Canon to NR or not to NR?

[obscura]

I know there have been similar threads but not current as far as I can find.

Darks.

I have read fridging the camera and taking darks at say -4 to +8C for the lights criteria incl. ISO, shutter etc. This seems to be a good idea except they need to be redone say annually due to sensor changes.

Canon NR. The downside is doubled imaging time and the darks are subbed at that time (shame). Upside is that darks are taken exactly under the same conditions as the previous light.

Are there any other cons wrt using the Canon NR option?

[JKB]

As far as I know you don't require as many darks as lights? Please correct me if wrong. So using NR essentially increases you imaging time. I would assume it would be better to get more light frames in the time available then do 20-30 darks.

Cheers

Jamie

[obscura]

There has been a number of posts concerning lights and darks and the average appears to be 30 of each tho' some run up to a hundred! At 30 apiece, something like 3 hours in total for say 30 x 2 x 2-3mins @ ISO800 after initial scope/mount set up.

Also depends on where the DSO is and where it'll be after a two ot three hours or so.

It'll come down to time available once conditions are more favourable.

Thanks for your input Jamie

[Photosbykev]

Turn off the in-built NR, grab as many lights as you can and then shoot 30-40 darks to control the noise. There is no reason why you can't shoot the darks on a cloudy night and gradually build up a dark library at differing temperatures.

[Dave Smith]

Agree with Kev. There is no point in wasting clear sky time taking darks. I usually set my camera doing the darks while I go to bed and I take the same number of darks as lights.

Dave

[obscura]

A good thought, I may well give that a try. Prefer it to the fridge and changing its temp.

Thanks Kev

[Psychobilly]

Over a period of time you will build up a large number of darks covering a whole range of temperatures .. and theres software out there to sort them out...

[obscura]

My train of thought.

I had been thinking that a light and a dark throughout some hours would be better temperature matched. Doing darks after the lights could mean a far lower temperature. Also, would the exposure be 2m, 3m, 5m? ISO 400, 800? NR method accomodates these variables at the time.

The sad thing is that the Canon dark is not saved seperately.

One reason for the question was 'is the Canon dark inferior to a seperate dark'. I have to reason - no, it isn't.

I recognise the general concensus is grab your lights while you can and I take this on board. I will muse further.

[roundycat]

For my money, although I don't use a DSLR, you can spend too long taking darks. Especially if the temp control is questionable. The idea of taking 100 darks is, to be honest, ridiculous. You are way into the realm of diminishing returns. Although I think the AIP book is out of date in many respects, they recommend 5 darks. If you look at the added noise that is not enough, ten is much better. Thirty is a good trade off between reducing the added noise from the master dark and spending half your life taking them.

If you can keep your background signal above the read noise by even a small amount good noise reduction software, judiciously applied, will make up for the 'missing' 20 darks if you only shoot ten.

Hot pixel filters, dithering and sigma reject combining all help to keep it clean.

Dennis

[Photosbykev]

My train of thought.

I had been thinking that a light and a dark throughout some hours would be better temperature matched. Doing darks after the lights could mean a far lower temperature. Also, would the exposure be 2m, 3m, 5m? ISO 400, 800? NR method accomodates these variables at the time.

The sad thing is that the Canon dark is not saved seperately.

One reason for the question was 'is the Canon dark inferior to a seperate dark'. I have to reason - no, it isn't.

I recognise the general concensus is grab your lights while you can and I take this on board. I will muse further.

A single Canon dark applied in-camera isn't going to be better than a number of darks taken in the appropriate conditions. You also have no control over the application of the Canon dark to the light so you get what you get. Using the built-in NR is certainly better than doing nothing at all.

[obscura]

The difference is then that a bundle of lights become a single master and a bundle of darks become a master and both masters with bias and flats are all bunged together to create a final image for processing. NR operates on a single frame and therefore not as effecive for camera noise, pixels etc. I had overlooked/forgotten this point.

[Shibby]

As far as I know you don't require as many darks as lights? Please correct me if wrong.

If you're going to use darks, it's just as important to get as many subs as possible because they introduce noise. But there does come a point where extra darks don't have much effect. There's a nice graph here: CCD Read Noise Summary

Using the built-in NR is certainly better than doing nothing at all.

I'm not sure about that, I think it depends how much noise you have in the first place. Sometimes (e.g. when you don't have many) it's best not to use darks at all, but I suppose with bad amp glow (like I had on the D70) darks are more necessary.

[obscura]

That link is interesting from which I can see why 30 darks are 'optimum' being on the lessening gradient. I think I am now convinced that darks to be taken at similar temp to subs @ 30 of each (+bias & flats). Not NR.

Next step is to make a cunning plan for DSO imaging.

Many thanks to everyone for their thoughts on what to me was a nagging problem.

[dph1nm]

The difference is then that a bundle of lights become a single master and a bundle of darks become a master and both masters with bias and flats are all bunged together to create a final image for processing. NR operates on a single frame and therefore not as effecive for camera noise, pixels etc. I had overlooked/forgotten this point.

It makes no mathematical difference whether you subtract a single dark from each light, or average up all those darks into a master dark first then subtract that from each light.

NigelM

[obscura]

Hi Nigel. I am going to leave that to DSS - until I get to grips.

Cheers

[roundycat]

It makes no mathematical difference whether you subtract a single dark from each light, or average up all those darks into a master dark first then subtract that from each light.

NigelM

If I understand you correctly this statement flies in the face of everything I have ever read or, come to that, proved mathematically. Aside from which, averaging single darks to make a master is pointless if any one of them contains a cosmic ray hit.

Take a look at the graph at the bottom of the page in the link Astrophotography

It looks much the same as all the other graphs I have seen and clearly shows the reduction in added noise you can expect to see when the number of darks goes up.

I think this boils down to something about single dark subtraction versus subtraction of the darks added in quadrature does not commute.

Dennis

[dph1nm]

Take 10 lights and 10 darks. Pair them up and subtract a different dark from each light. Then average the resulting dark-subtracted lights. This is exactly the same mathematically as taking the 10 darks, averaging them together to form one master dark, then subtracting this master dark from each light, then averaging the master-dark subtracted lights. Of course, the maths only works if you have equal number of darks and lights.

((L1-D1)+(L2-D2) ..... + (L10-D10))/10 is the same as (L1+L2 ... +L10)/10 - (D1+D2 ...+D10)/10 which is (L1+L2 ... +L10)/10 - MD which is ((L1+L2 ... +L10)- 10MD)/10 which is ((L1-MD)+(L2-MD) ... +(L10-MD))/10

where L1 ... etc are the 10 lights, D1 etc the darks, and MD the master dark.

There will be a slight difference if you don't average, but do some sigma clipping or somesuch to remove cosmic rays, but hardly noticeable (unless you have millions of cosmic rays in your darks - in which case I wouldn't use darks at all).

NigelM

[roundycat]

It is not as simple as (D1+D2...+D10) as the noise in the dark stack added to the noise in the light adds in quadrature. The noise in a single dark subtracted from a single light would add 41.4% of its noise to the light. You would have a stack of lights with an extreme amount of noise added to each one before you actually combine them. Make a master from ten darks and the added noise drops to less than 5%.

See AIP p135.

Apart from anything else, that is what everyone does.

Dennis