Go Back   IceInSpace > Equipment > Astrophotography and Imaging Equipment and Discussions

Reply
 
Thread Tools Rating: Thread Rating: 3 votes, 5.00 average.
  #1  
Old 02-09-2014, 07:43 AM
Peter.M's Avatar
Peter.M
Registered User

Peter.M is offline
 
Join Date: Sep 2011
Location: Adelaide
Posts: 947
shot noise, and long exposures

Hi guys

I know this has been discussed by Ray quite a bit and I see his argument that when shot noise dominates an image exposing for longer has no benifit. Logically thinking about this at work at the moment and I think that it will still be better, here is the reason.

A long exposure that is already dominated by shot noise will be averaging the shot noise if it exposes longer. This would mean that stopping the exposure and starting another so you can stack more images to reduce shot noise would effectively already be happening in a longer exposure.

If I have this wrong I'm happy to be told so, just trying to logically understand what's going on.
Reply With Quote
  #2  
Old 02-09-2014, 08:12 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Peter.M View Post
Hi guys

I know this has been discussed by Ray quite a bit and I see his argument that when shot noise dominates an image exposing for longer has no benifit. Logically thinking about this at work at the moment and I think that it will still be better, here is the reason.

A long exposure that is already dominated by shot noise will be averaging the shot noise if it exposes longer. This would mean that stopping the exposure and starting another so you can stack more images to reduce shot noise would effectively already be happening in a longer exposure.

If I have this wrong I'm happy to be told so, just trying to logically understand what's going on.
Hi Pete. Sorry, I guess I have rabbitted on a bit

I think we agree, but....

If by "exposing for longer" you mean "exposing subs for longer", there is no value in longer subs. For example there is no SNR difference between a 20 minute sub and the average of 2x10 minute subs if both are "sky limited". The differences are that more stars will be saturated in the 20 minute sub, (so you have lost dynamic range) and the chance of a dud sub is higher (eg due wind gust or clouds). The issue is not that long subs are inferior from an SNR perspective, just that once you have subs that are long enough to be "sky limited", the SNR after stacking is the same from short or long subs - but you will lose in other areas by going longer.

If by "exposing for longer" you meant overall exposure, there is obvious benefit from going longer.

Last edited by Shiraz; 02-09-2014 at 04:52 PM.
Reply With Quote
  #3  
Old 02-09-2014, 08:25 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
There's also a benefit in having a larger number of subs for the rejection algorithms to chew on. When you only have a small number of measurements it is more difficult to detect outliers.

Cheers,
Rick.
Reply With Quote
  #4  
Old 02-09-2014, 08:46 AM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by RickS View Post
There's also a benefit in having a larger number of subs for the rejection algorithms to chew on. When you only have a small number of measurements it is more difficult to detect outliers.

Cheers,
Rick.
guess it is also better for drizzle to have more subs.

Last edited by Shiraz; 02-09-2014 at 09:51 AM.
Reply With Quote
  #5  
Old 02-09-2014, 10:33 AM
Peter.M's Avatar
Peter.M
Registered User

Peter.M is offline
 
Join Date: Sep 2011
Location: Adelaide
Posts: 947
Quote:
Originally Posted by Shiraz View Post
Hi Pete. Sorry, I guess I have rabbited on a bit

I think we agree, but....

If by "exposing for longer" you mean "exposing subs for longer", there is no value in longer subs. For example there is no SNR difference between a 20 minute sub and the average of 2x10 minute subs if both are "sky limited". The differences are that more stars will be saturated in the 20 minute sub, (so you have lost dynamic range) and the chance of a dud sub is higher (eg due wind gust or clouds). The issue is not that long subs are inferior from an SNR perspective, just that once you have subs that are long enough to be "sky limited", the SNR after stacking is the same from short or long subs - but you will lose in other areas by going longer.

If by "exposing for longer" you meant overall exposure, there is obvious benefit from going longer.

I did mean exposing each subframe for Longer.
Reply With Quote
  #6  
Old 02-09-2014, 11:22 AM
Amaranthus's Avatar
Amaranthus (Barry)
Thylacinus stargazoculus

Amaranthus is offline
 
Join Date: Jan 2014
Location: Judbury, Tasmania
Posts: 1,203
Another benefit of lots of subs is if your image train is undersampled -- such that you need to Drizzle. More subs = better drizzling performance (and a finer drizzle mist is possible, up to the point where you arrive at what you consider to the the optimal image scale). So there is a real balance to be struck. Because I NEED to drizzle my current setup, I tend to expose a sub for as long as is required, but no longer
Reply With Quote
  #7  
Old 02-09-2014, 12:07 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Peter.M View Post
I did mean exposing each subframe for Longer.
then I think that we agree - If you integrate more within a sub (by making it longer), the SNR result after stacking a few of them is the same as integrating by stacking more shorter subs - but you get other advantages from shorter subs (that is if short and long subs are all sky limited). What matters is the total number of photons detected, not how they were collected.

Does that sound reasonable?
Reply With Quote
  #8  
Old 02-09-2014, 12:18 PM
Amaranthus's Avatar
Amaranthus (Barry)
Thylacinus stargazoculus

Amaranthus is offline
 
Join Date: Jan 2014
Location: Judbury, Tasmania
Posts: 1,203
There is another issue to consider, that I think Luc Coffier explains well:

Quote:
Are 100 x 1 minute and 10x10 minutes giving the same result?
Yes when considering the SNR but definitely No when considering the final result.

The difference between a 10 minutes exposure and a 1 minute exposure is that the SNR in the 10 minutes exposure is 3.16 higher than in 1 minute exposure.

Thus you will get the same SNR if you combine 10 light frames of 10 minutes or 100 light frames of 1 minute. However you will probably not have the same signal (the interesting part). Simply put you will only get a signal if your exposure is long enough to catch some photons on most of the light frames so that the signal is not considered as noise.

For example for a very faint nebula you might get a few photons every 10 minutes. If you are using 10 minutes exposures, you will have captured photons on each of your light frames and when combined the signal will be strong.
If you are using 1 minute exposures you will capture photons only for some of your light frames and when combined the photons will be considered as noise since they are not in most of the light frames.
So although you capture the same number of those precious rare photons in each case, unless you are careful, they can be averaged (kappa sigma clipped, etc.) out, as noise. So exposure depth is definitely required. As with most things in AP, there is not a firm rule!

Also, obviously, the read noise in the multiple shorter frames is a larger fraction of the total noise, but I think that is already assumed knowledge in this discussion.
Reply With Quote
  #9  
Old 02-09-2014, 01:40 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
I don't think that argument applies here Barry.

The basis of the idea seems to be that if there is a pixel with zero target electrons in most frames, any frames where that pixel includes one (or more) target electrons will be rejected by the stacking software because it is different - so you lose data.

However, with current sensors, no pixel will ever have zero or fixed signal - there will always be read noise, dark noise and sky noise electrons that produce frame-to-frame variations in pixel signal (of maybe ~10 electrons on average) that completely mask variations of 1 or 2 electron due to the occasional presence of a target photon. Thus, as I see it, you cannot lose signal because there is no way to design software that can distinguish between frames with and without a couple of photons of target signal (how nice would that be!). When the data is stacked, the noise will be uncorrelated from frame to frame and will integrate out, but the signal will be correlated and will accumulate - even if there is none in some frames there will be some in others.

Should we ever get sensors with zero read and dark noise and image through extremely narrow band filters, it may be that the argument applies - but not currently.

As you noted, Pete indicated that the discussion applies to shot noise limited conditions, so read noise is assumed to be an insignificant contribution to total noise.

Last edited by Shiraz; 02-09-2014 at 01:57 PM.
Reply With Quote
  #10  
Old 02-09-2014, 01:55 PM
Amaranthus's Avatar
Amaranthus (Barry)
Thylacinus stargazoculus

Amaranthus is offline
 
Join Date: Jan 2014
Location: Judbury, Tasmania
Posts: 1,203
That's an excellent point, Ray, thanks for the input.
Reply With Quote
  #11  
Old 02-09-2014, 02:16 PM
tonybarry's Avatar
tonybarry (Tony)
Registered User

tonybarry is offline
 
Join Date: Aug 2007
Location: Penrith, Sydney
Posts: 556
Shot noise refers to the random rate of influx of photons (e.g. from a dim star). The dimmer the star, the more the rate of incoming photons goes up and down due to random factors.

So for an exposure which nets you 10 photons, shot noise on one exposure might put that up to 12 the next time, or down to eight the time after, say a twenty percent change.

For an exposure which nets you 10,000 photons, shot noise will put that up to 10,002 ... which is close to zero percent.

To drown out shot noise, collect more photons.

Shot noise is averaged out across multiple exposures too. But then you introduce read noise, where each exposure, as it comes off the chip, has noise added by the CCD amplifier. So you have to balance the length of time for each sub (more = better), number of reads (less = better) with the quality of the mount and the guiding (longer subs require better mounts).

Hope this helps.

Regards,
Tony Barry
Reply With Quote
  #12  
Old 02-09-2014, 02:54 PM
Camelopardalis's Avatar
Camelopardalis (Dunk)
Drifting from the pole

Camelopardalis is offline
 
Join Date: Feb 2013
Location: Brisbane
Posts: 5,425
So...longer subs in a perfect camera but not so much in reality?

How does one estimate the ideal sub for SNR for a given setup? Or is it a factor of the target, pixel saturation and all that? I mean, I guess I'm asking how to get in the ballpark of optimal SNR when out in the field?

Apologies to the OP for the mini hijack.
Reply With Quote
  #13  
Old 02-09-2014, 03:04 PM
Geoff45's Avatar
Geoff45 (Geoff)
PI rules

Geoff45 is offline
 
Join Date: Jul 2006
Location: Sydney
Posts: 2,631
Quote:
Originally Posted by tonybarry View Post

So for an exposure which nets you 10 photons, shot noise on one exposure might put that up to 12 the next time, or down to eight the time after, say a twenty percent change.

For an exposure which nets you 10,000 photons, shot noise will put that up to 10,002 ... which is close to zero percent.
Regards,
Tony Barry
Shot noise depends on the photon influx in a simple way. If your average signal is N photons, the shot noise is sqrt(N) photons. When you gather more photons, the shot noise increases, but the important fact is that the signal increases faster, so the signal to noise ratio goes up. Gather on average 100 photons per sub and the shot noise will be 10 photons (square root of 100). The SNR will be 100/10=10. Gather on average 10000 photons per sub andthe shot noise goes up to 100 photons. The SNR will be 10000/100=100. Your SNR is 10 times better with the longer sub.
Just for the record: if the "true" signal is N photons, then saying the shot noise is sqrt(N) means that you can expect 68% of your subs to have values between N+-sqrt(N) and 95% of your subs to have values between N+-2*sqrt(N).
Geoff

Last edited by Geoff45; 02-09-2014 at 04:29 PM.
Reply With Quote
  #14  
Old 02-09-2014, 05:22 PM
avandonk's Avatar
avandonk
avandonk

avandonk is offline
 
Join Date: Aug 2005
Location: Melbourne
Posts: 4,786
The take home message is to collect as fast as possible.

Get a fast scope! And a detector with maximum signal to noise!

End of argument!

Bert
Reply With Quote
  #15  
Old 02-09-2014, 07:10 PM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Quote:
Originally Posted by avandonk View Post
The take home message is to collect as fast as possible.

Get a fast scope! And a detector with maximum signal to noise!

End of argument!

Bert
Not everyone can get a fast scope and a suitable detector, a lot of us have to make do with what we have.
making optimum use of what we have from where we can image is what one can do in most cases.
Speaking for myself, I'm still trying to learn and understand effects of noise and relation to exposure length. Watching with interest.

Alistair
Reply With Quote
  #16  
Old 02-09-2014, 11:09 PM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by avandonk View Post
The take home message is to collect as fast as possible.

Get a fast scope! And a detector with maximum signal to noise!

End of argument!

Bert
Some people care about image scale as well as imaging speed...
Reply With Quote
  #17  
Old 03-09-2014, 09:10 AM
strongmanmike's Avatar
strongmanmike (Michael)
Highest Observatory in Oz

strongmanmike is offline
 
Join Date: May 2006
Location: Canberra
Posts: 17,150
Always interesting reading these threads

So to put it simply and apart from taking up less disc space and putting the tracking and unexpected bump/wind/plane/satellite problems etc (which are indeed significant considerations) aside...under what conditions and with what sort of equipment is doing a single 30min sub a significantly better option than doing 6 X 5min subs?

Reply With Quote
  #18  
Old 03-09-2014, 09:28 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by strongmanmike View Post
...under what conditions and with what sort of equipment is doing a single 30min sub a significantly better option than doing 6 X 5min subs?

A good example would be a narrowband sub with a sensor like the KAF-16803. The signal (including sky glow) will be relatively small and in a 5 minute sub the read noise from the camera will swamp the shot noise from the target (and you'll incur that read noise 6 times.) A 30 minute sub will have much better SNR than a stack of 6 x 5 mins.
Reply With Quote
  #19  
Old 03-09-2014, 10:01 AM
alistairsam's Avatar
alistairsam
Registered User

alistairsam is offline
 
Join Date: Nov 2009
Location: Box Hill North, Vic
Posts: 1,837
Quote:
Originally Posted by tonybarry View Post
Shot noise refers to the random rate of influx of photons (e.g. from a dim star). The dimmer the star, the more the rate of incoming photons goes up and down due to random factors.
so Shot noise just refers to the rate of incoming photons or variation in the rate, more like jitter (in networking) and has nothing to do with sky background?? why is it referred to as "noise"? sorry to detract with the basics..
Reply With Quote
  #20  
Old 03-09-2014, 10:04 AM
Amaranthus's Avatar
Amaranthus (Barry)
Thylacinus stargazoculus

Amaranthus is offline
 
Join Date: Jan 2014
Location: Judbury, Tasmania
Posts: 1,203
Noise = variance (in this context!), signal = mean rate of photons

I say in this context because in the work I do, we separate process error (inherent variation stemming from stochasticity of dynamic systems) from observation/measurement error - the former is scientifically interesting, the latter is an annoyance that we seek to minimize,
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 04:50 AM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Testar
Advertisement
Bintel
Advertisement