Dynamic Range - The Maths

[Scooot]

I understand the gist of the noise & stacking issues although I haven’t studied all the above calculations. 

Back on my Canon 450D, at ISO 100 saturation/FWC is 26614e with read noise of 19.4. As it’s 14 bit my ADC is limited to 16384 levels. so to record the lower signals we need to round up 1.63 adu /e on my camera instead of 5 on yours. As read noise at that ISO is quite high, am I right in assuming this probably won’t be an issue, ie easily reached with random noise. If so ISO 100 could be useful if light levels allow, say on the moon, and I’d benefit from the better SNR at this ISO setting. Or have I totally misunderstood  

[vlaiv]

1 hour ago, Scooot said:

I understand the gist of the noise & stacking issues although I haven’t studied all the above calculations. 

Back on my Canon 450D, at ISO 100 saturation/FWC is 26614e with read noise of 19.4. As it’s 14 bit my ADC is limited to 16384 levels. so to record the lower signals we need to round up 1.63 adu /e on my camera instead of 5 on yours. As read noise at that ISO is quite high, am I right in assuming this probably won’t be an issue, ie easily reached with random noise. If so ISO 100 could be useful if light levels allow, say on the moon, and I’d benefit from the better SNR at this ISO setting. Or have I totally misunderstood  

Just looked up Canon 450D specs, and I'm not sure where you got those stats, this page

http://www.astrosurf.com/caron/canon_eos_perf.html

Shows different measured values to those that you are using. Maybe check it out?

As for usefulness of high well depth, that would depend on the way you shoot the moon. If you are going for a single exposure, then yes, maximizing SNR is good approach. Otherwise, if you are using scope and multiple exposures, then minimizing read noise is the way to go. If you want to freeze the seeing - you will need to use very short exposures, order of 5 ms. With any kind of serious magnification - moon image will be spread over large number of pixels and with such short exposure you simply will not get the chance to exploit 14bit ADC as level of signal per frame is going to be about 9 to 10 bits (probably less than 1000 electrons per pixel in single exposure)

[Scooot]

Here

https://web.archive.org/web/20180830114507/http://sensorgen.info/CanonEOS-450D.html

 

[Scooot]

21 minutes ago, vlaiv said:

Just looked up Canon 450D specs, and I'm not sure where you got those stats, this page

http://www.astrosurf.com/caron/canon_eos_perf.html

Shows different measured values to those that you are using. Maybe check it out?

As for usefulness of high well depth, that would depend on the way you shoot the moon. If you are going for a single exposure, then yes, maximizing SNR is good approach. Otherwise, if you are using scope and multiple exposures, then minimizing read noise is the way to go. If you want to freeze the seeing - you will need to use very short exposures, order of 5 ms. With any kind of serious magnification - moon image will be spread over large number of pixels and with such short exposure you simply will not get the chance to exploit 14bit ADC as level of signal per frame is going to be about 9 to 10 bits (probably less than 1000 electrons per pixel in single exposure)

Usually around 1/30 second is the right exposure at ISO 100 depending on its phase. ( at about F5)  I have a stack of about 20 I’ll have a look at it in PixInsight to see if I can tell how much of the ADC  it used.

[vlaiv]

1 minute ago, Scooot said:

Usually around 1/30 second is the right exposure at ISO 100 depending on its phase. I have a stack of about 20 I’ll have a look at it in PixInsight to see if I can tell how much of the ADC  it used.

1/30th of a second is about 33ms, and if you expose for 75% of histogram, with full well being 26614 that would mean that bright parts receive about 19960e. If you expose for 5ms, that would be about 6 times less, or ~3326e - 12bit range.  I'm guessing you are not using much of focal length to shoot it?

 

[Scooot]

22 minutes ago, vlaiv said:

1/30th of a second is about 33ms, and if you expose for 75% of histogram, with full well being 26614 that would mean that bright parts receive about 19960e. If you expose for 5ms, that would be about 6 times less, or ~3326e - 12bit range.  I'm guessing you are not using much of focal length to shoot it?

 

This is the one I was thinking of, it was actually a stack of 50ish , 1/20th second, at F10, focal length 600mm. Can’t remember where the histogram was but I suspect it was very near the left edge. I’m going to use my 10” dob on it’s eq platform next time, f5ish 1380 focal length. 

 

[Scooot]

Out of interest I put a single image and the stack in subframe selector. SNR is so good so ISO100 or 400 is probably academic and not noticeable. I always think the colours are better at the lower ISO but as the dynamic range is similar that’s probably just my imagination.