RE: Quick note on Disk 18a

[Tom Droege <tdroege@veriomail.com>]

Hey!  You guys are worrying about nits when it is raining boulders.

Yes it is true that most successive approximation ADCs have a momentary 
fairly well defined differential linearity error.  But is is going to 
change with the temperature and time of day and small variations in the 
power supply voltages.  It is not worth trying to calibrate.  I attempted 
to match the ADC quality to the data.  It is way overkill.  This ADC is 
always good to a few ADU.  The differential linearity is a 3 or 4 
ADU.   The integral linearity is somewhat worse, but it is normally better 
than the spec.

All this in a system where the noise from the system is 8 or so ADU 
RMS.  When looking at the sky, we have typically 40 or 50 ADU.  Even Arne 
has a bunch of ADU sky noise.

This is not the thing to worry about.  If you want something to worry about 
try dielectric absorption.  This is a serious problem in a wide range 
system like this.  The usual manifestation is that the sample following a 
high spike is larger that it should be.  This is because the best 
capacitors that you can get still remember the voltage on them for a 
while.   This killed a bunch of people in the early days of radio until 
everyone learned that you had to put a continuous short on a high voltage 
capacitor or it would develop a charge and kill you.  The same thing 
happens at a smaller scale on the integrator capacitor for the double 
correlated sample.

This is a pretty good ADC.  Don't try to calibrate it any more than a 
simple gain and offset calibration or you are just fooling yourself.  It 
they could be made better with calibration by a computer there would be one 
in the package.  (there may already be one in the package).

You may be able to guess that I have been through this a lot at Fermilab 
with the High Energy Physicists.

This is just a wonderful device for $20.  I would attempt to get a gain and 
an offset calibration.  One can expect both to move around at the 0.01%/C 
level.  This is what can be done unless you go to heroic measures, and then 
not at these speeds.  Besides you are just fooling yourself if you think 
you have a better calibrations as there are so many things going on.

Tom Droege

At 05:38 PM 5/21/01 -0700, you wrote:


> > -----Original Message-----
> > From: Stupendous Man [mailto:richmond@a188-l009.rit.edu]
> > Sent: Sunday, May 20, 2001 10:00 AM
><snip>
> >
> >   If one makes a histogram of the pixel values in the raw images
> > on Disk 18a, one finds
> >
> >        - the distribution of pixel values in I-band images shows
> >              a small peak every 8'th value, and a very large
> >              peak every 64'th value
> >
> >        - the distribution of pixel values in V-band images shows
> >              a small peak every 2'nd or 4'th value; the amplitude of
> >              the variations is smaller than those in the I-band images
> >
> >   Perhaps some sticky bits in the ADC?
>
>
>
>It looks like each ADC has it's own "signature".  Is this
>how you'd fix it?
>
>
>Normally you interpret a positive binary integer by doing
>
>      SUM( b[i] * 2^^i )
>
>where b[i] is the value of the bit (0 or 1) at location i.
>I wonder if you could histogram some random noise and determine
>values for c[i] and then replace the above by
>
>     SUM( b[i] * c[i] )
>
>So far so good.  But how to compute c[i]?  Is it based on just
>the difference between the observed fraction of 1s vs. the
>expected 0.5?  So we'd get values for C like 0.9998, 2.0003,
>3.9996, ...
>
>If what we are seeing really is an artifact of the ADC which
>remains constant over time we could built a hardware white
>noise source and spend a mount taking 10^10 samples in day light
>hours.  I've heard of people using a white hot lamp filament
>as a white noise source, so it could be cheap.
>
>If tom is right about this being a 0.03% problem it's still
>worth fixing if you want to look for stuff like transits.
>Those 0.03% fixes add up.
>
>
>This email message is for the sole use of the intended recipient(s) and may
>contain proprietary and confidential information. Any unauthorized review,
>use, disclosure or distribution is prohibited. If you are not the intended
>recipient, please contact the sender by reply email and destroy all copies
>of the original message. Thank you