This is an analysis of the experiment in which box 4 was removed, to compare fluxes 
with and without box 4.  The target was Mars, and the experiment was done on Jan. 10, 
in unchopped scan mode.

Additional data were collected on January 8th, using Jupiter, in Hertz-mode.  These 
have not yet been analyzed.

Note that the Mars fluxes given below were based on hand-written notes made in 
the instrument log, recording the peak flux.  This file will be updated later 
with a more careful analysis.  

*May 9 UPDATE: I have added a few more paragraphs at the end of this memo, based 
on Darren's recent analysis of these data.

__________________________
ANALYSIS
__________________________
                                                                                             
file 28147: SHARP fully installed; Mars signals are 170 and 176 for the two subarrays
file 28148: box 4 is removed and Mars moved to one subarray.  Signal is 561.
file 28149: box 4 is re-installed and signals are 175 and 208 for the two subarrays
                                                                                             
Ratio of signals with and without SHARP is 0.32, giving a 64 % efficiency
                                                                                             
But now we correct for changing gain, using Darren's recent posting:
                                                                                             
file 28147: G(H) = 1.434  G(V) = 1.460
file 28148: G(H) = 1.261  G(V) = 1.262
file 28149: G(H) = 1.442  G(V) = 1.468

The gain is 15 % lower with SHARP removed

So this means that the optical efficiency of SHARP is really 56 %

____________________________
DISCUSSION
____________________________
                                                                 
The throughput losses listed in Table 2 of our 2004 proposal (on the private page) 
would add up to a total loss of 23 % of our light, for an optical efficiency of 77 %.

So we should derate our performance specifications for the January run by a factor of 
1.38.

Possible reasons for the additional losses:
                                                                                             
(1) grid imperfections introduced during repair job in Fall 2005.
                                                                                             
(2) bad alignment of M4 and M5 (but note, this would also affect SHARC-II).
                                                                                             
(3) SHARP broadens the beam (above, I measured peak flux, not integrated flux)
                                                                                             
(4) HWP has more loss than expected (easy to test)
                                                                                             
(5) cross-polarization due to curved mirrors (ignored in initial analysis of loss)

(6) SHARP not installed at correct position (this was not checked as carefully in 
January as it was in August).

__________________________________
NOTES ADDED MAY 9:

Darren writes:
________________________
I found it helpful to disable pixel weighting, frame weighting, and deglitching 
within sharcsolve.  This is sometimes necessary for bright sources.  I did use 
"-rz" and aligned the images with -fa.

5 out of 6 images look good.  The V image for 28148 (which should have no source) 
converged to an interesting negative source, about 1/30 the amplitude of Mars.  I 
guess that gives some idea of systematic errors in the image reconstruction.

For the H array, with no correction for extinction or nonlinearity:

28147:  amp = 242.4 mV, FWHM = 13.62"
28148:  amp = 809.5 mV, FWHM = 11.37"
28149:  amp = 248.6 mV, FWHM = 13.75"

The ratio of amplitudes with and without Box 4 is 30.3% (no nonlinearity 
correction).

The ratio using integrated flux in the Gaussian fit is 43.9% (no nonlinearity 
correction). 
___________________

My analysis of Darren's numbers:

I make no extinction correction, as I don't think one is needed because of the 
sequence in which the data was taken, and the agreement between 28147 and 28149.

The 15% relative reduction in gain due to increased loading is important.  With 
this correction, the optical efficiency of SHARP becomes 0.85 * 2 * 0.439 = 75%. 
(this neglects differences between H and V, which however, look very similar in 
terms of width and peak intensity for all data taken in Jan. 2006).

Compare 75% with 77% and we get a very pleasant agreement between predicted and 
measured efficiency.  Note however, that the beam is very broadened, which seems 
to be a characteristic of the January run that was not seen in the August 
H-beam.

So we don't really have an efficiency problem, just a beam-size problem.  
Referring to the six possiblilites listed above, we now know that the answer is 
"(#3) SHARP broadens the beam".  As far as why SHARP broadens the beam, see 
explanations (#2) and (#6) above.  These should be easy to fix, which will get 
us back to our nice beam that we saw in our H-beam maps from August.

Note that the increase in beam size from August to January is almost certainly 
not due to SHARP alone because it seems approximately fixed in AZ and certainly 
not fixed in SHARP-space.  (This comes from an analysis Hua-bai and I did on his 
posted beam maps, by comparing direction of elongation with elevation.)  So I 
think we can rule out the X-grid wire-clumping and the hwp-a/r coat as the 
culprits (these are the only two things that we changed in the optics).

This means that the problem should go away once we achieve the good alignment of 
M4, M5, and (possibly) SHARP that we had in August.  Here I am assuming that we 
did not bump or move any of the mirrors in SHARP between August and January.  
If we wanted to check this assumption, in principle this can be done by re-doing 
Hua-bai's laser tests that he did at NU.  We could do this on a lab bench at 
CSO.