Uncertainty evaluations in
EMC measurements
Carlo Carobbi
Dipartimento di Elettronica e Telecomunicazioni
Università degli Studi di Firenze
Politecnico di Milano - 20 Feb. 2009
1
Non - reproducibility of radiated
emission tests
„
„
„
„
Presentation mainly focused on this
topic
The contribution due to test site
imperfections will be analyzed
Important subject per se
All the typical ingredients of EMC
measurement uncertainty evaluation
are involved
Politecnico di Milano - 20 Feb. 2009
2
Sources of non – reproducibility
of radiated emission tests
„
Imperfections of:
„
„
„
„
„
„
Test sites
Antennas
Receivers
Connections
Set-up (geometry)
EUT (intrinsic instability, layout of
cables and auxiliary equipment)
Politecnico di Milano - 20 Feb. 2009
3
Isolating the contribution due to
test site imperfections
„
„
„
Collaborative exercise performed: radiated
emission measurement repeated in several
different test sites
Same instrumentation involved in each site
(field source, receiving antenna, spectrum
analyzer, cables)
Same geometry, same measurement
procedure (pre-defined measurement
protocol), same personnel
Politecnico di Milano - 20 Feb. 2009
4
Test sites involved
„
„
Compact size fully anechoic rooms
14 nominally equivalent sites
investigated
„
„
„
30 – 300 MHz frequency range
Vertical polarization
Received power (dBm) is the measured
quantity
Politecnico di Milano - 20 Feb. 2009
5
Resolving power of the method
„
„
Ability to discriminate a site from
another
Limited by measurement non
repeatability
Politecnico di Milano - 20 Feb. 2009
6
Achieving repeatability
„
„
„
„
„
„
„
Stable field source, battery operated
Care of positioning (distance and mutual
alignment)
Well balanced receiving antenna
Weak coupling with and reflections from the
length of cable inside the room
High signal to noise ratio and numerical
averaging
Spectrum analyzer warm-up and self calibration
Automatic measurement
Politecnico di Milano - 20 Feb. 2009
7
Repeatability quantified
„
Spectrum Analyzer noise
± 0.2 dB (± 0.02 dB)
Spectrum Analyzer amplitude
resolution and repeatability
± 0.1 dB
Generator instability
(intrinsic + thermal fluctuations)
± 0.05 dB
Positioning uncertainty
± 0.02 dB
Inversion test
± 0.14 dB
Total
± 0.27 dB (± 0.18 dB)
Repeatability better than 0.3 dB or 0.2 dB (1 std.
dev.), depending on the signal to noise ratio
Politecnico di Milano - 20 Feb. 2009
8
Results: dispersion among sites
7
6
Standard
deviation 14 sites
4
3
2
Repeatability
1
300
270
240
210
180
150
120
90
60
0
30
dB
5
f (MHz)
Politecnico di Milano - 20 Feb. 2009
9
Results: mean received power vs.
prediction in ideal empty space
-60
Ideal
-80
-90
Mean
-100
Average noise level
-110
300
270
240
210
180
150
120
90
60
-120
30
p o w e r (d B m )
-70
f (MHz)
Politecnico di Milano - 20 Feb. 2009
10
0.5
0.4
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
„
Observed
0.16 dB std. dev.
Predicted
0.11 dB std. dev.
300
270
240
210
180
150
120
90
60
„
30
dB
Results: source + spectrum
analyzer observed instability
f (MHz)
Politecnico di Milano - 20 Feb. 2009
11
Questions
„
„
„
„
Sites investigated “equivalent”?
Deviation dominated by a minority of
bad performing sites?
Correlation with sites’ physical structure
possible?
Site correction factor?
Politecnico di Milano - 20 Feb. 2009
12
Deviations from the mean
15
#4
#5
10
5
0
-5
-10
#1
# 10
300
270
240
210
180
150
120
90
60
-15
30
dB
#6
f (MHz)
Politecnico di Milano - 20 Feb. 2009
13
Sites’ structural characteristics
l
a
r
g
e
Site #
Volume (m^3)
Absorbing lining
1
63
Ferrite
2
93
3
101
4
105
5
112
6
134
7
150
Ferrite + Pyramid
216
Ferrite + Pyramid 10/30/50 cm
321
Ferrite + Pyramid 45 cm
323
Ferrite
324
Ferrite + Pyramid 50 cm
371
Ferrite + Pyramid 50/200 cm
13
743
Pyramid 60 cm
14
1152
Pyramid 60 cm
8
9
10
11
12
m
e
d
i
u
m
s
m
a
l
l
Ferrite + Pyramid 45/60 cm
Ferrite + Pyramid 50 cm
Pyramid 60 cm
Ferrite + Pyramid 32/50 cm
Pyramid 30/55/65 cm
Politecnico di Milano - 20 Feb. 2009
14
Deviations from the mean
(removed small sites without ferrite)
15
#5
10
# 14
dB
5
0
-5
# 10
-10
-15
30
60
90
120
150
180
210
240
270
300
f (MHz)
Politecnico di Milano - 20 Feb. 2009
15
Rejecting outliers
# outliers
site #
6
5
4
14
„
3
10
„
2
13
2
1
2
7
1
8
1
2
1
12
1
11
„
„
„
Chauvenet’s rejection criterion
applied at each frequency
23 outliers distributed over 10 sites
Site #5 worst performing: 6 outliers
Decision to reject 3 measured
values: 2 (site 5) + 1 (site 10)
We are not rejecting blunders (risk
missing information)
Politecnico di Milano - 20 Feb. 2009
16
Dispersion:
2 sites removed and 3 outliers rejected
7
6
Standard deviation
12 sites
5
dB
4
3
2
1
Repeatability
0
30
60
90
120
150
180
210
240
270
300
f (MHz)
Politecnico di Milano - 20 Feb. 2009
17
Mean:
2 sites removed and 3 outliers rejected
-60
Ideal
power (dBm)
-70
-80
-90
Mean
-100
Average noise level
-110
-120
30
60
90
120
150
180
210
240
270
300
f (MHz)
Politecnico di Milano - 20 Feb. 2009
18
What we conclude
„
Need of inter-laboratory comparisons (both
collaborative exercises and proficiency tests)
„
„
„
„
„
„
Reproducibility quantified
Get physical insight
Uncertainty evaluation applied
Lab personnel involved in non-standard
experiments
Not expensive practice
If well designed can cover any type of EMC test
(RE, RS, CE, CS)
Politecnico di Milano - 20 Feb. 2009
19
A note on uncertainty calculations
in EMC
„
„
Quite large deviations
Extensive use of dB units
„
„
Specific asymmetric probability density functions
involved
„
„
„
A problem when mixing natural and logarithmic
quantities
Log-normal
Rice (weak signal plus receiver noise, strong multipath
interference)
All these analytical aspects dealt with in GUM
supplement 1
Politecnico di Milano - 20 Feb. 2009
20
Thank you for your kind attention
FM
Politecnico di Milano - 20 Feb. 2009
21