Quantity of darks v. exposure

[chris_todd]

I've read the DSLR_RAW workflow thread , and since I'm still learning about image calibration and the image aquisition sequence, I have a couple stoopid noobie questions related to the number of darks and the effect that has on the quality of the resulting calibration.

Is it absolutely critical to shoot your darks with the same exposure as your lights?

If darks should be the same exposure as lights, and you were limited to taking only 15 minutes of lights, would it make more sense to shoot 15 one minute exposures, or three 5 minute exposures?  The DSLR_RAW workflow thread emphasized that having at least 10 darks (and preferably more) would provide the best result.  Of course, in one minute lights, you're going to have a lot less signal than in 5 minute lights, so I guess my fundamental question is where is that balance, when you don't have unlimited time?

[astropixel]

Hi Chris. Not stoopid at all The workflow generalizes in this area. There will always be situations where the exception is the rule. The workflow is not a recipe and should be treated as a guideline only.

Originally, there was no comment about the numbers of calibration frames, or indeed, how they should be acquired. A few people however had difficulty with their data, I think, extending from acquisition techniques (including myself). A few pointers, illustrating the need for reasonable sampling rates of calibration frames, it was thought, may be helpful. If it's confusing though, I'm happy to amend it (I just did in places) .

Yours is an interesting conundrum. Of course, if you don't have time then work with what you have. However, the time invested in processing increases as the results become less satisfying, because of scrimping on acquisition time and calibration frames.

If you are in the suburbs 15 x 1minute subs depending on the ISO (say, 1600) would be more appropriate than from a dark site where 3 x 5 minute subs would be preferable. In any case, 15 minutes is not a lot at all.

I take your point, how many darks should you use? From the suburbs, 15 x 1 minute - same as the lights.

3 x 5 minute darks. Well you might get away with 3 or 5, but increasing the sample improves the chances of trouble free processing.

The sensor will probably get hotter with 5 minute subs than 1 minute subs. So, although the dark time is the same, I question the characteristics of the dark noise as a function of temperature, taking into account any cooling between dark subs. Generally, I would guess that 3 x 5 minute darks, taken in succession would be a lot 'warmer' than 15 x 1 minute subs with the same delay between shots.

I would tackle the time poor astrophotography problem this way. Dither all your light frames aggressively. Take bias and flats only (which is quick) and forget the darks. That way, you obtain 30 minutes of light frames (more SNR). If you dither, bad pixels, noise and other artifacts will either be eliminated or reduced significantly. It may not be the perfect image, but it will be a lot better than 15 minutes worth of data, for the sake of acquiring dark frames.

Perhaps someone else would like to comment on this.

[chris_todd]

Thanks for taking the time to write such a fantastic response!  Man, I love the SNR of this forum! 

Yours is an interesting conundrum. Of course, if you don't have time then work with what you have. However, the time invested in processing increases as the results become less satisfying, because of scrimping on acquisition time and calibration frames.

Perhaps some context would help.  I'm going to the Texas Star Party next month, and I'm planning on trying (emphasis on trying) to do a mosaic of the entire night sky, from the Western horizon at the end of astronomical twilight, all the way to the Eastern horizon just before the beginning of astronomical twilight, with a Canon T2i and a Tokina 11mm lens on an Astrotrac.  I've calculated that a 3x7 panel mosaic should do it (that lens has about an 86* wide FOV on that camera, so while I could do a 2x7, I really want to catch South and North, and I'm hoping the additional overlap will help me deal with some of the barrel distortion inherent to this lens).

I know this is an ambitious project, but life without challenges is boring. 

So in planning out the session, I guesstimated that I can take about one hour for each South to North row of three images (I'll have 8 hours of darkness to do seven rows, and I'll need some slack time for reframing between panels, etc.), so I was thinking about taking 15 minutes of lights for each panel (SW, W, and NW), followed by 15 minutes of darks.  So that's where my question came from.  I'm limited to 15 minutes per panel because the Earth keeps rotating, and TSP is not held in the dead of Winter.   

If you are in the suburbs 15 x 1minute subs depending on the ISO (say, 1600) would be more appropriate than from a dark site where 3 x 5 minute subs would be preferable. In any case, 15 minutes is not a lot at all.

I take your point, how many darks should you use? From the suburbs, 15 x 1 minute - same as the lights.

3 x 5 minute darks. Well you might get away with 3 or 5, but increasing the sample improves the chances of trouble free processing.

The sensor will probably get hotter with 5 minute subs than 1 minute subs. So, although the dark time is the same, I question the characteristics of the dark noise as a function of temperature, taking into account any cooling between dark subs. Generally, I would guess that 3 x 5 minute darks, taken in succession would be a lot 'warmer' than 15 x 1 minute subs with the same delay between shots.

The other challenge I may face is that the current weather patterns show daytime highs of about 80*F and night time lows of about 50-55*F, so I'll have significant ambient temp changes during the imaging session, which is why I thought of taking darks at the end of each row.

I would tackle the time poor astrophotography problem this way. Dither all your light frames aggressively. Take bias and flats only (which is quick) and forget the darks. That way, you obtain 30 minutes of light frames (more SNR). If you dither, bad pixels, noise and other artifacts will either be eliminated or reduced significantly. It may not be the perfect image, but it will be a lot better than 15 minutes worth of data, for the sake of acquiring dark frames.

Perhaps someone else would like to comment on this.

I get the idea behind dithering, and on the Astrotrac, I could aggressively dither without too much trouble (the camera mounts to a Bogen 410 geared head, which then mounts to the Astrotrac, so it's easy to move the camera a degree or so on each shot, though I've discovered that any time you touch the setup, you run the risk of throwing off the polar alignment).

If I took bias and flats only, given 60 minutes per row, that would still only give me 20 minutes of lights per panel, so is that enough benefit to justify ditching the darks?

And another newbie question - how can you take flats in the middle of the night?  The only way I know of to take flats is to wait for dawn, cover the lens with a taut T-shirt and shoot your flats using a fully illuminated sky.

Thanks again for the great info...

[astropixel]

To answer the flats first. You can take them at the end of the session. If you decide to take them at a lower iso, which is OK (not sure it's essential though), you will need bias frames to calibrate them - so 2 sets of bias frames, one for the lights and darks and one for the flats.

A T-shirt is OK.  An EL panel with some paper to reduce the brightness is a good method also.

Alternatively, a TFT screen (laptop). Open up a white frame in PS or something similar - perhaps with some A4 copier paper to protect the screen. You need to have the lens hood in contact with the paper (I never bothered with this, as the screen has a clear protector). Set the screen brightness to give you about 70% camera histogram at 1/10 - 1/8 second.

I get the idea behind dithering, and on the Astrotrac, I could aggressively dither without too much trouble (the camera mounts to a Bogen 410 geared head, which then mounts to the Astrotrac, so it's easy to move the camera a degree or so on each shot, though I've discovered that any time you touch the setup, you run the risk of throwing off the polar alignment).

Stopping and starting the Astrotrac between frames will achieve the same result. I am guessing, but 5 -10 seconds should be adequate.

[tsaban]

...If darks should be the same exposure as lights, and you were limited to taking only 15 minutes of lights, would it make more sense to shoot 15 one minute exposures, or three 5 minute exposures? 

I don't have hard facts, but would feel quite uneasy with 15 1-minute darks, because compared to a 15 minute light the read-noise to dark current ratio would be very different than in 15 minute darks. I personally always expose darks the same length as lights and try to get as many as possible.

Clear skies
Tahir