[CometGuy]

Originally Posted by Peter Ward

Alas, there are many sources of noise other than read noise.
Janesick (2001) notes the following on just thermal noise

depletion dark current
diffusion dark current
substrate dark current
surface dark current
backside dark current
dark shot noise
dark current non-uniformity
dark current spikes

Of the list of 8 above, 6 are sources of dark current which in turn generates dark shot noise and non-uniformity noise.

But all of these are time dependent, not readout dependent. It doesn't matter how we achieve that 600 second exposure, whether it is 1 single 600 second sub or 600 x 1 second subs - dark noise is the same. The only assumption here is that the total time the sensor is reading out is the same.

Despite zero read noise, an L3CCD sensor can't differentiate a thermally liberated electron from one generated by a photon

I might be missing something here, but how is this different to a conventional CCD? Why would you not use dark frame calibration like you would now?

, hence close attention to dark current minimization is essential.

Even in the darkest skies on earth, with narrowband filters, there is sky background illumination that generates photon shot noise. This will exceed the dark noise on a Kodak CCD cooled to say -40C.

There are other on chip noise sources

spurious charge
residual surface image
residual bulk image
clocking luminescence
pixel luminescence
diode luminescence
cosmic rays & radiation
excess charge
blem spillover
seam noise
cosmetic defects

Can you tell me which of these are readout dependent?. Lets take one - "cosmic rays & radiation" of which the effects are time dependent, and not dependent of number of times readout. In any case I would rather have a single short sub spoiled by a cosmic ray strike than a long exposure.

There are also off chip sources, eg light leak, pre-amp noise, ADC quantizing, clock jitter & EMF.

One of the advantages of L3CCD and "electron multiplication" ltechnologies ike Black Silicon is that they are actually far less effected by off chip sources.

While read noise may be minimal with L3CCD there are other technologies that are equally promising eg Clipper chips that also have sub electron read noise.

I am sure there are...I will have to read.

There is no doubt noise reduction is a good thing, but with very faint objects, the signal just isn't arriving quick enough for the above noise sources to be non trivial.....it would seem better practice to me to gather decidedly more flux (ie long exposure) than constantly run the gamut of noise sources with multiple short subs.

Agree, but most of the noise sources you listed are time dependent and not readout dependent. Correct me if I am wrong.