system identification approach applied to drift estimation
DESCRIPTION
Vrije Universiteit Brussel Faculteit Ingenieurswetenschappen Vakgroep ELEC Pleinlaan 2, B-1050 Brussels, Belgium. System Identification Approach applied to Drift Estimation. Frans Verbeyst Rik Pintelon Yves Rolain Johan Schoukens Tracy Clement. Motivation = phase calibration. - PowerPoint PPT PresentationTRANSCRIPT
System Identification Approachapplied to
Drift Estimation.
Vrije Universiteit BrusselFaculteit IngenieurswetenschappenVakgroep ELECPleinlaan 2, B-1050 Brussels, Belgium
Frans VerbeystRik PintelonYves RolainJohan SchoukensTracy Clement
2
Motivation = phase calibration
LSNA phase calibration
requires
requires
a calibrated oscilloscope
a calibrated HPR
3
Sampling oscilloscopes “reality”
drift
jitter
mismatches, connector saver
time base errors
distortiontk k.t
vertical errors
offset
nonlinearity
dynamics
measure,estimate,compensate
vertical cal plug-in
avoid: usesmall signal
measure and compensate
nose2noseEOS
4
Enhanced drift estimation
old:
new:
5
200 400 600 800freqGHz0.1
0.2
0.3
0.4
0.5varimx106 V2
freq (GHz)
var(im) (x10-6 V2)
200 400 600 800freqGHz0.1
0.2
0.3
0.4
0.5varrex106 V2
freq (GHz)
var(re) (x10-6 V2)
10 20 30 40 50 60 70 80freqGHz
0.4
0.2
0.2
0.4
re,im(re,im)
freq (GHz)
Circular complex noise ?
5000 realizations(jitter only, stdev 1 ps)
10 11 12 13 14remV
10.750.5
0.25
0.250.5
0.751
immVim (mV)
re (mV)
9.5 10 10.5 11remV
1
2
3
immV
re (mV)
im (mV)
5.5 6 6.5 7 7.5remV
4.5
3.5
3
2.5
2
1.5
1
immVim (mV)
re (mV)
1.5 1 0.5 0.5 1 1.5remV
1.5
1
0.5
0.5
1
1.5
immV
re (mV)
im (mV)
6
Covariance info: time - frequency
5000 realizations(jitter only, stdev 1 ps)Cov[y(t)] Cov[Y()]
Fourier transform, separate real and imag. part
10 20 30 40 50 60 70 80freqGHz
0.5
0.4
0.3
0.2
0.1
0.1
0.2
re,im
10 20 30 40 50 60 70 80freqGHz
0.5
0.4
0.3
0.2
0.1
0.1
0.2
re,im
freq (GHz)
uncertainty on LS estimate
construct WLS estimator
relevant value of cost
(re,im)
7
Comparison
0 1000 2000 3000 4000
2
1
0
1
2
K. Coakley and P. Hale, “Alignment of Noisy Signals,” IEEE Transactions on Instrumentation and Measurement,
Vol. 50, No. 1, February 2001
EstimatorRMS prediction error
95% conf. interval
naive cross-correlation ~ 0.25
complete cross-correlation ~ 0.15
naive LS 0.16 0.14 .. 0.19
enhanced LS 0.16 0.14 .. 0.19
enhanced WLS 0.08 0.07 .. 0.09
Case of moderate jitter and small additive noise
(arbitrary units)
8
Measurements
impulselaser
calibratedO/E
samplingoscilloscope
trigger2nd O/E
Ch1/3impulselaser
impulselaser
calibratedO/E
calibratedO/E
samplingoscilloscope
trigger2nd O/E
Ch1/3calibratedO/E
impulselaser
2nd O/E
samplingoscilloscope
trigger
Ch1/3
0.46 0.48 0.5 0.52 0.54 0.56 0.58 0.6tns
0.02
0.02
0.04
0.06
0.08
ytV
t (ns)
y(t) (V)
0.46 0.48 0.5 0.52 0.54 0.56 0.58 0.6tns
0.02
0.02
0.04
0.06
0.08
ytV
t (ns)
y(t) (V)
9
100 200 300 400 500
realizationindex
0.5
1
1.5
2
estimateddriftps
uncertainty reduced by factor of 2 !
Measurements
100 200 300 400 500
realizationindex
0.5
1
1.5
2
estimateddriftps
100 200 300 400 500
realizationindex
0.5
1
1.5
2
estimateddriftps
naive LS
enhanced LSenhanced WLS
estimateddrift (ps)
realizationindex
expected value WLS cost:2.M.K - p
p = 2.M + K
M = K = 500
expected value = 498500 ± 1997actual value = 500346
10
Conclusions
• System identification framework
• Outperforms any other published technique
• Proper weighting: relevant value of cost decreased uncertainty on estimated drift: factor 2
• Estimation of drift in presence of jitter