Re: Furthering the doom of astro CCDs...

Davoud <star@xxxxxxx> wrote in news:220120091959525187%star@xxxxxxx:

I'm very sorry to hear about your DSLR noise problem. Might I suggest a
Canon?

I own most Canon DSLRs and I can assure you that they all have a noise
issue...

there is no "major noise factor;" for _me_, getting the noise down to
an acceptable level is no more difficult than with a CCD image--dark
reduction and alignment and stacking.

A level that is acceptable to you, for your purpose. No doubt about that.
It's perfectly acceptable for my purpose as well.

I'm sorry to hear that your DSLR has a serious noise limitation...
Yet, your DSLR can produce pretty pictures. Go figure.

Yes, DSLRs do produce nice pictures.

WARNING: pictures produced by me, regardless of subject matter, and
whether produced with a DSLR or a cooled astronomical CCD camera, or a
pinhole camera, are not suitable for research purposes.

Research aims at reproductibility and best signal to noise ratio.

Thermal noise, as the name indicates, is related to temperature. CCDs and
CMOS are essentially electron counting devices. Photons hit them and
bang, here comes electrons. But temperature is essentially a measure of
how agitated atoms are. When they are agitated they collide and electrons
are unhappily kicked out of their orbits. If one electron is freed every
second, you have released 720 electrons in a 12 minutes exposure. About
18% of a loaded CCD electron well if its capacity is 40000 electrons or
13.5% a 60k electron well in a KAI-11002M... The bad news is that science
is usually interested in faint signals and wants a good signal to noise
ratio to be on the safe side. The good news is that the number of
electrons released goes down as you lower the temperature (roughly it is
halved everytime the temperature goes down by 6C, 7C is the value for the
KAI cip used in your SBIG). If the signal you are interested in generates
1 photon per second, that the quantum efficiency of the chip is around
33%, you'll have 240 electrons of signal. Not too good if you have 720
electrons of dark curent. Lower the temperature by 28C and you'll only
have 45 electrons of dark current. Much better! Remember, this is a
general principle, values may differ, but all other things being equal, a
cooled sensor will perform better than a warm one. Your Canon, which is
already very good, would be even better if it was cooled, in roughly the
same proportion. Of course, Canon may very well have improved on the
Kodak chip, and found ways to deal even more effectively with the dark
current - after all, dealing with dark current at room temperature is
what made digital cameras possible in the first place - but regardless of
their prowess, they would be even better if they were cooled.

You will tell me, of course, that dark current can be substracted, that's
what dark frames are for. Yes, but there is bad news here too, both for
signal to noise ratio and reproductibility. First of all, the
temperature/dark current relationship is not linear, which means the dark
frames should be taken at the same temperature as the light frame (or a
complete model of the temp/dark current curve should be taken into
account). Then, because the material of the sensor is not perfect, there
are different populations of pixels who respond differently to
temperature, the two extremes are "hot" and "cold" pixels, but you get a
range of in-betweens... Matching, as precisely as possible, the temp of
the light frames and the dark frames solves the issue. Again, it is quite
possible that Canon has developped advanced models of its sensors and
that it deals quite well with that issue, for practical purposes, but it
would be a black box, under an additional software black box (see the
recent Canon 5D MK II artifacts) and science doesn't like black boxes.

But that is not the end of it: there is of course the obvious Bayer
matrix and all its issues (a bit as if slightly different filters were
dropped on each and every pixels), the huge variations introduced by the
various interpolation techniques, the non linearity of the response (some
people even think it is impossible to do science with an anti-blooming
cooled CCD) etc... etc...

To cut a long story short, I believe Chris is interested in getting the
best signal to noise ratio possible, in the most reliable and
reproductible way and, in the current state of things, DSLR do not
deliver that. Nothing personal imho.

A nice cure showing how cooling can reduce dark current drastically

http://tinylink.com/?qSrXLrFUZx

Some input on pixel variations

http://www.mirametrics.com/tech_note_tempvar.htm

A word about blooming

http://learn.hamamatsu.com/articles/ccdsatandblooming.html

The 11002 specs

http://www.yankeerobotics.com/trifid/dnload/KAI-11002.pdf

.

Relevant Pages

  • Re: One upmanship and Canons claim
    ... read and thermal noise zero. ... best consumer cameras is already under 4 electrons ... dark current a long time ago, and DSLR CMOS dark currents are ...
    (rec.photo.digital.slr-systems)
  • Re: ISO 200000 ?
    ... The 20D for example is on the order of 3 to 4 electrons. ... photon noise similar to read noise. ... And if you have less than 40 photons?? ... More than an order of magnitude better than current dSLR devices - and that doesn't even begin to address the issue of QE or dark current noise. ...
    (rec.photo.digital)
  • Re: Why Nikons D50 is "the best"
    ... The D50 does have a much lower dark current generation than practically ... any other DSLR, it also has much lower amplifier glow. ... The read noise, for example is almost double the Canon ...
    (rec.photo.digital.slr-systems)
  • Re: Why Nikons D50 is "the best"
    ... The D50 does have a much lower dark current generation than practically ... any other DSLR, it also has much lower amplifier glow. ... The read noise, for example is almost double the Canon ...
    (rec.photo.digital.slr-systems)
  • Re: Why Nikons D50 is "the best"
    ... The D50 does have a much lower dark current generation than practically ... any other DSLR, it also has much lower amplifier glow. ... The read noise, for example is almost double the Canon ...
    (rec.photo.digital.slr-systems)