One more thing about light boxes.

[Tom Droege <tdroege@veriomail.com>]

You can't dim a tungsten lamp.  You have to run it at the right temperature 
to get something that has the right spectrum.  Now you have something that 
ages. Since the filament wears over time, it is a big project to adjust the 
illumination.  You have to do it with neutral density filters.  These have 
problems of their own.  Even so, you can't count on the spectrum being 
constant by just adjusting the brightness.  You really have to take a 
spectrum.  The whole thing gets really complicated at the 0.001 level where 
we want to work.

I think it may give better results to use several LEDs of different colors 
and pulse them.  These can be made pretty well behaved.  I like to drive 
them with 1-2 ns 100 volt pulses and put an ampere or so through 
them.  This makes a pretty uniform pulse.  Many of the problems you have 
with getting uniform light out of an LED go away when you pulse them with 
an avalanche transistor.  There are temperature effects and the like that 
are hard to control at low voltage.  Now the brightness can be adjusted by 
how many pulses one puts in the 100 second open time.  One can easily get a 
million to one dynamic range.

I have tested PM tubes this way at Fermilab, and got very nice linearity 
curves over 6 decades.  I was really testing that my electronics was 
linear.  One does this by running the PM tube into a cascode amplifier for 
the fans of such things.

This scheme allows one to make nice linearity curves.   Even Andrew has not 
yet asked for linearity data on the CCD.  We probably need to check 
linearity as we get near full well.  We could get it this way.

As you all can see, this is a topic with far more discussion than action.

Tom Droege