Optimal exposure length vs number of subs? (DSLR)

[MartinFransson]

I´m wondering how to best plan my imaging runs. What bothers me right now is the balance between exposure length and number of subs. What is the best way?

Let´s say I have 2 hours worth of imaging with an uncooled DSLR - would it be better to get 24 5-minute subs or 15 8-minute subs? The 5 minute ones reach about 1/3 of the way from the left on the histogram, the 8 minute ones reach almost half way.

See for example my 5 minute histogram. Would it be worth going for 8 minute ones despite getting fewer and probably raising the camera temps?

 

[MartinFransson]

I might add that LP is not the limiting factor. When imaging near zenith or to the north I can go 10 minutes at f/4 and ISO 800. Now I am imaging at f/5.

This is what one of the 5 minute subs looks like in BackYardEOS, if that´s any clue. It´s the one with the histogram from my first post.

[ollypenrice]

My real answer would be 'experiment to find out.' You will find theorists arguing both ways - for shorter or for longer. You'll even find people asserting that 100x1 minute = 10x10 minutes. It certainly doesn't do any such thing from our dark site using our deep-welled CCD cameras. However, as read noise goes ever downwards on modern DSLR sensors the results from many short subs are getting better and better, to the extent that this approach may become the best.

I honestly think that you have to try things to find out what works. Tonight I am shooting 30 minute Ha subs on a very faint target. I would be far more inclined to bump that up to an hour than to drop it to 15 minutes, but that is based on experiment.

Olly

[MARS1960]

I was pondering the same question a few weeks back so i did some research and came across this excellent website.

http://www.clarkvision.com/articles/astrophotography.and.exposure/

[Adam J]

No simple answer really, will depend on many factors including the ambient temperature and how faint the target is.

 

[StuartJPP]

There have been many discussions about this and the conclusion is...there is no conclusion. I prefer to do 5 minute subs and choose an ISO that balances the histogram, usually ISO, 400, 800 or 1600 with 800 being my preferred choice for my setup.

It would take one night for someone who lives under constant clear skies to put this to the test by taking a series of shots. For instance by alternating something like 20x30 second subs, 5x60 second subs and 1x300 second subs for the entire night on the same kit and provide this information to the community.

However like Olly says, things are changing all the time and it is better to try it out for yourself to see what you prefer.

[sharkmelley]

On ‎27‎/‎08‎/‎2016 at 21:55, MartinFransson said:

I might add that LP is not the limiting factor.

Can you be sure?  What are your figures for the noise contributions separately from light pollution, read noise and thermal noise?

Mark

[Helen]

I'm currently doing a project trying to measure faint comets and have to consider the whole noise issue.  I've just read this and thought it was helpful

http://www.astrometrica.at/papers/PointSources.pdf

Helen

[MartinFransson]

6 hours ago, sharkmelley said:

Can you be sure?  What are your figures for the noise contributions separately from light pollution, read noise and thermal noise?

Mark

No, you´re right. I can not be sure of this. What I meant is that I can go for long (10-12 minutes at ISO 800) exposures without saturating the histogram due to LP. I have no idea of the separate noise contributions. Is there a way to find that out? I´m guessing thermal noise is the worst since I´m running uncooled.

[sharkmelley]

2 hours ago, MartinFransson said:

No, you´re right. I can not be sure of this. What I meant is that I can go for long (10-12 minutes at ISO 800) exposures without saturating the histogram due to LP. I have no idea of the separate noise contributions. Is there a way to find that out? I´m guessing thermal noise is the worst since I´m running uncooled.

I would keep to the shorter (5 minute) subs because the noise from light pollution and dark current will almost certainly be swamping the read noise and, if so, there is nothing to be gained by increasing the exposure time.

I did write a PixInsight script for calculating these parameters for a DSLR:  http://www.cloudynights.com/topic/504543-pixinsight-script-for-calculating-dslr-sensor-characteristics/?p=6677614

It requires pairs of bias and flat frames plus a pair (or sequence of) dark frames and then a pair of light frames (of a fairly blank area of night sky).  At some stage I should turn this into a stand alone application without the dependency on PixInsight.

Mark

[MartinFransson]

8 hours ago, sharkmelley said:

I would keep to the shorter (5 minute) subs because the noise from light pollution and dark current will almost certainly be swamping the read noise and, if so, there is nothing to be gained by increasing the exposure time.

I did write a PixInsight script for calculating these parameters for a DSLR:  http://www.cloudynights.com/topic/504543-pixinsight-script-for-calculating-dslr-sensor-characteristics/?p=6677614

It requires pairs of bias and flat frames plus a pair (or sequence of) dark frames and then a pair of light frames (of a fairly blank area of night sky).  At some stage I should turn this into a stand alone application without the dependency on PixInsight.

Mark

I tried it on a few of my 300s subs from the other night but can not interpret what the values say:

=============================

DSLR Sensor Parameters v0.0.7

=============================

Camera: Canon EOS 550D

ISO speed: 800

Image Size: 5202 x 3465

Bias frames exposure time: 1/4098 sec

Flat frames exposure time: 1/49 sec

Dark frames exposure time: 300 sec

Light frames exposure time: 300 sec

Channel order: []

Gain per channel(e/ADU): [ 0.28, 0.279, 0.28, 0.277 ]

ISO for unit gain: [ 224, 223, 224, 221 ]

Read Noise(e): [ 3, 2.99, 3.01, 3 ]

Dark Current (e/pixel/sec) [ 0.0694, 0.0706, 0.0692, 0.0697 ]

Dark Current (e/pixel/sec) [ 0.0696, 0.0713, 0.0718, 0.0714 ]

Dark Current (e/pixel/sec) [ 0.0692, 0.0717, 0.0718, 0.0721 ]

Dark Current (e/pixel/sec) [ 0.0664, 0.0703, 0.0675, 0.0697 ]

Dark Current (e/pixel/sec) [ 0.0698, 0.0727, 0.0698, 0.0725 ]

Dark Current (e/pixel/sec) [ 0.0708, 0.0729, 0.0713, 0.0734 ]

Dark Current (e/pixel/sec) [ 0.0727, 0.0755, 0.0735, 0.0756 ]

 

Noise estimates for 300sec exposure

Read Noise(e): [ 3, 2.99, 3.01, 3 ]

Thermal Noise(e): [ 4.67, 4.76, 4.7, 4.76 ]

Skyfog Noise(e): [ 9.76, 8.76, 6.76, 9.74 ]

 

Mean values in ADU

Bias Frame 1: [2047.7, 2048.4, 2048.3, 2049 ]

Bias Frame 2: [2047.8, 2048.4, 2048.2, 2048.8 ]

Flat Frame 1: [3815.2, 2933.3, 3584.8, 3819.6 ]

Flat Frame 2: [3811.8, 2931.6, 3581.8, 3816.5 ]

Dark Frame 1: [ 2046.8, 2048.1, 2047.6, 2048 ]

Dark Frame 8: [ 2046.1, 2046.4, 2047, 2046.6 ]

Light Frame 1: [2387.1, 2323.6, 2211.1, 2391.5 ]

Light Frame 2: [2351.7, 2295.5, 2197.5, 2355.8 ]

 

DSLR Sensor Parameters v0.0.7 Completed.

[sharkmelley]

18 minutes ago, MartinFransson said:

I tried it on a few of my 300s subs from the other night but can not interpret what the values say:

 

That's a nice set of figures!  It tells us that for a 300sec exposure the read noise is 3e,  the noise from the dark current is around 4.7e (the thermal noise) and the noise from the skyfog is between 6.7e-9.7e depending on the channel.  The green channel is the 9.7e figure and the Blue/Red will be 6.7 and 8.7 but I can't be sure which way round. 

The sources of noise aggregate up in quadrature (like using Pythagoras' Theorem on a triangle) so the thermal noise and skyfog noise combined gives sqrt(6.7x6.7 + 4.7x4.7) = 8.1e  If we also add in the read noise we get the total noise of sqrt(6.7x6.7 + 4.7x4.7 + 3x3) = 8.7e  This indicates that the read noise is not really a major contributor to the overall noise and so your exposure times are quite adequate.

I'm assuming these exposures were taken in the Summer.  If so, then in the Winter the thermal noise can be expected to approximately halve for a 12C drop in temperature.  However the skyfog will still be dominant and you still won't notice any real difference in noise from using longer subs.  Longer subs will just cause the brighter stars to saturate more. 

Mark

 

[MartinFransson]

14 minutes ago, sharkmelley said:

That's a nice set of figures!  It tells us that for a 300sec exposure the read noise is 3e,  the noise from the dark current is around 4.7e (the thermal noise) and the noise from the skyfog is between 6.7e-9.7e depending on the channel.  The green channel is the 9.7e figure and the Blue/Red will be 6.7 and 8.7 but I can't be sure which way round. 

The sources of noise aggregate up in quadrature (like using Pythagoras' Theorem on a triangle) so the thermal noise and skyfog noise combined gives sqrt(6.7x6.7 + 4.7x4.7) = 8.1e  If we also add in the read noise we get the total noise of sqrt(6.7x6.7 + 4.7x4.7 + 3x3) = 8.7e  This indicates that the read noise is not really a major contributor to the overall noise and so your exposure times are quite adequate.

I'm assuming these exposures were taken in the Summer.  If so, then in the Winter the thermal noise can be expected to approximately halve for a 12C drop in temperature.  However the skyfog will still be dominant and you still won't notice any real difference in noise from using longer subs.  Longer subs will just cause the brighter stars to saturate more. 

Mark

 

Thank you for helping me! This camera is so new to me that I haven´t really got any more data with it. Here is a comparison from my other camera (1100D), taken in the winter as you suggested. Roughly the same sensor temps for both sublengths:

Noise estimates for 300sec exposure

Read Noise(e): [ 4, 4.11, 3.95, 3.99 ]

Thermal Noise(e): [ 1.53, 1.45, 1.44, 1.53 ]

Skyfog Noise(e): [ 19.5, 15, 16.26, 19.41 ]

Noise estimates for 599sec exposure

Read Noise(e): [ 3.99, 4.09, 3.93, 3.98 ]

Thermal Noise(e): [ 3.2, 3.17, 3.19, 3.12 ]

Skyfog Noise(e): [ 20.27, 18.13, 14.07, 20.16 ]

 

Does this say that I would be better off with twice as many 300s-exposures instead of the 600s ones? Better for me in that case :)

 

[sharkmelley]

27 minutes ago, MartinFransson said:

 

Does this say that I would be better off with twice as many 300s-exposures instead of the 600s ones? Better for me in that case

 

Your 300sec data and 599sec data do not look entirely consistent.  When exposure time is doubled, the thermal noise and skyfog noise would normally increase by a factor of sqrt(2) i.e. 1.414

It is likely that sky conditions and sensor temperature were different - note that sensor temperature, hence dark current and thermal noise typically increases quite a lot during a long imaging run of an hour or more.

If your skyfog+thermal noise is sufficiently larger than the read noise then increasing the length of you subs will not make any noticeable difference to the final image for the same total exposure time.  So you will not see any difference between twice as many 300sec exposures instead of the 600sec.  

Extremely long exposures (10 minutes and more) are typically used for cooled CCD narrowband imaging because the read noise is then frequently greater than the skyfog noise if the exposures are too short.

Mark

[MartinFransson]

12 minutes ago, sharkmelley said:

Your 300sec data and 599sec data do not look entirely consistent.  When exposure time is doubled, the thermal noise and skyfog noise would normally increase by a factor of sqrt(2) i.e. 1.414

It is likely that sky conditions and sensor temperature were different - note that sensor temperature, hence dark current and thermal noise typically increases quite a lot during a long imaging run of an hour or more.

If your skyfog+thermal noise is sufficiently larger than the read noise then increasing the length of you subs will not make any noticeable difference to the final image for the same total exposure time.  So you will not see any difference between twice as many 300sec exposures instead of the 600sec.  

Extremely long exposures (10 minutes and more) are typically used for cooled CCD narrowband imaging because the read noise is then frequently greater than the skyfog noise if the exposures are too short.

Mark

Thanks! The outdoor temp was probably lower for the 600s run. Sensor temps were +7 vs +4 for the 300 ones. Note however that these images were not of "blank" sky but rather of DSO targets. Might try a few blank ones some day.