Improved Noise Suppression for MSE Analysis

1
Improved Noise Suppression for MSE Analysis

• DNB Group discussion
• November 25, 2002
• New material 27 Dec 2002
• S. D. Scott
• PPPL
• Thanks to Gerrit Kramer and Fred Levinton, PPPL

2
20 KHz PEM Nearly equal Up and Down points
23 points up
22 points down
3
PEM 5 more down than up points
22 points up
27 points down
4
Example of PEM Drive with 5-volt values
Affected points
Time index ?
Both PEM drives are affected. Typically, about
50 data points are affected per-shot out of
130,000 0.05. Not a big deal unless it gets
worse.
Shot 1021108007
5
Phase of Signal 40 kHz Component
Phase constant with radius, except for
180-degree shift at innermost point
2nd DNB blip
1st DNB blip
EDGE MSE CHANNEL
CORE
Shot 1021023018
6
Phase of Signal 40 kHz Component
Phase constant with radius, except for
180-degree shift at innermost point
2nd DNB blip
1st DNB blip
D 180o
D 180o
EDGE MSE CHANNEL
CORE
Shot 1021023018
7
Phase of Signal 44 kHz Component
Phase constant with radius, except for 180-degree
shift at innermost point
2nd DNB blip
1st DNB blip
EDGE MSE CHANNEL
CORE
Shot 1021023018
8
Phase of Signal 44 kHz Component
Phase constant with radius, except for 180-degree
shift at innermost point
2nd DNB blip
D 180o
1st DNB blip
EDGE MSE CHANNEL
CORE
Shot 1021023018
9
Phase of noise 40 kHz Component
Time bin 1
3
5
2
4
6
EDGE MSE CHANNEL
CORE
Shot 1021023018
10
Phase of noise 40 kHz Component
Phase reasonably constant in radius within a
20-degree band near the edge, somewhat larger at
the core
20-degree bands
Time bin 1
3
5
D 180o
2
4
6
EDGE MSE CHANNEL
CORE
Shot 1021023018
11
Phase of noise 44 kHz Component
6
2
5
Time bin 1
4
3
EDGE MSE CHANNEL
CORE
12
Phase of noise 44 kHz Component
Phase constant in edge, but varies considerably
near the core
20-degree bands
6
2
5
Time bin 1
4
3
EDGE MSE
CHANNEL CORE
13
Noise at 40 kHz versus time
8
Interpolation For DNB
9 (core)
3
6
7
4
2
1
0 (edge)
5
Time index (6 before DNB, 7 after DNB)
14
Noise at 44 kHz versus time
Note factor 10 increase between points that are
used to provide Interpolation
Interpolation For DNB
8
9 (core)
7
4
3
5
2
0 (edge)
1
6
Time index (6 before DNB, 7 after DNB)
15
Noise at 40 kHz profiles vs channel number
Just before DNB
Last noise time point
EDGE MSE
CHANNEL CORE
16
Noise at 44 kHz profiles vs channel number
Last noise time point
Just before DNB
EDGE MSE
CHANNEL CORE
17
Present MSE Analysis Technique

• Current method Perform FFT on MSE signal for
  each channel, then
• fit a gaussian to the FFT amplitudes near 40
  and 44 KHz, then compute
• the area under the gaussian to obtain the
  amplitude at these frequencies.
• To account for noise, subtract the no-beam fft
  amplitudes on a frequency-
• by-frequency basis before fitting a gaussian
  to the net signal.
• Limitation of current method doesnt impose
  any requirement that the
• measured signals be synchronous with the phase
  of the PEM.
• Particularly in situations with weak
  polarization fraction and large noise, this
• could lead to over-counting the noise photons.
• Also, the current method requires that the time
  periods for the beam and
• no-beam signal acquisition be identical
  otherwise the beam and no-beam
• FFT arrays will be on a different frequency
  grid.

18
New MSE Analysis Technique

• New method construct a sine wave that is
  synchronous with the PEM
• square-wave drives.
• We observed that the MSE signals were
  consistently about 0.25 and
• 0.05 radians behind the PEM square-wave
  drives on both calibration
• shots and plasma shots having high
  signal-to-noise.
• So what we really do is construct sine waves
  that are about 0.25 and 0.50
• radians lagging the PEM drive.
• Multiply each MSE signal by the sine wave.
• Compute the FFT of the product, and take the
  lowest-frequency component.
• This component is the 40 or 44 khz amplitude.
• Calculate the angle from atan (40-khz amplitude
  / 44-khz amplitude) in
• the usual way.
• Advantage eliminate signals that are not in
  phase with the PEMs.

19
New Analysis doesnt affect calibration much
20
New Analysis less scatter in calibration data
Scatter taken from dividing beam pulse into three
parts (shot 1021004011 I think).
21
Old Analysis Possible problems with
q-measurement at high density? individual shots
Nel4 Series 4
0.46-0.58 Series 11 0.65-0.85 Series 18
1.01-1.10
Run day 1021017 Series 4 29, 30, 32
Series 11 20, 21, 28 Series 18 22, 26, 27
22
New Analysis Little Dependence of MSE Angle on
Density
Nel4 Series 4
0.46-0.58 Series 11 0.65-0.85 Series 18
1.01-1.10
Run day 1021017 Series 4 29, 30, 32
Series 11 20, 21, 28 Series 18 22, 26, 27
23
Shot-to-Shot Scatter in MSE angle is generally
reduced with new analysis technique but not
always.

• Examine 4 sets of shots each set contains 3
  identical shots.
• Data are shown in order of increasing density
• Set N NeL-4 0.65 e20 ? nebar
  0.90 1020m-3
• SET G NeL-4 0.92 e20 ? nebar 1.29
  1020m-3
• Set D NeL-4 1.00 e20 ? nebar
  1.39 1020m-3
• SET E2 NeL-4 1.15e20 ? nebar 1.59
  1020m-3
• Generally, new analysis technique does better
  in the core,
• where the noise is high.
• Dont understand why new technique doesnt do a
  better job
• consistently.

24
SET N ò nedl (ch4) 0.65 1020 m-2 nebar
0.90 1020 m-3
25
SET G NeL4 0.92e20 nebar 1.29 1020 m-3
26
SET D NeL4 1.00e20 nebar 1.39 1020 m-3
27
SET E2 NeL4 1.15e20 nebar 1.59 1020 m-3
28
Scatter in Real Pitch-Angle Degrees
29
Conclusions

• We should fix the PEM drive so that it has an
  equal number of on- and off-points, and so that
  the on-value is always the same value.
• We should determine the likely consequences of
  having the PEM drive sometimes hang up at 5 vs 15
  volts.
• Replacing the present analysis technique with one
  the numerically mimics a lock-in amplifier looks
  promising certainly it eliminates the spurious
  density dependence that is observed under the
  present analysis.

30
Ratio of signal levels R85.5 / R84.0
Issue the ratio typically hovers about unity
(within a factor of two or so) with the exception
of a some shots for which it varies by more than
an order of magnitude up or down. This holds
true for both the measured noise level and net
signal, i.e. signal background. Shot
10 11 12 14
21 23 net
signal 33. 1.0 0.5 0.5
0.12 1.0 Noise
17. 0.6 0.7 0.5
0.12 0.66 Data from November 8,
2002 Maybe this behavior is related to the
complaints by BES of large signal-strength
variability?
31
MSE channels closer to plasma core
32
Extract data from shot 010
33
Data shot 011 only
34
Data shot 021
35
Data shot 023
36
Extreme comparison shot 10 vs 21
37
Five consecutive shots varying density
38
Same data log scale
39
Old Analysis possible problems with
q-measurement at high density?
Nel4 Series 4
0.46-0.58 Series 11 0.65-0.85 Series 18
1.01-1.10
Run day 1021017 Series 4 29, 30, 32
Series 11 20, 21, 28 Series 18 22, 26, 27