Re: Linear range of the chip?

[Tom Droege <tdroege2@earthlink.net>]

Rob,

I did this a while back with my flat field box.  I think I wrote it up for 
the list, but I can't find it.  I did not make it a TN.  Sigh!  Should have.

OK, I just took the light box and made different length exposures.  Then I 
plotted the mean value of the image vs the length of the exposure.  The 
important thing is that you don't have to worry about short exposures.  If 
it is non-linear this is going to occur at the high end.  So you want to 
take exposures of 50%, 60, 70, 80, 90, 100, 110, ...  Then take a second 
set to study the area around the problem.  Sigh!  As I write this I can 
remember making the plots.  Don't fight the low end.  It is linear in 
theory, the ADC that measures it is linear in theory.  This is not the kind 
of trouble that we are going to have.  To measure the low end you will need 
techniques like #3 below.

I did it a second way, I think using stars.  I can't believe that I threw 
all this work away.  But I can't find it.

OK, the right way #1.

Take one of the register outputs on the Mark IV and put an LED on it.  Now 
arrange to turn the LED on and off for a fixed length of time with good 
timing software.  You should be able to do a millisecond reliably if you 
know how to do good timing.  The actual on time is not critical as long as 
it is repeatable.  Now shine he LED on a light box screen.  Now make 
exposures of equal length containing 1 flash, 2 flashes, 3 flashes ...  Up 
to 10,000 or so and now plot the mean pixel value.

The right way #2.

Build a avalanche transistor pulser and power it from a precision regulated 
supply.  Use a length of cable for the energy storage.  Now you just have 
to trigger it from the computer.  No timing is necessary.  Each pulse will 
be constant in energy.

The right way #3

This is like #2 but you have a photo transistor looking at the LED.  You 
double correlated sample the light put out by the LED and measure it with 
an ADC.  This signal is used to adjust the precision high voltage supply 
that drives the avalanche transistor.

Believe it or not I have done this sort of thing to measure PM tubes.  I 
once measured a million to one linearity range.  This is easy for a PM 
tube.  This was my trade, folks.

I think you don't have to study the linearity.  It is linear until it 
saturates.  It goes from linear to saturated pretty quickly so different 
length  exposures with a light box will tell the tale.

The short story is that everything in the system is designed to be much 
more linear than you can probably measure.  You just have to measure around 
the saturation point, which should be 32,000 or so counts above the bias 
level.  Measuring any better is beyond your ability to do it if you have to 
ask how.   ;^)  Even if you think you know how, you will learn a lot 
trying.  I assure you, I would learn a lot trying.

You might look at the thread "A Curious Result" from the March 2003 archive 
for some relevant information.  But this is not where I did the linearity 
check.  Perhaps someone remembers the data I put up.

Tom Droege

At 04:34 PM 6/7/03 -0600, you wrote:

>Tom,
>
>How does one do a linearity check?  One way I can think of would be to
>monotonically increase the exposure on a series of images, and graph the
>counts for some star in the image.  Other ways?
>
>And since I'm all positive, that would be 7000 to 42000 counts?
>
>Thanks,
>Rob
>
>On Sat, 07 Jun 2003 16:52:45 -0500
>Tom Droege <tdroege2@earthlink.net> said something like:
>
> > Rob,
> >
> > The specified linear range in MPP mode (which we use) is 80,000
> > electrons.  There are 2.5 e- per ADU so this is 32,000 counts.  So the
> >
> > linear range is the pedestal plus this.  That would put the linear
> > range at -25,000 to +7000.  It is usually better than this.  90 to
> > 100,000.  So -25,000 to +10,000 is a pretty safe range.  You will see
> > much higher readings, but had better do a linearity check.
> >
> > Tom
> >
> > At 03:47 PM 6/7/03 -0600, you wrote:
> >
> > >Hey Tom,
> > >
> > >I'm working with Doug W. on daophot setup, and realized I'm not sure
> > >what the linear range of the CCD's are.  I believe you indicated it's
> > >higher than you originally thought?  Do you have a number I could
> > >use?
> > >
> > >Thanks,
> > >Rob
> > >
> > >--
> > >O_
> >
> >
> >
> >
>
>
>--
>O_