Subject: Lowest noise CCD readout.
Date: Mon, 6 Jan 2003 03:07:48 +0000 (UTC)
NNTP-Posting-Date: Mon, 6 Jan 2003 03:07:48 +0000 (UTC)
User-Agent: tin/1.5.14-20020926 ("Soil") (UNIX) (Linux/2.4.20 (i686))
I'm trying to make a star-tracker, that can see stars at video rates.
(currently it's not good enough, using the stock electronics from a
surveliance camera, I need about 6.5 times lower noise.)
So far, I've picked up the following tips from web-searching.
1) Not resetting the pixel before sampling, so that the last pixel value,
rather than the reset acts as the reference, may reduce noise, as you don't
have reset noise. (if you don't get overflows.)
2) If the bandwidth of the amplifier (CCD and preamp) is much higher than
the sampling rate, multiple sampling, and doing the correlation digitally
can beat CDS by a relatively large margin. (or presumably doing the same
thing analoguely with a 4 quadrant multiplier and lots of good analog switches)
3) Reducing the readout rate helps lots, as you eliminate lots of HF noise.
Cooling a little can't hurt.
Now, some things I've thought of.
If you know what the image should look like roughly (4-20 stars), then
you can pull a lot of tricks.
Number 1 helps, as most of the background is more or less black, and also
means you can read out quite a bit faster than the spec, as the output
amp has longer to settle (or you have lower noise).
Numbers 3/2 means that if you reduce the sample rate to 1% of the normal
for 10 5*5 pixel areas, you reduce noise by around 10 times, if you combine
this with 1, you may get around 15 times. All without increasing
readout time much.
Both of these add complexity in the timing, sampling and processing stages,
as you no longer get a simple image from the output of the CDS stage, but
one where the noise varies across the image.
I'm slightly less clear when it comes to the best way to reduce the preamp/CCD
Cooling a little won't hurt, but practical amounts won't help much.
I want to minimise the noise both at 10-30Mhz or so, for normal rate sampling,
and 10-500Khz or so, for low rate.
Do I try to match the CCD output impedance, or use a jfet or other buffer?
(the CCD is the ICX255AL, I don't know what the output circuit looks like)
Cost isn't that much of an issue, both power and space are.
http://inquisitor.i.am/ | mailto:firstname.lastname@example.org | Ian Stirling.
He who lives in a glass house should not invite he who is without sin.