CCD-world: cross-talk in multi-channel CCDs

[Tom Droege]

The following was posted to CCD-world:

Roy and CCD-world,

There are about a million ways that you can get that level of cross talk.
I have seen this level of cross talk from 1/4" of printed circuit trace in
the wrong position.  

The first thing to do with a problem like this is to make a very careful
diagram of where all the currents go.  I find it is very hard to get people
to do this.  Often, if you study the circuit long enough and draw it with
enough care the answer will pop out at you.  I would concentrate on the
output current path from the chip.  Are there any common traces?  i.e. the
ground return from the pull down resistor on the CCD output.  You might
change this to a JFET that draws constant current so that big signals do
not move the ground around.  (Careful here, some of these devices are
noisy.) The same holds for the later stages.  No common current paths.
Move the signals from place to place with differential drivers and
receivers.  Move local ground with the signal.  Remember we are talking
about 1/4" differences in printed circuit traces here.  You do not quote
your digitization rate, but I assume it is MHz rate or you would be using a
16 bit ADC.  Lots of things can add problems at high frequencies.  There
are large peak currents that bounce the ground around.  One needs very
short leads between the driving amplifier and the ADC.  I assume you are
using surface mount devices.  Really work on the layout to get the leads
very short.  

However, the non-linear part gets me.  Are you sure?  Or is it just coming
up out of the noise.  

Congratulations on what sounds like a very interesting system.

Tom Droege  

At 09:37 AM 3/22/00 -0700, you wrote:
>The following was posted to CCD-world:
>
>Dear Fellow Photon-Foragers,
>
>	I am participating in the up-grading of the GONG
>(http://www.gong.noao.edu) video data system which includes a new camera
>(Silicon Mountain Designs 1M60-20) with a Thomson THX7887A imager
>(frontside 1024 square array, four channels of 1024 x 256, 14 um pixels, 60
>frames per second, 12 bits per pixel). Our data acquisition algorithm
>consists generally of co-adding images for 60 seconds into three or six
>accumulation buffers in synchrony with the angular orientation of rotating
>optical elements in the system. Co-adding 600 or 1200 12-bit numbers
>produces approximately 23-bit results.
>
>	We are noticing a low-level (far below the 12-bit resolution of an
>individual image) cross-talk between the four output channels of the
>camera. The phenomenon is non-linear in that it does not appear until a
>threshold luminance level is reached. Our initial hypothesis that there
>were insufficient bypass capacitors associated with the video amp power
>supply rails and CCD bias voltages appears to have been disproven by an
>experiment involving the installation of larger caps. There appeared to be
>no beneficial effect.
>
>	Although not a serious problem, we would like to reduce or eliminate the
>effect if possible. Before proceeding further with our testing, we thought
>it might be wise to consult the CCD engineering community to see if anyone
>else has seen such an effect and if they could indicate to us if it is due
>to the CCD support electronics or an inherent property of this particular
>CCD imager or multi-channel CCDs in general.
>
>	Thank you very much for your attention.
>
>					Best regards,
>					  Roy Tucker
>
>
>- -- CCD-world -- --
>CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
>Standard replies will go to the list; address personal replies manually.
>For more information, please go to:  http://www.not.iac.es/CCD-world/
>
- -- CCD-world -- --
CCD-world is fully moderated. Send posts to CCD-world at astro.ku.dk
Standard replies will go to the list; address personal replies manually.
For more information, please go to:  http://www.not.iac.es/CCD-world/