Some plots from March 2007 tests related to bolometer PSF

Transcription

1 Some plots from March 2007 tests related to bolometer PSF D.Lutz May 3, Introduction Document number PICC-ME-TN-020 This is a collection of sparsely commented plots from a quick analysis of some of the PSF related data, taken on March 9/10 and subsequent days, for the purpose of discussion and daily briefings. The problem of a severely corrupted bolometer PSF in phase 2 of the FM ILT, that was the motivation for this plot collection, has later been traced to protective Kapton foils left on the attenuation filters on the cryostat filter slider ( G.Jakob April 23, 2007). This note is kept as a cursory documentation of some of the analysis steps taken. 2 Focus sequence, ad-hoc tests on March 9 This analysis is based on files FILT PhotFocus filb Aper1mm 02.tm to 14.tm which give the focus sequence from z=-12mm to 12mm in steps of 2mm, done with the 1mm diameter hole and the blue filter B. All beams at all chop throws appear elongated, the lower (large x) ones more, with a tendency to be double. This non-gaussianity and partially resolved double structure is not a bad pixel effect - it is seen in all 4 lower beams. The more pronounced double structure seen later in the red and green filters reinforces the conclusion that we are not seeing a well-behaved elongated gaussian (which might occur for example due to cutting of the beam in one dimension plus diffraction), but something more resembling a superposition of several (almost diffraction limited?) beams at different intensity ratios and offsets, depending on position on the array. Changing the cryostat filter slider from FLE filter to Quartz filter and later re-inserting the FLE still produced elongated beams. Switching off the chopper controller still left elongated beams - not an effect of high frequency chopper crosschop oscillations. Such oscillations would also not explain variations over the array. This second argument also applies to a putative XY stage mechanical jitter. Elongation is seen with different apertures, this as well as the PSF variation when placing the source at different places on the array rules out a problem with the hole shape. The following sequence of plots shows the results of fitting all 8 beams (2 chop positions, 2 nod positions, 2 chop throws) for each observation at a given z (focussing) coordinate of the XY stage. Fits werde made with an elongated 2D gaussian (plus constant background). Peaks are coded by same color in all figures. Most quantities show clear trends with the z position in the focus sequence, but the highly unsual beam shapes and the inconsistent results from minima for x/y widths and maxima for peak height prevented adoption of a new focus value. The value z=5 adopted in December 2006 was kept. 1

2 Figure 1: Example of chopped/nodded observation for focus. 1mm hole. Small chop throw part Figure 2: Example of chopped/nodded observation for focus. 1mm hole. Large chop throw part 2

3 Figure 3: Peak height of gaussian fit 3

4 Figure 4: Pixel x coordinate of peaks 4

5 Figure 5: Pixel x coordinate of peaks with mean subtracted 5

6 Figure 6: Pixel y coordinate of peaks 6

7 Figure 7: Pixel y coordinate of peaks with mean subtracted 7

8 Figure 8: X Gaussian sigma of peaks in pixels 8

9 Figure 9: Y Gaussian sigma of peaks in pixels 9

10 Figure 10: Tilt angle of elongated 2D Gaussian fit 10

11 3 Rasters Several 17x31 1.5mm step rasters were done on March 9 to study the variation of the PSF over the array and for different chopper positions. All of them used the blue filter (B). Only the scan with 2mm gives good results on the red side as well. In detail, the files are: FILT PhotPSF 17x31simpleraster 1mm tm Done at chopper position CPR=664 (optical axis). Called Center 1mm in the titles of plots below FILT PhotPSF 17x31simpleraster 2mm tm Also at optical axis but with larger aperture. Called Center 2mm in plots. FILT PhotPSF 17x31simpleraster 1mm CPR tm Chopper set to CPR=-6700 such that the full bolometer stays just inside the open FOV but close to one of the calibration source. Called m6700 1mm below. FILT PhotPSF 17x31simpleraster 1mm CPR tm Chopper set to CPR=+8000 such that the full bolometer stays just inside the open FOV but close to one of the calibration sources. Called p8000 1mm below. The figures below again show peak height, x and y gaussian sigma and tilt of the elongated gaussian. To give an impression of the spatial distribution over the arrays,these are plotted as a function of the x and y pixel coordinates. Clear trends are seen with position on array (mostly with x) for both blue and red. Blue and red images seem inconsistent with the assumption that they show the same structure with just wavelength-dependent smoothing! This is seen from the sigma in x direction, which is similar in pixel coordinates despite the different pixel scales. Another example is the second case of the direct comparison images taken from both the 2mm 17*31 raster at specific XY positions and the Mar 10 slews (see below). The integrated flux (approximated by gaussian peak*xsig*ysig) does not change significantly with position on the array for the blue filter 1mm hole, central chopper position case where this was tested. Remaining differences are consistent with flatfield effects. This was not tried for the green/red channels with their yet more non-gaussian PSF. This integrated flux is about half of the one for the 31*61 raster on Dec 22 with same filter and hole diameter. ExtBB was 1000K on Dec 22 and Mar 9/10 according to the logs. According to Marc Sauvage/ Koryo Okumura, consitent bolometer settings were used durign the two periods, i.e. measured fluxes can be compared meaningfully. The results for Chopper positions 664 and -6700, and look similar to first order. There is no gross variation of the effects with chopper position, despite the fact that different regions of windows etc. are sampled outwards of the chopper. 11

12 Figure 11: Figure 12: 12

13 Figure 13: Figure 14: 13

14 Figure 15: 14

15 Figure 16: Figure 17: 15

16 Figure 18: Figure 19: 16

17 Figure 20: Figure 21: 17

18 Figure 22: Figure 23: 18

19 Figure 24: Figure 25: 19

20 Figure 26: Figure 27: 20

21 Figure 28: Figure 29: 21

22 Figure 30: Figure 31: 22

23 Figure 32: blue - position 1 - March 9. 2mm aperture 4 Direct comparison of beams in different filters for same XY stage position The plots in Fig. 32 to 37 show direct images in blue (Filter B) and red from the raster with 2mm aperture (FILT PhotPSF 17x31simpleraster 2mm tm). The numbers in the figures are from left to right x and y (stage), x and y pixel coordinate (from gaussfit, contiguous pixel coordinates despite matrix gaps ), and x and y sigma (from gaussfit). As noted above, the red double beam for position 2 is clearly not just the diffraction-smoothed version of the blue beam at the same XY stage position. No green filter data were taken on March 9. Figures 38 to 40 show data taken at position 2 during line scan exercises at 1mm/sec. (FILT ExtBB1mm slew1mmsec filb tm and FILT ExtBB1mm slew1mmsec fila tm). The green (filter A) observations are more clearly split into a double source than the blue ones. Note that the blue elongated source in fig. 38 was taken AFTER the green one but looks consistent (considering the different aperture) with the data taken the day before. The blue/green difference is thus not simply an artefact of a trend with time, but rather appears to be real. The blue structure looks more complex than a simple superosition of two peaks (additional shoulder?). Ignoring this for the moment, the rough length is in blue 5 small pixels and centrally peaked, in green a double peak separated by 6 small pixels, in red a double peak sperated by 3.5 big pixels. The Dec 22, 2006 tests suggested a scale of xy stage mm per blue pixel and xy stage mm per red pixel. With a simplified assumption of 1.25 pixels per matrix gap, the observed y (horizontal) pixels separations in the plots below are consistent with that scale and the respective xy stage y position difference of 21mm. The xy stage x difference of 7.5 between position 2 and 3 correspond to expected 11.6 blue and 6.2 red pixels (pixel gap in blue only!). This is pretty well consistent with the measured offsets for the centroids measured by the gaussian fits - see figures!. I.e. for the split red beam of position 3, none of the individual maxima is at the expected location - the centroid / the minimum between them is! This can be put on a more quantitative basis by comparing the actual XY stage positions with the ones reconstructed from the measured centroids on the array and application of the preliminary (FM 1 0) SubarrayArray and PhotArrayInstrument calibration files derived from the Dec 22 Tests. The next three figures show such comparison where a constant y offset of 1.3mm has been removed. Note that the deviations have been amplified by a factor 2 for clarity. In the blue filter, centroids agree with expected positions to typically half a mm or better with no clear trend. In the red, there seems to be a systematic difference of 1pix between zones at x 2 and x 8 23

24 Figure 33: red - position 1 - March 9. 2mm aperture Figure 34: blue - position 2 - March 9. 2mm aperture 24

25 Figure 35: red - position 2 - March 9. 2mm aperture Figure 36: blue - position 3 - March 9. 2mm aperture 25

26 Figure 37: red - position 3 - March 9. 2mm aperture Figure 38: blue - position 2 - March mm aperture 26

27 Figure 39: green - position 2- March 10. 1mm aperture Figure 40: red - position 2 - March 10. Low S/N because of 1mm aperture! 27

28 Figure 41: Comparison of collapsed beam profiles for the three channels. Blue and green ar from the March 10 1mm hole slews with 1mm/arcsec, Red is from the March 9 raster with 2mm hole. The relative registration of red vs. the short channels is just by pixel number, i.e. slightly incorrect. 28

29 Figure 42: Difference between actual and reconstructed peak position. Chopper on optical axis, 1mm hole, blue filter. NOTE: The conversion routine used works on integer pixel coordinates which will create scatter of order half a pixel in this and the next two plots. Figure 43: Difference between actual and reconstructed peak position. Chopper on optical axis, 2mm hole, blue filter 29

30 Figure 44: Difference between actual and reconstructed peak position. Chopper on optical axis, 2mm hole, red filter 30

31 Figure 45: Spatial correlation analysis of green filter data. The images show median correlation coefficients between timestreams of detectors in the same column but different rows of the blue array. Bottom left in each image: row 0 vs. row 0 (corresponding to top of array in QLA) Top right: row 31 vs row 31. Left image: Scan with 1mm hole taken on march 10. Right image: Laser line speckle FOV scan taken on March Laser speckle FOV tests and correlation analysis A way to have the PACS bolometer observe a structured field outside the instrument proper, that is not the XY source with its hole, was used during the laser tests: setting the OGSE test optics to have PACS look at the integrating sphere while injecting a strong laser line of appropriate wavelength. This produces a speckle pattern that can be observed in the style of a FOV scan using the PACS chopper. The relevant files are: /FM ILT data/ /filt FOVscan CSs IScold Laser tm /FM ILT data/ /filt FOVscan CSs IScold Laser tm /FM ILT data/ /filt FOVscan CSs IScold Laser tm with useful speckles for blue, green, red in that sequence. The green one has best contrast. Note that laser power varies a lot during scans. QLA and slightly more elaborate IDL movie scripts show a pattern of speckles, consistent with the PACS diffraction limit, passing across the FOV. There is no obvious indication of elongated or binary structure. Because of the complex pattern lacking single outstanding peaks, this Rorschach test may not be fully conclusive, though. The obvious solution of spatially autocorrelating the resulting images is doomed: Too few pixels, and the correlation signal searched for varies rapidly with array x coordinate. Instead, a different approach was tried motivated by the search for structures that are elongated/double in the short array dimension: For each pixel, extract a part of the FOV scan that covers most of the free FOV but cuts away the CSes. Subtract the mean of these timestreams. Correlate the timestream of a given pixel with the timestreams of all pixels at same y i.e. in a column along the short array axis Take the median of all correlation coefficients comparing pixels from two given rows along the long axis (same x) Produce an row n vs row m image showing these average correlation coefficients. Fig. 45 shows the result for a hole and external BB XY scan and a speckle FOV scan. The hole scan shows the expected correlation pattern, with the main diagonal=1 but secondary structure fanning out in the upper right part of the image, because of the double peaks in the lower part of the array. Note this signal is superposed on a clear correlation between all pixels, possibly due to variations in the laboratory background. This is suppressed in the scaling of the figure but defers a more quantitave comparison of correlation coefficients to the future. 31

32 Figure 46: Change of beam profile when moving the test cryostat filter slider. The position of the xy stage and hole size is the same in all cases. Green filter (A). FILT PhotRaster4x12 Aper1mm tm etc. The speckle scan correlation signal can be described as: (i) It does not show the same elongated secondary structures as the hole scan (the glass is 2/3 full). (ii) The brightest peaks off the main diagonal nevertheless are on part of that structure (the glass is 1/3 empty). The qualitative impression from the simple QLA-type view of the speckle scans - lack of similar double structure as in the hole scans - is thus confirmed. Caveats remain about the broad band vs. monochromatic nature of the signal in the two cases, given the mechanism of double peak production is currently not understood. The speckle correlation analysis would obviously benefit from more S/N. My gut impression is that for the 118micron line we are limited by the number of speckles over the FOV, not detector S/N, i.e. observing longer won t help but more FOV scans of different very strong laser lines could, both in view of S/N and possible dependence on the specific monochromatic wavelength. 6 Ad-hoc tests on March 30 Several additional tests were done on March 30. A shading experiment was done with the XY stage placed at a position where a clear double image was visible on the blue array (using green Filter A). An occulting strip just outside the cryostat window was slowly moved manually across the source in vertical and horizontal direction while the image was monitored in QLA. To the accuracy achievable, the sources seemed to disappear and reappear simultaneously, indicating an origin inside the occulting strip, i.e. at the cryostat entrance window or further inside. The relevant files are FILT PhotFLEfilter ManualActions 01.tm and FILT PhotFLEfilter ManualActions 02.tm A large number of tests was done with the cryostat filter slider at various nominal and offset positions for both the FLE and Quartz filters. Each observation took the form of a small raster allowing proper background subtraction out of the dataset itself, even for non-nominal slider positions. The relevant files are FILT PhotRaster4x12 Aper1mm tm to FILT PhotRaster4x12 Aper1mm tm. For identical XY stage positions, the beam shape changed drastically when moving the filter slider (Fig. 46). This 32

33 was observed for both FLE and Quartz. This strongly pointed towards an origin of the problem on the slider and clearly outside PACS, as later confirmed by the identification of the left-over wiggly Kapton foils on both filters. 7 Document change record Version Date Initials Comment DL 10: DL 13: DL DL collapsed profiles added DL position added, total flux, speckle correlation DL Added tests moving the filter slider, Doc number 33