints.
Published results [50] define the orifice plate, and provide the discharge coefficient as a
function of orifice-to-pipe diameter ratio; however, in our model the ‘pipe’ diameter is not
defined, only specified to be much larger than the orifice diameter. Predicted results should
therefore be compared to published results with very small diameter ratios. Published results
for the discharge coefficient are shown in Figure 127, the predicted results in Figure 128.
Predictions up to Re = 2000 (Re0.5= 43) were performed using the laminar flow model;
predictions beyond Re = 2000 used the RNG k − ε . The discharge coefficient is plotted
against Re0.5 because their relationship is linear at very low Re (see section 4.4.2).
192
Comparison of predicted and published results at small diameter ratios shows excellent
agreement for all Reynolds numbers.
Figure 127: Published experimental results for the orifice plate discharge coefficient [50].
0.8
0.7
0.6
0.8
0.7 0.5
0.6
Cd
0.4
0.5
0.3
Cd
0.4
0.3 0.2
0.2
0.1
0.1
0 0
0 1 2 3 4 5 6 7 0 20 40 60 80 100 120 140 160
Re 0.5
Re0.5
Figure 128: Predicted results for the orifice plate discharge coefficient.
193
6.3.6.2 Validation by Comparison to Measured Results
Laboratory measurements of the discharge coefficient of Straight, L-shaped, and Z-shaped
CT silencers, and a slot ventilator, were conducted at high Re, and the results were presented
in section 5.5.5. Predictions were made for these same geometries. The upstream volume
height (1 m) was ten times the silencer height (0.1 m). The downstream volume height (3 m)
was 30 times the silencer height; increasing the downstream volume height further resulted in
no change to the predicted discharge coefficient. All measured and predicted results
correspond to Re>15,000; therefore, as shown in Table 28, all results are independent of Re
and directly comparable to each other (see section 4.4.2).
As seen in Table 28, there is modest agreement between the predicted and measured results,
with disagreements of up to 20%. The predictions and measurements do, however, show
better agreement with respect to the relative performance of the different silencers types.
Predictions show the L-shaped and Z-shaped silencers to be 38% more restrictive to airflow
than the Straight silencer; measurements show the L and Z shapes to be 38 and 47% more
restrictive, respectively. It is possible that a major source of the disagreement between
measurement and prediction is because the silencers were measured with a porous liner.
Some amount of flow could pass through the liner, increasing the flow rate. Further work
should implement a porous liner in the flow model to increase prediction accuracy.
Table 28: Measured and predicted silencer discharge coefficients – high Re.
Straight L Z Slot
Cd - Predicted 0.64 0.39 0.40 0.64
Cd - Measured 0.79 0.49 0.42 0.66
194
results will conclude with general comments on the Re dependence of the discharge
coefficient. Figures showing the velocity and pressure fields in each CT silencer at high-Re
are provided in Appendix C.2.
Through observation of the velocity and static pressure profiles in the flow it is possible to
hypothesize the cause of the increase in Cd with increasing silencer length at high Re. In
Figure 130, showing the flow velocity in the 1 m Straight silencer at high Re, a vena-
contracta is present immediately following the entrance to the silencer, which results in high
velocity flow near the top of the duct, and recirculating flow near the bottom. The high
velocity flow is associated with a high dynamic pressure and, as predicted by Bernoulli’s
equation (see section 4.2), the corresponding static pressure in the flow must decrease. Figure
131 shows that the static pressure does indeed decrease in the region of the vena-contracta.
Moving beyond the vena-contracta, where the bottom boundary layer attaches and the
velocity profile becomes more uniform, there is a recovery in static pressure which
corresponds to the reduction in dynamic pressure. One might expect the length of the region
of pressure recovery to be associated with the duct entrance length ( l e = Dh ⋅ 4.4 Re1 6 [46]);
however, the recovery actually occurs much more rapidly. At Re = 27000, the entrance
length ( le ) is 24 times the hydraulic diameter; however, in Figure 131, and in results
published by Lichtarowicz [51], the pressure recovery is does not extend beyond two or three
times the hydraulic diameter. Additionally, beyond the silencer outlet there is effectively no
195
recovery of the dynamic pressure. In Figure 132 and Figure 133, the velocity profile in a 0.05
m long Straight silencer, there is a vena-contracta which will cause an increase in the
dynamic pressure near the inlet, like the 1 m Straight silencer; however, as there is no length
of duct in which the dynamic pressure can recover, there is no pressure recovery, which
directly explains the lower discharge coefficient for very short Straight silencers.
Recovery of dynamic pressure, beyond the dynamic pressure associated with the vena-
contracta, can be obtained using diffusers [47]. The use of diffusers to optimize ventilation
openings silencers should be considered for future optimization work.
196
end, has a smaller discharge coefficient than either the 0.5 or 1 m L-shaped silencer
alone. Clearly it is not possible to consider the effect of elbows as additive.
197
0.9
0.05 m
0.8
0.3 m
0.7 0.5 m
1m
0.6
0.5
Cd
0.4
0.3
0.2
0.1
0
1 10 100 1,000 10,000
Re
Figure 129: Predicted Straight CT silencer discharge coefficients as a function of Re.
198
Figure 132: Predicted flow velocity in 0.05 m Straight CT silencer – Re = 22,000.
0.9
0.6
0.5
Cd
0.4
0.3
0.2
0.1
0
1 10 100 1,000 10,000
Re
Figure 134: Predicted discharge coefficients of 0.3 m CT silencers as a function of Re.
199
0.9
0.5
Cd
0.4
0.3
0.2
0.1
0
1 10 100 1,000 10,000
Re
Figure 135: Predicted discharge coefficients of 0.5 m CT silencers as a function of Re.
0.9
0.8 1m Straight
1m L
0.7 1m U
1m Z
0.6
0.5
Cd
0.4
0.3
0.2
0.1
0
1 10 100 1,000 10,000
Re
Figure 136: Predicted discharge coefficients of 1 m CT silencers as a function of Re.
200
1
0.9
0.8
0.7
0.6
0.5
Cd
0.4
0.3
0.2
0.1
0
-6 -5 -4 -3 -2 -1 0 1
10 10 10 10 10 10 10 10
∆p [Pa]
Figure 137: Predicted discharge coefficients as a function of pressure loss for all silencer configurations.
Black – 0.05 m, blue – 0.3 m, green – 0.5 m, red – 1 m. ––––– Straight, – – – – – L-shaped, ········· U-shaped, – · – · – · Z-shaped.
201
6.4 CT Silencer Modeling Summary
Methods have been developed for predicting the diffuse-field transmission coefficient and
the Re-dependent flow performance of ventilation openings. Only 2D geometries have been
modeled; however, nothing in the proposed method prevents the modeling of 3D geometries.
In this section the OAR of the CT silencers, for which acoustic and flow performance
predictions have been made, are presented. Because attempts to validate the transmission loss
predictions in comparison to experimental results were inconclusive, the OAR results should
not be considered conclusive. This summary discussion therefore only compares the relative
performance of various CT silencer types. Table 30 presents the specific equivalent open area
and open area ratio results from all CT silencer predictions, assuming high Re. Figures 138
and 139 show the OAR as it varies with pressure loss and Reynolds number for all CT
silencer types. General conclusions about the silencer OAR performance are:
• Silencer length is dominant in determining the OAR, as it is highly related the
acoustic performance; interestingly, over the range of silencer shapes and lengths
predicted, increasing the length also tends to improve airflow performance.
• The Straight silencer has the highest OAR of any silencer shape, because it has the
best airflow performance; acoustic performance is largely independent of silencer
shape.
• The Z-shaped silencer has the lowest OAR due to high airflow restriction.
202
10
2
OAR
0.5
0.2
0.1
-6 -5 -4 -3 -2 -1 0 1
10 10 10 10 10 10 10 10
∆ p [Pa]
Figure 138: Predicted open area ratio as a function of pressure loss for all silencer configurations.
Blue – 0.3 m, green – 0.5 m, red – 1 m. ––––– Straight, – – – – – L-shaped, ········· U-shaped, – · – · – · Z-shaped.
203
10
2
OAR
0.5
0.2
0.1
0 1 2 3 4
10 10 10 10 10
Re
Figure 139: Predicted open area ratio as a function of Reynolds number for all silencer configurations.
Blue – 0.3 m, green – 0.5 m, red – 1 m. ––––– Straight, – – – – – L-shaped, ········· U-shaped, – · – · – · Z-shaped.
204
Chapter 7: Conclusions
To conclude, the results are summarized and recommendations for future work are presented.
7.1 Summary
The work presented in this thesis is now summarized by directly addressing the research
objectives stated in section 1.2.
Ventilation opening silencer design for acoustical performance cannot be done independently
of the design for airflow performance, as many of the possible modifications to improve the
acoustic performance are to the detriment of airflow performance. To address this, an
optimization parameter, the open area ratio (OAR), has been developed as a single number
performance metric (see Chapter 2). OAR is the ratio of the equivalent open area for airflow
(EOAf) to the equivalent open area for sound transmission (EOAs). Normalizing the
equivalent open areas to the ventilator area gives the specific equivalent open areas for
airflow (SEOAf) and sound transmission (SEOAs). Both SEOAf and SEOAs approach unity
for a plain, large aperture; OAR is also unity for any ventilation opening with a combined
performance equivalent to a large aperture.
The SEOAf is directly related to the discharge coefficient (Cd), and the effective leakage area
defined by ASTM E779-10 [27] assuming a high Reynolds number.
205
Objective 2: Provide an understanding of the factors that affect speech privacy in a naturally
ventilated building.
A diffuse field model was developed to predict the speech privacy, in terms of SII, between
two rooms separated by a common partition (see Chapter 3).
Independent of the partition construction, privacy was shown to increase with room volume
and background noise. The effect of surface material acoustic absorption, which alters the
reverberation time, is complicated. If background noise is very low, increasing reverberation
reduces privacy; however, if the background noise is not low compared to the speech signal,
then a moderate reverberation time corresponds to the lowest level of privacy. Previous work
by Hodgson and Nosal has addressed this issue more completely for speech intelligibility in a
single room and drawn similar conclusions [44].
For this work it was necessary to measure the acoustic and airflow performance of ventilation
openings in naturally ventilated buildings in a laboratory environment and in real buildings.
Standardized methods exist for measuring acoustic transmission loss of building elements in
laboratory environments [56], and in buildings [55]. Measuring ventilation openings
buildings, however, is challenging because the ventilation opening represents only a small
portion of the partition between two spaces – the energy transmitted through the ventilation
opening must be distinguished from that transmitted through the rest of the partition and by
other flanking paths. The standard method requires the elimination of all flanking paths, to
206
overcome this challenge [55]; this approach is not realistic here. An alternative method was
developed for this work. The transmission loss was measured with the ventilator open and
again with the ventilator acoustically blocked (see section 5.1.1.2); the difference between
the open and closed transmission is the transmission through the ventilator alone. This
approach is effective; however, it is only able to provide a lower bound for the performance
of the ventilation opening.
A standardized method [27] was also adopted and modified to measure the airflow
performance of the ventilation openings (see section 5.1.2). A blower door – typically used to
measure building envelope air tightness – was used to introduce a known flow rate into the
room containing the ventilation opening. The corresponding pressure differential across the
ventilation opening, due to the air exiting the room through it, was measured. In order to
isolate the flow rate through the ventilation opening from other exit flow paths, the flow rate
and pressure differential were measured with the ventilator open and again blocked in a
similar fashion to the transmission loss measurements. Subtracting the two measurements
provided the flow through the ventilator alone. This method has been used successfully for
both laboratory and building measurements. A limitation of this method was that it is not able
to test the ventilation opening flow at low pressures – as a result, only high Re results can be
obtained.
Analytical prediction techniques, such as the one implemented in section 6.1, exist for
determining the transmission loss of propagating modes of silencers; additionally, finite
element and boundary element methods have been used to predict the transmission loss of
duct silencers. While these methods are useful in investigating the attenuation mechanisms
within the silencer, they are unable to predict the effect of diffraction into the ventilator,
diffuse sound field incidence on the ventilator, or the effect of mounting location on the
partition. These factors have been discussed theoretically in section 2.3, and investigated
using FEM prediction in section 6.2.2.1.2. Prior to this work, no methods existed for
predicting the diffuse field transmission loss of a ventilation opening. A FEM based
207
technique, motivated by standardized measurement methods [56, 66], has been developed
and implemented using COMSOL to predict the diffuse field transmission loss of ventilation
openings and silencers (see section 6.2). Validation of the model was inconclusive and needs
further work; however, the FEM prediction technique was able to provide valuable insights
into the mechanisms of acoustic transmission through silencers (see section 6.2.4).
ANSYS Fluent CFD has been used to model airflow through ventilation openings (see
section 6.3). As with the FEM acoustic model, validation of the model proved inconclusive
and needs further work; however, the CFD predictions are able to provide valuable insights
into the mechanisms of airflow through, and pressure loss across, ventilation openings and
silencers (see section 6.3.7). Limitations of the flow modeling were that the porous domain
was not modeled, and the surfaces were all modeled as smooth.
The techniques developed for predicting the acoustic transmission loss and airflow properties
of ventilation openings are of great value for optimization and characterization work, as
laboratory measurements are time consuming and facilities are often not available.
Additionally, FEM and CFD techniques, as compared to conventional physical
measurements, can provide greater detail and insights into the mechanisms that govern
silencer performance by providing a solution for the entire acoustic and flow domain.
Four different sections of this thesis provide measurement or prediction results for the
acoustical and airflow performance of ventilation openings, along with exact descriptions of
their geometry, to inform designers of configurations available for use:
• In situ measurements of the acoustic and airflow performance of ventilation openings
in naturally ventilated buildings were performed for 15 ventilation opening
configurations in five different buildings; the results have been presented in section
5.3.
208
• Laboratory measurements have been completed, in a purpose-built testing facility, for
19 different ventilation opening configurations; the results are presented in section
5.5.
• Results from a BRE research project on interior natural ventilation opening silencers
[12] have been analyzed using the techniques developed in this work. Acoustic and
airflow performance results for 12 different cross-talk silencers, measured by the
BRE, are presented in section 5.4.
• Numerical techniques have been used to predict the performance of 11 different
cross-talk silencers, as presented in section 6.2.4.
Objective 6: Provide best practice guidelines for designing successful interior natural
ventilation openings.
Throughout this thesis, a number of best practices for successful ventilation opening silencer
design have been identified:
• Non-acoustical grilles should be avoided. In situ measurements (section 5.3), and a
review of measurements by the BRE (section 5.4), show that adding grilles to a
ventilation opening halves the flow rate for a given pressure loss. All measurements
showed that grilles have negligible effect on acoustic transmission.
• The addition of a fiberglass absorptive liner to the ceiling above a slot ventilation
opening has been shown, in laboratory (section 5.5.1) and in situ (section 5.3)
measurements, to increase the opening’s STC by 3 to 6 dB.
• The acoustic baffle, a novel silencer concept presented in section 5.5.3, was shown to
be an effective silencer for situations where the depth of the silencer is restricted.
• Cross-talk silencers, as studied in situ (section 5.3), in the BRE measurements
(section 5.4), in the laboratory measurements (section 5.5.5), and in the numerical
modeling section (sections 6.2.4 and 6.3.7), are capable of the highest combined
acoustical and airflow performance when compared to any other type of silencer
tested. Their acoustical performance can be improved by increasing the length of the
acoustically-lined flow path; within reasonable limits, this increase in length does not
further restrict, and can actually increase, air flow.
209
• Reducing the aspect ratio of a CT silencer to one, if the silencer is lined on all four
sides, can increase the attenuation of the fundamental acoustic mode by up to twice
that of an otherwise identical silencer with an aspect ratio greater than 10. This
conclusion is based on an analytical investigation presented in section 6.1.
• Elbows in the L-, U-, and Z-shaped CT silencers increase the transmission loss at
high frequencies; however, the overall acoustic performance in these silencers (STC
or SEOAs) is limited by low frequency transmission. Elbows will increase the
transmission loss if the wavelength is shorter than the duct height. As elbows increase
the flow losses but do not increase the overall acoustic performance (EOAs, or STC),
the Straight CT silencer has the best performance of any CT silencer shape tested.
This result was observed in laboratory measurements, (section 5.5.5), and in
modeling predictions (sections 6.2.4 and 6.3.7).
• In order to be effective at attenuating speech frequencies, the acoustic liner in a
silencer liner should be at least 50 mm thick (see sections 6.1 and 5.5.1).
• The attenuation rate in straight sections of lined duct is maximized when the
wavelength is equal to the duct height (see sections 6.1 and 6.2.4). As the wavelength
becomes smaller than the duct height the attenuation rate decreases. For this reason, it
is best to use a duct height no greater than the wavelength of the highest critical
frequency; a 0.1 m tall duct provides good attenuation for A-weighted speech sounds.
• EOAf and Cd are both generally dependent on Reynolds number; however, they are
constant if the Reynolds number is sufficiently high. In mechanical ventilation
systems, high Re flow is a good assumption, and Cd is considered constant with
respect to flow rate for duct fittings. To investigate the Reynolds number dependence
in the context of natural ventilation openings an analytical discussion has been
provided in section 4.4.2, and a discussion of in situ measurement and modeling
results is given in sections 5.3.6 and 6.3.7.3. It was shown that the EOAf and Cd
become independent of Re when the Reynolds number is above some ‘transitional’
Reynolds number. The transitional Reynolds number is geometry-dependent and
exists in the range of 100 – 4000. As shown in section 5.3.6, ventilation openings in
naturally ventilated buildings cannot be assumed to operate at high Re. Also, because
Cd decreases with decreasing Re below the transitional Reynolds number, using the
210
high Reynolds number discharge coefficient to design a ventilation opening for low
Reynolds number flow could result in an underestimation of the flow losses and an
undersized system (section 4.4.2). In most cases, however, if the flow rates are low
enough such that Re<4000, the pressure loss through the ventilation opening is at
least two or three orders of magnitude less than the 10 Pa often assumed available in a
naturally ventilated building (sections 5.3.6 and 6.3.7.3). As a result, the errors
associated with using a high Re discharge coefficient to determine the losses through
a ventilation opening at low Re may not be of practical concern. These same errors,
however, will be of concern if either the silencer is very restrictive (i.e. small Cd), or
if the hydraulic diameter is very small (low Re at high flow rate), as significant
pressure losses would be observed at low Reynolds numbers. Grilles are likely the
most common example of low Re flow leading to significant errors in predicted
pressure loss because they are often implemented and, due to small fin spacing, have
a small hydraulic diameter.
211
Laboratory measurements were limited due to the low partition transmission loss, small room
size, and inability to measure at low-Re flow rates. Physical scale modeling may be a useful
method to improve the measurements while making use of the present facilities. Scale
modeling can:
• increase the noise-isolation effectiveness of the partition
• provide a diffuse sound field at lower frequencies
• increase the ventilator pressure drop with respect to Re, allowing measurements of
airflow performance at relatively low Re.
Calculation of the airflow measurement results should also be modified. Because the
assumption of high Re flow is used, it is also assumed that n from Eq. (10), section 2.1.2, is
equal to 0.5. Allowing n to take any value is inappropriate and will result in an
underestimation of the actual uncertainties in the measured result.
Numerical predictions in this work were limited primarily by a lack of confidence in the
validation results. Before the proposed numerical models are used for optimization work,
their accuracy, or regions of validity, must be established. In the acoustical model the porous
absorber is modeled empirically; in the airflow model, it is not modeled at all. The porous
absorber should be investigated as a possible cause of disagreement between measurement
and prediction for both acoustical and airflow results. Additionally, to provide improved
mesh independence and solver stability, a structured mesh should be implemented in the
airflow model.
212
speech sounds. Modifications for increased acoustic performance should target these
frequencies. Increasing the liner thickness will improve the low frequency performance;
however, reactive elements should also be investigated. To improve the airflow performance,
diffusers should be considered. Diffusers will, however, present a compromise between
increased airflow performance and reduced acoustical performance. The diffuser will act as
an impedance-matching section and reduce the sound attenuation due to end-reflections
caused by the radiation-impedance.
213
References
[3] M. H. de Salis, D. J. Oldham and S. Sharples, "Noise control strategies for naturally
ventilated buildings," Building and Environment, vol. 37, 2002, pp. 471-484.
[4] D. J. Oldham, M. H. de Salis and S. Sharples, "Reducing the ingress of urban noise
through natural ventilation openings," Indoor Air, vol. 14, October 2008, pp. 118-
126.
[5] M. Hodgson and A. Khaleghi, "Design and evaluation of noise-isolation systems for
the natural-ventilation system of the UBC Liu Institute for Global Studies," UBC
Acoustics and Noise Research Group: Nov 2007.
[6] C. Field, "Acoustic design in green buildings," ASHRAE Journal, September 2008,
pp. 60-70.
[8] M. Hodgson, "Occupant satisfaction with the acoustical environment: 'green' office
buildings before and after treatment," in 26th Conference on Passive and Low Energy
Architecture. Quebec City, 2009.
[10] P. F. Linden, "The fluid mechanics of natural ventilation," Annual Review of Fluid
Mechanics, vol. 31, 1999, pp. 201-238.
[12] C. Hopkins, "A prototype ventilator for cross ventilation in schools: Sound insulation
and airflow measurements," Building Research Establishment Ltd., Dec 2004.
214
[13] Z. Nunes and A. Rickard, "Acoustic attenuators desined specifically for natural
ventilation, controling noise break in as well as cross talk across partitions," in
Internoise 2010. Lisbon, 2010.
[14] Z. Nunes and A. Rickard, "An assessment of cross-talk attenuation with natural
ventilation," Acoustics Bulliten, vol. 36, no. 5, Oct 2011, pp. 32-40.
[16] C. R. Chu and Y. W. Wang, "The loss factors of building openings for wind-driven
ventilation," Building and Environment, vol. 45, 2010, pp. 2273-2279.
[17] F. Flourentzou, J. Van der Maas and C. A. Roulet, "Natural ventilation for passive
cooling: Measurement of discharge coefficients," Energy and Buildings, vol. 27,
1998, pp. 283-292.
[19] C. Short and M. Cook, "Design guidance for naturally ventilated theatres," Building
Services Engineering Research and Technology, vol. 26, 2005, pp. 259.
[20] C. Ghiaus and F. Allard, Natural Ventilation in the Urban Environment: Assessment
and Design. Earthscan/James & James Ltd., 2005.
[21] I. L. Ver and L. L. Beranek, Noise and Vibration Control Engineering, 2nd ed. John
Wiley & Sons, 2006.
[23] Methods for calculation of the speech intelligibility index, ANSI Standard S3.5-1997,
June 1997.
[25] Classification for rating sound insulation, ASTM Standard E 413-10, 2010.
[26] H. Park, J. Bradley and B. Gover, "Evaluating airborne sound insulation in terms of
speech intelligibility," JASA, vol. 123, no. 3, 2008, pp. 1458-1471.
215
[27] Standard test method for determining air leakage rate by fan pressurization, ASTM
Standard E779-10, 2010.
[30] M. Hodgson, "Acoustical evaluation of six 'green' office buildings," Journal of Green
Building, vol. 3, no. 4, 2008, pp. 108-118.
[32] L. Cremer and H. A. Muller, Principles and Applications of Room Acoustics. , vol. 2,
Applied Science Publishers, 1982.
[34] R. V. Waterhouse, "Interference patterns in reverberant sound fields," JASA, vol. 27,
no. 2, 1955, pp. 247-258.
[35] H. J. M. Steeneken and T. Houtgast, "A physical method for measuring speech
transmission quality," JASA, vol. 67, no. 1, 1980, pp. 318-326.
[39] J. S. Bradley and H. Sato, "On the importance of early reflections for speech in
rooms," JASA, vol. 113, no. 6, June 2003, pp. 3233-3244.
[40] B. N. Gover and J. S. Bradley, "Measures for assessing architectural speech security
(privacy) of closed offices and meeting rooms," NRC Canada, Tech. Rep. NRCC-
47039, Dec 2004.
[41] T. Cox and P. D'Antonio, Acoustic absorbers and diffusers: Theory design and
application., 2nd ed.London and New York: Taylor and Francis, 2009.
216
[42] M. Hodgson, G. Steininger and Z. Razavi, "Measurement and prediction of speech
and noise levels and the Lombard effect in eating establishments," JASA, vol. 121, no.
4, 2007, pp. 2023-2033.
[45] Acoustic performance criteria, design requirements and guidelines for schools, ANSI
S12.60-2002, June 26 2002.
[46] D. Manring Noah, "Hydraulic Control Systems," John Wiley and Sons Inc., 2005.
[50] F. Johansen, "Flow through pipe orifices at low Reynolds numbers," Proceedings of
the Royal Society of London, Series A, vol. 126, 1930, pp. 231-245.
[52] A. Ellman and R. Piché, "A two regime orifice flow formula for numerical
simulation," Journal of Dynamic Systems, Measurement, and Control, vol. 121, 1999,
pp. 721.
[53] W. Borutzky, B. Barnard and J. Thoma, "An orifice flow model for laminar and
turbulent conditions," Simulation Modelling Practice and Theory, vol. 10, 2002, pp.
141-152.
[55] Standard test method for measurement of airborne sound attenuation between rooms
in buildings, ASTM E 336-10, 2010.
217
[56] Standard test method for laboratory measurement of airborne sound transmission
loss of building partitions and elements, ASTM E90-09, 2009.
[59] Ventilation for acceptable indoor air quality, ANSI/ASHRAE Standard 62–2007,
2007.
[61] M. L. Munjal, Acoustics of ducts and mufflers with application to exhaust and
ventilation system design. Wiley-Interscience, 1987.
[62] H. Kuttruff, Acoustics: An introduction. New York: Taylor and Francis, 2007.
[63] K. U. Ingard, Notes on sound absorption technology, New York: Noise Control
Foundation, 1994., 1994.
[64] Standard test method for impedance and absorption of acoustical materials by
impedance tube method, ASTM Standard C384-04, 2004.
[65] S. Daltrop, M. Gosselin and M. Hodgson, "Measurement and prediction of the airflow
and acoustical characteristics of porous materials," UBC Acoustics and Noise
Research Group, 2011.
[66] Standard test tethod for laboratory measurement of airborne transmission loss of
building partitions and elements using sound intensity, ASTM Standard E2249-02
(reapproved 2008), 2008.
[67] E. Geddes and J. Porter, "Finite element approximation for low frequency sound in a
room with absorption," JASA, vol. 83, no. 4, 1988, pp. 1431-1435.
[68] R. Tomiku, T. Otsuru, N. Okamoto and Y. Kurogi, "Direct and modal frequency
response analysis of sound fields in small rooms by finite element method," JASA,
vol. 123, no.5, 2008, pp. 3092.
[69] R. Tomiku, T. Otsuru and Y. Takahashi, "Finite element sound field analysis of
diffuseness in reverberation rooms," Journal of Asian Architecture and Building
Engineering, vol. 1, 2002, pp. 2-39.
[70] A. Craggs, "The application of acoustic and absorption finite elements to sound fields
in small enclosures," Finite Element Applications in Acoustics, 1981, pp. 1-19.
218
[71] A. F. Seybert, R. A. Seman and M. D. Lattuca, "Boundary element prediction of
sound propagation in ducts contating bulk absorbing materials," Transactions of the
ASME, vol. 120, 1998, pp. 976-981.
[72] A. Craggs, "A finite element model for rigid porous absorbing materials," Journal of
Sound and Vibraion, vol. 61, 1978, pp. 101-111.
[74] T. J. Cox and P. D Antonio, "Determining optimum room dimensions for critical
listening environments: A new methodology," in AES Convention 110, 2001.
[80] T. Cox, "Schroeder diffusers: A review," Building Acoustics, vol. 10, 2003, pp. 1-32.
[81] J. A. S. Angus, "Using grating modulation to achieve wideband large area diffusers,"
Applied Acoustics, vol. 60, pp. 143-165, 2000.
219
[86] Q. Chen, "Comparison of different k-ε models for indoor air flow computations,"
Numerical Heat Transfer, Part B Fundamentals, vol. 28, 1995, pp. 353-369.
220
Appendices
221
Appendix A: Manufacturer Data Sheets
Manufacturer data sheets are provided for further information on the fiberglass absorber,
acoustical louver, and air filters used in the laboratory measurements.
222
223
A.2 Kinetics Noise Control Acoustical Louver (2’x2’, 6” Val/1)
224
225
A.3 Filter Data Sheets
Pink air filter: “Red Excel” by B.G.E.
226
White air filter: Series 400 Pleated air filter by Aerostar (2” Std. Cap)
227
228
Appendix B: MATLAB Code
Code is provided for the diffuse field model, the analytical model for attenuation in a lined
duct, and the code used to process typical FEA results.
%Olsen
normal_F = [26, 37, 48, 47, 42, 49, 50, 48, 46, 42, 43, 42, 38, 36, 38,
40, 36, 36, 34];
normal_M = [48, 43, 48, 52, 51, 53, 54, 52, 46, 45, 47, 44, 40, 41, 41,
38, 24, 35, 32];
normal_C = [24, 29, 42, 51, 47, 47, 53, 52, 49, 44, 43, 43, 41, 38, 39,
40, 38, 36, 37];
229
speech_A_norm = [-26.7, -22.7, -17.4, -14, -11.9, -8.8, -7.1, -7.3, -
9.4,...
-11.2, -12, -13.6, -15.6, -19, -19.9, -21.6, -25.5, -28.2, -30.1];
% % 1
% floorAbs_1 = concrete;
% wallsAbs_1 = concrete;
% ceilingAbs_1 = concrete;
% floorAbs_2 = concrete;
% wallsAbs_2 = concrete;
% ceilingAbs_2 = concrete;
230
% %2
% floorAbs_1 = concrete;
% wallsAbs_1 = drywall;
% ceilingAbs_1 = drywall;
% floorAbs_2 = concrete;
% wallsAbs_2 = drywall;
% ceilingAbs_2 = drywall;
% % 3
% floorAbs_1 = carpetNoUnderlay;
% wallsAbs_1 = drywall;
% ceilingAbs_1 = drywall;
% floorAbs_2 = carpetNoUnderlay;
% wallsAbs_2 = drywall;
% ceilingAbs_2 = drywall;
% 4
floorAbs_1 = carpetNoUnderlay;
wallsAbs_1 = drywall;
ceilingAbs_1 = acousticTile;
floorAbs_2 = carpetNoUnderlay;
wallsAbs_2 = drywall;
ceilingAbs_2 = acousticTile;
%
% %5
% floorAbs_1 = carpetWithUnderlay;
% wallsAbs_1 = drywall;
% ceilingAbs_1 = acousticTile;
% floorAbs_2 = carpetWithUnderlay;
% wallsAbs_2 = drywall;
% ceilingAbs_2 = acousticTile;
%% Partition description
231
%PartitionSTC = 30;
%Scilencer Insertion Loss, IL_S [dB], in octave bands from 125 Hz to 8 kHz
IL_S = [0, 0, 0, 0, 0, 0, 0];
IL_S_1_3 = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0];
%% Source and background levels
%Source power level, Lw_1 [dB], in octave bands from 125 Hz to 8 kHz
Lw_1 = [60, 60, 60, 60, 60, 60, 60];
%Room Volumes
vol_1 = L_1*W_1*H_1;
vol_2 = L_2*W_2*H_2;
%Surface areas
floorS_1 = L_1*W_1;
floorS_2 = L_2*W_2;
wallsS_1 = 2*L_1*H_1+2*W_1*H_1;
wallsS_2 = 2*L_2*H_2+2*W_2*H_2;
232
ceilingS_1 = floorS_1;
ceilingS_2 = floorS_2;
Stot_1 = floorS_1+wallsS_1+ceilingS_1;
Stot_2 = floorS_2+wallsS_2+ceilingS_2;
if (Flag_reverbCalc == 1)
Atot_1 = 0.16*vol_1./RT_1;
Atot_2 = 0.16*vol_2./RT_2;
end
Aavg_1 = Atot_1/Stot_1;
Aavg_2 = Atot_2/Stot_2;
R_1 = Atot_1./(1-Aavg_1);
R_2 = Atot_2./(1-Aavg_2);
if (Flag_TL_P == 1)
TL_P = PartitionSTC+[-16, -7, 0, 3, 4, 4, 4];
end
TL_E = -10.*log10((10.^(-TL_P./10).*(s_P-s_V)+10.^(-
IL_S./10).*s_V)./(s_P));
TL_E_1_3 = -10.*log10((10.^(-TL_P_1_3./10).*(s_P-s_V)+10.^(-
IL_S_1_3./10).*s_V)./(s_P));
Lp_1 = Lw_1+10*log10(4./R_1);
Lp_2 = Lp_1-TL_E+10*log10(s_P./Atot_2);
SNL = Lp_2-Lp_BG;
NI = Lp_1-Lp_2;
233
if(SNL_AI(1, i)<0)
SNL_AI(1, i)=0;
disp('SNL_AI increased to 0dB');
end
if(SNL_AI(1, i)>30)
SNL_AI(1, i)=30;
disp('SNL_AI decreased to 30dB');
end
end
AI = SNL_AI*AI_weight';
%frequency weighting
SII_bandWeight = [0, 0.0617, 0.1671, 0.2373, 0.2648, 0.2142, 0.0549];
234
%SII = sum(BAF); sum seems to not want to work
SII = 0;
for i=1:length(BAF)
SII = SII+BAF(1, i);
end
SRI_E_speech = -10*log10(sum(10.^(SRI_E_weighted_spA/10)));
SRI_P_speech = -10*log10(sum(10.^(SRI_P_weighted_spA/10)));
SRI_S_speech = -10*log10(sum(10.^(SRI_S_weighted_spA/10)));
%This function takes in 1/3rd octave TL data from 125-4kHz and returns the
%STC value
cont = 1;
while cont == 1
total = 0;
STC_curve = STC_curve+1; %raise STC curve
%Check if deficiency exceeds 8dB
for i=1:16
temp = STC_curve(1, i)-TL_round(1, i);
235
if(temp == 9)
cont = 0;
disp(['STC limited by 8: Temp = ', num2str(temp)]);
end
%sum deficiencies
if(temp >= 0)
total = total+temp;
end
end
%check if sum of deficiencies exceed 32
if total > 32
cont = 0;
disp(['STC limited by 32: Total = ', num2str(total-16)]);
end
end
STC_curve = STC_curve -1;
STC = STC_curve(1, 7);
disp(['STC = ', num2str(STC)]);
clear all
%% Inputs
T = 20;
c = 331.3*(1+(T+273.15)/273.15)^1/2; %Speed of sound
rho = 1.204; %Density of air
Z0 = c*rho; %Characteristic Impedance
f = 90:5:11310; %Frequencies
k=2*pi*f/c; %Wavenumber
sigmay = 40000;
sigmax = 20000;
236
Ry = sigmay*dy/Z0; %Normalized flow resistance (flow resistance *d/Z0)
Rx = sigmax*dx/Z0;
%Rx = 5; %Normalized flow resistance (flow resistance *d/Z0)
%Ry = 5;
semilogx(f, ax);
title('Liner Absorption (liner on x)')
ylabel('Normal Incidence Absorption')
xlabel('Frequency Hz')
xlim([90 11310])
Qx=j*k.*hx*Z0./Zsx;
Qy=j*k.*hy*Z0./Zsy;
ikxa=1/hx*((2.47+Qx+((2.47+Qx).^2-1.87*Qx).^0.5)/0.38).^0.5;
ikxb=1/hx*((2.47+Qx-((2.47+Qx).^2-1.87*Qx).^0.5)/0.38).^0.5;
ikya=1/hy*((2.47+Qy+((2.47+Qy).^2-1.87*Qy).^0.5)/0.38).^0.5;
ikyb=1/hy*((2.47+Qy-((2.47+Qy).^2-1.87*Qy).^0.5)/0.38).^0.5;
237
for i=1:length(f)
% Initial guess
oldx = ikxa(1, i);
oldy = ikya(1, i);
newx = oldx-subs(gx,kxi,oldx)/subs(d1gx,kxi,oldx);
newy = oldy-subs(gy,kyi,oldy)/subs(d1gy,kyi,oldy);
oldx = newx;
oldy = newy;
Exa = 100*abs((oldx-newx)/oldx);
Eya = 100*abs((oldy-newy)/oldy);
iteration = iteration+1;
end
%Define k and E @ f
kxa(1, i) = newx;
kxa(2, i) = Exa;
kya(1, i) = newy;
kya(2, i) = Eya;
end
for i=1:length(f)
% Initial guess
238
oldx = ikxb(1, i);
oldy = ikyb(1, i);
newx = oldx-subs(gx,kxi,oldx)/subs(d1gx,kxi,oldx);
newy = oldy-subs(gy,kyi,oldy)/subs(d1gy,kyi,oldy);
oldx = newx;
oldy = newy;
Exb = 100*abs((oldx-newx)/oldx);
Eyb = 100*abs((oldy-newy)/oldy);
iteration = iteration+1;
end
%Define k and E @ f
kxb(1, i) = newx;
kxb(2, i) = Exb;
kyb(1, i) = newy;
kyb(2, i) = Eyb;
end
if ((max(Exa)>0.1)||(max(Eya)>0.1)||(max(Eya)>0.1)||(max(Eya)>0.1))
disp('Solution did not converge within 0.1%')
end
%% Calculate wavenumber in z
kzi = zeros(1, length(f));
239
Lh = -8.686*imag(kzi)*hy; %Normalized attenuation coefficient
eta=2*hy*f/c; %Ratio of channel height to wavelength
figure (2)
semilogx(f, TL)
title('Transmission Loss')
ylabel('TL dB/m')
xlabel('f Hz')
xlim([90 11310])
%ylim([0 10])
figure (3)
loglog(eta, Lh);
title('Normalized Attenuation')
ylabel('Lh dB')
xlabel('2hf/c (Channel depth / Wavelength)')
%xlim([0.01 5])
%ylim([0.1 10])
Script 2: Delaney-Bazley
%% Delany-Bazley model (ref Cox, Acoustic absorbers and diffusers, 2009,
%% pg 173)
sigmax = Rx*Z0./dx;
240
sigmay = Ry*Z0./dy;
Xx = rho*f./sigmax;
Xy = rho*f./sigmay;
No_bands=length(Lp_Sv(1, :));
No_frequencies=length(Lp_Sv(:, 1));
tao_Sv_third=zeros(1, No_bands);
tao_Win_third=zeros(1, No_bands);
TL_Sv_third=zeros(1, No_bands);
TL_Win_third=zeros(1, No_bands);
241
%Generate sampled frequencies
frequencies=zeros(No_frequencies, No_bands);
frequencies(1, :)=[88.39, 111.36, 140.31, 176.78, 222.72,...
280.62, 353.55, 445.45, 561.23, 707.11, 890.9,...
1122.46, 1414.21, 1781.8, 2244.92];
for i=1:No_bands
for s=2:No_frequencies
frequencies(s, i)=frequencies(1, i)+(frequencies(1, i)*2^(1/3)-
frequencies(1, i))...
/(No_frequencies-1)*(s-1);
end
end
%duct area
Vc_h=0.05;
A=343./frequencies(1, :)*(3*4^(1/3)-1.4)*Vc_h;
242
tao_Sv_third(1, i)=temp/(log10(frequencies(end, i)/frequencies(1,
i)));
TL_Sv_third(1, i)=-10*log10(tao_Sv_third(1, i));
end
No_bands=length(Lp_Sv(1, :));
No_frequencies=length(Lp_Sv(:, 1));
tao_Sv_third=zeros(1, No_bands);
tao_Win_third=zeros(1, No_bands);
TL_Sv_third=zeros(1, No_bands);
TL_Win_third=zeros(1, No_bands);
A=0.1; %duct height
243
%Generate sampled frequencies
frequencies=zeros(No_frequencies, No_bands);
frequencies(1, :)=[1414.21, 1781.8, 2244.92, 2828.43, 3563.59, 4489.85,
5656.85, 7127.19, 8979.7];
for i=1:No_bands
for s=2:No_frequencies
frequencies(s, i)=frequencies(1, i)+(frequencies(1, i)*2^(1/3)-
frequencies(1, i))...
/(No_frequencies-1)*(s-1);
end
end
244
C.1 Predicted CT Silencer Transmission Loss
25
20 0.3 m Straight
0.3 m L
0.3 m Z
15
Transmission Loss [dB]
10
-5
-10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 140: Predicted transmission loss of 0.3 m CT silencers. 95% confidence intervals shown.
40
35 0.5 m Straight
0.5 m L
30 0.5 m U
0.5 m Z
25
Transmission Loss [dB]
20
15
10
-5
-10
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 141: Predicted transmission loss of 0.5 m CT silencers. 95% confidence intervals shown.
245
90
80
1m Straight
70 1m L
1m U
60
1m Z
Transmission Loss [dB]
50
40
30
20
10
-10
-20
125 250 500 1000 2000 4000 8000
Frequency [Hz]
Figure 142: Predicted transmission loss of 1 m CT silencers. 95% confidence intervals shown.
246
Figure 145: Predicted flow velocity in 0.5 m Straight CT silencer – Re = 25,000.
247
Figure 149: Predicted flow pressure in 0.5 m Straight CT silencer – Re = 25,000.
248
Figure 151: Predicted flow velocity in 0.3 m Figure 152: Predicted flow velocity in 0.5 m Figure 153: Predicted flow velocity in 1 m L-
L-shaped CT silencer – Re = 15,000. L-shaped CT silencer – Re = 15,000. shaped CT silencer – Re = 18,000.
249
Figure 154: Predicted flow pressure in 0.3 m Figure 155: Predicted flow pressure in 0.5 m Figure 156: Predicted flow pressure in 1 m
L-shaped CT silencer – Re = 15,000. L-shaped CT silencer – Re = 15,000. L-shaped CT silencer – Re = 18,000.
250
Figure 157: Predicted flow velocity in 0.5 m U-shaped CT Figure 158: Predicted flow velocity in 1 m U-shaped CT silencer – Re = 21,000.
silencer – Re = 20,000.
251
Figure 159: Predicted flow pressure in 0.5 m U-shaped CT Figure 160: Predicted flow pressure in 1 m U-shaped CT silencer – Re = 21,000.
silencer – Re = 20,000.
252
Figure 161: Predicted flow velocity in 0.3 m Figure 162: Predicted flow velocity in 0.5 m
Z-shaped CT silencer – Re = 15,000. Z-shaped CT silencer – Re = 12,000.
253
Figure 164: Predicted flow pressure in 0.3 m Figure 165: Predicted flow pressure in 0.5 m
Z-shaped CT silencer – Re = 15,000. Z-shaped CT silencer – Re = 12,000.
254