Long subs vs many subs

[mikey2000]

Now I've finally got a reasonably stable guiding setup, my thoughts wandered towards the issue of the optimum length of subs.  Apologies if this has been discussed ad-infinitum but it's new ground for me. 

 

In simple terms, I was wondering: Is 20x3mins better than 3x20mins?

 

I stumbled across this article....

http://dslr-astrophotography.com/long-exposures-multiple-shorter-exposures/

It's fairly mathematical but if my understanding is correct, the conclusion is that under typically light polluted skies, once you get to 2 or 3 minute exposures, it's just the total integration time that's important; stretching to 10+ minute exposures confers no SNR benefit for the same total integration time.  (However, under a properly dark sky, longer exposures still give more benefits)

 

So, I'm happy to stick to 3 mins and reduce the risk of wrecked subs :-)   (I live under mild LP - I never saw the milky way from my garden....)

 

 

The only spanner in the works might be that I use a LP filter.  Does that move me into dark sky territory????   I feel an experiment coming on:  16x2mins vis 8x4mins vs 4x8mins then stack 'em up and compare noise :-)

 

Wish me luck or am I wasting my time?

[AndyW2008]

Hi Mikey

As I understand it, the standard approach is that fewer longer subs should always be better than a lot of shorter subs. For the same overall exposure time you can achieve the same SNR with a larger number of shorter subs, but you cannot improve the overall signal. So 3 x 25 min subs will have the same SNR as 300 x 15s subs, but the signal component of the 25 min subs will be 100 times higher than that of the 15s subs (using an example from Joseph Ashley's book). Stacking cannot overcome this, otherwise what would be the point of spending so much on precision equipment in order to take long exposure photos? However, light pollution is also a signal, so increasing exposure time also increases the signal from light pollution. Therefore if an image signal is lost to light pollution in your backyard, you will not be able to overcome this by taking longer exposures. Hence the draw of dark sky sights. So the real answer to your question is: It Depends!  But mostly it depends on the level of light pollution where you are.

Incidentally, in the discussion after the article you quoted, someone asks about the impact of light pollution filters, and the authors comment was that it 'shifts you toward dark sky conditions'. So I think a good experiment would be to see what the impact is with and without the filter for your conditions.

Hope I am right - seems quite complex!

Andy

[Guest Tuomo]

6 hours ago, AndyW2008 said:

Hi Mikey

As I understand it, the standard approach is that fewer longer subs should always be better than a lot of shorter subs. For the same overall exposure time you can achieve the same SNR with a larger number of shorter subs, but you cannot improve the overall signal. So 3 x 25 min subs will have the same SNR as 300 x 15s subs, but the signal component of the 25 min subs will be 100 times higher than that of the 15s subs (using an example from Joseph Ashley's book). Stacking cannot overcome this, otherwise what would be the point of spending so much on precision equipment in order to take long exposure photos? However, light pollution is also a signal, so increasing exposure time also increases the signal from light pollution. Therefore if an image signal is lost to light pollution in your backyard, you will not be able to overcome this by taking longer exposures. Hence the draw of dark sky sights. So the real answer to your question is: It Depends!  But mostly it depends on the level of light pollution where you are.

Incidentally, in the discussion after the article you quoted, someone asks about the impact of light pollution filters, and the authors comment was that it 'shifts you toward dark sky conditions'. So I think a good experiment would be to see what the impact is with and without the filter for your conditions.

Hope I am right - seems quite complex!

Andy

Thanks for the info. Thats the way I understand it too. Can you help me out with one question tho?

As I understand it all comes down to histogram and about 20-30% from the left hand side of it. Does it mean, that if I go over this 20-30% mark of histogram, imagel is mostly saturated by light pollution and not usable signal from the stars? I mean there is lots of room for signal in histogram before everything turns to white noise.

[wornish]

Really great presentation here.  That I think answers your question.

 

Warning its a large file.

 

Everything You Want to Know About Deep Sky Imaging - Craig Stark

[frugal]

2 hours ago, Tuomo said:

As I understand it all comes down to histogram and about 20-30% from the left hand side of it. Does it mean, that if I go over this 20-30% mark of histogram, imagel is mostly saturated by light pollution and not usable signal from the stars? I mean there is lots of room for signal in histogram before everything turns to white noise.

You are imaging the night sky, so most of the image should be black (far left of the histogram) except for stars (to the right) and nebula (mostly to the left). However you are also taking an image of the light pollution which will push everything to the right, as will the noise from the camera itself. Having the peak of the histrogram about 1/3 of the way in from the left is a good indicator that you have lots of signal without over saturating the stars. If you see a spike on the very right hand edge of the histogram then you have overexposed the stars.

Ideally you want to expose for as long as possible until you hit one of these limiting factors:

1 - Saturating stars (sharp peak on the right hand column of the histogram)

2 - You have reached the limit of your guiding / polar alignment, and your stars are starting to become ovoids

3 - The sky glow is pushing the histogram too far to the right, and you are saturating everything.

Which of these you hit first will depend on your setup. I used to have (2) as my limiting factor I could only manage 20-30 seconds before the polar alignment error became obvious. After I moved to guiding my limiting factor became the accuracy of my guiding (I managed to get up to 5 minute exposures). Now my guiding has got better and (3) is now my limiting factor; I can now guide for 15 minutes, but much more than that and the sky glow swamps everything; I do get a little bit of (1), but only in the cores of the brightest stars and I am willing to accept that to get at the dark details in the nebula I am actually targeting. 

Sometimes the target you are imaging will change your limiting factor. If I image the faint parts at the edge of the Orion Nebula my limiting factor is sky glow (10-15 minutes); if I image the bright core of the nebula then my limiting factor is the saturation of the core (about 30 seconds).

[Guest Tuomo]

6 minutes ago, frugal said:

You are imaging the night sky, so most of the image should be black (far left of the histogram) except for stars (to the right) and nebula (mostly to the left). However you are also taking an image of the light pollution which will push everything to the right, as will the noise from the camera itself. Having the peak of the histrogram about 1/3 of the way in from the left is a good indicator that you have lots of signal without over saturating the stars. If you see a spike on the very right hand edge of the histogram then you have overexposed the stars.

Ideally you want to expose for as long as possible until you hit one of these limiting factors:

1 - Saturating stars (sharp peak on the right hand column of the histogram)

2 - You have reached the limit of your guiding / polar alignment, and your stars are starting to become ovoids

3 - The sky glow is pushing the histogram too far to the right, and you are saturating everything.

Which of these you hit first will depend on your setup. I used to have (2) as my limiting factor I could only manage 20-30 seconds before the polar alignment error became obvious. After I moved to guiding my limiting factor became the accuracy of my guiding (I managed to get up to 5 minute exposures). Now my guiding has got better and (3) is now my limiting factor; I can now guide for 15 minutes, but much more than that and the sky glow swamps everything; I do get a little bit of (1), but only in the cores of the brightest stars and I am willing to accept that to get at the dark details in the nebula I am actually targeting. 

Sometimes the target you are imaging will change your limiting factor. If I image the faint parts at the edge of the Orion Nebula my limiting factor is sky glow (10-15 minutes); if I image the bright core of the nebula then my limiting factor is the saturation of the core (about 30 seconds).

Finally some good info about the matter at hand

 

About number 3. How much right in your opinion is too far right side? 75%

[Adam J]

Depends on your definition of 'better' 3x20min subs that have satellite trails, clouds passing over and guidance errors are not going to be better than 20x3min subs of which you have had to chuck out a couple. 

However, generally longer subs are better if your tracking and conditions allow. 

You also need to consider dynamic range as with very long exposures you will find you will reach saturation. Also you will end up with little star color and large stars. 

Everything is a balance and hence target dependent. 

 

[frugal]

It doesn't help that different applications display the histogram in different ways. Most astronomical programs will display the histogram as a linear graph from 0 to the maximum value for the camera (16,000 for my 14 bit Canon 60D). Image processing tools like photoshop will display the same image as a logarithmic histogram from 0 to 1 where 0 is black and 1 is white.

So when I captured the images last night SGPro showed the peak of the histogram at about 35-40% of the way in. Photoshop shows the same image at about 60%, and PixInsight shows tight over to the left as it is processing my 14bit file as a 32 bit image without any scaling of the data.

[Adam J]

30 minutes ago, frugal said:

It doesn't help that different applications display the histogram in different ways. Most astronomical programs will display the histogram as a linear graph from 0 to the maximum value for the camera (16,000 for my 14 bit Canon 60D). Image processing tools like photoshop will display the same image as a logarithmic histogram from 0 to 1 where 0 is black and 1 is white.

So when I captured the images last night SGPro showed the peak of the histogram at about 35-40% of the way in. Photoshop shows the same image at about 60%, and PixInsight shows tight over to the left as it is processing my 14bit file as a 32 bit image without any scaling of the data.

I agree, the best way is to stretch the sub frame and look at it. Can you see any hints of the faintest component you want to be able to draw out. Yes? keep this settings, No? increase exposure. Simple, histograms of the entire frame can be misleading anyhow when you are looking at a small object within that frame. That is why I love the ability to perform a quick curves adjustment on a sub frame in Backyard EOS pro. 

[mikey2000]

Fascinating discussion!  Lots for me to take in here...

 

is there an an objective way of measuring the signal to noise ratio of a single raw file or a DSS output .fts file?

[Adam J]

Not sure about objectively.

Problem is that you will be dealing with many different noise sources contributing to the image and for very short exposures lots of photon noise / read noise that will reduce in the stack along with other noise sources. In longer exposures thermal noise and noise caused by light pollution will dominate.  Examining subs for noise is not going to give you the full story. 

But ultimately you are asking the wrong question anyway. 

Problem is there is no such thing as S/N ratio for the entire image or rather it is a useless measurement, if you take a image of M42 then you will have a very high signal to noise ratio around the bright central trapezium, however, you will have poor signal to noise ratio in the wispy bits between M42 and the running man......some areas of the image have more signal and hence better S/N than others and hence no single measurement of signal to noise ratio across the image is meaningful.

What you really care about is the S/N ratio in the small part of the image containing the faintest data that you want to be able to pull out of the image in processing. Hence, ill say it again, histograms of the entire image don't mean much.  

[frugal]

1 hour ago, mikey2000 said:

is there an an objective way of measuring the signal to noise ratio of a single raw file or a DSS output .fts file?

These days I look at the images and throw out any with obvious guiding errors (eg. dumbbell shaped stars). Then I use FWHM and Eccentricity as my measures for a good image. I use the Subframe Selector script in PI to exclude anything that is not good enough (FWHM > 4.5 or Eccentricity > 0.55))

[mikey2000]

it seems the objective measure of a pic SNR is a bit impossible then.  I noticed that DSS assigns a score to each pic (it seems to concern itself with star shape, mostly) (If I remember correctly)  There are also columns for FHWM and Sky Background.

 

Last night I tried a very quick test.  With a half moon up, I aimed at the Whirpool and compared 12x2mins with 6x4mins.   After stacking and PP to get similar looking results, the 6x4mins is noticeably noisier.     I didn't take any darks though.  Maybe that would make a difference.   I was going to get 4x6mins as well but then eqmod crashed, wrecked my PEC/worm syncronisation.  PHD2 got all wobbly after that so I gave up a put the whole lot away in a huff.

 

I suppose a better test would be to get at least an hour of data then repeat with different sub lengths.   Then do the same on a moonless night, a half moon night and finally a full moon night.  Then I'll know my own personal skies a bit better.

 

 

Having never seen the milky way from my garden, I have a sneaking suspicion I'll be looking at 2 to 3 mins as optimum.   I get a fair few spoiled subs due to planes flying over and fairly frequent PHD2 hiccups (I only have a 'cheap' eq3pro and it's loaded up to max weight - I've only got myself to blame!)

[Guest Tuomo]

31 minutes ago, mikey2000 said:

it seems the objective measure of a pic SNR is a bit impossible then.  I noticed that DSS assigns a score to each pic (it seems to concern itself with star shape, mostly) (If I remember correctly)  There are also columns for FHWM and Sky Background.

 

Last night I tried a very quick test.  With a half moon up, I aimed at the Whirpool and compared 12x2mins with 6x4mins.   After stacking and PP to get similar looking results, the 6x4mins is noticeably noisier.     I didn't take any darks though.  Maybe that would make a difference.   I was going to get 4x6mins as well but then eqmod crashed, wrecked my PEC/worm syncronisation.  PHD2 got all wobbly after that so I gave up a put the whole lot away in a huff.

 

I suppose a better test would be to get at least an hour of data then repeat with different sub lengths.   Then do the same on a moonless night, a half moon night and finally a full moon night.  Then I'll know my own personal skies a bit better.

 

 

Having never seen the milky way from my garden, I have a sneaking suspicion I'll be looking at 2 to 3 mins as optimum.   I get a fair few spoiled subs due to planes flying over and fairly frequent PHD2 hiccups (I only have a 'cheap' eq3pro and it's loaded up to max weight - I've only got myself to blame!)

Check the connections.

[frugal]

1 hour ago, mikey2000 said:

it seems the objective measure of a pic SNR is a bit impossible then.  I noticed that DSS assigns a score to each pic (it seems to concern itself with star shape, mostly) (If I remember correctly)  There are also columns for FHWM and Sky Background.

Things like FWHM in arc seconds and star eccentricity are objective scores, regardless of the camera or scope they should be the same.

[ollypenrice]

Given the magnificently reliable lack of concesus on this question I sugget that experimentation will provide the only reliable answer! 

For what it's worth, I image from a very dark site and like long subs. I use large format CCD cameras with deep wells and low QE. I don't worry about over exposing stellar cores because you can fix that in post processing. When going after faint details I use 30 minute luminance subs and am entirely satisfied that 20x15 does not go as deep as 10x30. Your mileage may vary...

The low read noise CMOS cameras don't behave in the same way. With those, many short subs seem to work very well.

Olly

[frugal]

To follow up the conversation with regards to objective and subjective quality of image statistics, I was having a look over some of the data I have captured for M101. Going Clockwise from the top left

Top left: 300s, FWHM 3.6

Top Right: 600s, FWHM 3.6

Bottom Right: 900s, FWHM 3.6

Bottom Left: 900s, FWHM 6.2

All the images have had an automatic screen stretch applied to bring the apparent exposure up to about the same level. The 900s 3.6 image has more definition than the 300s one. It is interesting that the bottom left image which was captured under poorer conditions has the least signal despite being a 900s image. 

The background noise is horrible in all of them, but DSLR, what can you do...