Jump to content

SkySurveyBanner.jpg.21855908fce40597655603b6c9af720d.jpg

Guidance on light frames, ISO & exposure time?


imakebeer

Recommended Posts

I've been making my first steps into DSO imaging over the last few weeks and made decent progress so far but want to continue improving. One thing I'm uncertain about (and I realise there's probably no firm answer) is how to know how many light frames are enough? And within that how to decide on the exposure time and ISO?

I posted some images the other day of M31 & M33 (SW150PDS + HEQ5 + Nikon D5500 + BYN):

  • For M31 I took 60 x 15s lights at ISO 800, plus 10 darks and I'm reasonably happy with the result (I did also take 10 bias frames, and next time I'll do flats properly too).
  • For M33 I was hurrying and only did 30 x 15s lights at ISO 1000, plus 10 darks & 10 bias. Next time I'll plan better so I can take my time.

When I compare with similar images from others it seems like they are maybe taking more frames, and definitely at considerably longer exposure times (2mins+ ?). At the moment the best I can manage is about 60s before I get star trailing, but I can work on improving alignment and I'm working on getting tracking set up so this will hopefully enable longer exposures if they're needed.

So how do I know when I have enough frames, are there any rules of thumb? This would be help me plan the imaging session better ahead of time. And assuming the ISO stays low enough that there isn't lots of associated noise/graininess, what are the pros & cons of more images with short exposure but higher ISO, vs. fewer images with longer exposure and lower ISO?

BYN shows a histogram - besides the peaks not being all the way to the left or right, what specifically are we looking for?

In contrast to M31, M33 is much fainter - so how does that change the game? Is it simply a question of getting more lights, or do you need to go to higher ISO/longer exposures (I guess the latter means capturing more photons, right?)

Thanks in advance 👍

Link to comment
Share on other sites

Noise goes down only as the square root of the exposure time so that makes it a classic case of diminishing returns.  The same is true though, of every step of the way between capture and final image. Processing, for instance, involves a host of tiny steps, none of which is particularly spectacular in itself.

What I can say, though, is that using 'normal' telescopes I always budget for between 10 and 20 hours per target. Using super-fast systems like the RASA 8, which is about 4.5x faster than my previous fast refractor, we are usually happy with 3 hours.  Like most people, I started off taking three images in a night whereas now it's more like three nights for one image. I certainly won't be the first to say this.

I don't use a DSLR so can't help with the ISO. Exposure time is likely to be limited by light pollution in the UK. Experimentation is everything but don't make a habit of using different exposure times. Once you've found the optimum it will be far less target-specific than you might expect.

Olly

Link to comment
Share on other sites

Histogram: The commonest pixel value is of background sky, so that should be the biggest peak. If you place that peak at 1/4 to 1/3 of full scale, you should acquire even the dim nebulosity in each sub-exposure.

ISO: DLSRs differ, but many of them are "ISO-invariant", that is, within a certain range an image shot at one ISO and "stretched" to be as bright as one shot at double or quadruple the setting will appear basically identical. Totally backwards to what we're used to in terrestrial photography, but honest, it's true. So just use the longest exposure time that reliably yields good subs (i.e. small round stars), and dial the ISO up or down to get the histogram peak where it should be.

Totally concur with Olly that it's counterproductive to tweak subexposure times and ISOs. Find a pair of values that yield reasonably good results and just stick with that. Because...

...Integration time matters way, way, way more than the number of photons a sub receives. WAY more. Yes, there are corner cases and read noise to minimize and entire books on the subject, but four hours of 1-second exposures will beat a half-hour of 10-minute exposures every time. Again, not intuitive, but true. The question of whether 10 one-minute exposures are better than four 2.5-minute exposures is a red herring. Find a basic combination of exposure and ISO that yields reasonable results with your equipment and then just hammer it with as many minutes/hours/days of integration time as you can stand. (Among other things, it greatly simplifies processing to only need one set of calibration frames.)

The Deep-Sky Imaging Primer has a really excellent chapter explaining all this stuff, covering signal/noise in detail and explicitly running examples of short vs. long subs. You don't need a ton of math to follow it but if you like math, he shows his work. Hard to recommend that book highly enough for anyone starting out.

I don't aim quite as loftily as Olly most of the time. My rule of thumb is more like "One hour marginal, two hours OK, four hours better". You can look at our image galleries and set your own level of "acceptable".

Edited by rickwayne
  • Thanks 1
Link to comment
Share on other sites

Thanks @rickwayne , that's helpful I think, as long as I've understood correctly...

I was having another crack last night at M33 - now as far as I'm aware my D5500 is ISO invariant...

Initially I did 90 x 60s at ISO 800, plus 10 each darks, bias and (this morning) flats.

Since it wasn't that late I figured have another go so I tried 45 x 120s at ISO 400 (+ 10 darks, bias & flats again) - this was cut short to about 35 lights as the scope hit one of the tripod legs! 😖

But what you're saying is it kinda doesn't matter which approach I use, and is anything more shorter subs is better, right?

(which would suit me because star trailing at present is negligible at 60s but getting noticeable at 120s).

So I'm better off just doing lots of 60s lights, right? And just tweak the ISO (with reason at least) to get the histogram in the sweet spot?

Is that about right?

Cheers 👍

Link to comment
Share on other sites

I take it you're relatively new to imaging, that's not aimed at being insulting, it's just when we all start we all did the same thing, stack a few subs, get a result that excites you, and so it should, that's why we all do it ehh..

As time goes on you see what you can do to improve the image, for sure that's stacking more subs as the aim of stacking is to get the signal( what you want) above the noise( what you don't)  signal improves with stacking, noise averages out..so in essence the worst noise you get is in a single sub, if you stack 10 subs you get an improvement, add a 100 you get a further improvement, and say 1000 you will get a improvement even more, while adding more signal... That's the reason behind stacking, signal to noise ratio... Generally there's 2 types of noise, 1 that comes from your sky quality, so less noise in a dark sky location and more noise in a heavy light polluted sky... The other is noise created by you camera, thermal noise, read noise.. these are why we use calibration frames, to reduce noise or cancel them out...

You say you get star trails at 120 secs, and don't at 60 secs... So there's your exposure length that you need to capture at with your current setup... There maybe room for improvement on your sub length such as better polar alignment, better balance, maybe adding guiding etc as these contribute towards longer subs

Seems like you're asking the right questions so continue to do so

  • Thanks 1
Link to comment
Share on other sites

8 minutes ago, newbie alert said:

I take it you're relatively new to imaging, that's not aimed at being insulting, it's just when we all start we all did the same thing, stack a few subs, get a result that excites you, and so it should, that's why we all do it ehh..

How very dare you, I've never been so insulted, I've been imaging DSOs 20+ years and consider myself an authority on the subject...

Nah, just kidding 🤣 Yeah I'm a newbie, only really been doing it since October, which if you factor in Cloudvember means only a few days of actual practice. So all advice is welcome and very much appreciated 👍

12 minutes ago, newbie alert said:

if you stack 10 subs you get an improvement, add a 100 you get a further improvement, and say 1000 you will get a improvement even more

Yep, all the stuff about improving S/N ratio is clear. But it goes as the square root or something doesn't it? That's what @ollypenrice was saying above? So 1000 frames isn't 10 times better than 100, but only 3.16x better noise reduction?

15 minutes ago, newbie alert said:

You say you get star trails at 120 secs, and don't at 60 secs... So there's your exposure length that you need to capture at with your current setup...

The other night 120s wasn't too bad actually, I think polar alignment and backlash might be playing into it. I do actually have a guide scope and camera but haven't fitted them yet, it's on the to-do list. But I think these should help too (as well figuring out how to stop the scope clattering into the tripod on longer runs 😖🤣)

Link to comment
Share on other sites

You are asking 2 very good questions!

1) What total exposure time should I aim for.

Olly has explained this perfectly.  I think what nearly all of us have done over the years is start with relatively small amounts of data.  Doing this on the typical A-list celebrity targets such as M42 and M31 will show the bright parts of the target very nicely but you will discover that the fainter outer regions are very noisy.  You will then decide at a later date that you want to revisit this target and "go deeper" by making those dim areas smoother.  So, save your existing data so that you can add to it in the future.

2) What parameters should I use for my sub exposures.  

When you combine sub exposures the signal noise reduction is directly related to the overall exposure time.  It is reduced by by the stacking process in exactly the same way as if you had taken one long exposure.  However, there is another kind of noise which arises when the voltage information from the pixels in the camera chip is converted to digital pixel value.  This is called read noise and isn't eliminated by stacking.  Fortunately, the background sky glow will blot out the read noise provided the sub exposure is long enough.  Read noise is a serious issue with CCD chips  where all pixels are read by a single analogue to digital converter.  CMOS cameras (like all DSLRs these days) have lots of analogue to digital converters so the read process is quicker and consequently there is much lower read noise.  If you increase the iso  you further decrease the read noise but at the expense of dynamic range.  The key message is that the quality of your data is dependant on the number of photons being captured by your chip.  The ISO has no impact on this.  Unless you have a very dark sky, with less sky glow to mask out the read noise, it is likely that 1 minute subs at iso 400 will be a good choice.  Obviously you can experiment.  The lower the iso the greater the processing headroom you give yourself.

So, enough waffle, try this -  60 second subs, iso 400, total subs 60.  That's one hour of total data.  That should give you something to enjoy!!

 

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

Thanks @MartinB , it's reassuring to know I'm more or less on the right path with what I'm doing at the moment. I took a few test pics on M33 the other night, 60s at both ISO 400 & 800 but went with 800 in the end. That said I think there's enough "stretch" in the histograms (if that's a phrase?) that ISO 400 wouldn't be too dark so I can give that a whirl next time.

I can also compare my D5500 vs. 450D and see if either is particularly better - on paper I suppose it should be the Nikon as it's newer less old and it seems much quicker to send images & video from the camera to the laptop, but who knows.

I've been processing the images of M33 from the other night and comparing with my previous efforts - it's certainly heading in the right direction so I'm pleased, but then I see other folks' efforts and I'm just blown away, they're in a totally different league! 😱😬😂 I'm starting to be able to tease out some of the the cloudy wisps of nebulosity on M31, but for M33 I'm still miles off (fainter target I suppose???).

Link to comment
Share on other sites

7 hours ago, imakebeer said:

Nah, just kidding 🤣 Yeah I'm a newbie, only really been doing it since October, which if you factor in Cloudvember means only a few days of actual practice. So all advice is welcome and very much appreciated 👍

Phew🫡🫣....

Yeah but you have cloudcember,and clouduary etc yet to come...

Most if not all of us probably started with m42 as it's pretty bright, easy to find and you can get a pretty decent image regardless of short exposure lengths...  It's sort of the same as m31 where it's easy but hard at the same time... Hard because as you have really bright areas and dim parts in the same target.. m31 has a bright core and dim lanes, m42 has a bright core( the trapezium) and dim cloud areas at the edges... Some do 2 sets of exposures, a few secs for the bright parts and as long as you can for the dim parts ,then merge together in processing

Anyway, you seem to be heading in the right direction

What mount do you have?

  • Thanks 1
Link to comment
Share on other sites

M42 and M31 are very Zen targets -- some of the easiest to image in the night sky, and yet with endless challenge for the experienced imager.

If you can get 60 seconds, that's probably enough for just about anything you're likely to encounter starting out. (Criminy, one night when my autoguiding didn't work I got this with 60-second exposures through narrowband filters. APOD? No. Recognizably the Veil? Absolutely.) Do go by the histogram and not by the appearance of the sub-exposure, though. Ideally you want no or very few pixels at 0, and a small number of maxed-out ones as well. No amount of stacking will recover out-of-range detail on either end (yes yes, if a photon hits once every ten frames that's recoverable...work with me here, OK?). And, as newbie alert points out, you can use HDR techniques to expand the dynamic range if needed.

Mathematically speaking, the best SNR you can obtain comes from the smallest number of exposures -- that is to say, longer is better. But as you've already discovered, there are practical limits to that. Even with a perfect mount, if you did a two-hour exposure you'd still have the problem of blowing out the bright areas (most frequently seen as a loss of star color, since for most DSOs they're the brightest thing in the frame), and transient issues such as satellites and aircraft. If you have 200 exposures and one has a bright line through it, the software can statistically eliminate it with little loss. If you have one long one...not so much.

The question of diminishing returns crops up over and over again in astro, whether it's total integration time, aperture, setpoint for cooled sensors, number of dark frames... You can do the math to figure out exactly how long your exposures should be to make the read noise below a given percentage of the total noise, but that percentage is arbitrary to begin with. So...good enough is good enough. Integration time rules.

Edited by rickwayne
  • Thanks 1
Link to comment
Share on other sites

2 hours ago, newbie alert said:

Some do 2 sets of exposures, a few secs for the bright parts and as long as you can for the dim parts ,then merge together in processing

Ooh, clever! I hadn't heard of that.

2 hours ago, newbie alert said:

What mount do you have?

HEQ5 👍

Link to comment
Share on other sites

1 hour ago, rickwayne said:

M42 and M31 are very Zen targets -- some of the easiest to image in the night sky, and yet with endless challenge for the experienced imager.

Out of interest, where does M33 fit on that scale? Is it much more challenging would you say?

M42 was one of my first targets with the 150PDS + HEQ5, I should revisit it, along with M45 too.

1 hour ago, rickwayne said:

transient issues such as satellites and aircraft. If you have 200 exposures and one has a bright line through it,

Funnily enough that's exactly what I had the other night in just 1 image in 90!

1 hour ago, rickwayne said:

Integration time rules.

Dumb question maybe, but by integration time you mean the total time the shutter is open, right? Regardless of whether you do 1 x 3600s exposure or 60 x 60s exposures?

Link to comment
Share on other sites

12 hours ago, imakebeer said:

Out of interest, where does M33 fit on that scale? Is it much more challenging would you say?

M42 was one of my first targets with the 150PDS + HEQ5, I should revisit it, along with M45 too.

Yeah m45 is another fantastic target.. that's reminded me to pay another visit..

Link to comment
Share on other sites

No dumb questions! Yes, that is what integration time means in DSO jargon, the sum of all subs' exposure times.

M33 is much dimmer, I have never tried to capture it without tracking but I reckon it would be much harder.

Conveniently enough, M42 happens to host a lot of satellites this time of year from my location. Thanks Elon! (Only a little bitter there, since many satellites contribute much more to overall human well-being than my astrophotographs do.)

  • Thanks 1
Link to comment
Share on other sites

On 19/12/2022 at 22:23, imakebeer said:

Out of interest, where does M33 fit on that scale? Is it much more challenging would you say?

M33 is an easy target to process, once you get enough data to get a meaningful amount of stuff in it. That is the harder part because while the numbers suggest its a bright target it is actually quite diffuse because it occupies such a large area in the sky and so you will need to invest a few hours to get a decent image out of it. But from a processing perspective its quite easy as there is no obvious extreme brightness difference as you have with M31/M42 so you could just simply stretch it and be done with it.

M81 and M82 on the same field of view is another great beginner target that is not that difficult to capture or process. Should become shootable in the coming months for most imagers in northern europe. Another one later on in spring would be M101 which is kind of like M33 in terms of imaging difficulty but maybe requiring a little bit more time to finish. Think at least 5h rather than a couple and the more the merrier.

  • Thanks 1
Link to comment
Share on other sites

On 18/12/2022 at 21:14, rickwayne said:

Histogram: The commonest pixel value is of background sky, so that should be the biggest peak. If you place that peak at 1/4 to 1/3 of full scale, you should acquire even the dim nebulosity in each sub-exposure.

ISO: DLSRs differ, but many of them are "ISO-invariant", that is, within a certain range an image shot at one ISO and "stretched" to be as bright as one shot at double or quadruple the setting will appear basically identical. Totally backwards to what we're used to in terrestrial photography, but honest, it's true. So just use the longest exposure time that reliably yields good subs (i.e. small round stars), and dial the ISO up or down to get the histogram peak where it should be.

Never seen this so well explained in writing! I've never got decent images when the subs have their histograms leaning to, or crossing over the center. As you get  guiding going, it's easy to get gready. Dont. The histogram rule rules them all! As for signal/noise, my experience is that in a given time window a bigger number of subs gives the best results, as long as the left foot of the individual histogram is well clear of the egde. I usually try to get 100+ subs on a session, and adjust exposure time and ISO accordingly, as long as I'm within the parametres above.

 

  • Thanks 1
Link to comment
Share on other sites

On 14/12/2022 at 11:21, imakebeer said:

One thing I'm uncertain about (and I realise there's probably no firm answer) is how to know how many light frames are enough? And within that how to decide on the exposure time and ISO?

Have you seen this? I know its not about DSLRs but it will give you a bit of background on the topic as the sensors are pretty much the same on both DSLR & Astrocams.

 

  • Thanks 1
Link to comment
Share on other sites

On 14/12/2022 at 11:21, imakebeer said:

When I compare with similar images from others it seems like they are maybe taking more frames, and definitely at considerably longer exposure times (2mins+ ?). At the moment the best I can manage is about 60s before I get star trailing, but I can work on improving alignment and I'm working on getting tracking set up so this will hopefully enable longer exposures if they're needed.

I am only 2 years old in this hobby, so not an expert by any means. As I understand it, Increasing exposure time is not necessarily the right answer. It just saves disk space on your computer. At the end of the day, the more the photons captured the better the image and the SNR.

And its also a lot down to post processing. As an example if I were to give my data to folk here, they would end up getting a much better image from the very same data that I used for my image 🙂

Edited by AstroMuni
  • Like 2
Link to comment
Share on other sites

  • 4 weeks later...

Little update - I posted a recent pic of M33 over here.

As noted there it was the first chance I'd had to put into practice the tips everyone has kindly given in this thread. Still a way to go but I'm pleased with the improvement over previous images of M33.

I think 120 x 60s has made a big difference compared with earlier efforts using only 30-60 light frames. Also it seems ISO 400 is plenty good enough and leaves plenty of leeway for messing with the histogram 👍

Question: I "only" did 10 each darks, flats and biases (vs. 120 lights). Would there be any benefit to doing more of these, say, 20?

Link to comment
Share on other sites

I have the Nikon D5600

ISO invariant as you note. The sensor captures the same amount of photons regardless of ISO.  ISO is an in-camera signal amplification and the D5500 and D5600 work best at ISO 200 or 400.  I shoot at ISO 400 for better dynamic range, rather than 800 or 1600. I usually use 120 or 180s exposures but you will need guiding for long exposures. Also depends a lot on sky background light pollution

With guiding you can dither and get rid of walking noise from drift.

The histogram is useful but I found I could get good images even if the histogram appeared hard left, as long as the exposure time was sufficient (I was in dark sky location though)

Long exposure thermal noise was an issue for me with ambient temperatures of 30 to 40C. Much less of an issue in winter northern climates

For targets with high dynamic range like Orion, I would take shorter exposures for the core and longer for the nebulosity. Like 12 x 30s and 64 x 150s for nebulosity. Darks can be a bit hit and miss with DSLRs because of thermal matching, but flats are required if stretching, and bias are easy to do. 

 

 

 

Link to comment
Share on other sites

On 21/12/2022 at 15:50, AstroMuni said:

Have you seen this? I know its not about DSLRs but it will give you a bit of background on the topic as the sensors are pretty much the same on both DSLR & Astrocams.

This is a very good video, some very useful tips there. Thnkas for posting up

 

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue. By using this site, you agree to our Terms of Use.