Can anyone give me some advice regarding minimising the amount of noise that i am getting (see unprocessed test pic): ISO 1600, 2min, Canon 400D. Is this temperature related? should i be reducing the ISO??
Noise is definitely proportional to temperature. Not much you can do if it's hot weather. Wait until the early hours of the morning, try a fan, get a cooler box...
Reducing iso is one way. Cooling is another and more effective. Dark noise doubles with every 6C rise in temperature. Conversly reducing temperature.
Signal to noise ratio can be improved by selecting a suitable iso and exposure time. It could be that the exposure is too short, but, at iso1600 , it's probably not, so temperature is a good guess. Try iso800 and 3 - 3.5 and see what it looks like. Get more signal into the image.
I found from testing with my 400D that signal to noise ratio will be maximised at ISO1600, with the sub exposure time then chosen based on the temperature (see attached, which shows the SNR of different settings all relative to ISO800 / 5 mins @ 20C). I then take LOTS and LOTS of subs (I now aim for at least 100+ subs). Suggest that you dither between subs as well if you can.
Richard your data is interesting. I must try this out. My only concern is that SNR does not necessarily equate to a properly exposed image or does it? At iso800, 5 mins,under my sky, detail is obliterated and the image over exposed/bright. 3 - 3.5 is much better. Signal is signal, whether overdone or not, it contributes to SNR. Dynamic range is also a consideration and is inversely proportional to iso - is it significant between 400 and 1600? Does it matter? Are slower integration times better under polluted skies?
A fix for over exposed areas is combining shorter subs, which means adding data with significantly lower SNR - M42 trapezium.
Intuitively I would have said that the sky noise (such as from light pollution) is fixed for a given location, and so signal and sky noise will be captured together regardless of the sensor gain. Dynamic range will obviously be reduced as you increase gain (ISO), but I would have thought that provided the sensor is not being clipped at the sub time chosen you could use a higher ISO (such as ISO1600 as suggested by my plots)...?
Intuitively, yes, agreed, the difference appears to be the rate/gain at which photons are accumumlated. That being true, detail and SNR is a compromise, if overexposure is signal.
In that case my technique has been to preserve detail and integrate slower for increased dynamic range. Regulated DSLR cooling adds another layer of noise reduction and it is tempting to increase exposure time or gain to increase SNR - sky and object brightness are limiting.
Hi Richard,
Just curious about the temperatures on your graphs, are these ambient temperatures or is your cam also modded with cooling so it's the sensor temperature as reported in the exif data?
Cheers.
Last edited by MrB; 03-02-2014 at 02:32 PM.
Reason: damn Nexus autocorrect....
