NIKA3 Offline Processing Results (2011 October run)
NIKA3 Offline Processing Results (2011 October run)
- Private results
- Summary of Data reduction results
- FXD, 24 Oct 2011 - Photometric quality
- RZ, 25 Nov 2011 - Pointing & focus effects
- RZ, 30 Nov 2011 - Beam maps: FoV coordinates
- RZ, 30 Nov 2011 - Beam maps: Flat Field
- RZ, 30 Nov 2011 - Beam maps: Variable beam size across the FoV
- RZ, 26 Apr 2012 - Response (fluxes) of the pixels
- RZ, 30 Nov 2011 - Beam maps: Average beam sizes
- RZ, 1 Dec 2011 - The ''plateau''
- RZ, 5 Dec 2011 - Instabilities, nonlinearities ...
- RZ, 9 Feb 2012 - Different apparent synchronization of 1mm and 2mm data
- RZ, 10 Feb 2012 - Comparison of R and PF signals
- SL, 07 Nov 2011 - Expected field distortion
- FXD, 21 Nov 2011 - Technical note on kid frequencies
- AM,MC 24 Nov 2011 Spectral transmissions
- SL, 30 Nov 2011 - Calibration sources with PdB flux references
- MC RF calculations
- FXD 16 Dec 2011 Happy new year to the NIKA team : DR21 and DR21(OH) Preliminary data reduction (Sky noise decorrelation and then High-pass filtering by one subscan length)
- Nico 21 Dec 2011: Double resonnance separation + focal plane characterization
- FXD 24 feb 2012 OnOffJitter Animation
- FXD, JMP 6 Mar 2012 Atmospheric Opacity calibration
List of offline processing tasks
Summary of Data reduction results
FXD, 24 Oct 2011 - Photometric quality
Here are the maps obtained on one OTF scan with IRC 10216. Sky noise decorrelation is used.
The flux per kid is shown here using the calibration on Mars Scan #55. This is an indication that the new camera with its dIdQ modulation scheme provides a big photometric improvement with respect to 2010 campaign. The average flux is compatible with the range of fluxes measured on this (variable) source.
RZ, 25 Nov 2011 - Pointing & focus effects
To derive the source positions I used the pointing maps available in the IRAM Multi Beam FITS (IMBFits) format. However, currently over 20% of the pointings are missing. The statistics shown below is therefore represantative but by far not complete.
Source positions in the eq offsets. Positions of planets are not shown. In red are shown positions of sources used for the pointing model, in black after the installation of the pointing model. The pointing run was apparently performed with Nasmyth offsets of EMIR ! The pointing offsets after the installation of the pointing model are still very large, even 3-5 times larger than the normal values.
As above in the hor system; positions of planets are also shown.
Effect of defocusing: the source moves by a couple of arcsec in azim & elev. Note the larger scatter of the elev-positions.
Both, the pointings and foci were done by mapping. However, to correctly process maps the correct beam parameters, i.e. the pixel positions and their flat field are mandatory. The effective FoV geometry, the flat field and their errors will come next.
Origin of the large pointing errors
All the errors and effects shown above and below (FoV geometry): scatter of the pixel positions, effect of the focus, imperfect pointing model a.s.o., cannot explain the very large (>10arcsec) pointing offsets. There must be still another problem I could not identify.
Update Feb 9, 22; May 22:
To search for the source of the large pointing errors (and bimodal distribution in elev) I analysed also the pointings obtained with the heterodyne receivers. The poitings during the campaign are of mediocre quality but indicated some strange arcs in the plane (azCorrection,elCorrection). This finding is further confirmed by the (much clearer) examples below.
The pointing behaviour of the telescope may change rapidly, as show the examples below. This also happened during the NIKA run and explains the bimodal distribution of the pointing corrections.
Further information gives the focus behaviour of the telescope (as the source position depends also on the focus, see above): the focus values show an unusual bimodal distribution around the mean value, which (in the example below) never appears as the true focus value. The difference between the mean focus values in the two branches of the distribution is >0.7mm !
Important to note: such a pointing and focus behaviour of the telescope is new !
RZ, 30 Nov 2011 - Beam maps: FoV coordinates
To determine the pixel positions, the effective beam shapes and the flat field I used 9 Mars maps observed from Oct 17 to 22. Maps observed later on are of too bad quality. I could use sources weaker than Mars (Uranus, Neptune, ...) with sky noise filtering to determine the pixel positions & beams, but without the line-of-sight tau-corrections could not calculate the flat field. The pixels which show a clear cross-talk within the main beam were excluded.
The figure above left (scatter of pixel positions) shows the derived pixel positions in each of the maps. Different markers and colors correspond to different maps. The scatter of the positions is larger in the centre and in the upper(left) part of NIKA, i.e this cannot be an atmospheric effect (anomalous refraction). For the few examples I checked, the changed positions were due to a cross-talk. The resulting average pixel positions are shown above right (fig. average FoV 1mm).
Centre of Nasmyth rotation
One very important point cannot be properly checked using the available data: the centre of the Nasmyth rotation. This centre is defined by the rotation axis of the mirror M3, but its determination depends also on the pointing behaviour. The ultimate proof would be if all the usable pixels would always show the source at one position, i.e. independent of the elevation and azimuth. For all coordinate transformations I had to assume this centre is close to pixel #21 (22 in my plots), as given by FXD.
RZ, 30 Nov 2011 - Beam maps: Flat Field
Update Dec 7: flat field for the beams in the near field.
Two flat fields are necessary for the data processing: for the far field (i.e. main beam = astronomical sources, MBFF) and for the near field (sky noise, i.e. the correlated signal, CSFF). The correct main beam flat should use integrals over the main beams. However, because the reason(s) for the beam broadening (see below) are not clear to me, I used just the peak fluxes, i.e. a flat field (PFFF) valid only for on-offs observed with not filled receiver arrays like MAMBO, LABOCA or SCUBA. Currently I also did not use any extinction corrections. The peak flux flat field shown below is therefore just indicative.
To calculate the flat for the beams in the near field in higher than first order approximation, just the slopes of the correlation distributions in data of weak sources are sufficient, i.e. when the assumption noise=data can be made (see "5 Dec 2011 - Instabilities, changed flat field, ..." below). Note that the calculation of the CSFF does not need any beam sizes.
Whereas each of the flat fields depends on the pixel response to the incoming signal, the ratio of the flat fields does not. In the no-problem case the ratio of the flat fields shows simply the change of the coupling of the receiver to the telescope. It may vary, e.g. due to a distortion of the FoV, but this variation cannot explain the observed ratios of >200%.
RZ, 9 Feb 2012 - Flat Field variations = calibration accuracy = stability of the system
The two figures below compare the beam parameters derived from the Mars map #54 and the Uranus map #149 (both on Oct 18) with the average beam parameters derived from 9 Mars maps. The upper inlet in both figures shows the difference in beam positions of each pixel (RPO = Receiver Pixel Offset). The differences of positions are within the possible accuracy. The flat field (FF) however shows very large discrepancy. Whereas the FF derived from the Mars map agrees within +-20% with the average FF, the FF derived from the Uranus map differs at the edges by a factor of up to 4.
Further, for sources weaker than roughly Uranus the FF depends also on weather conditions, i.e. for the same source the derived FF is different for different sky loads. Below another comparison of the FF derived from an Uranus map observed during bad weather (#239 on Oct 20) with the same average FF as in the figures above.
This demonstrates that the calibration per pixel varies by a factor of up to 4.
RZ, 30 Nov 2011 - Beam maps: Variable beam size across the FoV
Strange beam broadening is visible at 1mm (roughly a factor of 2) and much less at 2mm (below 50%). Two examples for Mars and Neptune are shown below.
Pixel #1 is not reliable. Note also the larger and elliptical beams of pixels #29 and 40 for Neptune. This is an effect of the cross-talk.
Towards the centre the beam profiles deviate stronger from a Gaussian than at the edges. This is most easily visible in a comparison of the maxima in the maps with the peaks of a Gaussian fitted to point-like sources.
The beam broadening is visible also in I & Q (below) and at 2mm (not shown).
For comparison, below are shown the beam sizes @1mm during run 2.
SL, 23 May 2012 - For comparison: FoV during run #2
Just to clarify black on white on this page what I showed several months ago:
The grid distortion foreseen by the optics simulation is ~15% at max in the corners (see the document "Estimation of the NIKA v2 optics grid distortion ...:, version 20/05/2011, section: SL, 07 Nov 2011 - Expected field distortion). What we see in the far field geometry plots above is compatible with these simulation, i.e. with run 2 (see the FoV geometry plots below). This distortion effect is mainly due to the projection of a curved image plane onto a flat detector array; depending on the focus we expect also up to 15% beam broadening difference between the center and the edge pixels. However, what we see only in run 3 but not in run 2 nor in the simulation is up to 100% beam broadening. So again: grid distortion is explained by the optics architecture, but the excess beam broadening is not.
Here are the geometry plots from run 2 produced by Robert and presented in July 2011 (see http://www.iram.fr/~leclercq/NIKA/NikaPage.html for more information on the run 2 analysis by Robert). The distortion at 1mm is not as evident as in run 3 and as at 2mm only because the 1mm FoV in run 2 was considerable smaller.
RZ, 9 Dec 2011 - Beam broadening as function of pixel number
The beam broadening shows also a quite clear dependence on the pixel number. Below are shown the average FWHM of two first Mars maps @1mm.
We can therefore ask: is the beam broadening a function of the pixel number, the FoV distribution therefore just a result of the choice of the pixel position, or vice versa ?
RZ, 26 Apr 2012 - Response (fluxes) of the pixels
The two figures below show the peak and the integrated flux of all usable pixels in 9 Mars maps @1mm as function of the pixel number (1-116). Whereas the peak fluxes do not show any slope with the pixel number the integrated fluxes do show the same slope as their FWHM (see above Update Dec 9: beam broadening). This is not what one expect if the beam broadening was an optical effect: in this case the incoming power should be preserved, i.e. the peak fluxes should show a variation corresponding to the variation of the beam widths. Also important discussing effects of the optics: the deviations of the beam profiles from Gaussians (see above the figure 'beam deviation from Gaussian @1mm').
In the coordinates of the FoV this slope shows the same picture as the beam broadening and the deviations of the beam profiles from Gaussians. Note further that this effect is not centered within the FoV.
Left: flat-field derived using the peak fluxes, centre: flat-field derived using the total fluxes, right: the ratio of both.
RZ, 30 Nov 2011 - Beam maps: Average beam sizes
For pointing sources weaker than roughly Neptune the errors of individual beams are too large to allow the check of the beam change across the FoV. Only the average beam shape can be fitted with sufficient accuracy. The figure below shows the major (triangles) and minor (circles) FWHM for all pointing sources available in the IMBFits format. There is a clear change of the observed beam size after the pointing run, as after the pointing run the sources appeared at different region of NIKA (recall: the pointing run was performed with the Nasmyth offsets for EMIR). This change is therefore an effect of the variation of the FWHM across the FoV. Within the errors there is no dependence of the FWHM on the source flux.
Example beam maps obtained on Mars (~160Jy), Uranus (~33.5Jy) and Neptune (~12.6Jy) with NIKA1mm are show below. The contour levels are at 0.4, 0.8, 1.6 ... 25.6 and 50%, for Mars also at 0.1 and 0.2% (in yellow). The averaged FWHM is in all cases ~15arcsec.
RZ, 1 Dec 2011 - The ''plateau''
Update Dec 7: figures @2mm
Details of the plateau. The intensity scale is Jy/beam. The contour @1mm is at 9.5Jy/beam, @2mm 6.7y/beam.
Intensity profiles @1mm per row (i.e. along the azimuth) and column (i.e. along the elevation) of pixels #22 (in red), #80 (in green), #102 (in cyan) and #105 (in blue).
Intensity profiles @2mm per row (i.e. along the azimuth) and column (i.e. along the elevation) of pixels #11 (in red), #104 (in green), #105 (in cyan) and #91 (in blue).
The Mars map 20111019s52 in I and Q.
RZ, 3 May 2012 - The ''plateau'' of extended sources: the case of Crab
The signal of extended sources is dominated by the plateau. The figure below shows the 2mm data of all 116 pixels. Note the difference to the maps of point-like sources, e.g. above the map of Mars, scan 52 on Oct 19.
The processed maps of Crab @1 and 2mm in equatorial coordinates (only reliable pixels used) are shown below (left and centre, figures Crab@2mm & Crab@1mm). These maps do not show any clear structure. This is because the plateau shows up in the receiver coordinates, its signal is therefore smeared across an area corresponding to the extent of NIKA creating a kind of a halo (compare with the figure Mars beamMap 1mm above).
After filtering out the correlated signal some structure appears (above right, figure Crab@1mm filtered), i.e. a fraction of the plateau was successfully filtered out. It is still quite fuzzy, as the signal of the plateau does not correlate to 100%.
RZ, 9 May 2012 - The variation of the ''plateau'' signal btw the pixels
The plateau in the off-resonance pixels @2mm is stronger than in the working pixels. The two figures below show maps of Mars and Crab. From the signal of each pixel an estimate of the plateau and linear baselines were subtracted. In case of Mars the estimate of the plateau was the signal of pixel #2 scaled down to 60%, in case of Crab the average signal of pixels #2, 24, 45, 54 and 86. Clearly visible that the remaining rest-plateau in the signal of working pixels changes from pixel-to-pixel and it shows the same drift with pixel # (i.e. change across the FoV) as the beam broadening. The plateau of the off-resonance pixels correlates much better.
RZ, 5 Dec 2011 - Instabilities, nonlinearities ...
The correlation plots below show few examples of instabilities, changed flat field and nonlinearities of the 1mm signal
RZ, 9 Feb 2012 - Different apparent synchronization of 1mm and 2mm data
The 1mm data show a different apparent synchronization than the 2mm data. This is visible as a zig-zag of the source position from one subscan (i.e. a row in the map) to another. The two figures below demonstrate this. At 1mm the pixel #22, at 2mm the pixel #11 is shown. To remove the zig-zag visible in the 1mm data the time of the data has to be shifted by roughly -0.084sec, i.e. two samples. The 2mm data do not show this problem.
RZ, update Mar 28
The different apparent synchronization of the 1mm and 2mm data was caused by wrong times in the NIKA files and is visible in the R signals alone. The new files with corrected times do not show this discrepancy in the PF signals, i.e. both the 1mm and the 2mm show a zig-zag corresponding to roughly -0.084sec.
RZ, 10 Feb 2012 - Comparison of R and PF signals
The calculated signals R and PF (which are supposed to represent the same incoming signal) differ considerably. Two examples - for a pixel in the centre (#10) and at the edge (#90) - @2mm are shown below. The ratio R/PF is not constant (i.e. the source shape is different in R & PF), it depends on the position within the FoV and is further band dependent (not shown).
The plots for the first NIKA FITS file with the R and PF signals provided by FXD were created with the FITS Viewer fv.
In the centre the linearity of R(PF) breaks down at roughly 0.5 of the max signal, i.e. sources will appear broader in R than in PF. At the edge this effect is not visible. Qualitatively (but not quantitatively) a similar behaviour of R(PF) is visible also in the 1mm data (not shown).
This comparison shows there should be a beam broadening in the 2mm R data, but this is contrary to what is seen in the beam maps in R.
Obviously, in this example the R and the PF are simple inconsistent !
See below also "MC 30 Nov 2011 RF calculation on Mars", ....
RZ, 22 Mar 2012 - Variation of the beam size and the flat field across the FoV
The beam broadening and the variation of the Flat Field FF are clearly software effects. However, the question arises, is this *also* or *only* a software effect. I do not see how this could be answered for sure using just the astro data. Below I show the FWHM of the beams, the ratio of the beam angles and of the Flat Fields as derived using the R and PF signals for the first map with the PF signals provided by FXD; and at the bottom a standard comparison plot for another map.
Comparison of receiver pixel parameters (RPPs) derived from R and PF. "IN" values were derived using PF, the values stored in the file were derived using R. FF = Flat Field, RPO = Receiver Pixel Offsets, beamAng = beam angle.
RZ, 28 Mar 2012 - Nonlinearity of PF(R): maps
The two figures below show the signal PF as function of R (PF(R), "phase 1" = R, "phase 2" = PF) for two consecutive scans on Oct 18. Whereas for scan 54 PF(R) is almost linear (R is slightly stronger than PF), for scan 55 PF(R) is highly nonlinear for most of the pixels. Note also the changed relative response of some pixels, i.e. the change of the Flat Field for these two maps.
Further, the difference PF-R is time variable. The two figures below show an on-on (more precisely position switch) observation. In the first figure is shown in red the R, in black the PF signal as a function of time. In the second figure is shown the difference PF-R. Note also the negative signals given by the pixels in the neighbourhood of #63.
RZ, 2 Apr 2012 - Nonlinearity of PF(R): on-ons
The figures below show the results of two on-on (more precisely position switch) observations, i.e. images in eq system. The numbers give the signal of Mars seen by the individual pixels. The negative signal (see also above) is an effect of the algorithm used to calculate the PF. Using the same Jy/counts calibration factor the on-ons give a factor of two lower peak flux for Mars than the maps. Note further the very bad shape of Mars, i.e. the "beam shape", slightly different in the positive and the negative signal. Such a bad beam shape is not visible in maps observed prior to these on-offs with the same focus setting. Both, the lower flux and the bad beam shape cannot be explained neither by focus effects nor by the rotation of the pixels across Mars. The reason for both problems is most probably also a software effect, i.e. an effect of calculation of the PF signal.
SL, 07 Nov 2011 - Expected field distortion
grid distortion (updated Dec 1: 2 coefficients of the grid equation were switched)
FXD, 21 Nov 2011 - Technical note on kid frequencies
FXD, 24 Nov 2011 updated v2.1 (v2.1 Modulation explanation added, v2 Fits format exchange, v1 Basic frequencies) FXD, 24 Jan 2012 updated v2.2 (v2.2 Polynomial fit explained, fits header inputs from Albrecht)
Kid frequencies, fits format, polynomial fit
AM,MC 24 Nov 2011 Spectral transmissions
SL, 30 Nov 2011 - Calibration sources with PdB flux references
Calibrators fluxes for Oct 2010 (run #2) and Oct 2011 (run #3), fitted on PdB database (updated Dec 14, 2011)
MC RF calculations
MC 30 Nov 2011 RF calculation on Mars
Blue line is the RFdIdQ as calculated in the standard way. Red, green and black are the results obtained using a 2-d polinomial fit of grade 2, 3 and 4 respectively.
MC 30 Nov 2011 dIdQ modulation on strong sources
The blue line is the trajectory on the IQ plane when observing mars (approx 7kHz signal). The red line is the observed DeltaI, DeltaQ, averaged on 100 points (please do not consider it. It doesn't have a lot of sense. I'll update.). Interesting are the green arrows, whiwh are the dI dQ obtained by the LO modulation: near the peak of the source, where we get far from the resonances, they get farther from the tangent of the IQ trajectory... to be investigated!
MC 07 Dec 2011 More comments on RF calculations
The RF variation calculated by CAMADIA can differ from that obtained by a 2d fit used to find a relation F(I, Q). This in any case generally happens only for very strong signals (Mars, essentially...) and only for the pixels that happen to be already slightly distorted or in any case biased really to their limit. The images show 2 different pixels during a Mars observation. It can be seen that the one where there is a disagreement between RFdIdQ as evaluated by CAMADIA (blue line) and the one by the fit (red line) is the pixel with a distorted resonance (kid 5).
FXD 16 Dec 2011 Happy new year to the NIKA team : DR21 and DR21(OH) Preliminary data reduction (Sky noise decorrelation and then High-pass filtering by one subscan length)
Nico 21 Dec 2011: Double resonnance separation + focal plane characterization
Here is a brief note on detectors decoupling and focal plane parameters reconstruction. Results are summarized in .fits files, one per configuration. I've only run the pipeline on two nights of observation for now but will update the list shortly. Comments on this preliminary work very much welcome. Merry christmas to you all !
FP_quicklook_111212.pdf d_2011_10_18_00h29m35_Uranus_333_FP_AB.fits d_2011_10_19_23h14m23_Uranus_220_FP_AB.fits
Nico 08 Feb 2012: update on focal plane characterization and distortions
Looking at 4 different OTF_Geometry scans throughout the calibration campaign, I compare focal plane reconstructions : they are very stable. I also look at the beam distortion pattern and show that it correlates very well to optical distortions. Comments welcome. Focal plane geometry fits files are available on the Neel Institute ftp website under the directory /Archeops/FPG.
FXD 24 feb 2012 OnOffJitter Animation
Animated data reduction of OnOff sequences
FXD, JMP 6 Mar 2012 Atmospheric Opacity calibration
Analysis of frequency variations as a tool for the total power