glitches in pacs photometer
DESCRIPTION
Glitches in PACS photometer. BA & BA. What are glitches?. Anomalous signal readouts caused by: Charged particle hits to bolometer pixel. “standard” glitches Charged particle hits to electronics. “dropouts” Erratic behavior in the readout electronics. “Standard” Glitches. - PowerPoint PPT PresentationTRANSCRIPT
26 January 2010 NHSC DP workshop 1
Glitches in PACS photometer
Glitches in PACS photometer
BA & BA
26 January 2010 NHSC DP workshop 2
What are glitches?What are glitches?
• Anomalous signal readouts caused by:– Charged particle hits to bolometer pixel.
• “standard” glitches
– Charged particle hits to electronics.• “dropouts”
– Erratic behavior in the readout electronics.
26 January 2010 NHSC DP workshop 3
“Standard” Glitches“Standard” Glitches
• Thanks to the small cross section of the bolometer, the cosmic ray rate in very low.– <10 hits / array / readout [<0.5%]
• Most are short-lived (1-2 readouts)• Can be both positive or negative (wall)
• Mitigation– MMT deglitcher module
• scan-map
– Sigma clipping • point-source
26 January 2010 NHSC DP workshop 4
The MMT deglitcherThe MMT deglitcher
• Based on ISOCAM wavelet source finder by Stark (1998, PASP, 110, 193)
• Segregates the various temporal scales in the time streams in wavelet space.
• A range of wavelet scales coupled with significance threshold define “glitch” in the data.
26 January 2010 NHSC DP workshop 5
Relevant MMT deglitcher parameters
Relevant MMT deglitcher parameters
• scales – Number of wavelet scales. This should reflect
the maximum number of affected pixels.
• nsigma – The significance threshold in wavelet space.
Values above nsigma*RMS are considered glitches. Where RMS is measured per wavelet scale.
• Other parameters determine the properties of how MMT scales are calculated.
26 January 2010 NHSC DP workshop 6
MMT usageMMT usage
• The default values of the parameters do not provide a good default for all scientific cases.– Usually requires “tuning” for optimal results.
• The 'scales' parameter is the most relevant for a good performance of the method. – scales=3 provides the best result for scan-maps at 20”/s
scan speed– Scale=1 at high speed (parallel mode)
• By default MMT deglitchers replace glitch with an interpolated value in dp.pacs 2.0, feature removed in dp.pacs 3.0
– This feature can be turned to ‘mask’ only, • maskOnly=True
26 January 2010 NHSC DP workshop 7
An example of MMT deglitching at work. Blue points show glitches in both signal space and corresponding wavelet space.
Bolometer behaves fortunately much better in space than in these radiation tests !
26 January 2010 NHSC DP workshop 8
Sigma ClippingSigma Clipping
• MMT deglitching is not a good choice for chopped observation. – chopper plateaus are incorrectly identified as
glitches. • For point source (chopped) AOT observations.• Simple sigma clipping per chopper plateau per
dither position.
26 January 2010 NHSC DP workshop 9
Signal stepsSignal steps
• Discontinuous signal “jumps”• Much rarer• Likely, charged particle hit causes
electronic offsets to “jump”– Even rarer: a full matrix upset
• Mitigation:– With high pass filtering, only a few
readouts near the discontinuity are affected.
– Show up as anomalies in final maps.– Mask manually.
DC offsets casesDC offsets cases
Line dropout Line up
26 January 2010 NHSC DP workshop 10
26 January 2010 NHSC DP workshop 11
Erratic readoutsErratic readouts
• Signal is row 11 sometimes appears abnormal for a few readouts
• Signal fluctuates between two values.
• Causes not understood.
• Mitigation:– Examine row 11– Examine time-streams with mask
viewer– Mask manually, as needed.
2nd level deglitching2nd level deglitching
• HIPE also provides an alternative sigma-clip based de-glitcher.– it finds, for each pixel of the map, the list of all
cube pixels that contribute to it and performs sigma clipping on this list to identify outliers.
• However,– Current implementation is extremely memory
intensive and slow• The data and projected cubes must be kept in
memory
2nd level deglitching2nd level deglitching
• Do a sky projection without creating a map– photProject(frames, deglitch = True)
• Perform sigma clipping on projected values– index = photProject.getValue("index")– from herschel.pacs.spg import IIndLevelDeglitchTask– from herschel.ia.numeric.toolbox.basic import Sigclip– deg = IIndLevelDeglitchTask()– s = Sigclip(10,10, outliers = "both")– img = deg(index, frames, mask=True, map=True,
algo=s)