Tons of Noise - Flame and Horse

[Fordos Moon]

All advice welcomed!

Unguided ED80 on HEQ5 with unmodded 1100d.

21 lights ISO800 x 100 seconds

12 darks

12 bias

No flats ( working on it)

1 light ISO800 x 4 mins ( experiment)

1 dark as above

Last night I thought I would give these targets a go. Following stacking in DSS the image has come out super noisy:

I daren't turn up the brightness as exaggerates the noise even more!

I am now redoing in DSS without the 4 min exposure to see if makes difference.

What should I do to reduce noise? Longer exposures and lower ISO?

[Digz]

Looking good. Some nice detail showing through.

Most people with DSLRs run at around ISO800 and dropping this will mean less data captured for a given sub length thus an increase in sub length to compensate. This will introduce further 'problems' such as increased amp glow for example, it may even push your sub length too long for your sky conditions.

To reduce noise you really need to get more subs, the more the better.

Keep going though, its looking good.

[Fordos Moon]

Thanks Digz your response is music to my ears confirming I'm not doing a lot wrong and can certainly take a lot more exposures! Much appreciated!

[cosmojaydee]

Reducing noise is a fine balance using a DSLR especially with this target, ISO 800 is about right , you could try to reduce the exposure times but at the same time you will need to increase your fame count in all area's lights , dark's , dark flats .....

Good Luck

Jeff

[StuartJPP]

It is a good start, however you need a lot more data. First thing I'd do is generate some flats, they really do help and without them you start to eat into the data, especially using some gradient removal tools.

Because of my setup I tend to shoot at ISO1600, and the maximum exposure time I can get without trailing is about 90 seconds. I haven't had much experience due to the weather but in what limited experience I have got I found that ISO1600 was the best compromise for me, and that you need plenty of data, especially with this target and also with an unmodded camera.

What does your histogram look like per sub? My 90 seconds with an f/4 lens and ISO1600 from a darkish location is about 1/3 the way through the graph. You can clearly see the outline of both the Flame and the Horsehead in each sub on the LCD on the back of the camera though it is faint. If you can get 4 minute subs without trailing and you aren't clipping, that would be the best way to get there.

In my gallery there is a picture of Horsehead and Flame with details on the total number of subs, darks, bias and lights etc. I couldn't have got this without the flats, I tried processing the data for hours until I generated some flats and the image literally popped out. I created the flats using my LCD monitor with a piece of paper in front of it with a white screen.

I have subsequently modded a Canon 1100D but haven't been able to image this again since then due to the weather, hopefully I'll get a chance to one day. If the difference with this target using a modified camera is the same as I got with Rosette/NGC2244 then I can't wait to image Horsehead again.

Good luck!

[cp255]

Looks right for the exposure length, ISO, and number of pictures. Take more pictures and you will get less noise. Flats would also help.

[Fordos Moon]

It is a good start, however you need a lot more data. First thing I'd do is generate some flats, they really do help and without them you start to eat into the data, especially using some gradient removal tools.

Because of my setup I tend to shoot at ISO1600, and the maximum exposure time I can get without trailing is about 90 seconds. I haven't had much experience due to the weather but in what limited experience I have got I found that ISO1600 was the best compromise for me, and that you need plenty of data, especially with this target and also with an unmodded camera.

What does your histogram look like per sub? My 90 seconds with an f/4 lens and ISO1600 from a darkish location is about 1/3 the way through the graph. You can clearly see the outline of both the Flame and the Horsehead in each sub on the LCD on the back of the camera though it is faint. If you can get 4 minute subs without trailing and you aren't clipping, that would be the best way to get there.

In my gallery there is a picture of Horsehead and Flame with details on the total number of subs, darks, bias and lights etc. I couldn't have got this without the flats, I tried processing the data for hours until I generated some flats and the image literally popped out. I created the flats using my LCD monitor with a piece of paper in front of it with a white screen.

I have subsequently modded a Canon 1100D but haven't been able to image this again since then due to the weather, hopefully I'll get a chance to one day. If the difference with this target using a modified camera is the same as I got with Rosette/NGC2244 then I can't wait to image Horsehead again.

Good luck!

Stuart many thanks for your time with this one. With the 100 secs ISO800 each sub gave a histogram pretty narrow about a fifth of the way in from the left. No sign at all of the nebulae on the screen.

Wondering if worth trying ISO1600 x 90 seconds next and lots of subs? I will also get some flats. Or try longer subs? Will check out your images thanks again!

[StuartJPP]

I'd try whatever is necessary to get the histogram further away from the left-hand side of the graph. Ideally longer exposures, I can't do that so I up the ISO to a maximum of 1600.

[ollypenrice]

Your biggest problem is surely the very short subs. 100 seconds is very short indeed. If you can make them longer without trailing you should certainly do so. And then you just need a lot more.

Good to see Alnitak split as a double. If you do get a deeper dataset and find that Alnitak starts to bloat just use this version's Alnitak as a layer underneath and erase to over exposed top layer.

Olly

[IanL]

The "higher ISO = more data" myth again?

The amount of signal you record is not affected by ISO (another word for gain). Whether you use ISO 400 or ISO 1600 makes no difference to the result; your camera receives the same amount of photons from the sky regardless of ISO, and there is no magical setting that makes the sensor chip more sensitive to that light. Yes the image looks brighter 'out of the box' (i.e. on the LCD screen or maybe capture software preview), but that is just a combination of amplification of the signal prior to digitising it (useful) and mathematical manipulation of the data once it has been digitised (not so useful).

Basically, for a given exposure length you can achieve the same result by using ISO400 and stretching the image in post-processing as you can by using ISO1600.

So what are the reasons for/against high iSO?

Against:

- Using a higher ISO reduces the potential dynamic range of your image. If you shoot at ISO400 you can expose for much longer than you can at ISO1600. If you find you are saturating your image (you have pixels that are at the maximum value for your camera) you have saturated and lost detail (bad thing). This will usually be bright stars, galaxy cores or the sky background if you have heavy light pollution.

- Don't bother with silly ISO numbers like 3200 +. On consumer grade Canons, up to ISO1600 the gain is partly determined by on-chip amplifiers and partly by multiplication after the image is digitised. So there is a physical effect on the read-out of the sensor. Above ISO1600, the same amplifier settings are used as at ISO1600 and using a bigger multiplier after the image is digitised. it makes the image look brighter, but it is just a marketing gimmick, nothing more, since you can do the same (with more control) in post processing.

For:

- Higher ISOs can help overcome read noise in the image. The brightness of your sky background also comes in to play. Ideally you would use different ISO settings for your target exposure length and sky background and measure which gives you the best signal to noise ratio. If you don't have time, as a rule of thumb go for about ISO800 for an 1100D, whereas for (say) a 500D it typically better around ISO400.

- Higher ISOs can let you see the results on the back of your camera or in preview software more readily. Nothing wrong with using a high ISO for frame and focus operations, but turn it back down to the optimal ISO (above) once you are on-target and focussed. If you are using laptop capture software, it doesn't make much odds since the preview on screen should auto-stretch the image anyway.

Doesn't Matter:

- People think that higher ISOs make images more noisy. They don't, they just make the noise more apparent but it is the same amount of noise (save for the point about read noise above where using too low an ISO will make the image more noisy!)

So really you should choose an ISO that lets you expose for as long as possible without saturating the target or background, and that gives you the best read noise performance. A really high ISO may be a convenient part of your frame and focus routine, but there is no good reason to use it for actual imaging (and at least one really good reason not to).

As Olly says, longer subs are the key if you can guide for that long (and turn down the ISO!). Once you have reached the limits of your guiding, the only solution is then more subs.

[Fordos Moon]

Wowsers thanks Ian that's some quality low down. My plan is to stick with my favoured ISO800 and more capture time be it with more subs or trying to get longer exposures without trailing.

Being unguided at the moment I'm sure I can up the 100sec subs a fair bit before this happens.

I appreciate your time.

[StuartJPP]

Sometimes I wonder why I bother trying to offer advice on forums when there are so many experts around.

IanL, if you had bothered to actually read any of my responses instead of jumping on the "ISO myth" bandwagon, you would see that I actually offer advice. This is real world advice from my own experiences, which is exactly what the OP was asking for (the very first line of this thread), your experiences may differ which they obviously do. If people choose to ignore my advice, fine, I am no expert, but then again I never claimed to be.

You said that "Whether you use ISO 400 or ISO 1600 makes no difference to the result" however you also say "amplification of the signal prior to digitising it (useful)" so which is it? It either has an effect (useful) or not (not useful). I know that the same amount of photons hit the sensor no matter what the ISO setting is, however it is what you do with those photons that count.

Until someone from Canon's imaging sensor department comments exactly how the ISO algorithm is actually implemented, I will do what I find is best for me, not what someone else tells me what is best. I have actually tried imaging at ISO800 and find that ISO1600 is better for me under these types of short exposure circumstances, perhaps it is down to my processing method that differed between the two or magic pixies I really don't care.

With regards to your ISO1600 vs. ISO400 statement being identical, for the same exposure time, how I understand it, the amplification of the signal prior to digitisation will beat amplification after digitisation when operating on data near the noise floor (and all else being equal). If the values captured were real numbers instead of integer values then I'd say things would be equal cranking up the exposure by 1 stop after the event. However they aren't real numbers, so any cranking up loses detail in the lower bits of the original ADC value. In fact it could artificially make the data seem cleaner because the lower bits are all zero after amplification which gives the impression of being clean. We don't know what Canon really get up to in the exposure triangle regarding ISO, what the amplifier gains actually do, what noise they introduce, so as I said until Canon comment I will keep on using what I use because it works for me.

If the OP or anybody can take longer exposures and get more of them, then that is the way to do it, which I said even before Olly did. I would love to image at ISO800 for 8 minute exposures, but I can't and from the OP's original post it looked like he couldn't either being unguided. However given the choice of exposing for 90 seconds at ISO1600 or 90 seconds at ISO 800, I will choose the one that gets me away from the left hand side of the histogram as that is where the data is. I'll repeat...that is where the fine data is. The OP has already stated that his histogram is barely 1/5th of the way into the graph, to me this is not acceptable, but what do I know?

[IanL]

My comment wasn't a personal attack on you, and I'm sorry if you feel that it was. I'm not going to respond any further to this thread as I'm surprised stating a few basic facts (based on both theory and empirical evidence from my own tests and those carried out by people far more expert in the field) has resulted in an unexpectedly hostile response.

[Digz]

Chill out guys, there is good information in this thread please don't let it degenerate.

[StuartJPP]

Chill out guys, there is good information in this thread please don't let it degenerate.

I am chilled Lee.

[Welrod50]

Fordos, has anyone suggested turning the camera off for a few minutes every ten subs or so to allow the sensor to cool? I always do this, and even with my Pentax Kx (which is superb at high iso work) it makes a difference!

HTH

[Luke]

A while back I found this page useful (thanks, RobH! ):

http://www.middlehillobservatory.co.uk/articles-primers/Signal%20to%20noise.htm

I was really wowed by the image link at the end of the page, showing how the noise goes down with more images in the stack! I didn't think it would make such a massive difference!

I think more total exposure time would bring the noise down.

[Fordos Moon]

Fordos, has anyone suggested turning the camera off for a few minutes every ten subs or so to allow the sensor to cool? I always do this, and even with my Pentax Kx (which is superb at high iso work) it makes a difference!

HTH

Cheers mate I've not heard that, just been having the ten seconds between exposures but I will certainly give it a go!

[Fordos Moon]

A while back I found this page useful (thanks, RobH! ):

http://www.middlehillobservatory.co.uk/articles-primers/Signal%20to%20noise.htm

I was really wowed by the image link at the end of the page, showing how the noise goes down with more images in the stack! I didn't think it would make such a massive difference!

I think more total exposure time would bring the noise down.

Hey Luke that is pretty cool, thanks for sharing!

[Jessun]

I think your image is surprisingly noise free considering the exposure. This may be because you're not guiding.

Strange as that may sound, in my experience guiding can introduce a bucket load of noise if done without dithering. To dither is to allow the mount to move a fraction between subs so that each pixel on the camera doesn't capture the exact same area of the target. A perfect guiding with some drift due polar alignment 'error' can often introduce streaks. Something to remember when you start guiding .

Letting the camera cool down between the subs seem to be good idea too. No one knows for how long, but 20-30 seconds pops up as a familiar figure.

I think two 'identical' cameras side by side will behave slightly different, so experimenting with ISO and exposure times can never be wrong.

All in all I'd be jumping up and down with joy with a start like that on the HH. I have probably 20 hours worth of capture of that target with a cooled CCD, but due to light pollution it just won't materialize in the stack.

Well done!!

/Jesper

[Fordos Moon]

Wow thanks Jesper that's great to hear! And thanks for the tips I have not heard about dithering before so it is definitely something I will take on board, as well as the 20-30 second cool down period! :-)

[Fordos Moon]

Had a little go with the Bubble but its only in the dark sky for a couple hours before disappearing. I assume its fine to capture subs on other clear nights or does it matter if there are days between clear nights for alignment of sub groups?

This was 32 x 2 minute lights with darks:

[Jessun]

Doesn't matter at all!

Many imagers turn back day after day - year after year - to add data! That's why focus and framing is so important. The bubble for instance is beautiful but perhaps not where you expected it to be in the frame. Next time you add to it, you may frame it differently which means the two sets of data will only overlap in a section of the image - so it will shrink over time....

Nail focus, nail framing, keep your data! Then go back and add more.

What I've seen here in the HH and Bubble is remarkable.

Never worry about noise. Noise means that you haven't got enough subs to average it out (or that you like me try to image from an impossible location). Your camera looks like a keeper btw! I had loads of amp glow on my Canon 600D.

/Jesper

[StuartJPP]

Agree with Jesper, I recently did two sessions nearly two weeks apart on NGC2244/Rosette (link below). As long as you frame correctly (within reason), you can combine from numerous sessions.

http://stargazerslou...erent-sessions/

I know of people who collect and combine data over many years.

[Fordos Moon]

Thanks chaps! I am chuffed with my camera, and the astronomik filter. Looking forward to more clear nights to get out there and get more subs!