Calibration of Photometry from WFCAM

Author: Simon Hodgkin
Document Number: VDF-TRE-IOA-00008-0010
Version Number: Draft 0.2
Date: 26th October 2005

Contents

  1. INTRODUCTION
  2. UKIDSS CALIBRATION GOALS
  3. MANUAL CALIBRATION
    1. THROUGHPUT
    2. RELATIVE DETECTOR SENSITIVITIES
    3. EXTINCTION AND NIGHT-TO-NIGHT STABILITY
    4. NIGHTLY CALIBRATION
    5. WFCAM VS UFTI
    6. SPATIAL SYSTEMATICS
  4. PIPELINE CALIBRATION
    1. TECHNIQUE
    2. CALIBRATION OF UKIRT FAINT STANDARDS
    3. CALIBRATION OF T DWARFS
    4. EXTINCTION
  5. 5 TOWARDS A SYSTEM OF WFCAM SECONDARY STANDARDS
  6. 6 SUMMARY
  7. Appendix A LIST OF STANDARDS
  8. Appendix B SYNTHETIC PHOTOMETRY

1. Introduction

I have analysed observations of standard stars and normal target fields from semester 05A.
Figure 1: A histogram to show the number of standards observed on each night in semester 05A. The main spike represents a couple of nights when large numbers of standards were targetted. It is only towards the end of the semester that observing settled into a pattern of observing around 10 standard fields per night (i.e. roughly hourly).

2. UKIDSS Calibration Goals

Requirement is photometric calibration to 2%, with a goal of 1%.

3. Manual Calibration

3.1 Throughput

The total throughput of the system is relatively easy to measure (but much harder to quantify where in the system the actual losses are occurring). I have assumed the following: Thus the table below gives the estimated number of photons that would be incident on the primary mirror, assuming no atmosphere, multiplied by the relative transmission of the filter in each band. These are converted to detector counts using an average gain of 5.1. This would then give the zeropoint of the system in each filter if there were no losses due to the atmosphere, telescope and the instrument. Comparison with the measured zeropoints then give us the throughput for WFCAM+UKIRT+atmosphere.

Filter Photons
(mag=0)		 Counts
(mag=0, gain=5.1)		 ZP
(100% throughput)		 ZP
(Measured)		 Implied
Throughput
Z 3.8e10 7.4e9 24.80 22.98 19%
Y 3.0e10 5.9e9 24.55 22.81 20%
J 2.9e10 5.6e9 24.50 23.00 25%
H 3.0e10 5.8e9 24.53 23.24 30%
K 1.6e10 3.1e9 23.83 22.55 31%
Table 1: Estimate of WFCAM throughput

Still to do: Compare to predicted throughput (Hewett et al. 2005)

3.2 Relative detector sensitivities


Figure 1:The top panel shows the detector-to-detector variation in sensitivity (really it's gain*QE), measured with respect to the mean of all 4 detectors, from twilight flatfield counts (ZYJHK). Chip 1 is the most sensitive, while chip 4 is the least. Chip 3 shows the largest colour dependent scatter with a 10% change in relative sensitivity between Z and K (decreasing to the red). Chip 4 shows an increase to the red of about 8%. The bottom panel shows the detector-to-detector gain variation measured post-pipeline processing, i.e. the residuals after gain correction (using the twilight flats). It is measured from the standard stars (their zeropoints). It illustrates that the gain correction calibrates each chip to within 2% of the mean, across all filters.

The four detectors are not uniformly sensitive. Figure 1 shows the variation in sensitivity between the detectors as measured in the twilight flatfields. The assumption is that the twilight sky is flat. The detector-to-detector sensitivity is measured from the average counts on each detector. The ZPs derived for the night of 20050408 for each chip from the UKIRT faint standards are summarized in Table 2.

Filter chip 1 chip 2 chip 3 chip 4
ZP err ZP err ZP err ZP err
Z 22.982 0.056 22.962 0.059 22.989 0.050 23.002 0.055
Y 22.810 0.018 22.803 0.016 22.821 0.015 22.810 0.020
J 23.003 0.021 22.998 0.020 23.005 0.026 23.005 0.021
H 23.241 0.017 23.238 0.016 23.226 0.019 23.246 0.014
K 22.557 0.016 22.547 0.021 22.535 0.021 22.557 0.016
Table 2: per-chip ZPs for 20050408

3.3 Extinction and night-to-night stability

Two nights have sufficient standards to enable a good stab at measuring the extinction. The spread in airmass is not quite ideal. There is coverage up to X=1.5 on April 8th and X=1.7 on April 18th. All 4 detectors are combined to constrain a fit to

mMKO – Minst = ZP – kX

with the following results:


Figure 2: Extinction diagrmas for UT20050408 and UT20050418. All four chips have been combined onto one figure. A better approach would be to fit the data from all four chips simultanously, allowing ZP to vary but fixing EXT

20050408 20050418
Filter ZP ± k ± ZP k
J 23.003 0.015 0.036 0.012 23.004 0.055
H 23.237 0.024 0.031 0.021 23.236 0.033
K 22.548 0.024 0.050 0.020 22.535 0.024
Table 3: ZPs and extincxtions derived on two nights

These values are at the lower end of site testing results. Note the very good agreement between the ZPs derived for these two nights.


Figure 3: Figure taken from Krisciunas et al. 1987. Our measurements overlaid

3.4 Nightly calibration

In this section I plan to discuss the calibration we hope to get on a typical observing night. I will do this by analysing a couple of nights where we have observed hourly standards.

3.5 WFCAM vs UFTI

Figure 4 illustrates that there is essentially no colour term between the UFTI based UKIRT faint standard magnitudes, and the same objects measured through the WFCAM MKO filters and detector, as expected.

Figure 4: A plot of the offset between catalogue magnitude and (extinction corrected) instrumental magnitude for UKIRT faint standards on UT20050408 (JHK)

3.6 Spatial Systematics

3.6.1 From 2MASS

A simple analysis compares the WFCAM measured photometry against the 2MASS catalogue as a function of position. WFCAM sources are matched against 2MASS. The 2MASS photometry is converted to the WFCAM system using the colour equations listed elsewhere in this document. Only objects brighter than J=15.5,H=15,K=15 are used, they are also required not to be saturated in WFCAM. The offsets are combined for a whole night and averaged over a 8x8 grid in WFCAM pixel space. Each of these jumbo-pixels contains of order 100 stars. The rms scatter in each jumbo-pixel is about 0.05-0.1 mags and the standard error on the mean is about 0.005-0.01 mag. The figures below show the results for the analysis in the J, H and K Bands.


Figure 5: Spatial systematics in WFCAM in the J-band. For each chip this is the spatially-dependent difference between 2MASS and WFCAM magnitudes, binned into a 8x8 grid to improve statistics. The analysis is the stack of a whole night of observations (20050408). The scale is in magnitudes, and is WFCAM-2MASS, i.e. white=positive=the objects are fainter in WFCAM than 2MASS. The colorbars indicate the range of delta magnitudes for each chip.


Figure 6: Spatial systematics in WFCAM in the H-band.


Figure 7: Spatial systematics in WFCAM in the K-band.

It's hard to see from this whether anything very significant is going on. An alternative way of investigating the data is to look at WFCAM-2MASS magnitudes as a function of distance from the rotator centre. the following 6 plots illustrate this analysis on 2 nights.

Figure 8: Each of the 6 plots is split into 6 panels. The top 4 panels (labelled 1-4) show the per-chip variation in WFCAM-2MASS magnitude as a function of offaxis angle. The bottom two panels are the median over all 4-chips. The second of these removes the expected radial distortion contribution to the photometry (see http://www.ast.cam.ac.uk/~wfcam/docs/reports/astrom/index.html).

At the corners of the field WFCAM has photometry systematics due to this effect of at most ~1.4% relative to the centre. This is clearly seen in the above figures. There are also some low level chip-to-chip variations - but any residual systematics are at about the 1% level and will need further analysis to really tie down.

3.6.2 From mesostep experiment

We observed a 5x5 grid in two standard fields, offsetting the detector by 1/5 of a detector width each time. We repeated this pattern in each of ZYJHK filters. I present here the analysis for the J filter in one field only. Unfortunately this field is a little sparse. It would be useful to repeat these observations in a very crowded region. Briefly, I stacked all 25 frames to make a master image, and from this generated a master catalogue. This is an input list used to drive the photometry. I then generate a light curve for every object in the master list. The RMS diagram for these lightcurves is shown below.
Figure 9: The rms diagram for the mesostepped Field-1 in the J-band, based on 25 images. Only objects with >=10 data points are shown in this diagram. Estimated noise contributions are shown from sky (dash-dot), photon counts (dash) and a systematic fudge factor of 3 millimags (dot)

I split each chip into an 8x8 grid. In each cell, I extract the object lightcurves. For each object (with more than 10 datapooints in the lightcurve), I then found the median flux of the object, and the difference between its measurement in this cell, and the median. This is repeated for all cells and all chips. The results are shown in the figure below, and again illustrate that there are no systematics at the >1% level. One could infer this directly from the RMS plot. For the brightest stars, nothing has an RMS>2%, and most stars with J<13.0 have RMS values below 1%. I plan to extend this analysis to the remaining filters, and the second field (which is rather less crowded than the first).

Figure 10: Mesostep analysis for Field-1, chips 1-4 (clockwise from bottom right) in the J-band. the XY plots are slices along row 4 and column 4.

4 Pipeline Calibration

4.1 Technique

The pipeline photometric calibration is currently based on 2MASS, via colour equations to convert to the WFCAM instrumental system. 2MASS solutions for every catalogued frame are generated and allow monitoring of effective ZPs at the ~few % level. The 2MASS-WFCAM colour equations are generated from a large number of catalogued 2MASS stars observed on the night of UT20050418. The solutions are average ones, and take no account of, for example, luminosity class. This has still been done via visual inspection. 

ZWFCAM = J2MASS  + 0.95(J-H)2MASS		
YWFCAM = J2MASS  + 0.50(J-H)2MASS
JWFCAM = J2MASS - 0.075(J-H)2MASS		(dwarfs: -0.067 giants: -0.003)
HWFCAM = H2MASS + 0.040(J-H)2MASS		(dwarfs: +0.080 giants: +0.065)
KWFCAM = K2MASS - 0.015(J-K)2MASS		(dwarfs: -0.023 giants: +0.032)

The bracketed values are from Steve Warren's analysis of synthetic colours generated from template spectra (see Hewett et al. 2005) and Appendix B, which is a copy of an email circulated by Steve Warren. The CASU derived system zero-points (corrected to unit airmass) for the main passbands are shown in the next table. Note also that in deriving these zero-points, all detectors have been gain-corrected to the average detector system. The CASU pipeline assumes a default extinction of 0.05 mags/airmass. Thus frame-to-frame variations in zeropoint include real variations in extinction.

Zero Point Z Y J H K
WFCAM counts/s 22.8 22.7 23.0 23.3 22.6
WFCAM e-/s 24.4 24.3 24.6 24.9 24.2
UFTI e-/s - - 24.5 24.7 24.2
Table 4: WFCAM Zeropoints. The conversion to electrons assumes an average gain of 5.1.

4.2 Calibration of UKIRT faint standards

For the nights of 20050408 and 20050418 the difference between the published UFTI photometry for the UKIRT faint standards in the MKO system has been compared to the pipeline calibrated photometry. The idea is to see how good a job this first-pass photometric calibration is doing (Figure 11). The average offsets (and associated standard deviations) are given in table 5. The two most significant findings are:
  1. There is a residual offset in the H-band measurements for the standards at the 4% level. A closer look at the H_WFCAM to H_2MASS conversion (http://www.astro.caltech.edu/~jmc/2mass/v3/transformations/) shows that this effect was already apparent in the 2MASS colour equations (see also Carpenter 2001, but note there is no transformation to MKO, only to the old IRCAM3 system). Since writing this document, the CASU calibration has altered to include the 4% offset in H. Figures have not been updated yet to reflect this change
  2. The scatter in all filters is rather larger than one might hope for. At 2-4% it may well represent the limit to how accurately the WFCAM photometry can be calibrated with 2MASS. Residual spatial systematics and chip-to-chip variation may also be contributing to this scatter

Figure 11: The differences between pipeline calibrated and UFTI magnitudes for UKIRT faint standards measured on 8th April 2005. J-band in green, H-band in yellow and K-band in red. Horizontal lines show the mean offset between the CASU and UFTI photometry.

Figure 12: The differences between pipeline calibrated and UFTI magnitudes for UKIRT faint standards measured on 18th April 2005. J-band in green, H-band in yellow and K-band in red. Horizontal lines show the mean offset between the CASU and UFTI photometry.

20050408 20050418
FILTER WFCAM-MKO ± WFCAM-MKO ±
J -0.004 0.022 0.002 0.036
H 0.036 0.022 0.034 0.028
K 0.016 0.020 0.006 0.023
Table 5: UKIRT faint standards calibrated by the WFCAM pipeline using 2MASS; differences with published photometry.

Still to do: direct comparison between UFTI and 2MASS systems for UKIRT faint standards. Is there an inherent offset in the 2MASS photometry? Now done - see comment above. Add in this effect and update colour equations

4.3 Calibration of T dwarfs

Figure 13 shows the differences between UFTI and WFCAM photometry for a series of L and T dwarfs. These stars were observed over a series of nights. Table 13 summarises the numbers. As with the previous section, these stars were calibrated via the WFCAM pipeline using 2MASS stars measured simultaneously.

Figure 13: The differences between UFTI and WFCAM pipeline photometry for a series of L and T dwarfs measured as part of SV, as a function of colour (top) and J magnitude (bottom). J-band is green, H-band is yellow, K-band is red.

Principally, there may again be an offset at H, and perhaps at K (see comments above). However these results are dominated by the very large scatter in all filters. It's not clear why the scatter should be as large as 5-8%, given that these stars are not significantly fainter than the standards. However there are a couple of factors which may come into play:

  1. They are observed across a number of nights. It's important to exclude potentially non-photometric data from this analysis (not yet done).
  2. They have spectra that are significantly different from the standard stars.
  3. There is a trend for increasing scatter with increasing magnitude (do we have any estimate for the errors on the original photometry?)

FILTER WFCAM-MKO ±
J 0.014 0.049
H 0.075 0.078
K 0.048 0.061
Table 6: L and T dwarfs calibrated by the WFCAM pipeline using 2MASS; differences with UFTI photometry

4.4 Extinction

In this section I will attempt to measure the extinction from the 2MASS data to compare with my derivation from the UKIRT faint standards.

5 Towards a system of WFCAM secondary standards

This section will describe how the WFCAM secondary standards will be calibrated from the UKIRT faint standards and collected together to form a community resource. Ultimately a paper.

6 Summary

6.1 Remaining uncertainties

6.2 Proposed changes to observing strategy?

Appendix A. List of Standards

Appendix B. Synthetic Photometry


Hi Mike

Here are some comments on your WFCAM-2MASS colour equations, based on
an analysis of the synthetic colours. I get mostly pretty good
agreement for dwarfs, but with a few odd effects to note.

Steve

I confined myself to objects of luminosity class III and V in the
BPGS, as well as the additional M dwarfs from Sandy, in Hewett et
al. (2005).

1. Regarding the Z equation you got:

      Z_wfcam  - J2 =  0.95*(J2 - H2)

where J2, H2 are 2MASS. Below I have plotted dwarfs as crosses and
giants as green circles. The red line is your relation, which is
mostly a good fit. But two comments i) the relation goes very badly
wrong for M dwarfs cooler than M3 (become way redder in Z-J2), ii) you
can see from the plot that it seems to turn down very sharply at A0,
so this is something to watch out for. Some of these colours seem
quite odd e.g. A stars with negative Z-Y or Y-J:

    Z-Y     Y-J    class  BPGS no.  name
  -0.069  -0.120   B9V        13  HD189689
  -0.027  -0.094   A0V        14  THETA-VIR
  -0.023  -0.084   B9V        15  NU-CAP
  -0.056  -0.021   A2V        16  HR6169
  -0.022  -0.010   A1V        17  HD190849A
  -0.013  -0.044   A2V        18  69-HER
  -0.009   0.037   A3V        19  HD190849B
  -0.001  -0.050   A0V        20  58-AQL
  -0.029  -0.025   B9V        21  78-HER
  -0.015   0.127   A7V        22  HR6570
   0.028   0.009   A2V        23  HD187754
   0.012   0.083   A5V        24  THETA1-SER
   0.020   0.087   A5V        28  HD190192


2. For the Y band you got

  Y_wfcam  - J2 =  0.50*(J - H)

The analogous plot is below, and you see the same behaviour i.e. cool
M dwarfs very red in Y-J2, and a turndown at zero colour. However in
this plot the scatter looks worse, and the dwarfs and giants may
follow different relations.

I think it would be justified to worry about the usefulness of the
synthetic analysis for the Y band which is probably the band where the
BPGS calibration is worst.  Nevertheless there is a hint that the
relation for dwarfs is somewhat steeper than your value.


  3. J band, you got

    J_wfcam  - J2 =   -0.075*(J2 - H2)

I get somewhat different behaviour for dwarfs and giants

    J_wfcam  - J2 =   0.01-0.067*(J2-H2) dwarfs
    J_wfcam  - J2 =   -0.01-0.003*(J2-H2) giants


4. H band, you got

    H_wfcam  - H2 =  0.075*(J2 - H2)

I get pretty good agreement with that (0.080 for dwarfs and 0.065 for
giants)

5. K band, you got

      K_wfcam  - K2 = - 0.015*(J2 - K2)

The K band is the band where the giants and dwarfs differ the most.

I get
K_wfcam  - K2 = - 0.023*(J2 - K2) (dwarfs)
K_wfcam  - K2 =  -0.01 + 0.032*(J2 - K2)    (giants)

These relations are somewhat non-linear, and so the colour term
depends on the colour range selected.