Sunteți pe pagina 1din 1

Evaluation of Mismatch Uncertainty in Microwave Power Sensor

Calibration Using Monte Carlo Method


Thomas Y. Wu, S. W. Chua
National Metrology Centre (NMC)
Agency for Science, Technology and Research (A*STAR), Singapore
email: thomas_wu@nmc.a-star.edu.sg

Overview
(1  RS  I S )[(1  RD RG  I D I G )2  ( RD I G  I D RG )2 ]
2 2

The mismatch uncertainty in microwave power sensor MM 


(1  RD  I D )[(1  RS RG  I S I G )2  ( RS I G  I S RG )2 ]
2 2

calibration is analyzed. Propagation of uncertainties from the


6
reflection coefficients in real and imaginary format are U RI  k C ' 2
i u ' 2
i
considered. Monte Carlo method (MCM) is used to estimate the i 1

probability distribution function (PDF) of the mismatch factor,


which is found to be non-Gaussian for small reflection coefficient
Mismatch Uncertainty Obtained Using
values. The mismatch uncertainties obtained by MCM for various Monte Carlo Method
reflection coefficient values are larger than those obtained by the Monte Carlo method (MCM) is used to propagate the PDF
conventional method recommended in the Guide to the of various reflection coefficients to obtain the PDF of ΓG and
Expression of Uncertainty in Measurement (GUM). This study mismatch factor MM. Mismatch uncertainty is then obtained
shows that MCM gives more reliable estimation of the mismatch from the PDF of MM. The PDF of MM obtained by MCM is
uncertainty than a GUM uncertainty framework with first-order
compared to that obtained by GUM uncertainty framework with
terms in microwave power sensor calibration.
first-order terms for three cases of data.
Histogram Histogram

Mismatch Uncertainty In Microwave Power 0.02 0.02


Histogram of the samples
Histogram of the samples Histogram of a Gaussian distribution
0.018 Histogram of a Gaussian distribution 0.018

Sensor Calibration 0.016

0.014
0.016

0.014

Probability density

Probability density
0.012 0.012
The mismatch factor (MM) is a key uncertainty component in 0.01 0.01

microwave power sensor calibration. It is computed from several 0.008 0.008

0.006 0.006

reflection coefficients through a nonlinear function. 0.004 0.004

0.002 0.002

0 0
-0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06
Power Meter with PMS Equivalent source reflection coefficient (real part) Equivalent source reflection coefficient (imag part)

Power Sensor
PMD
(Monitoring)
Figure 2. The PDF of the real and imaginary part of ΓG obtained by
S
S PS
MCM using case 1 data. The PDF of a corresponding Gaussian
distribution with the same mean and variance of RG and IG is also
Standard Power
Sensor meter plotted in red line.
Histogram
Resistive Test port 0.05
Power splitter GUM
0.045 MCM

G 0.04
Power
DUT 0.035
meter
Probability density

0.03
Microwave D
Synthesizer D PD 0.025

0.02

0.015

0.01

Figure 1. Block diagram of a direct comparison microwave 0.005

power sensor calibration system. 0.995 1


Mismatch factor
1.005

(a) Case 1 data


Histogram Histogram
0.02 0.02

PD PMS (1  S )( 1 - D G )
2 2 GUM GUM

D  S
0.018 MCM 0.018 MCM

PS PMD (1  D )( 1 - S G )
2 2 0.016 0.016

0.014 0.014
Probability density
Probability density

0.012 0.012

S12S 23
G  S 22 
0.01 0.01

0.008 0.008
S13 0.006 0.006

(1  S )( 1 - D G )
2 2 0.004 0.004

MM  0.002 0.002

(1  D )( 1 - S G )
2 2
0 0
0.94 0.95 0.96 0.97 0.98 0.99 1 1 1.05 1.1 1.15 1.2
Mismatch factor Mismatch factor

(b) Case 2 data (c) Case 3 data

Mismatch Uncertainty Obtained Using GUM Figure 3. Comparison of the PDFs of MM obtained by GUM
Uncertainty Framework With First-Order Terms uncertainty framework with first-order terms and MCM, using
three cases of reflection coefficient values.
Propagation of uncertainties from the magnitude and phase
of the reflection coefficients has statistical difficulties. Thus MM Table I
has been derived from the real and imaginary parts of the 95% coverage interval length of the mismatch factor, obtained
complex reflection coefficients and the mismatch uncertainty is by the GUM uncertainty framework with first-order terms and
obtained using GUM uncertainty framework with first-order terms. the MCM for three cases of data
95% Coverage Interval Length
Three cases of reflection coefficient values with magnitude Method
Case 1 Case 2 Case 3
ranging from 0.003 to 0.33 (based on real data at 1 GHz, 40 GHz
GUM 0.0007 0.0207 0.0664
and 50 GHz) have been used to evaluated the mismatch
MCM 0.0021 0.0248 0.0854
uncertainties.

S-ar putea să vă placă și