Whats the best ISO for Astro DSLRs?

[Bassnut]

I understand the ISO selection on a DSLR simply selects amplifier gain. From the results ive seen, I cant see any reason to not use the highest ISO available (on the 300, 350 and 400D say, 1600iso) except for short exposures on high brightness objects eg M42 to layer on deeper exposures to stop core blowout. Deep space exposures always require streching anyway, so why add noise streching lower ISO settings when higher ISOs need less of that with higher gain cam settings?.

From the pics at 1600ISO ive seen, the extra streching at a lower ISOs just adds more noise.

[joshman]

i would recommend an ISO of 400 or 800

as i understand it, your signal to noise ratio gets worse as you bump up the ISO. you get heaps more noise for very little signal increase. ISO1600 is useful for testing settings like focus and alignment, but i wouldn't use it for imaging.


of course the best way to do it is to shoot RAW and mess with all that later.

[vash]

I mainly shoot with 1600 but thats only because I couldn't get enough exposure time so I needed all the help I could get, now that I have autoguiding I'm most likely going to use 800 at first till I get a bit more experienced then maybe drop down again.

I've shot some objects with 1600 and used 32 exposures and they where still noisy, I'm sure that the same number of exposures with 800 and a longer time would yield better results but I haven't tried it yet.

[h0ughy]

i can get away with 1600 because my camera is cooled but get better results from iso 800 and 400

[acropolite]

This article (http://www.northlight-images.co.uk/article_pages/cameras/canon_1ds3_noise.html)on partial ISO stops may be of interest. It seems that ISO's 100,200,400,800 etc are implimented in hardware, those in between are software created and won't necessarily deliver optimal noise performance.

[leon]

From my experience, the best ISO for my Particular camera the Canon 5D is 500, for wide field images, just take more subs.

Leon

[Bassnut]

Interesting range of experiences, ive only used 1600. My point was I guess, that in the end any way to increase signal, also increases noise, its just which way gives the best S/N ratio. Lower ISOs need longer exposures (or more subs) or more stretching, but high ISOs are noisier because of higher amp gain. It seems perhaps it depends on the cam, and object brightness (or Houghys cooled cam, no brainer ;-).

I still think that for the dimmest objects 1600 overall would be necessary, especially if exposures need to get much past 5 mins at lower ISOs, given thermal noise would be getting overwhelming.

[turbo_pascale]

Since this post went up, i've been digging around where I saw some figures. I'm not sure if it was on Christian Buill's site (I can't find it there now), but for the various recent flavours of Canon, there was a table with the optimal ISO for each. I know that for the 350D, it was 800.

I thought I may have found reference to this table here on IIS, but again, my searching has come up empty.

The figures were derived in a very scientific manner IIRC, based on the chip sensitivity and amplification characteristics. Sorry I can't find reference to it.

Turbo

[citivolus]

This may be what you were after:

http://www.astrosurf.com/buil/eos40d/test.htm

There is a table about half way down that shows base readout noise for each ISO setting on the 40D. All things being equal, the noise was lowest when divided by gain at ISO 1600, unless I am misinterpreting the data.

Eric

[Alchemy]

i use iso 800 and exposures of 6 mins , this has been brought about by trial and error but it works for me. brighter objects less of course,

i note that very few dslr exposure runs go beyond about an hour, that might be a limiting factor, diminishing returns perhaps.

i think the really faint objects are in the realm of cooled ccd cameras that can do longer low noise exposures , you can do them with a dslr ... but not a superb job, for the bright objects i reckon dslr is just fine.

[Bassnut]

Eric, that link is very interesting, it seems the 14bit dynamic range and lower noise of the 40D sure makes it look attractive for Astro. From what I can tell tho, it seems a shame he didnt show/test an image at 1600 ISO.

The 16bit dynamic range, high QE and low noise of cooled astro CCD cams is what makes them de riguer for serious astro, but a cooled moded 14bit 40D, despite the relatively low QE, would make them serious contenders for astro pics against CCDs for a lot less money, especially given the hi res. Res like that cost a bomb in CCDs. We live in interesting times, the next gen hi res 16bit CMOS DSLRs will be very attractive, if they can lower the noise even more (and without cooling, thats hard with CCDs) and up the QE. Cooled multi mega pixel CCDs arnt that flash in the QE department anyway (30% typical).

[Bassnut]

Alchemy, 1hr expoures?, umm, uncooled?, are you sure?, I havent seen much past 10-15 mins. 1hr is pushing it with astro CCDs ;-).

Yes, from what ive done and seen, 6 mins is about optimum for a DSLR, 10min at a push at lower ISOs.