Further to my earlier post, here are some posts made by Scott Rosen in a Yahoo group I'm part of....
Quote:
If you have a moment, read through the entire message between Roger Clark and me. While I was asking my questions of him, I made a point of explaining my best understanding of the theories involved, and I was quite happy to find that Roger confirmed that my general understanding was correct. As such, I think its a good guide to REALLY comprehending what we're trying to accomplish when capturing our photons.
As to the 100 x 1 minute versus the 10 x 10 minute - if we assume that 1 minute is long enough to bring the faint parts of the image above the read noise, then I don't particularly agree with the conclusion. Yes, the SNR in each of the 10 minute frames will be 3.16x higher than a corresponding 1 minute frame. However, there will only be 1/10th as many frames. Ergo, you'll correct for the lower SNR by stacking 10 times as many low SNR 1 minute frames resulting in a 3.16x restoration in SNR.
Now, real world, 1 minute subs are often not long enough to bring a separation between the left axis of your histogram and the data mountain (even at a high ISO). This may be the case where you have a reasonable amount of light pollution, but this wouldn't be so in most dark skies. So, there may very well be some difference between 100 x 1 minute and 10 x 10 minute.
I shoot from a pretty dark sky (SQM typically around 21.4). I usually shoot my RGB images at ISO 1600 and 3 to 5 minute subs. This gives me a little bit of separation between the left axis of the histogram and the data mountain - right around 15%. I choose these relatively short subs (particularly for a dark sky) because they give me the histogram I want. Any higher, and I just risk more clipping of my highlights. Additionally, if I have relatively short 5 minute subs, each of them is 1/2 as likely as a 10 minute sub to be ruined by an airplane flying through the FOV, or some guiding issue, etc. This way, when the airplane strikes, I've only wasted 5 minutes of my dark time and not 10 minutes. Moreover, I have 5 minutes more of photons to improve my SNR.
I've probably posted this image as a demonstration more times than it deserves, but here is an image of M101 where I captured 7 hours of 45 second subs (I hadn't retrofitted my mount for autoguiding yet):
http://www.astronomersdoitinthedark....hp?c=157&p=409
Now, I can assure you that my histogram for this image was pegged against the left axis. This is NOT optimal. But, the image still has surprisingly good SNR in the faint parts of the galaxy. My 9 hour image taken a year later with 3 minute subs doesn't look that much better:
http://www.astronomersdoitinthedark....hp?c=113&p=438
Keep in mind, too, that the latter image had the benefit of being reprocessed using the DSLR-LLRGB workflow, whereas the earlier image did not. This probably accounts for more of the improvement in chromatic noise than the longer subs would.
Now, I don't recommend 45 second subs (unless that is your unguided limit), but I think this helps to show that we may be giving too much credit to long subs. Long subexposures DO look better individually than their short sub equivalents, but their aggregate stacks don't look that much different.
|
and....
Quote:
If you want to skip the back and forth and find the crux of the matter, look at the end of the message for the last paragraph or two where it says:
>
>
>
> --- BEGIN QUOTATION ---
>
> >
>
> > But, at the shadow end of an image, am I correct that the rule
>
> > of thumb to have the back of the camera histogram showing
>
> > a separation of the data "mountain" from the left axis of
>
> > 15% to 50% along the axis is based upon the principle of
>
> > ensuring that you've separated the faintest parts of the image
>
> > from the read noise of the camera?
>
>
>
> Certainly separation, but once separated, it matters little if at 15%
>
> versus 50% as the sensor is linear.
>
>
>
> --- END QUOTATION ---
>
>
>
> The important point here is that there's no great value in taking longer exposures once you're able to separate the faintest parts of the image from the read noise (i.e., once you have some separation of your histogram from the left axis). In fact, one of the down sides to longer exposures once you achieve this minimum separation is that you're going to be at risk of clipping more of your highlights (typically the cores of the bright stars).
|
but....
Quote:
|
Now, if you were trying to capture narrowband data (i.e., Ha), you'll probably need the longer sub length and higher ISO.
|