3 minutes vs

[toml42]

I've found I'm able to push my unguided exposures up to 3 minutes - if I'm willing to sacrifice 80% of the frames to what appears to be periodic error.

Is it worth persevering with those odds, or would i be better off taking the safe option and dropping back to 90 seconds or so?

say i have a time budget of half an hour, i can either take 10 3 minute exposures and discard 8, or take 20 90 second exposures and only have to throw away about half of them.

so it's 2x180 seconds vs 10x90 seconds.

any thoughts?

[ollypenrice]

I'm sure it should be 10x90. Per frame the S/N will be worse but the stacking algorithms need a good number of subs to do their thing.

Olly

[toml42]

ah, i went for the 3 minute subs in the end, i took it as a test of manliness

the odds weren't as bad as i initially guessed, it was more like discard 2/3 or so, and i got nearly an hour of data.

http://stargazerslounge.com/imaging-deep-sky/151111-first-attempt-m51.html

You're probably right about the S/N olly, next time i have a clear night i'll take the opposite approach on the same target and compare the results.

[ollypenrice]

That's pretty good signal but you still have considerable trailing. I'd speak to Santa about an autoguider! They're getting cheaper...

Olly

[toml42]

Yeah, another result of me pushing it with the exposure length... the stars were round up to about 2 minutes, i think i'll settle there as a compromise next time.

I'm going to experiment with guiding when i get my hands on the eqdirect cabling (hopefully soon!)

[roundycat]

your M51 will clean up quite a bit and the posted version is miniature. There must be plenty of scope for getting the original to look a bit better.

I would go for more time spent on polar aligning. Although you cannot align on the pole star you can go through the motions in daylight. Then use the longest exposure you can trust and lots of them.

Make sure you use good alignment software.

Dennis

[pete_l]

so it's 2x180 seconds vs 10x90 seconds.

any thoughts?

I read somewhere that the darker the sky, the longer the sub. You get the biggest benefit from long subs where you have dark skies.

[toml42]

Dennis, What steps did you use to improve the image? was that just tweaking the levels or did you use any other processing steps?

cheers

[roundycat]

Tom,

your image was a bit on the dim side inasmuch as none of the stars had reached level 255 or peak white. You don't always want that but you also don't want them stuck at 195 or whatever as it just looks dim. Always maximise your dynamic range but preferably without saturating everything in sight.

So, Curves to brighten, High Pass filter under a star mask to accentuate the galaxy detail and Neat Image to reduce the noise present because of the shortness of the exposure. A final touch of Curves to smarten up the contrast/brightness. None of this works very well on a jpeg.

Dennis

[Daniel-K]

id be chuffed at that pat on the back for you

[blackparticle]

This link appeared in another thread.

Astrophotography Test: Canon 20D/40D - ISO Setting - Sub-Exposure Length | philhart.com - Astronomy & Photography

I'd love to hear the SGL experts view on this as it suggests that many short subs = exactly the same image as a few long ones and it's the total duration of imaging time that makes the difference.

With that in mind then maybe it is better to reduce the exposure times so that you never waste a sub at all?

[FraserClarke]

many short subs = exactly the same image as a few long ones and it's the total duration of imaging time that makes the difference.

Yes -- with a few caveats that depend on the characteristics of your camera and your site.

All cameras have an intrinsic noise (read noise) which is added to the image once for every sub. That's a fixed amount though, and independent of sub length. Other noise sources, which do depend on sub length, are dark current and sky background (the main one).

Basically, as long as the read noise is lower than the dark current and/or sky background, there is no point using longer subs. You might as well just take more subs.

Exactly where that trade-off happens depends on your site (how bright is the sky) and camera (how much is the read noise). For most people using broad-band (LRGB) images, it will happen pretty quickly (a minute or two). If you have a really dark site, or are using narrowband filters, it will be worth making each sub longer.

[roundycat]

I'd love to hear the SGL experts view on this as it suggests that many short subs = exactly the same image as a few long ones and it's the total duration of imaging time that makes the difference.

With that in mind then maybe it is better to reduce the exposure times so that you never waste a sub at all?

Many short subs will never equal one long one. In the first place read noise will become a real problem with lots of short subs (find out what the read noise of your camera is) and there is the added problem of short exposures, by definition, never being deep enough to 'see' the low level detail you are doubtless after.

Lots of exposures (at least eight) help no end if you can stack using a sigma reject algorithm. Nigh on all of the bother from sat and con trails, intermittent dodgy guiding, a lot of the hot pixels and so on can be automatically rejected leaving you with the inescapable conclusion that throwing subs away is daft.

One of the tests I did some while back was to take sets of different exposure length. These were combined in the normal way (MaxIm, manual two star alignment, SD Mask) and the resultant masters were studied carefully. Although the masters were arranged to have the same total exposure the one with the longest subs was clearly superior to the others. Almost certainly down to read noise. And that using an ST10 with double co-related sampling.

Dennis

[dph1nm]

there is the added problem of short exposures, by definition, never being deep enough to 'see' the low level detail you are doubtless after.

This is (mostly) a myth - it doesn't matter what you can or cannot see in the sub. If you stack enough, the detail will appear.

NigelM

[Sam]

This is (mostly) a myth - it doesn't matter what you can or cannot see in the sub. If you stack enough, the detail will appear.

NigelM

I'm not sure about that - if enough photons don't hit the sensor then it won't register a signal then you end up stacking lots of nothing. The longer exposure increases the chances of enough photons hitting the sensor and resulting in a signal being recorded.

[blackparticle]

A definitive test of this this would be to find two identical cameras and scopes, put them on a side-by-side plate on the same mount and do an imaging session.

One camera could be set for a long exposures (yet still within the mounts capabilities so no subs are dropped due to PE) and the other to much shorter exposures (but long enough to achieve maximum sensor saturation of at least one star in the image frame)

While the results of this test probably wouldn't be relevant to those with a more advanced setup where guiding is used, for astroimagers on a tight budget I think the results could be very useful.

One of the first images I did was M51 as well. I spent 5 nights gathering data with subs in the 20 - 45s range on an unguided mount.

At the time, I was using Nebulosity for the stacking process but had to manually "grade" each image by eye as I had rotated the camera between each session to account for any differences in bias and read-noise (plus I couldn't actually see the galaxy so everything was done on a best-guess positioning). The Nebulosity "auto-grade" function was giving me incorrect results as I think it is based on FWHM and the number of stars in the image.

After getting DSS, I went back to the subs and re-stacked them. This time I just used the FWHM figure as a guide and excluded at least 50% of them due to dubious PE. It turns out that I retained most of the shorter ones with only a few of the longer exposures making it through.

The result I got was far superior in detail and sharpness although it lacked the tonal contrast of the one done in Nebulosity.. but that was to be expected as I'd used less than half the data.

This leads me to think that if I'd kept all the subs at 20s, I would have achieved a far superior image rather than trying to push the capabilities of the mount and traded single-frame exposure time for total exposure time.

[dph1nm]

I'm not sure about that - if enough photons don't hit the sensor then it won't register a signal then you end up stacking lots of nothing. The longer exposure increases the chances of enough photons hitting the sensor and resulting in a signal being recorded.

OK, so I said "mostly" because there is the issue of quantisation. I think you have to have your gain set so that you sample your noise well enough (in luminosity). If your noise has a value of 10 e- and you sample at a gain of 10e- per ADU then I admit you are in trouble.

What constitutes 'well enough'? I read a paper on this which claimed, if I remember right, that it is our old friend Nyquist again - i.e. you only need a couple of ADU covering the width of the noise. Not sure what kind of picture this would result in mind you! But I suspect most people sample the noise much better than this anyway.

Note that (from what I read) it is sampling the noise that matters, not the signal from the object of interest. I think the reason is that that even 1e- will add to the background signal and occasionally skew its value by one ADU, even if the gain is such that 1e- on its own would not trigger an ADU.

Perhaps an easy way to imagine it is just to make sure you have a gain of least 1ADU per electron. Then you will not lose anything from quantisation.

On a slightly different tack, even if your object gives less than 1e- per sub (on average), this is still not a problem. Some subs will have an electron and some won't - if you stack enough it will average out to the correct rate.

NigelM

[martin_h]

I have 2 ED80's side by side and 2 QHY9's attached - one mono and one colour, if I bin 2x2 both, thus getting rid of the matrix would this serve as a test rig? If anybody is interested, suggest a suitable target and imaging times and I'll give it a go on a cloudless night (as if!).

[blackparticle]

I think the bayer filter would skew the results as its effectively blocking photons. The cameras would have to be the same for it to be a valid test.

If someone does try it I'd suggest a target with a high degree of nebulosity as it's the tonal contrast in the wispy bits that will really show the difference.. or not.. as the case maybe.

The Iris Nebula would probably be a good one as we have Olly's image as a reference and it's located fairly near Polaris so will be an easier target for an unguided mount.

IC 1396 (Elephants Trunk) or IC 1848 (Soul Nebula) would also be other good candidates. Anything around there really as they are on their way up during the evening so should provide a good amount of imaging time with the right weather window.

It might be one for doing at a star party where a couple of people could get together for it. At least at the end of it, the images could be combined to produce a single, finished picture.

[martin_h]

Thats why I suggested binning the CCD's thus negating the bayer matrix in the one shot colour...just a thought.

[blackparticle]

Would that actually negate it though?

As the sensor has a layer of tiny coloured filters over each pixel I think the light transmission with and without the bayer matrix is going to be slightly different. It may only be a difference of a few percent when the RGGB is combined but I think that difference would invalidate the test.

It's also probable that the bayer matrix will exclude, or at least attenuate certain wavelengths that the unbayered sensor might pick up.

There is a diagram at the bottom of this page to show you what I mean.

Filter Curves