Advice on subs and stacking.

[Astro Noodles]

I am hoping to get 8 hours or so of integration time on a specific target. Obviously this will take me several sessions to achieve at this time of year. My method is to collect a single night's data with darks taken before the session and stack these with a master bias. Then, I will stack the separate night's files together in DSS to form a composite of each individual night's imaging.

Is this the correct way to do it, or should I forget about the darks and just save all the subs and stack them together in one go?

[Xilman]

What is your camera?

With my Starlight Xpress CCD a master dark lasts for months and I generally build one in daylight hours. A master bias is generally taken then, and flats each time the image train is adjusted.

Accordingly, each image is auto-calibrated by MaximDL as it is taken and I stack the complete series in one go --- after rejecting any with guiding errors, egregious satellite trails, etc.

[Astro Noodles]

Camera is a stock Canon 1100d.

[Xilman]

5 minutes ago, Astro Noodles said:

Camera is a stock Canon 1100d.

In that case, you should pay much more attention to those who reply (if any) who are familiar with CMOS DSLR cameras. They can't be cooled to a standard temperature so master darks are likely trickier unless you can create one for each temperature at which you observe.

I picked up a couple of very cheap Canon cameras on eBay, one very cheaply because it can't take exposures under 30 seconds or thereabouts, but haven't yet used them on a telescope so I will also be paying attention to subsequent responses.

 

[Xilman]

13 minutes ago, Xilman said:

They can't be cooled to a standard temperature so master darks are likely trickier unless you can create one for each temperature at which you observe.

Thinking about this: temperatures generally change during the night so perhaps darks should be taken at start and end of the session, and perhaps part way through as well?

Keep in mind that I don't really know what I am talking about ...

[Martin Meredith]

I don't have a cooled camera so my approach is to collect darks at various points during the observing season (certainly not every time I observe) and note the ambient temperature (and duration, of course). It doesn't take too long to build up a darks library in this way and so long as you also note the ambient temperature for your lights you can reuse such darks many times.

Martin

[Astro Noodles]

Thanks Xilman. I have had that thought myself. I take the darks before the session whilst I am setting up my kit as it's going to be the warmest. I don't know whether that is the correct thing to do.

[rickwayne]

Definitely shoot bias (or flat darks, if you prefer) and flats. Darks for DSLRs are not divinely ordained -- some need them, some do not. My Pentax's sensor has so little dark current at cool ambient temperatures that darks actually impair the image. Not saying you DON'T need them for your  camera, just that you should experiment and see.

Flats IMO are must-have, and you need either bias or flat darks to calibrate with those (the latter are just darks that you shoot at the same settings/exposure as the flats).

As for stacking: while the math for averaging pixel values works out the same whether you stack each session and then stack the integrations or do one massive stack of all sessions' frames, many other things do not. In particular, rejection algorithms for outlier values (satellites, airplanes, cosmic rays), as well as background normalization, are statistical and work much better for larger data sets. If you're just doing  simple  averaging then no, it  doesn't matter.

[Xilman]

20 minutes ago, rickwayne said:

Flats IMO are must-have, and you need either bias or flat darks to calibrate with those (the latter are just darks that you shoot at the same settings/exposure as the flats).

Absolutely essential, IMAO, and for all types of cameras. You might get away without them if all you want is a somewhat pretty picture and there isn't too much crud on your optics but I personally don't feel lucky. If you want something of scientific interest, meaning subsequently measurable, flats are absolutely essential.

Even if all you want is a pretty picture you may want to measure your images in the future, so take flats and keep all your subs for posterity. You never know when they will come in useful and storage is almost (but not quite) free. Which raises another topic on which I may post and offer software I have writted, but that is for another forum.

 

Edited July 10, 2021 by Xilman
add "all" in first sentence

[vlaiv]

41 minutes ago, rickwayne said:

As for stacking: while the math for averaging pixel values works out the same whether you stack each session and then stack the integrations or do one massive stack of all sessions' frames, many other things do not. In particular, rejection algorithms for outlier values (satellites, airplanes, cosmic rays), as well as background normalization, are statistical and work much better for larger data sets. If you're just doing  simple  averaging then no, it  doesn't matter.

Not quite so.

Take simple example:

5,5,4 and 6,5

average of 5,5,4 = 14/3

average of 6,5 = 11/2

Average of those two is:

(14/3 + 11/2) / 2 = ((28 + 33) / 6) /2  = 61/12 = 5.08333

But if you add them together( 5+5+4+6+5)/5 = 5

It will work out the same only if you average same number of subs on each night, but if you have different number of subs on each night - you'll get different result if you stack the stacks versus stacking each sub.

 

[Xilman]

1 hour ago, vlaiv said:

Not quite so.

Take simple example:

5,5,4 and 6,5

average of 5,5,4 = 14/3

average of 6,5 = 11/2

Average of those two is:

(14/3 + 11/2) / 2 = ((28 + 33) / 6) /2  = 61/12 = 5.08333

But if you add them together( 5+5+4+6+5)/5 = 5

It will work out the same only if you average same number of subs on each night, but if you have different number of subs on each night - you'll get different result if you stack the stacks versus stacking each sub.

 

I think you just described the desirability of a weighted average 🙂

Some software (SWarp, for example) keeps a weight map for this purpose. Stacking with sum, rather than mean, also preserves this information.

Ideally sum-stacking or weighted-average should be used. If all you want is a pretty picture it doesn't really matter in practice. As long as you keep the original subs you can re-process when it does matter.

[alacant]

On 10/07/2021 at 15:37, Astro Noodles said:

Is this the correct way to do it,

1100d?

I wouldn't recommend dark frames of any kind.

Bias and flat frames, a big dither in-between each light frame followed by stacking with a decent implementation of a clipping algorithm is the best we can get with Canon DSLRs. Especially the older models.

Try with and without to see for yourself? 

To have more control over calibration and stacking, try Siril. It also has a good dark optimization routine if you find that your sensor responds to it.

HTH

Edited July 11, 2021 by alacant

[callisto]

7 minutes ago, alacant said:

1100d?

I wouldn't recommend dark frames of any kind.

Bias and flat frames, a big dither in-between each light frame followed by stacking with a decent implementation of a clipping algorithm is the best we can get with Canon DSLRs. Especially the older models.

Try with and without to see for yourself? 

To have more control over calibration and stacking, try Siril. It also has a good dark optimization routine if you find that your sensor responds to it.

HTH

????????????????????????????????????????????????????????????????????????  I might sell all my AP gear an stick to visual 🤪

[Xilman]

10 hours ago, alacant said:

1100d?

I wouldn't recommend dark frames of any kind.

Bias and flat frames, a big dither in-between each light frame followed by stacking with a decent implementation of a clipping algorithm is the best we can get with Canon DSLRs. Especially the older models.

Try with and without to see for yourself? 

To have more control over calibration and stacking, try Siril. It also has a good dark optimization routine if you find that your sensor responds to it.

HTH

I respectfully disagree.  As I said earlier in this thread, you can get away with quite a few shortcuts if all you want is a reasonably pretty picture.

If you want something which is of scientific value, and therefore measurable, you should be much more careful. AAVSO has a very good document on DSLR photometry which goes into detail on how and why to calibrate your images.

Please keep in mind that you may well only want a pretty picture now but subs are (essentially) forever and you may well want to measure them further down the line. Don't deny yourself that opportunity is my view.