WAW V.3 PHOTOMETRY STATUS REPORT

Version 3 Photometry Status

Bill W.

Reverse time-ordered, most recent at beginning.

17 July 2001 @ 09:00
1. Have added some scans from 980403n run with vmflor=0 and vmscal=1 to the VM Floor Test Page, as Run 15, at the bottom.
  The full night 980811s is also processing and should be ready shortly.
2. Important result: the knee and bifurcation we had earlier have been removed in all the VM floor tests, ever since the 1000:1 clip on the total variance was removed. There has been no sign of this in test runs in the past week.
  The multiple sub-branches of the Sig(PSFMag) Vs Mag curves remain for bright sources, and are associated with particular PSF/VM combinations.
3. Possible phenomenological fix-up program for the low S/N for bright sources:
  1. By study of StdDev of (PSFMag-StdApMag) and StdDev of cal field mags, determine realistic uncertainty for bright-source limiting photmetric magnitudes as function of seeing shape.
  2. For each PSFID, adjust VM floor and scale factor to get this desired bright-source limiting uncertainty (plot #2 -- essentially set by scale factor) while keeping chisqure (plot #1) distribution ~ 1 as a function of Mag.
  3. Keep an eye on the linearity (plot #10) to make sure it does not disintegrate as this process is carried out.
  Main problem is iterative character, mainly due to lack of understanding of S/N limit for different PSFs.
4. Thus have been trying to understand the high Sig(PSFMag) we have been getting for some PSFs and Variance Maps. I have been trying to identify those points in the PSF/VM that have the largest influence. Main conclusion so far is that south H-band PSF/VM 09505 is clearly grossly anomalous due to very large "pull" values far out in the wings. I am working towards a sharper, more quantitative result, hoping to remove the need to run scans with each PSF in order to determine the bright-source limiting S/N, which is the main operational problem I see to implementing the variance map floor & scaling technique.
7 July 2001 @ 19:40
1. First four test runs 1-4 with VM floor show following striking features:
  - Complete independence of minimum PSF Sigma(Mag) on floor value, over factor of 30× in VM floor (Plot 11).
  - Sharp variation in minimum PSF Sigma(Mag) as function of PSF ID, giving two branches expecially striking for H-band (plot 11).
  - No or very slight dependence of PSF Magnitude on VM floor value, and no sign of non-linearity dependence on VM floor (plot 10).
  - Substantial decrease in chisquare for bright sources as VM floor increases (plot 1).
2. Evidently we can rescale the Variance Map to bring the chisquares for bright sources back up to 1.0 per DOF. This will decrease the PSF sigmas in the bright source limit, and raise the S/N.
  "When do we stop?" raising the VM floor is obvious question. Answer seems to be "when bright-source S/N reaches actual observed RMS scatter of bright source measurements".
3. Gene generates a set of empiracle scaling factors to remove the chisquare droop in the chsisquare Vs Mag plots (plot 1). This can be done in PROPHOT with a single scale factor for each namelist only for a single shape; but can be done correctly for all shapes by rescaling the VMs themselves later.
  Scale factors determined from runs 1-4 are:
```
              run       J      H        Ks
              1      0.9440  0.8548   0.9598
              2      0.7892  0.6554   0.8391
              3      0.6697  0.5554   0.6507
              4      0.5459  0.3845   0.4750
```
4. Bill hacks PROPHOT to take yet another namelist parameter, vmscal; default is 1.0. If the VM floor (vmflor) is .gt. 0, then (a) VM is clipped to floor from below and (b) vmscal factor is applied.
  Agree to add one additional run, with VM floor 3× higher than run 4 (ie, 1/10th peak), scale factor same as that determined from run 4. New runs are 5-9; 5-8 correspond to runs 1-4, with scale factors applied, and run 9 is with the additional higher vmflor values.
5. Discover that "desens" parameter previously installed was not turned off in previous runs (1-4) done on night of 5 July. Need desens = .false. (default is .true.) in PROPHOT namelist. Decide that we can make runs with Gene's scale factors without redoing runs 1-4 first.
6. Runs 5-9 successfully completed overnight of July 6-7. (The desens parameter was set .false. for these runs.)
  Runs 1-4 also redone, with "desens" parameter set .false. No obvious difference between old runs 1-4 and new.
7. New set of runs (1-9) still show some further VM scaling needed, as chisquare droop for bright sources not entirely removed. Gene has generated a new set of vmscal values which we can apply to this end.

5 July 2001 @ 20:11

All well on first test run, except "min" should be "max" of course. Redeliver and tell ops to try again.
Data are in dino:/b1/prophot/980811s/s026_run1/image/ etc., with links in dino:/b1/prophot/run1/980811s/s065/image/ etc.
We appear to be ready to go. Have transmitted all this to Ron Beck; await news.

We have decided to do 4 test runs on the scan 980811s scan 065, each with a different variance map floor. For the 3 bands there are then 12 values in all.

                           J          H         K

             Run 1:     7.5e-7     1.5e-6    6.0e-7

             Run 2:     2.5e-6     5.0e-6    2.0e-6

             Run 3:     7.5e-6     1.5e-5    6.0e-6

             Run 4:     2.5e-5     5.0e-5    2.0e-5

These are based on numbers from Gene of peak VM values (table below). Peak we define as 99%-tile value. For S/N of 1000:1 we would go down to 0.1 of peak, and then choose three points bracketing that, #X above and below. Ken & Gene suggest also one much larger value as a sanity check, so that gives four altogther.

Eg, for H 09595, peak is ~5e-4, so pick 5e-6, 15e-6, 15e-7, and 5e-5.

                  VM PEAK VALUES for H/W 5 (SOUTH)

            >    dafoe:/2massd/data/prophot/psf_v3/s:
            > 
            >     file               99% 
            >  j09155_varpsf     2.479831e-04
            >  j09315_varpsf     2.438355e-04
            >  j09535_varpsf     2.520324e-04
            >  j09695_varpsf     2.714418e-04
            >  j09875_varpsf     2.178488e-04
            >  j10075_varpsf     1.962586e-04
            >  j10285_varpsf     1.691410e-04
            >  j10515_varpsf     1.376337e-04
            >  j10705_varpsf     1.470259e-04
            >  j10885_varpsf     1.313128e-04
            >  j11095_varpsf     1.141054e-04
            >  j11355_varpsf     1.246987e-04
            >  j11495_varpsf     1.241760e-04
            >  j11695_varpsf     1.193265e-04
            >  j11905_varpsf     1.265008e-04
            >  j12085_varpsf     1.087160e-04
            >  j12325_varpsf     1.092625e-04
            >  j12505_varpsf     1.130419e-04
            >  j12715_varpsf     1.033413e-04
            >  j12905_varpsf     1.190428e-04
            >  j13095_varpsf     8.906872e-05
            >  j13275_varpsf     9.156325e-05
            >  j13525_varpsf     1.063702e-04
            >  j13755_varpsf     9.489244e-05
            >  j14045_varpsf     8.056053e-05
            
            > 
            >  h09115_varpsf     3.269817e-04
            >  h09325_varpsf     4.573306e-04
            >  h09505_varpsf     4.661087e-04
            >  h09715_varpsf     2.131906e-04
            >  h09895_varpsf     1.843887e-04
            >  h10085_varpsf     2.048687e-04
            >  h10305_varpsf     1.931631e-04
            >  h10495_varpsf     1.530646e-04
            >  h10675_varpsf     1.740548e-04
            >  h10905_varpsf     1.656977e-04
            >  h11085_varpsf     1.594721e-04
            >  h11285_varpsf     1.833933e-04
            >  h11475_varpsf     1.502254e-04
            >  h11705_varpsf     1.899950e-04
            >  h11915_varpsf     1.653447e-04
            >  h12105_varpsf     1.259305e-04
            >  h12345_varpsf     1.196331e-04
            >  h12615_varpsf     1.872011e-04
            >  h13025_varpsf     1.157461e-04
            >  h13415_varpsf     1.632834e-04
            >  h13735_varpsf     1.446952e-04
            >  h14075_varpsf     1.013879e-04
            > 
            >  k09355_varpsf     2.422367e-04
            >  k09545_varpsf     2.122380e-04
            >  k09705_varpsf     1.965934e-04
            >  k09905_varpsf     1.911027e-04
            >  k10095_varpsf     1.842425e-04
            >  k10315_varpsf     1.946744e-04
            >  k10505_varpsf     1.751701e-04
            >  k10715_varpsf     1.531904e-04
            >  k10915_varpsf     1.844551e-04
            >  k11105_varpsf     1.456687e-04
            >  k11315_varpsf     1.407111e-04
            >  k11485_varpsf     1.358051e-04
            >  k11725_varpsf     1.471140e-04
            >  k11885_varpsf     1.466195e-04
            >  k12125_varpsf     1.427828e-04
            >  k12305_varpsf     1.568644e-04
            >  k12515_varpsf     1.154470e-04
            >  k12665_varpsf     1.193139e-04
            >  k12865_varpsf     1.223882e-04
            >  k13055_varpsf     1.735870e-04
            >  k13295_varpsf     9.382548e-05
            >  k13555_varpsf     1.317628e-04
            >  k13705_varpsf     1.346597e-04
            >  k14055_varpsf     1.322175e-04
            >

To set the variance map floor, use variable "vmflor" in prophot namelist. This is set to -1. in PROPHI before namelist is read, and if it is found to be greater than 0 afterwards, the variance map "Ve" is set to min(vmflor,Ve) whenever a new VM is read in subroutine getpsf.
Executable pixphot_tst is now in
rodan:/2massc/dev/src/pixphot_tst
so everything is ready except for the namelist I think.
Find PROPHI routine (called once at beginning) reads the namelist; PROPHO routine (called for each batch of candidates) gets PSF & VM in routine getpsf.
Remove 1000:1 dynamic range clip from PROPHOT (solve.f, subroutine findm); verify with Ken.
Talk to Gene, look at Sig Vs Mag plots and agree H-band in south is good (ie, especially awful) for shape ~0.95 or so. Decide to run 980811s, scan 065 as our test case for that.
Had meeting with the team (Roc, Ken M, Gene, Davy, Schuyler) on the situation. Agree I will make a special PROPHOT test version that will remove the 1000:1 variance floor, and put a floor on the Variance Map at the point it is read in. The floor value will be read from the namelist.
Have updated www page on spider with del-BG info. Page is http://spider.ipac.caltech.edu/staff/waw/2mass/v3_photom/delta-bg/index .html. Quit that for now.

Go To:
Up - | IPAC Home Page | JPL

e-mail: Bill Wheaton
URL: http://spider.ipac.caltech.edu/staff/waw/index.html
Last modified: 17 July 2001.