Single Long Exposure or multiple shorter ones?

[Rossco72]

Hi all,

I know this might be a bit of a "how long's a piece of string" questions, but as a general principle is a single long guided exposure better or worse than a stack of shorter exposures for Nebulas or Galaxies?

I am using a Canon 450D which has had the canon filter removed and I have various filters to try out UV/IR, Ha, SII, OIII etc.

Any guidance, pardon the pun!, welcome.

Cheers

[ashenlight]

My instinct on this is that a stack of shorter subs is better. Perhaps this way your guiding doesn't have to be perfect - a long shot would show trails etc? Perhaps shorter subs would reduce noise too. I can see that Catanonia is viewing this thread so hopefully you'll get some good advice there Plenty will be along soon... be good to see if my instinct was right!

[jgs001]

I've seen this one going around and around. I've got good results including Ha neb showing on a non modded camera using 40s subs... ok, I shot over 150 of them to do it. It's a fine balance in terms of sub length between what you gear and local conditions will allow and how many subs... of course, I'm not sure there's such a thing as too many subs. From what I understand, a single long sub is going to not have as good a SNR as many subs stacked. Also, if you do get a cloud pass through, or satellites, or a plane, they seem to happen a lot, then you will have to manually edit them out. With stacking, the stacking process can remove them from the end result anyway.

[roundycat]

Don't know about your particular camera but there are many general principles that apply here.

Theoretically, and I stress that, one sub will do the job but in practice you have to deal with a host of things that muck up a single exposure. Exposure per se has to be a balance between the amount of signal you can detect through your local sky and the amount of noise the camera will pick up or generate internally.

Typically, skyglow, aircraft trails, satellite trails, cosmic ray hits and other stuff will come at you through the camera lens. Thermal noise, hot pixels, read noise and fixed pattern signal (often called FP noise) are introduced by the camera/optical system.

Some of this can be cleanly removed by data reduction (calibration) some of it you are stuck with. In essence you have to accept the compromise. Too little exposure will lead to a signal that struggles to make itself seen and too much will saturate the chip with skyglow/light pollution. Also you have to be aware that the chip can be simply overloaded by too much signal in which case it will go non-linear and bright objects will bloat. Imagine it as the stars getting so big they join up.

Short exposures may seem to make guiding less of a problem but at the expense of too much noise or, in a stack, co-related noise which looks even worse. Whichever way you cut it shorter subs do not reduce noise. Quite the opposite. A stack of relatively short subs will not give a signal noise ratio better than a single exposure but bear in mind the caveat about long exposures above.

The actual minimum exposure will depend on object brightness, camera sensitivity (don't confuse with ISO setting), f ratio for extended objects and any filter factors that have to be taken into account. The main limit for exposure time is skyglow, the chip full well depth and the normal problems of guiding accurately.

I would suggest starting at around 1-2 minutes @ 400ISO for a set of ten subs, combine them using an outlier rejection algorithm and then carefully stretch the data and see what you have got.

Noting your filters I would stay with either none or Ha on a bright nebula (NGC7000) until you have a feel for what is happening.

Dennis

[dph1nm]

If you are stacking, then you should try and make sure your individual subs are not dominated the camera read-noise, otherwise stacking will give worse s/n than one single long exposure.

If you have significant sky glow then read-noise is probably not going to be an issue - where it can become a problem is if you have a very low sky signal - particularly the case in narrow-band imaging, or if you do very short exposures.

NigelM

[Ford Prefect]

As decribed above, its all to do with signal to noise ratios and local conditions.

I found this artice very informative when trying to get my head around things

Signal to Noise: Understanding it, Measuring it, and Improving it (Part 1) - Article

Cheers Jon

[ncjunk]

Stunning link there Jon, but I am going to have to re-read it as it was too much information for one read!

Neil C

[Ford Prefect]

I think I read it 4 times and still didn't get it all

Jon

[Whippy]

I know how you feel Jon! I get to the point where I know theoretically I can do x, y & z to get a great picture but most of the time it's just not possible so I have to sit back and just look at the image and say 'is it a nice image or not?' if it's yes then great, if it isn't then I'll have another crack at it at some point as it normally down to not having enough exposure time. I can wait for another time.

Tony..

[narrowbandpaul]

i think we could talk all day about whether longer is better.

i have found that in low signal applications, such as narrowband imaging, longer is always better. There reasons are quite simple.

Shot Noise.

this noise is created simply by detecting photons themselves, its nothing to do with the sensor or its temp. This scenario is the best we can ever do.

and your camera will be shot noise limited if....

you collect sufficient signal where the shot noise is greater than the read noise.

Most cameras exhibit around 7e- read noise. To be shot noise limited you would need more than 49 signal electrons. Most astro images have these.

if you are in the shot noise region (you can guage this by taking a small box, say 10x10 pixels on an area of object that has an almost constant signal, ie that area of the object emits the same amount of light. If the standard deviation (noise) is approx the square root of the average, then you are shot noise limited), going longer will bring better results. In fact, assuming the object outputs a constant amount of light, true of almost every object, then the signal to noise ratio will increase by the square root of the time increase.

eg, you 4x longer subs, each image will have twice the SNR, you might say twice as good.

So, shot noise limited...longer is better.

if you are in the read noise region, ie very low signals, then SNR actually increases linearly with signal. Doubling exposure time will double the SNR.

interestingly, and I only just realised this, if you stack say several frames that are read noise limited then the SNR will increase only as the square root of the number of images. Therefore it is more efficient to increase exposure length.

if you have lots of signal, then your sensor may enter the fixed pattern noise regime. Bad things now happen.

all other noise sources mentioned are random, and stacking improves the noise. But fixed pattern noise is fixed. Stacking will not improve the SNR, nor will exposing for longer. You are only wasting your time.

you may be FPN limited if you do RGB/lunar/solar/planetary imaging, where high signals are encountered. This may also happen if you image from a bright, high LP sky.

you can remove FPN, simply by applying a flat field image. Correct application of flats yields shot noise behaviour, and life is once again good.

last scenario....

high LP....most people think that you should stop exposing once you start seeing the background. But the sky background actually adds more shot noise to the image. And shot noise limited images are improved by longer exposures. You need longer subs to overcome the extra shot noise.

bottom line.....you need several frames to reduce the noise, but you want long enough subs so that you get good signal from the target and each sub has a good SNR.

i would say decide on how long you will image the object for. Divide it in to say 10 equal sized pieces. The result may or may not be do-able depending on your mount/tracking capablilities, but...

go for as long as you can without introducing tracking errors......

if you can do 5mins.....12x5min would be good for an hour of subs. so would 6x10, or 10x6mins.

for two hours, 12x10, or 24x5, 20x6mins, but

LONGER IS BETTER

hope that helps

paul

[Rossco72]

wow, a lot of info. I guess experience and a lot of reading will help but i also think a lot of practice and experience will be more telling!

To answer a few questions raised above, i have a decent dark sky location (i have to travel to observe or image!!) I have only just purchased a guide cam and scope so my guiding is still questionable until i get more practice.

At the moment my difficulty is that the sky never reaches "astronomical darkness" as the sun is only just below the horizon and there is permanent sun glow covering about 1/3 of the sky but during winter this will improve, i hope...

Anyway, thanks for the good info, i appreciate your help and views.

Cheers

[Sam]

There was something posted a while back - sorry can't find the link - that mentioned that when using a DSLR the noise generated by the camera made it pointless going longer than about 5mins. I think it was something to do with it being not cooled and the thermal noise building up to quickly after about 5mins - to the point where darks don't really fix it much.

Practically I have found this with both the Canon 400D and 1000D and now limited exp to 5 mins (when the conditions permit).

Sam

[jgs001]

I didn't know that Sam. My M51 image is a sequence of 10 and 20 minute subs with the 450d, I didn't use darks on that image.

[brendant]

A lot of good advice already given but from practical experience with a 400D, I agree with Sam (Lightbucket) that 5-6 minute exposures are about optimum. Light pollution is generally the limiting factor for most of us even from relatively dark skies.The longer you go the more demands you place on your guiding, the more chance of satellite trails etc.. ruining your subs.

Generally the more subs you have with DSLRs the better as you get a smoother result, if it was a choice between 3 x 20 Min subs or 10 x 6 Min's I would always take the latter (10 x 6 Mins). With a DSLR you will always require more integration time over a CCD because they are not as sensitive, aim for a minimum of 1 hour total exposure time on each object for starters.

Another thing to note the sweet spot on most Canons is ISO800, not sure if it applies to the 450D but it will not be a bad starting point in the balance between noise and sensitivity.

Rather than try and start guiding with 20 min subs, why not try with say 3 Min @ ISO800 and work up from there.

Brendan

[roundycat]

I would say that too much is made of astronomical darkness. This time of the year I often start at 10:30 and finish at 03:15, the only stars visible in the sky to me are the very brightest ones, my camera manages to see a lot more.

Dennis

[narrowbandpaul]

i agree dennis

[Catanonia]

All valid points.

One thing to remember as well is disk size and processing time.

The other night I broke my shutter release cable and was limited to 30 second subs. Now 70 mins worth equated to 140 subs each of 8mb in size. That was a lot of data and the stacker program needed massive amounts of disk space to process it all. About 93GB worth of space.

Luckily I had my spare server from work and could transfer the data there for processing. Otherwise I would have been left with massive amounts of data and no way of processing it.

I tend to go for 8 mins subs with a LP filter. Gives enough time, reduces cloud cover and satellites and doesn't flood the CCD in the modded 350D i use.

Experiment is the key. If your subs / raws are past pink / brown in colour then you may be going to long.

[roundycat]

If you have too many subs to process at once do them in batches of, say, ten and combine them using an outlier rejection algorithm. Then combine the batch masters as an average.

Dennis

[Catanonia]

If you have too many subs to process at once do them in batches of, say, ten and combine them using an outlier rejection algorithm. Then combine the batch masters as an average.

Dennis

mm, always wondered if you could do this. Can you give simple steps on how to achieve this in DSS please ?

[roundycat]

Sorry, I use Maxim, never even seen anything else in action.

Dennis