Re: Help Wanted from CCD Experts

[hjohnson@pluto.njcc.com (Herbert R Johnson)]

Here's some feedback as you requested. - H

On Sat, 10 Oct 1998 16:55:02 -0500, Tom Droege <droege@wwa.com> wrote:
*>First, I have set the gain at 1.7 e-/ADU.  This is to match the expected
*>MPP full well to full scale. 
*>.....the ADC.  This is a $30 device, and
*>only claims "16 bits with no missing codes".   They claim +/-3 LSB INL.

*>Running with the camera cable disconnected, I get sigma of 10.2 e-

Fair enough. 3 LSB = 8 counts ==> 13 e-, within specifications.

*>Running with the camera connected, but without the CCD plugged in, I get a
*>sigma of 11.9 e-.
*>
*>Most of this would seem to be due to the ADC.

Would seem to be the case. One source of noise will be whatever transistor
you use to buffer the CCD output.

*>Next I put in a chip, and a Vertical program that runs backwards, i.e.
*>shifts charga away from the output serial register.  This gave a sigma of
*>23.8 e-.

At what temp? and does this figure change with CCD temperature and integration
time? That's where you can determine dark current.

*>Next I set the Vertical program to run forward, and got a sigma of 40.8 e-.
*> This for a 50 second time between the start of the readout and the start
*>of the previous clear. (Note that these two measurements indicate a dark
*>current of about 33 e- /sec.  Not so far from the number below.  It depends
*>on how you decide to subtract the square of 23.8 from the square of 40.8) 
*>
*>Profiles taken across the chip in the vertical direction look even except
*>for a small effect at the start of the scan.  A horizontal profile has both
*>front end and larger back end effects.  It could be stray light, though I
*>have things pretty well covered up.  But why only on the vertical edges?
*>OK, I can believe a transient at the start of each line, but how can there
*>be one at the end?  Nothing special happens there.  This effect is about
*>800 counts above the mean at the edges.  It is shorter, possibly persists
*>for 100 counts at the beginning of the line.  The effect at the end of the
*>line starts rising a few hundred counts from the edge and climes to the 800
*>at the edge.  

Change the horizontal readout rate and see if the effect changes. See if
it changes with temperature too. Maybe you need to play with the reset pulse
height or width.

*>And now the curious bit when I try changing the time of the dark exposures:
*>
*>Exposure            Sigma               Mean
*>Seconds             ADU                  ADU
*>
*>60                       28                    -22140
*>120                   212                    -21884
*>240                   339                    -21388
*>
*>Note that everything is in ADU ad 0.57 ADU per e-


*>I would expect sigma to increase as the SQRT of the mean.  Assuming we
*>start near zero, and increase linearly.  I had planned this experiment to
*>determine the ADU value for zero exposure???  But it does not make sense to
*>me??   i.e. sigma is increasing too fast.  I have looked at the
*>distributions and there is nothing obviously funny.  The noise is going up,
*>and at the level sigma indicates.  

Keep in mind your readout time is pretty LONG. How long does it take
to get to the LAST line vs. the FIRST line? Compute a per-line statistic
and see if your curve looks better. IF this is an issue, the curve would
"staighten out" with longer exposures as exposure time dominates over
readout time.

Herb Johnson

Herbert R. Johnson              http://pluto.njcc.com/~hjohnson
hjohnson@pluto.njcc.com         voice 609-771-1503, New Jersey USA

             amateur astronomer and astro-tour guide
     classic S-100 computers restoration & parts as "Dr. S-100"
   rebuilder/reseller of classic Macs for your computing pleasure
          and senior engineer at Astro Imaging Systems