JANUARY2831,2013

SANTACLARACONVENTIONCENTER

DeterminingPCBTraceImpedancebyTDR:
ChallengesandPossibleSolutions
Session6TA4
byIstvan Novak,YingLi,Eben Kunz,SarahPaydavosi,LauraKocubinski,KevinHinckley,
AlexanderNosovitski,NathanielShannon,JasonMiller,GustavoBlando
 Acknowledgement,Credit

The authors wish to thank Karl Sauter, Mike Freda,

Gene Whitcomb and Bill Couillard of Oracle, Mike
Resso and OJ Danzy of Agilent, Mike
Tranchemontagne of Tektronix, Joe Beers of Gold
Circuit Electronics, Martin Neusch of ISU-Petasys
and Brian Butler of Introbotics for their help and
expert advice and for supporting the project.

6TA4:DeterminingPCBTraceImpedance 2
 Outline

Introduction and Background

Performance of Uniform Transmission Lines

Manufacturing Variability

Modeling and Potential Compensation

Conclusions

6TA4:DeterminingPCBTraceImpedance 3
 Outline

Introduction and Background

Performance of Uniform Transmission Lines

Manufacturing Variability

Modeling and Potential Compensation

Conclusions

6TA4:DeterminingPCBTraceImpedance 4
 TimeDomainReflectometry (TDR)

High-speed interconnects must be designed for

specified impedances
Validation requires measuring the impedance
Popular option to measure impedance is TDR

Short cable
Oscilloscope DUT

Step-generator

Launched test wave

TDR
instrument
Reflected wave(s)

6TA4:DeterminingPCBTraceImpedance 5
 TDRResponseofUniformLosslessLine

Zref Z0, tpd

Vin
VS

Source DUT

Z 0 Z ref
Roundtripdelay
Z 0 Z ref

Lu L
Zo ; t pd _ u Lu Cu ; t pd LC
Cu C

6TA4:DeterminingPCBTraceImpedance 6
 PCBTraceValidation

Specification:45ohms+/ 10%

Is this
trace OK?

6TA4:DeterminingPCBTraceImpedance 7
 Outline

Introduction and Background

Performance of Uniform Transmission Lines

Manufacturing Variability

Modeling and Potential Compensation

Conclusions

6TA4:DeterminingPCBTraceImpedance 8
 UniformLossy Line

Conductor:

R(f)
R L(f)
L

Insulator: C(f) C R
G(f)

R ( f ) j L ( f )
Z0 ; ( f ) ( R ( f ) jL ( f ))(G ( f ) jC ( f )) ( f ) j ( f ) j t pd
G ( f ) j C ( f )

6TA4:DeterminingPCBTraceImpedance 9
 RLGCParametersofUniformLossy Line
C(f) and L(f) [F, H] R(f) and G(f) [Ohm, Siemens]
2.08E-11 8.E-8 1.E+03 1.E+0
C(f) G(f)

1.E-1
2.04E-11 7.E-8
1.E+02
1.E-2
2.00E-11 6.E-8

1.E-3
1.E+01
1.96E-11 5.E-8 R(f)
1.E-4
L(f)

1.92E-11 4.E-8 1.E+00 1.E-5

1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11
Frequency [Hz] Frequency [Hz]

Wide-band DeBye model used

With increasing frequency:
Capacitance and inductance drop
Resistance and conductance grow
6TA4:DeterminingPCBTraceImpedance 10
 ImpedanceandDelayofUniformLossy Line

Impedance magnitude, phase [Ohm, deg] Group delay and phase delay [s]
58 2 1.2E-09

0
56
-2 1.1E-09
Phase
Phase delay
54 -4

-6 1.0E-09
52 Group delay
-8
Magnitude

50 -10 9.0E-10
1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11
Frequency [Hz] Frequency [Hz]

Wide-band DeBye model used

As a function of frequency Z0 has a shallow
bath-tub shape
With increasing frequency delay drops
6TA4:DeterminingPCBTraceImpedance 11
 SimulatingaUniformLossy Line

Wtest1in20out220N=1RLGCMODEL=W1L=20
+INCLUDEGDIMAG=YES
W Element RLGC Model +INCLUDERSIMAG=YES
*Wlinemodel,valuesareperinch
.param Charimp1=50
Zref Z0, tpd Zref .param Delayperinch1=160p
.param Lunit1='Delayperinch1*Charimp1'
Vin .param Cunit1='Delayperinch1/Charimp1'
VS Vout
.param Rdcunit1=0.2
.param tandelta1=0.01
Source Interconnect Load .MODELW1WMODELTYPE=RLGC,N=1
+Lo=Lunit1
+Co=Cunit1
All TDR responses should settle +Ro=Rdcunit1
at Zref + RDC +Go=0
+Rs ='Rdcunit1/10000'
+Gd ='Cunit1*tandelta1'
6TA4:DeterminingPCBTraceImpedance 12
 UniformLossy Line,ResistanceSweep

20-inch 50-ohm line 20-inch 50-ohm line

Ideal dielectric, Df = 0 Medium dielectric loss, Df = 1%
TDR impedance [ohm] TDR impedance [ohm]
120 120
Max [Ohm] Rdc_u Rdc_u
Max [Ohm]
112 111
100 83 2 100 2
82
67 1 1
67
57 0.5 56.7 0.5
80 53.6 80
0.2 53.3 0.2
51.8 51.6
0.1 0.1
50.72 50.58
60 0.05 60 0.05
0.02 0.02
2*tpd
40 40 2*tpd

20 20

0 0
0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08
Time [s] Time [s]

Resistance creates upslope.

AC resistance produces overshoot.
Dielectric loss reduces impedance!
6TA4:DeterminingPCBTraceImpedance 13
 UniformLossy Line,DC/ACResistance
20-inch 50-ohm line
Medium dielectric loss
Df = 1%

TDR impedance [ohm] TDR impedance [ohm]

90 85.0
Rdc/Rs
80 1k
80.0
2k
70
75.0 5k
60 10k
20k
50 70.0
Rdc/Rs 50k
40 1k 100k
65.0
2k
30 5k
60.0
10k
20
20k
55.0
10 50k
100k
0 50.0
0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09 1.0E-08
Time [s] Time [s]

AC resistance increases apparent impedance and creates a changing slope.

With no AC resistance, TDR response climbs linearly to Zref + RDC
6TA4:DeterminingPCBTraceImpedance 14
 UniformLossy Line,DielectricLoss

20-inch 50-ohm line 20-inch 50-ohm line

Ideal conductor Rdc = 0.2 ohm/inch
TDR impedance [ohm] TDR impedance [ohm]
51 58

56
50

Df(1GHz)
Df(1GHz)
54
0.001
49 0.001 0.002
0.002 52 0.005
0.005 0.01
48 0.01 0.02
0.02 50
0.05
0.05 0.1
0.1
47 48
0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09 1.0E-08 0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09 1.0E-08
Time [s] Time [s]

Dielectric loss lowers apparent impedance.

After a fast initial drop, impedance drops slowly.
6TA4:DeterminingPCBTraceImpedance 15
 UniformLossy Line,Bandwidth

20-inch 50-ohm line 0.5-inch 40-ohm line

Typical parameters Typical parameters

Lower bandwidth reduces horizontal resolution and slows response edge.

6TA4:DeterminingPCBTraceImpedance 16
 UniformLossy Line,LengthSweep

50-ohm line
Typical parameters

TDR impedance [ohm] TDR impedance [ohm]

60 58.0

Length ["]
50
20
56.0
10
40 5
Rdc [Ohm] Length ["] 2
4 20 Rdc [Ohm]
1
30 2 54.0 4
10 0.5
1 2
5 0.2 1
0.4
2 0.1 0.4
20 0.2
0.1 1 0.2
0.04 0.5 52.0 0.1
0.02 0.04
10 0.2
0.02
0.1

0 50.0
0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09 1.0E-08
Time [s] Time [s]

Envelope response can be used for any length of DUT

6TA4:DeterminingPCBTraceImpedance 17
 UniformLossy Line,Impedance

20-inch line
Typical parameters

TDR impedance [ohm] TDR impedance [ohm]

70

60
62.0

50
57.0
40
57.5 Ohm
57.5 Ohm
30 55.0 Ohm
52.0 55.0 Ohm
52.5 Ohm
52.5 Ohm
20 50.0 Ohm
50.0 Ohm
47.5 Ohm 47.0 47.5 Ohm
10 45.0 Ohm
45.0 Ohm
42.5 Ohm
42.5 Ohm
0 42.0
0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09 1.0E-08
Time [s] Time [s]

Response shape is independent of DUT characteristic

impedance as long as loss parameters stay the same
6TA4:DeterminingPCBTraceImpedance 18
 Outline

Introduction and Background

Performance of Uniform Transmission Lines

Manufacturing Variability

Modeling and Potential Compensation

Conclusions

6TA4:DeterminingPCBTraceImpedance 19
 DeviceUnderTest,Connections

Ground vias

Semirigid probe
Total length: 17

Signal vias

Precision 0402
50-ohm resistor Differential
wafer probe
Solder braid

Signal vias

6TA4:DeterminingPCBTraceImpedance 20
 Repeatability
Same board, same trace measured multiple times.
Same day with same calibration.
Next day with old calibration.
Next day with new calibration.
Very little variation was seen.
TDR impedance [Ohm] TDR impedance [Ohm]
52 52

50 50

48 48

46 46

44
Traceismeasured
44 Traceismeasured
board6-SW6-P board6-SW4-P
twicefromleft twicefromright
board6-SW4-P-repeat
board6-SW6-P-repeat
42 42

40 40
0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09 0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09
Time [s] Time [s]

6TA4:DeterminingPCBTraceImpedance 21
 BoardtoBoardVariation
Six boards from two vendors
The same trace pair is measured
Same calibration, same probe
All traces follow the simulated signatures
TDRImpedance,LeftN[Ohm] TDRImpedance,LeftP[Ohm]
54 54

52 52

50 50

48 48

46 46

Nlegofdiffpairis Plegofdiffpairis
44
measuredfromleft 44
measuredfromleft

42 42
0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09 0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09

6TA4:DeterminingPCBTraceImpedance 22
 PN,LeftRightVariation
Two boards from the same vendor

TDRImpedance,Board1,VendorA[Ohm] TDRImpedance,Board2,VendorA[Ohm]
54 54

PandNmeasuredfromleft PandNmeasuredfromright
52 52
Responsehaslowerslope Responsehaslowerslope
50 50

48 48

46 46
Board1 board1-right-N Board2 board2-right-N
board2-right-P
44 PandNmeasuredfromright
board1-right-P
44 PandNmeasuredfromLEFT
board2-left-N
board1-left-N
Responsehashigherslope
board1-left-P Responsehashigherslope
board2-left-P

42 42
0 1E-09 2E-09 3E-09 4E-09 5E-09 6E-09 0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09
Time[s] Time[s]

6TA4:DeterminingPCBTraceImpedance 23
 PN,LeftRightTrends
Two boards from the same vendor
Different slopes from left-to-right vs. right-to-left!!!

TDRImpedance,Board1[Ohm] TDRImpedance,Board2[Ohm]
52 52

51 51

50 50

49 49

48 48

47 47

board1-right-N board2-right-N
46 46
board1-right-P board2-right-P
board1-left-N board2-left-N
45 45
board1-left-P board2-left-P

44 44
0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09 0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09
Time[s] Time[s]

6TA4:DeterminingPCBTraceImpedance 24
 VNAMeasurements,Ports

Total length: 17

P P
Port 1 Port 2

Left Right

Port 3 Port 4
N N

6TA4:DeterminingPCBTraceImpedance 25
 SingleEndedReflections
P P
Port 1 Port 2

Left Right

Port 3 Port 4
N N

S11 S33 magnitude [dB] S22 S44 magnitude [dB]

0 0

-5 -5

-10 -10

-15 -15

-20 -20

-25 -25

-30 -30

-35 -35

-40 -40
1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11
Frequency [Hz] Frequency [Hz]

Up to 10 GHz only small differences

6TA4:DeterminingPCBTraceImpedance 26
 DifferentialReflections
P P
Port 1 Port 2

Left Right
1 2
Port 3 Port 4
N N

Sdd11 Sdd22 magnitude [dB] Scc11 Scc22 magnitude [dB]

0 0

-5
-5
-10
-10
-15
-15
-20

-20 -25

-30
-25
-35
-30
-40
-35 -45

-40 -50
1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11
Frequency [Hz] Frequency [Hz]

5-10dB difference in the 1-10 GHz range

6TA4:DeterminingPCBTraceImpedance 27
 Transfer,ModeConversion
P P
Port 1 Port 2

Left Right
1 2
Port 3 Port 4
N N

Sdd21 Scc21 [dB] Sdc21 Scd21 magnitude [dB]

0 0

-10 -10

-20 -20

-30 -30

-40 -40

-50 -50

-60 -60

-70 -70
1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 1.E+07 1.E+08 1.E+09 1.E+10 1.E+11
Frequency [Hz] Frequency [Hz]

Up to 10 GHz only small differences

6TA4:DeterminingPCBTraceImpedance 28
 Delay,Skew
P P
Port 1 Port 2

Left Right
1 2
Port 3 Port 4
N N

Propagation delay [s] Propagation delay (left axis) and skew (right axis) [s]
3.15E-09 2.86E-09 2.0E-11
1>>2 SE
2.85E-09 3>>4 SE
3.10E-09
DM 1.8E-11
2.84E-09
3.05E-09 CM
2.83E-09 Delta_tpd SE
Delay 1.6E-11
3.00E-09
2.82E-09

2.95E-09 1>>2 SE 2.81E-09 1.4E-11

3>>4 SE
2.80E-09
2.90E-09 DM
1.2E-11
CM 2.79E-09
2.85E-09 Skew
2.78E-09
1.0E-11
2.80E-09
2.77E-09

2.75E-09 2.76E-09 8.0E-12

1.E+07 1.E+08 1.E+09 1.E+10 1.E+11 0.0E+00 4.0E+09 8.0E+09 1.2E+10 1.6E+10
Frequency [Hz] Frequency [Hz]

6TA4:DeterminingPCBTraceImpedance 29
 CrossSections,Dimensions
a

b f i k

c h m
d g j l

Percent deviation from average:

Location: b d c f g k l m i j
1 2.43 5.45 -2.10 2.07 -1.17 3.10 4.00 -0.89 0.86 0.00
2 2.43 1.93 -2.10 -0.46 2.74 -0.18 2.13 -0.89 0.86 0.79
3 -0.83 -4.42 8.41 -0.46 -2.22 -0.18 -5.82 3.56 -0.06 -1.05
4 2.43 0.99 -2.10 1.15 1.69 3.10 0.49 -0.89 -0.06 0.79
5 -0.83 -2.54 -2.10 -2.30 -2.22 -2.56 -2.31 -0.89 -0.75 0.00
6 -2.01 -2.54 8.41 2.07 1.69 -1.37 -3.01 3.56 4.32 0.79
7 -4.38 -1.83 -2.10 -2.30 -3.26 -2.56 -1.38 -0.89 -2.59 -2.11
8 -2.01 -2.54 -2.10 0.23 2.74 -0.18 -0.44 -0.89 -2.59 0.79
9 2.43 0.05 -2.10 -1.13 -1.35 -0.18 1.43 -0.89 -4.89 -1.16
10 0.36 5.45 -2.10 1.13 1.35 1.01 4.93 -0.89 4.89 1.16

Random!
6TA4:DeterminingPCBTraceImpedance 30
 CrossSections,GlassWeave

Copperplane Coppertrace Glass(darkcolor) Resin(lightcolor)

Highercapacitanceduetoglassbundle. Morecouplingduetoglassbundle.

Same differential pair at two different locations.

Glass-weave effect changes local Dk.

6TA4:DeterminingPCBTraceImpedance 31
 EstimatedDeviation

TDR impedance and estimated deviation [Ohm, %]

54 8

6
52
4
50
2

48 0

-2
46
-4
44
-6

42 -8
0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09
TDR time from left [s]

Little correlation!

6TA4:DeterminingPCBTraceImpedance 32
 RecalculateDCResistance

Measurementlinesmust
beplacedacrossminima.

Even at DC, surface roughness matters.

Current will not (fully) penetrate the peaks.
For DC resistance purposes, copper thickness must be measured
between lines across minima.

6TA4:DeterminingPCBTraceImpedance 33
 DCResistanceMeasurements
TDR impedance [Ohm] TDR impedance [Ohm]
60 60

50 50

40 40

30 30

20 20
5.37Ohm 5.31Ohm
10 10

0 0
0.E+00 2.E-09 4.E-09 6.E-09 8.E-09 1.E-08 0.E+00 2.E-09 4.E-09 6.E-09 8.E-09 1.E-08
Time [s] Time [s]

Reasonable correlation is achieved only with copper

thickness measured between lines across minima.

6TA4:DeterminingPCBTraceImpedance 34
 CascadedLines
TDR impedance [ohm] TDR impedance [ohm]
70 65.0

60
60.0
50

55.0
40 42.5-57.5 ohm
45-55 ohm 42.5-57.5 ohm
30 47.5-52.5 ohm 45-55 ohm
50.0
50-50 ohm 47.5-52.5 ohm
20 52.5-47.5 ohm 50-50 ohm
55-45 ohm 45.0 52.5-47.5 ohm
10 57.5-42.5 ohm 55-45 ohm
57.5-42.5 ohm
0 40.0
0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09
Time [s] Time [s]

Each cascaded section follows its own trend.

The only interaction among sections is due to edge slow-down.

6TA4:DeterminingPCBTraceImpedance 35
 TaperedLines
TDR impedance [ohm] TDR impedance [ohm]
70 62.0
50-55-ohm taper
60 60.0 50-45-ohm taper
50-45-ohm low-loss taper
50 58.0

40 50-55-ohm taper 56.0

50-45-ohm taper
30 54.0
50-45-ohm low-loss taper
20 52.0

10 50.0

0 48.0
0.0E+00 5.0E-09 1.0E-08 1.5E-08 2.0E-08 2.5E-08 3.0E-08 0.0E+00 2.0E-09 4.0E-09 6.0E-09 8.0E-09
Time [s] Time [s]

Even if specified as a uniform trace, its characteristic impedance

could change or vary within the specified limits.
Simulations with ten cascaded uniform lines.
The characteristic impedance values step up or step down.

6TA4:DeterminingPCBTraceImpedance 36
 Outline

Introduction and Background

Performance of Uniform Transmission Lines

Manufacturing Variability

Modeling and Potential Compensation

Conclusions

6TA4:DeterminingPCBTraceImpedance 37
 ApproximatingtheSlope

TDRImpedance,LeftN[Ohm]
54

52

If the interconnect is uniform

50 and its parameters are known,
we know the TDR response.
48
In long resistive traces the
response has a positive tilt.
The positive tilt can be
46
approximated and subtracted
from the measured response.
44

42
0.E+00 1.E-09 2.E-09 3.E-09 4.E-09 5.E-09 6.E-09

6TA4:DeterminingPCBTraceImpedance 38
 LinearApproximations
A linear estimate ignores the A linear estimate of the actual
impedance increase due to AC response still has large errors close
resistance. to the excitation edge.

TDR impedance [ohm] TDR impedance [ohm]

51 51

49 49

Simulated
measured
Linear (Simulated)
47 47
Simulated
measured
[9] estimate y = 1.00E+09x + 4.53E+01

45 45
0.E+0 1.E-9 2.E-9 3.E-9 4.E-9 5.E-9 0.E+0 1.E-9 2.E-9 3.E-9 4.E-9 5.E-9
Time [s] Time [s]

6TA4:DeterminingPCBTraceImpedance 39
 PolynomialApproximations
A sixth-order polynomial captures Subtracting the estimated response
the time-dependent impedance from the measured TDR curve.
with sufficient accuracy. Trace is within +1.5 -0.5% of
Lines are undistinguishable. nominal.
TDR impedance [ohm] TDR impedance [ohm]
52 52 2.00

50
50 1.00

48

48 0.00
Simulated
46 measured
Fitted simulation data
46 measured -1.00
44 Fitted simulation data
Deviation

42 44 -2.00
0.E+0 1.E-9 2.E-9 3.E-9 4.E-9 5.E-9 0.E+0 1.E-9 2.E-9 3.E-9 4.E-9 5.E-9
Time [s] Time [s]

6TA4:DeterminingPCBTraceImpedance 40
 Conclusions
Long and/or resistive traces have significant difference between actual and
apparent impedance.
Apparent impedance is increased with increasing DC and AC resistances.
AC resistance creates overshoot beyond the DC resistance value.
Apparent impedance is decreased by dielectric loss.
DC resistance is increased by surface roughness.
Thin 20-40 traces may have apparent impedances go outside the
tolerance window.
In PCB manufacturing, resistive panel coupons need compensation.
In full-board validations, TDR results need compensation.
A sixth-order polynomial is suggested to compensate for losses.
6TA4:DeterminingPCBTraceImpedance 41
 ClosingRemarks
TDR has been used for decades to check trace impedance.
Why was this not a problem earlier?
TDR impedance [ohm]
58

1ou
56
0.5ou
54 5mil
Df = 3%, Rdc = 0.05 ohm/inch 10mil
`
Df = 0.5%, Rdc = 0.2 ohm/inch
52
Because we used
50 lossier dielectrics, wider
and heavier traces and
48
0.E+0 2.E-9 4.E-9 6.E-9 8.E-9 1.E-8
Time [s]
maybe shorter etches
6TA4:DeterminingPCBTraceImpedance 42
 THANKYOU!

AnyQuestions?

6TA4:DeterminingPCBTraceImpedance 44