Re: Curious Result

[Tom Droege <tdroege2@earthlink.net>]

I have done one more experiment.  I set the flat field box to produce a 
nearly saturated image.  I then scanned VVH throught it's complete range 
and let it run overnight.

The result was a very narrow range over which the expected flat field image 
was produced.  For low values, the leading edge (top) of the image was 
saturated, falling off toward the bottom of the image.  For high values, 
again the same image was produced.  In a narrow range, the expected flat 
field image was produced.  The peak value was 15000 counts or so.  With a 
lower level of -26000 counts, this indicated a full scale range of 41000 
counts, or 102500 electrons range.  Well above the guaranteed linear range 
of 80000 electrons.  I did not measure whether the signal was linear in 
this range.

The range of VVH over which an image was obtained was 3.7 to 4.1 volts.  I 
have been running at 3.9 volts for the last year and a half.  This is 
exactly in the center of the possible range.

It would appear that the experiment now nearly two years ago produced a 
near optimal value for VVH.

I still want to do the mag sigma vs mag experiment at various VVH values, 
but I do not expect it to show anything other than I have set VVH at the 
right value.

Tom Droege

At 06:45 PM 3/21/03 +0000, you wrote:
>Andrew and all,
>
>I have finished the first round of investigation of this.  As you say, 
>once you make VVH large enough it appears to make very little difference.
>
>Since there are clouds, this was done looking at a light box field at 
>about 40% and 80% of full scale.  As near as I can tell, there is no 
>difference in the gain at the top and the bottom of the frame as long as 
>one is somewhat back from saturation.  There is a real problem doing this 
>at saturation since there is all that extra charge and it goes 
>someplace.  Once you get close to saturation, the data is trash.  With 
>stars, you can look at points at various positions on the image without 
>flooding the CCD with charge.  This allows saturated measurements to be made.
>
>I have log book data that says that there is a big difference in the top 
>and bottom saturation level.  That is you can have a different saturation 
>level at the top and bottom of the frame.  By fussing with VVH one can get 
>the saturation level at the bottom of the frame to be nearly the same as 
>the saturation level at the top of the frame.  From my data, it is always 
>higher at the top of the frame.  I have been running with a relatively low 
>VHH (3.9 instead of mfgr's recommended 6) as determined by this experiment.
>
>It should be noted that the 80,000 full well (32000 count) linearity 
>specification for MPP mode is achieved at any possible VVH.  My log says 
>that anything over 3 volts will do it.  We have been setting the 
>saturation limit at 20,000 counts.  Because of the 16 bit data used, for 
>practical purposes we can rarely get over 30,000 count signals.  All these 
>things point to this not being the problem with the noise floor.
>
>This earlier measurement was done after some advice from the experts at 
>CCD_World.  Remember the full well limit for this device is something like 
>200,000 e- in BCM where it is only 80,000 e- in MPP mode.  We go to MPP 
>mode to get the lower leakage and thus much less noise for us.  We trade 
>off dynamic range.  This is not a bad trade for us since the bright stars 
>are less interesting since they are well studied.
>
>Now there is this curious effect that there are more stars found at the 
>top of the frame than at the bottom.  There is also the observed effect 
>that the brighter stars hit a noise floor.  So I am wondering whether the 
>two things are related.  The coincidence is too strong, and I think it 
>will have to be investigated.
>
>The plan is to take sets of images at different VHH levels.  Then I will 
>plot mag sigma vs mag for each image set and see if the noise floor is 
>related to VHH.  Sigh!  There is not a clear day in sight.
>
>Tom Droege
>
>At 11:52 PM 3/21/03 +0000, you wrote:
>>On Thu, 20 Mar 2003 07:22:11 +0000, Tom wrote:
>>
>> >Those of you worrying about this sort of thing (Andrew, Michael) might
>> >consider whether one can apply a top to bottom gain fit that would make a
>> >correction.
>>
>>Yes.
>>
>>Unless voltages are so badly off that charge is leaking
>>all over the place. It has never been that bad in the
>>past ...
>>
>>I seem to remember that you played with this before and
>>so long as the voltages were not grossly wrong it made
>>damn all difference? Apparent saturation levels changed
>>a bit but I was never convinced that you were actually
>>changing the charge transfer efficiency appreciably. Yes -
>>the gradients are real. But they calibrate out nicely.
>>
>>Andrew Bennett, Avondale Vineyard
>
>