3nm Exposure times.

[ian_bird]

Hi

I've been taking 3nm HaII images for the past year or so. I'v been using 30 minute exposures.

Recently someone asked me the question - why 30 minute exposures? Why not 6 x 300 or 3 x 600?

I didn't really have an answer.

Anyone have an opinion on this?

The best comment I have found was one on CloudyNights by Jon Rista:

Longer subs in general are better, not just with narrow band. With RGB/OSC/LRGB your sub lengths are usually limited by light pollution or airglow, however if you did not have those limitations, longer would still be better. The fundamental reason for that is read noise. Without read noise, it wouldn't matter how long each sub was, the total aggregate noise in the final integration would be the same: SQRT(signal). Because of read noise, however, the integrated noise is: SQRT(SubCount + (Signal + ReadNoise^2)). Read noise compounds as you integrate. It's a constant amount of noise added to each sub, but since it compounds, the more subs you integrate, the more read noise you have. As such, using the longest exposure you can reasonably get away with (i.e. without signal clipping anywhere you cannot afford it, and within the limits of your tracking ability) is going to be best. Then you have fewer subs to integrate, and thus you will have less read noise in the end.
 
If you can go for 20-30 minutes, or even longer for fainter objects, do so. It will only benefit your results in the end, and you'll pick up fainter details as well. 
 

Thoughts?

Regards

Ian

[ollypenrice]

My thoughts are that he is dead right. I'm not a theorist, I'm an experimenter. I've read all sorts of stuff on the net suggesting that 100x1 minute (in CCD, not CMOS) equals 10x10 minutes but, in my opinion, only someone who had never taken an astrophoto could come up with such a view.

I think the sub length/number of subs thing is target specific. For faint go for longer.

And what I really think is that you should experiment for yourself under your own specific conditions.

Olly

[Uranium235]

The quote you have is fairly accurate, but you also need to consider the camera you are using (bit depth, full well capacity), and the photograophic speed of the optics. What might look pretty good at f5 (in Lum), say for 20min - might not be suitable for f2.

The only thing I might question is the role of read noise on a stacked image, since properly calibrated data shouldnt have any read noise present. The only point where I think it becomes relevant is at the capture stage - when you need any faint signal to overcome the read noise in each sub in order for it to be better recorded.

 

[ian_bird]

Thanks Olly

30 minutes works for me. I will stick with it.

3nm Narrow Band is about all I can do from my location.

Somewhat amazed that I can do 30 minute subs using a Celestron CG-5! (My CGEM-DX is having some TLC done to it).

Thanks again. Your comment pretty much confirms what I felt.

Cheers

Ian

[ian_bird]

Another question I get asked is "Wont a Satellite trail ruin a sub and waste 30 minutes".

Well no - stacking will take care of that.

But the odd thing is - I have never seen a  Satellite trail in any single 3nm HaII 30 minute sub.

Is anyone else seeing this? Is there an explanation? Like perhaps Satellites don't emit HaII? ;-)

Cheers

Ian

 

[steppenwolf]

I love all the technical papers on subjects like this but there is nothing like some real life experimenting and this will show pretty convincingly that longer exposures and fewer of them with a CCD camera will outperform shorter exposures but more of them with the same camera. Of course, in an ideal world you would take lots of long exposures! What you do find is that 'smoothness' comes with multiple exposures (short or long) but depth and detail require long exposures.

With your 3nm Ha filter, 30 minute exposures are a great place to be.

[steppenwolf]

7 minutes ago, ian_bird said:

But the odd thing is - I have never seen a  Satellite trail in any single 3nm HaII 30 minute sub.

Is anyone else seeing this? Is there an explanation? Like perhaps Satellites don't emit HaII? ;-)

I get plenty of satellite hits with Ha!

[ian_bird]

Steve - interesting.

Do they stack out OK?

Cheers

Ian

[tony210]

Hi- I think your quote is spot on! -longer for feint targets and reduce read noise if LP allows-strict rules can be very  misleading- I have found that binning for example reduces resolution even though my sky conditions should 'not' allow the detail I can extract from subs with careful processing using theoretical arc second per pixel calculations- so no substitute for personal experience and experimentation - Tony

[ian_bird]

WO 72mm Megrez. 430 mm focal length.

Asrtodon 36mm 3nm HaII filter.

Atik 4000

Celestron CG-5 Mount. Yes- really.

2 x 30 minutes exposure.

It doesn't bear close examination - but I'm amazed what you can get with this combination.

[wxsatuser]

2 hours ago, ian_bird said:

Steve - interesting.

Do they stack out OK?

Cheers

Ian

I get loads in 10minute 5nm subs.
Have been doing 3hours, 18 subs and no sign of sat or plane trails after stacking.

My location is right on the Gatwick inbound and outbound routes so loads of planes between
8pm and midnight, it tails off after midnight but I don't worry about it anymore.

[ian_bird]

I'm in Windsor - under the Heathrow flight path. Similar problem.

Agreed - doesn't seem to be a problem.

[wxsatuser]

Have only just started my narrowband journey and my first experiment was with 5minute subs at f/2.
This was ok but when I tried 10minutes it made such a big difference and very surprised how easily it defeats LP.

This is IC1396 18x600secs @ f/2, the neb was quite lowdown to start with, in my NE lightdome.
When I saw the subs come in I could'nt believe how good they were, a dark site will do a lot better, no doubt.
Now I look forward to going out and doing some real imaging.

[vlaiv]

Few longer exposures will always yield better SNR than more shorter ones for same total integration time - provided you are working in regular regime, so no saturation, linear response, linear dependency of dark current to time, etc.

Question is only where you cross the line of diminishing returns, and that depends on so many factors (read noise, dark current, LP, target brightness, filter band, aperture, resolution, ....). Taking many exposures does have some benefits - choice of stacking algorithms increase (sigma clip benefits from many subs, as well as dithering), lower probability of discarding the sub due to wind, guiding problems, any mechanical glitches, ...

Additional benefit of having shorter exposures is that it allows for much higher count of dark frames - this has significant impact if camera has dark current / large read noise.

This is where SNR calculators come in handy. I've created a spreadsheet as a guide, where I can change parameters and see how it will affect final SNR. For my conditions for example, most of the time, there is no need to go over couple minutes of exposure (depends on camera/scope combination, for some I use 1 minute and for other 4 minute exposures). Longer exposures in there cases improve SNR only minimally - less than 1% for doubling exposure time.

You could do the same if you manage to obtain all the information needed for such calculation (most is straight forward - read noise, dark current, resolution, aperture, light losses in optical train - you can guesstimate that rather well, sensor QE, but others are more difficult - you need to estimate target brightness, and sky brightness - you really need to measure later by taking couple of exposures of sky itself on average night to get the idea for your location and setup).

[RichLD]

I tend to use 1200s subs @ f/3.6 with a 7nm HA filter on a KAF8300, so I'd definitely be going for 1800s or longer with a 3nm - spot on! 

[swag72]

I did an experiment last year..... Richard Crisp had been in contact with me and did some sums and what not with my camera..... he said that I only needed to do 10m subs to get above the noise floor. I did an image with only 10m subs and I have to say that the processing was much more difficult than with 30m subs. The 30m subs are sublime to process and there is almost nothing to do with them!! 

So I reverted back to 30m subs and having tried the experiment won't change my methodology.

[wxsatuser]

49 minutes ago, swag72 said:

I did an experiment last year..... Richard Crisp had been in contact with me and did some sums and what not with my camera..... he said that I only needed to do 10m subs to get above the noise floor. I did an image with only 10m subs and I have to say that the processing was much more difficult than with 30m subs. The 30m subs are sublime to process and there is almost nothing to do with them!! 

So I reverted back to 30m subs and having tried the experiment won't change my methodology.

What works, works.
No amount of fiddling with formulae or theory will change it.

We all have different situations and am sure we find out fairly quickly what works with our setup and location.

[steppenwolf]

9 hours ago, ian_bird said:

Do they stack out OK?

Yes they do using SD Mask as the stacking algorithm although Sigma Clip and Kappa Sigma will also work well but all three will require an absolute minimum of four subframes for the 'outlier pixel' routine to kick in and the more subframes the better.

[ollypenrice]

1 hour ago, swag72 said:

I did an experiment last year..... Richard Crisp had been in contact with me and did some sums and what not with my camera..... he said that I only needed to do 10m subs to get above the noise floor. I did an image with only 10m subs and I have to say that the processing was much more difficult than with 30m subs. The 30m subs are sublime to process and there is almost nothing to do with them!! 

So I reverted back to 30m subs and having tried the experiment won't change my methodology.

I concur.

Regarding sat and plane trails you can rely on Sigma Clip if you have sufficient subs but you can also give Sigma Clip a helping hand by running a cosmetic 'remove line' filter on the individual subs before stacking. The one in Astro Art is a work of genius. You just click on each end of a line and go to Filter, Cosmetic, Remove Line and it either vanishes or is greatly diminished.

It's nice to see the long sub gaining ground because for a long time it came under persistent attack from the students of spec sheets! The long sub works. As Sara says, the processing is a peach. It was Yves who said to me once, ''Go on, let's try thirty minutes in luminance". It was fantastic! (It still is...)

Olly

[sharkmelley]

I've come across so-called exposure calculators that give nonsense results.  An ideal exposure needs to be skyfog limited i.e. the noise from the skyfog should swamp the read noise of the camera.  This happens when the skyfog noise is approximately 5x the read noise.  It's actually very easy to test by experiment:  simply measure the noise in a bias frame and then the noise in the background of a light frame.   For a typical CCD cameras  on typical scopes at a dark site this will always result in ideal exposure times of 30minutes or more, even for luminance, much more so for narrowband.

With low read noise CMOS cameras you can certainly get away with much shorter exposures.

Mark