Mosaic Photometry vs PFCCD photometry

The Problem:

A comparison of magnitudes between two images of the same field should give consistent magnitudes, within observational errors. This is a example of this for two on-band images with PFCCD:

n1.31.30.ps

Mosaic also shows this behavior:

However, between the PFCCD and the Mosaic for Virgo IPN Fields 1 and 7, we do not see this behavior. The plots show apeture photometry with apeture size of different FWHMs:

fwhm1-1.5.ps

fwhm1.7-2.0.ps

There is a systematic offset in magnitudes between the PFCCD and the Mosaic magnitudes, with a "knee-like" feature occuring at an approximate m(PFCCD) magnitude of 24.0. The magnitudes of the same stars in the Mosaic can be up to 0.5 mags fainter than the relation found for bright stars.

A typical star (f1=1623.93,744.75 f7_b2=3063.42,2504.660) that shows this effect has a peak value of 740 ADU in the PFCCD image, and 940 ADU in the Mosaic image. The sky value is about 570 ADU in the PFCCD exposures, and 830 ADU in the Mosaic exposures

The effect occurs for both the on-band and the off-band images:

on-off.ps

The effect does not seem to differ by chip type, observing season, or Mosaic reduction steps. The following image compares the magnitudes of a raw Mosaic frame to Field 1 and Field 4, which were both taken by the PFCCD, but in different seasons

The size of the effect

The difference in flux between above and below the knee is large, about 700 ADU in a 2.7 pixel radius apeture for Mosaic. This corresponds to 30.56 ADU per pixel.

amount.mosaic.ps

The magnitude refers to the Field 1 instrumental magnitude.

non-linearity tests

Mosaic

For the first Mosaic linearity test, I compared obj068[im7] to obj074[im7]. obj068 is a 60s offband exposure and obj074 is a 900s offband exposure. I ran DAOFIND on obj068, then did photometry on both images, with a apeture size equal to the FWHM. Then, I compared the flux to both of them, using a linear model, ignoring the bright stars in the fit. I found a relation of obj074_flux = 16.5* obj068_flux - 613. I subtracted this relation from the measured data, and plotted the residuals below:

Here's the residuals, in terms of the precentage error in the flux:

There appears to be a non-linearity in the data, at a obj068 flux of about 20,000-40,000 ADU. The peak values of the stars in this range are 7257, 9255, and 14249 ADU respectively, on obj074. Assuming the default gain of 3.2 e- per ADU, this is well below the limit of 70,000 e- given in the Mosaic manual.

Results of new tests:

Following George's advice, I re-did the tests, but I went through and cleaned up the object list, and used apetures that were 4 times larger than the FWHM. The results changed quite a bit:

PFCCD

I performed a similar test on the PFCCD data between a 1200s exposure and a 3600s exposure (both onband). The function found was flux_long = 3.52* flux_short - 1450. Below is the difference between the model, and the flux in the long data:

Results of new tests:

Following George's advice, I re-did the tests, but I went through and cleaned up the object list, and used apetures that were 4 times larger than the FWHM. The results changed quite a bit:

pfccd.linear.fwhm4.ps

pfccd.lin.resid.fwhm4.ps

From inspection, it turns out that the only star that has a non-linearity has a peak value of 47,451 ADU, though that non-linearity. I checked the start thoroughly to see if anything could have been messed up (sky, cosmic rays), and found nothing. Adjacent to the bright star in the plot is a galaxy that has the same flux, but a much lower peak value (22293 ADU). It lies very close to the linear regime. So, there is a non-linearity.

New Mosaic data:

I took new Mosaic data in April, 2001, of Field 7, once I had assured myself that the Mosaic chip was linear to about 2% below a certain threshold. I then found compared magnitudes between the PFCCD data, the "old" 3/99 Mosaic data, and the "new" 4/01 Mosaic data. If one of the chips were at fault, it should be apparent in the new data.

The Field 1 comparison (f1.88.ps) clearly shows the "knee" feature, strongly indicating that the PFCCD data is bad somehow. However, the average shift is about 0.3 mags (see dashed line), smaller than the 0.5 mags we need.

Looking at the "old" and "new" Mosaic data (f7.88.ps), the two are in closer agreement. However, there is a smaller (0.2 mag) offset between bright and faint objects.

Working Hypotheis - There are two effects, that add together to make the 0.5 mag effect. 1) A non-linearity in the PFCCD data (0.3 mag), and 2) A smaller, 0.2 mag in the Mosaic data.

The effects togtether can be seen at: colors.ps They add up to an offset of 0.5 mags

Conclusions

 

Last Modified: April 23 2001 Web Page by John Feldmeier johnf@eor.astr.cwru.edu Department of Astronomy Case Western Reserve University