Stub Mandrel

OK, I'm flying by the seat of my pants here too... On the plus side of using 800 compared to 1600, there will be: half the read noise Not necessarily - my camera has less read noise at 1600 ISO than ISO800 and read noise is independent of exposure time (that's why you use bias frames) don't confuse it with thermal noise that builds up through the exposure. each pixel will accrue twice as many photons and create a signal at source twice as large. Even though this is amplified with a gain half that of the 1600 case, surely this must be better statistically ISO 1600 counts the photons with twice the resolution. Its like measuring a distance in inches or half that distance to in half inches - the number is the same. will the sensor noise be doubled? If we are following the sensor with half the gain, why would the sensor noise double? Wouldn't the effects of reduced gain and increased sub length balance out? Double exposure, double thermal noise (that's why we use darks). the signal is comprised of wanted signal and skyglow signal. Both will double but the difference between the two components will double so would be statistically easier to separate. That's why 1600 may be better there would be twice as many steps between signal and skyglow for any given real difference. Also say I take twice as many shorter subs with ISO 1600 than with ISO 800. Twice as many subs doesn't double read noise, it increases by root two - 1.4 times (because it is random). Same with thermal noise, so if one 30 second sub has 'n' thermal noise, a 60 second sub has 2n thermal noise but two 30 second subs have 1.4n. This is why long exposures (10 minutes + etc.) use cooled cameras. So you are actually better getting your total exposure from lots of short subs rather than a few short subs EXCEPT that if they are too short you may lose details (except a few people like Emil Kraikamp are showing you can get results with 1s exposures and fast scopes!) So my understanding is use the ISO that has the best level of read noise and use the longest exposures you can get without clipping and get as many subs as you can. My experience is that the difference between 800 and 1600 ISO is hard to call with my camera but MORE subs always improves the result. As for exposure I always aim to see a trace of my target DSO on the preview as that means I will always get something when I stretch!

Try Noel Carbineris's actions his remove large/small blue violet haloes works really well on that. I've run both on it, but nothing else:

I think its because the read noise is a much larger proportion of the small signal at 100 ISO.

Yes, I'm afraid! Change from 1600 to 800, you double the number of subs but you would need the same number of them, so twice the duration.

No, because the read noise stays the same regardless of exposure and its better at ISO 800/1600 than, say, at ISO 400. By reducing the exposure I'm reducing the dark noise, whatever ISO I use. They shouldn't really call it ISO; over the last month or so I've unpicked enough to see its completely different from ISO with print films - the print film effect is closer to adding pixels together to get more sensitivity but less resolution. DSLRs don't change resolution, it's more like push processing. Here's a test with three 30-second darks at ISOs 100, 800 and 1600, all stretched twice, first linear stretch to get the same exposure, then an identical gamma stretch (like when you process to show a DSO) to reveal the noise. 1600 is marginally better than 800, but 100 is useless.

Not quite, I experiment to get the 'best' ISO (still not sure if its 1600 or 800!) but whichever I'm going with I change the EXPOSURE to get an image with detail but hopefully no clipping.

Sorry I was sloppy in my explanation. Bit depth and dynamic range aren't the same. My point is that a higher ISO does results in increased dynamic range IN THE FINAL IMAGE - if you compare the captured image and neither image is clipped. I.E. I am talking OUTPUT dynamic range not INPUT dynamic range. So in my example the ISO200 image would have a dynamic range of 140 compared to 70 for the ISO100 image. Obviously higher ISOs clip before lower ISOs so they can deal with less dynamic range in the subject My argument ONLY applies when there is no clipping. Remember old tape recorders? They had a VU meter so you could turn up the gain (= increase the ISO) to maximise the signal to noise ratio and dynamic range of the recording, but at a high gain you were more likely to get clipping on a loud passage.

If you avoid clipping using a higher ISO increases the dynamic range in the image. For example, an 8-bit mono image can have up to 256 levels or shades o grey, 0 to 255. Say a mono 8-bit image at ISO 100 covers the range 0 to 70, it will contain 71 shades of grey (not fifty!) At ISO 200 it would cover the range 0 to 140, and have 141 shades. At ISO 400 it would require the range 0 to 280, but 280 is more than 255. In this case the 200 ISO image would clearly capture the most detail without losing data. Bear in mind this is true for astro images that typically start near the black point (0) and will only approach high values at the centre of the brightest stars. This also shows why a 14-bit DSLR like my 450D (16,384 levels) shows much more detail than my 10-bit 10D (1,024 levels) at the same ISO setting. This is why, at face value, it makes sense to use the highest possible ISO rating without clipping. But working against this is noise, both read noise and thermal noise. If noise doubles from, say, ISO800 to ISO1600 then because the signal doubles as well the signal to noise ratio doesn't change and the extra detail revealed gets lost in the extra noise. Curiously the noise/ISO relationship appears to be different for every type camera or sensor, and perhaps in detail for every individual camera. My reading of various websites is that you need to find the 'sweet spot' for your own camera using online data as a guide to find the likely range (ISO800 and 1600 are quoted as 'best' for the 450D by different websites) and then do some experiments. My own experimentation on this is inconclusive, for my camera any given exposure 800 is less noisy than 1600, but if you stretch an 800 image to match a 1600 one, the 1600 version shows marginally less noise. What to do?

It's important to realise that for most images the bulk and left hand side of the histogram is the empty background, right of the peak you generally have your DSOs and as the peak fades down it moves into being the outer edges of stars with the cores of the brighter stars forming the long, low tail to the right. A good test is to use photoshop's dust and scratches tool and watch the histogram, watch as stars disappear and see the change in the histogram.

If its a fairly recent Nikon it should have a facility to highlight any overexposed parts of the image by flashing them.

A good start Your stars are round but hard to judge focus with short data as no tiny stars. Might be worth getting/making a bhatinov focussing mask as a precaution. I found my M31 kept getting better with more subs, right up to three hours, but if you can get an intervalometer (cheap as chips) and work up to half an hour or more you will see big improvements.

Thinking about it you must be right. Diffraction is caused by edges not thickness. An inch-diameter steel bar across an aperture has the same amount of edge as a spider silk the same length!

Y0ou will lose detail, mostly from JPEG being 8-bit as much as from the compression, but you should still get something usable from 500 frames.

Ah, I'd guessed the basic idea but not understood exactly what it was used for.

I keep reading posts from people who are spending ages 'building their models' I assume these aren't airfix kits, so what are they? Mapping errors in the mount?

I've tried some experiments performing the same linear stretch on 30-second subs at ISO100, 800 and 1600 from my 450D. The basic assumption is that you need to stretch 800ISO twice as much as 1600, and 100 ISO 16 times as much. To simulate this I made three 30-second RAWs and converted to 16-bity tiff without any changes. First I left 100 ISO alone. I used curves to scale 100 ISO so 16-->256 and 8-->128 (second point needed to get a straight line) and similarly to make 800ISO 128-256 After this the 800 and 1600 darks looked pretty much identical but the 100ISO, as expected was not too good! I then applied a gamma correction of 3, which approximates to enhancing faint details without overblowing stars. This brings out the noise pretty much to the level you might see in a single sub, processed. I thinks this 'empirical' test is perhaps more useful than mathematical tests as it shows what will happen if you stretch, then process data at each ISO in exactly the same way after an initial stretch to scale the data identically. Converted to jpegs,I think you can see that ISO1600 probably just has the edge over ISO800. These are crops out of the middle of each image. Bear in mind the ISO800 image has been scaled twice as much as ISO 1600. It's interesting that there is less 'pattern noise' (horizontal lines) at ISO100. Perhaps the ultimate test would be to do this with, say, 50 darks at each ISO. I'll leave that to Nigel :-) ISO 1600 ISO 800 ISO 100

HELLO ALL! If you image using a 130P-DS on a lighter mount, such as an EQ3, you probably use subs of 30 seconds to a couple of minutes. The Alt-Az imagers have recognised the need for some more topics to provide places to discuss the equipment and techniques opened up by more sensitive, lower-noise modern cameras. We also need more showcase threads, like this, so that instead of telling beginners not to bother trying imaging unless they have masses of expensive kit, they can see the possibilities of imaging on light and alt-az mounts. If you have a view on this, please share it here:

HELLO ALL! If you image using a 130P-DS on a lighter mount, such as an EQ3, you probably use subs of 30 seconds to a couple of minutes. The Alt-Az imagers have recognised the need for some more topics to provide places to discuss the equipment and techniques opened up by more sensitive, lower-noise modern cameras. We also need more showcase threads, like this, so that instead of telling beginners not to bother trying imaging unless they have masses of expensive kit, they can see the possibilities of imaging on light and alt-az mounts. If you have a view on this, please share it here:

HELLO ALL! If you image using a 130P-DS on a lighter mount, such as an EQ3, you probably use subs of no more than 60 seconds to a minute. The Alt-Az imagers have recognised the need for some more topics to provide places to discuss the equipment and techniques opened up by more sensitive, lower-noise modern cameras. We also need more showcase threads, like this, so that instead of telling beginners not to bother trying imaging unless they have masses of expensive kit, they can see the possibilities of imaging on light and alt-az mounts. If you have a view on this, please share it here:

Definitely getting there!

it's really making progress and I think that the dust clouds are starting to emerge in the first one. How much data have you got the patience to accumulate? This is exactly the sort of multiple short subs on hard target example we need to develop the processing tools for. I hope you don't mind but I've done a rough stretch and you can see there's more detail that will become less noisy with more subs. I've watched a (very long) video and there's a better way of getting this faint stuff to show with less noise you might want to try in photoshop: Create a duplicate layer, run make stars smaller (noel's actions or any other way) several times over. use spot heal tool to get rid of all bright stars (just set it to a circle bigger than a typical big star and click on them - its magic!). Use dust and scratch to eliminate the remaining small stars, setting of about 15 radius? Then blur to remove all noise and use curves to highlight the nebulosity. Correct the black point. Now mix with the original layer using 'screen' mode and it will brighten the nebulosity without overcooking the stars. You can add in extra sharpening and noise removal. I think this might work on your image. These are the sorts of things we need to be experimenting with IMHO!

Thanks Wim, I've done a stack with no control frames, and it's JUST there. As its a straight, bright lkine I wonder if it's a reflection off an edge inside the camera?:

I don't know perhaps it's a flat issue :-( I've had a go at spot healing it out of the RGB and using horizontal band reduction in the LUM, which seems to have helped:

I posted this in the EQ3 challenge thread as well, but I think its one of my best yet, so please excuse the cross posting:

Here's last night's result, the Pacman Nebula NGC 281. This picture was a stack in DSS of the best 85% of 96 frames taken last night, then processed in photoshop following advice gleaned on this forum. I took had to lose about 20 others that were badly obscured by cloud. Each frame was 120 seconds at 800 iso. I used flats, darks and bias frames and no doubt the cold temperature helped keep the noise down. The telescope is a 130P-DS. My camera is an astro-modded Canon 450D, but this object is bright enough you should be able to image it on most DSLRs.