limit of dslr exposure times

[coatesg]

So, my take on this is that I try and follow the approach of ensuring that sky noise dominates read noise. Once the read noise is drowned out in the total noise term, increasing the exposure doesn't really help much over just taking lots of subs. 

In bright skies, and/or with broadband (as opposed to narrowband), this means that you can happily take a large number of relatively short subs and get effectively the same result as fewer longer subs. This helps in eliminating issues with airplanes, cosmic ray hits, etc. and hot pixels get sorted with dithering.

In very dark skies, or using narrowband filters, it means you may need much longer sub exposures - here's where you might struggle with tracking accuracy, etc. 

For ISO, I choose the lowest value where you move into the ISO invariant region. For the 350d I have, this is at 400ISO. The subs don't look as bright straight off the camera as at 800/1600, but the dynamic range is greater, and the S-N ratio doesn't further increase at higher ISO (but you'll clip bright objects more quickly). Post processing brings the data out. A good article is this one : http://dslr-astrophotography.com/iso-dslr-astrophotography/

Cheers

 

[Andywilliams]

51 minutes ago, tooth_dr said:

I have a modded 1000d and a modded 40d too! Great thread this 

my 40d is modded.....unfortunately my 1000d is not.

Im glad I asked the question because it has given forth some very valuable info......so thanks to you all for your knowledge and do keep it coming !! 

[tooth_dr]

If I'm using my canon for H-alpha, 10mins at ISO1600, should I be doing the same length at ISO800, and stretch it more in PS, or do I need to do 20mins at ISO800 to get equivalent?

[coatesg]

5 hours ago, tooth_dr said:

If I'm using my canon for H-alpha, 10mins at ISO1600, should I be doing the same length at ISO800, and stretch it more in PS, or do I need to do 20mins at ISO800 to get equivalent?

Which Canon?

[tooth_dr]

6 minutes ago, coatesg said:

Which Canon?

Either 40d or 1000d, both modded, dual rig.

[SteveNickolls]

I hope this link is allowed. It discusses ISO and worth reading, especially if you suffer light pollution.

Best Regards,
Steve

https://www.cloudynights.com/topic/500991-is-dslr-unity-gain-useful/

[Stub Mandrel]

On 10/01/2018 at 21:58, bobro said:

I don't know where the camera sensor is on the 1000D, but my camera doesn't register a temperature rise with long exposures.

If it's like the 450D (probable) it will be some way away from the sensor. I can have thick hoar frost on my cold finger but the sensor reports temperatures above zero.

Personally I judge by the quality of the subs! I have far better ways to ruin my subs than mere thermal noise...

[coatesg]

5 hours ago, tooth_dr said:

Either 40d or 1000d, both modded, dual rig.

So according to the graphs at http://dslr-astrophotography.com/iso-values-canon-cameras/, 1600 is in the ISO invariant region for both, so in theory the signal to noise is the same for both (but greater dynamic range at lower iso - note both cameras appear to have different recommended values...). But, you need sky noise >> read noise for this to all hold.  For narrowband, background sky is usually close to black, and so here you might choose to use high ISO (ie less read noise) to make sure this is true in a relatively shorter integration time. 

You'll probably have less issue with clipping highlights in narrowband, except possibly the very brightest stars, so lower dynamic range isn't so much of an issue as for broadband.

[tooth_dr]

52 minutes ago, coatesg said:

So according to the graphs at http://dslr-astrophotography.com/iso-values-canon-cameras/, 1600 is in the ISO invariant region for both, so in theory the signal to noise is the same for both (but greater dynamic range at lower iso - note both cameras appear to have different recommended values...). But, you need sky noise >> read noise for this to all hold.  For narrowband, background sky is usually close to black, and so here you might choose to use high ISO (ie less read noise) to make sure this is true in a relatively shorter integration time. 

You'll probably have less issue with clipping highlights in narrowband, except possibly the very brightest stars, so lower dynamic range isn't so much of an issue as for broadband.

Thanks for replying! I’m not sure I know what the iso invariant region is?  Sky noise? 

[Alien 13]

I am a bit lost with my camera given the figures in the link, I guess an ISO of 100 would be the best option...

Alan

[coatesg]

12 hours ago, tooth_dr said:

Thanks for replying! I’m not sure I know what the iso invariant region is?  Sky noise? 

The ISO invariant region is the ISO range where an increase in ISO value results in no increase in signal to noise ratio. In this region, you effectively get the same result in boosting a lower ISO shot in post processing, rather than shooting at a high ISO (It's to do with the downstream read noise getting swamped by the upstream read noise - see (eg): https://petapixel.com/2017/03/22/find-best-iso-astrophotography-dynamic-range-noise , https://photographylife.com/iso-invariance-explained). Some cameras are invariant over almost the whole range, some only go invariant above certain values.

All this ignores things like fixed pattern noise as well - it's very camera dependent, and most sites do say you should test to see what the best result is.

 

"Sky noise" is the shot noise associated with skyglow, light pollution, moonlight, etc. Generally, this goes as the square root of the number of incident photons. For narrowband, the background level is very low (lots of it blocked by the filter) and therefore the shot noise is also v low.

[coatesg]

12 hours ago, Alien 13 said:

I am a bit lost with my camera given the figures in the link, I guess an ISO of 100 would be the best option...

Alan

For long exposure work, I'd go with 200 on the 80d myself, but you need to test...

[tooth_dr]

1 minute ago, coatesg said:

The ISO invariant region is the ISO range where an increase in ISO value results in no increase in signal to noise ratio. In this region, you effectively get the same result in boosting a lower ISO shot in post processing, rather than shooting at a high ISO (It's to do with the downstream read noise getting swamped by the upstream read noise - see (eg): https://petapixel.com/2017/03/22/find-best-iso-astrophotography-dynamic-range-noise , https://photographylife.com/iso-invariance-explained). Some cameras are invariant over almost the whole range, some only go invariant above certain values.

All this ignores things like fixed pattern noise as well - it's very camera dependent, and most sites do say you should test to see what the best result is.

 

"Sky noise" is the shot noise associated with skyglow, light pollution, moonlight, etc. Generally, this goes as the square root of the number of incident photons. For narrowband, the background level is very low (lots of it blocked by the filter) and therefore the shot noise is also v low.

Thanks for clearing this up for me Graeme.  So basically I should only be using ISO800 with either camera?

[coatesg]

40 minutes ago, tooth_dr said:

Thanks for clearing this up for me Graeme.  So basically I should only be using ISO800 with either camera?

I'd start testing with 800 on the 40d. The 1000d isn't quite as clear cut - going by the DXOMark graph for DR vs ISO, I'd personally try 400 and 800 (https://www.dxomark.com/Cameras/Canon/EOS-1000D---Measurements). As with it all, you'll need to try it out and see if you get any adverse effects (banding, etc). 

This assumes broadband - you may find narrowband behaves differently and benefits more from the reduced read noise at higher ISO values.

[Stub Mandrel]

I read lots of online tests for the 450D, which show great variation in what is suggested as the optimum ISOP - I suspect because the tests are not done at the same temperature? The most frequent suggested ISO is 800.

My conclusion is that the best ISO may depend on individual sensors and your workflow etc. and the best thing to do is experiment.

I ended up putting some of my real world data through AP Lab (Astrophotography Lab). I found this less than intuitive to use and difficult to interpret, but nonetheless after some effort I concluded that IS0800 was, indeed, the best setting for my 450D camera.

 

[Stub Mandrel]

5 minutes ago, coatesg said:

This assumes broadband - you may find narrowband behaves differently and benefits more from the reduced read noise at higher ISO values.

Interesting observation.

[coatesg]

37 minutes ago, Stub Mandrel said:

Interesting observation.

I'm kinda basing that suggestion on the way that modern CMOS DSO imaging is going - high gain (-> low read noise) and shorter exposures.

I can't try this myself (unsuitable filters), but would be interested in the results - it might be different! 

The background is vlow though, therefore read noise has more impact on the total noise term than it would from "broadband in a not-very-dark location". Reducing this term may well prove beneficial to the overall S-N ratio of a subs of a given length, and if you can live with the dynamic range reduction and there's no banding/systematic errors induced. High ISO *might* be the way to go there...

[tooth_dr]

Thanks. I use 1600 for NB and 800 for colour currently. 

[Stub Mandrel]

Shifting to 1600 ISO is something I will try for Ha; looking at a single ISO 800 600s exposure of the Rosette Nebula the only sign of nearing clipping is the central pixel of the brightest star and the histopgram is well to the left so I don't think dynamic range would be a problem, even at 1200s exposures.

[coatesg]

Cool - Let us know how it goes!