How many darks?

[alcol620]

Hi folks

Now the Atik 383 has been serviced I am in the process of producing a "master dark" library. 

I want to be in a position to take subs of up to 1200 seconds. Obviously taking darks to match is a long process. How many 1200s darks would be adequate to make a master dark? I know, more is better, but how many are adequate?

Thanks

[Davey-T]

A law of diminishing returns applies to calibration frames, after about 16 / 20 the improvement is small, some folks take loads but I've not seen any improvement over twenty, may depend on how good your data is to start with I guess.

Dave

[Tinker1947]

Sort out the summer temperature and the run off as said around 20 that should do you for a long time, i run my 414EX at -12C which it should get down to in the summer.....

[alcol620]

Thanks folks.

I have currently taken 11, that's over 3 hours worth. I have also accumulated 20 to 30 subs of 120s, 300s and 600s darks. Probably did wrong by taking them all at -25C, which the camera can just get to at these current mild temperatures. I am beginning to realise that for summertime evenings I might need some at lesser reduced temperatures or I won't be able to cool down to the -25C darks.

[Tinker1947]

Thanks folks.

I have currently taken 11, that's over 3 hours worth. I have also accumulated 20 to 30 subs of 120s, 300s and 600s darks. Probably did wrong by taking them all at -25C, which the camera can just get to at these current mild temperatures. I am beginning to realise that for summertime evenings I might need some at lesser reduced temperatures or I won't be able to cool down to the -25C darks.

Its all a learning curve, fortunately its not wasting money at this stage.....that comes later....

[Starlight 1]

This may sound strange but in the world of mathematics & logarithms odd numbers always stack better than even numbers, so 19 23 27 will always give you better averages than 20 25 30

[MartinB]

Shouldn't be a problem this time of year to get 20-30 (or 21 to 31!).  

[gnomus]

I have taken 25 darks at various exposure lengths - all done at minus 20 degrees.

[alcol620]

Hi gnomus

25 at 1200s? takes along time

Regards

Alec

[Tinker1947]

Hi gnomus

25 at 1200s? takes along time

Regards

Alec

I did mine with the CCD on my desk, running from a Tracer plus USB leab to  a PC, with the Temperature set @-12c, the ATik comes with a screw in blank to shut all light out from the sensor, do them in day light this way.....

[Davey-T]

This may sound strange but in the world of mathematics & logarithms odd numbers always stack better than even numbers, so 19 23 27 will always give you better averages than 20 25 30

   

Dave

[cfpendock]

Hi gnomus

25 at 1200s? takes along time

Regards

Alec

It does indeed.  Approximately eight and half hours.  Many years ago I used to test things for various people for certain lengths of time, e.g. a 1000 hour test.  It always surprised me how many people would ring up after a week or so:

"Have you finished that 1000 hour test yet?"

"No"

"Why not?"

"Because a 1000 hour test  takes 1000 hours (at least)........."

"Oh......"

Chris

[Stub Mandrel]

This may sound strange but in the world of mathematics & logarithms odd numbers always stack better than even numbers, so 19 23 27 will always give you better averages than 20 25 30

I assume this is to minimise artefacts caused by recurring patterns (e.g. moire-type effects). I have no idea if this is true or not, but if that's the case we should use prime-numbers?

So 19 or 23, but 29 not 27.

[alcol620]

I did mine with the CCD on my desk, running from a Tracer plus USB leab to  a PC, with the Temperature set @-12c, the ATik comes with a screw in blank to shut all light out from the sensor, do them in day light this way.....

Does the front cap on the Atik shut out all light? Is it possible light could get in via the plug holes or the fan in and out slots?

[Tinker1947]

Does the front cap on the Atik shut out all light? Is it possible light could get in via the plug holes or the fan in and out slots?

This is a 1200 darl @ -12C converted from a TIFF to a JPEG so it could be loaded on here, the white specks are dead pixels (i think) there the same place in other dark's taken at the same time, i don't know to much about if its good or bad but if i use Levels om it and move the sliderers right over to the left on my PS, it goes pure white....maybe somebody that knows can comment....

DARK 1200 seconds @ -12C

[ollypenrice]

Darks can be quite invasive and destructive. You don't have to use them at all. I haven't used them for several years and this has improved my stacking and simplified my life. And this despite the fact that I can't dither at capture because I run a dual rig on which we use different exposure times in each scope. If you can dither then darks become even less attractive.

If you can dither guide (and you really should) then you can use a master bias as a dark and the dither will do the rest.

In my case I can't dither so I follow AstroArt's instructions, as follows: I use a master bias as a dark. I create a long exposure master dark and clip 2000 ADU off the black point. I call this a bad pixel map and put it in the bad pixel map box they provide. I also run a very aggressive hot pixel filter.

Using this method I have never, ever, had such clean stacks, free from pixel level cleaning, in my life. Honetly, I've done multi panel mosaics without touching a single pixel at the end.

Plus, plus (!!!!) I can stack an entire run from the night before while using exactly the same calibration files for everything from that camera. On winter nights when I might have 24 hours of data* from three rigs to calibrate that is no small matter.

Olly

*Exceptional but not unkown. 

[opticalpath]

.....  I use a master bias as a dark. I create a long exposure master dark and clip 2000 ADU off the black point. I call this a bad pixel map and put it in the bad pixel map box they provide. I also run a very aggressive hot pixel filter.

Using this method I have never, ever, had such clean stacks, free from pixel level cleaning, in my life. Honetly, I've done multi panel mosaics without touching a single pixel at the end.

Interesting.  I have always just followed conventional wisdom and done the dark subtraction, but this is certainly worth an experiment! 

Adrian

[steppenwolf]

I'm with Olly on this - although I use what is conventionally known as a 'noisy' sensor (KAF-8300), I don't take darks but I do dither between subframes and an SDMask stacking protocol removes all the hot pixels providing I have in excess of 5 subframes to stack.

Darks certainly have their place and I still use them with my one shot colour camera where I use 22 subframes (although anything in excess of 16 should be fine) for historical reasons.

[alcol620]

Will definitely give the "no dark" option a try. I dither with PHD so this may do the trick.

Thanks folks

[johnrt]

This may sound strange but in the world of mathematics & logarithms odd numbers always stack better than even numbers, so 19 23 27 will always give you better averages than 20 25 30

I have heard this said before, and I did look in to this by stacking various numbers of subs and comparing statistics in Pixinsight's statistics tool. The conclusion was that it made no difference whatsoever beyond the incremental benefit of stacking a single or couple of extra subs.

The only rule is the more subs the better, keeping in mind the law of diminishing return after 20 or so subs, but with darks and the months of clouds, we have all the time we like to put a really hefty number together.

[MartinB]

I suppose I'm a bit of a dinosaur and do use darks, usually. I often run 30 min subs and have no problem with collecting 40 darks. Easy with an obsy, less so if not. It's only a problem during the summer and then I tend to resort to recently created dark masters. This seems to give nice clean data but I haven't done a more scientific assessment. I have an idea that dark current creates more noise than just hot pixels and that bad pixel mapping and dithering doesn't entirely elimate this but I'm probably talking out of my bottom!

[opticalpath]

I suppose I'm a bit of a dinosaur and do use darks, usually. I often run 30 min subs and have no problem with collecting 40 darks. Easy with an obsy, less so if not. It's only a problem during the summer and then I tend to resort to recently created dark masters. This seems to give nice clean data but I haven't done a more scientific assessment. I have an idea that dark current creates more noise than just hot pixels and that bad pixel mapping and dithering doesn't entirely elimate this but I'm probably talking out of my bottom!

Me too.  There's no shortage of cloudy nights and If you're fortunate enough to have an observatory it's no great hardship to build up a library of darks.

I suppose that dark calibration is intended to remove major defects caused by hot and warm pixels, but also to correct for small variations from pixel to pixel in how dark current builds up.  The method Olly described deals effectively with the hot pixels, but I think would ignore the much smaller pixel to pixel variation; dithering would smooth out those small variations.

Just for interest, I looked closely at a long exposure master dark (20 x 1 hour) taken with my QSI 583 at -20C.  There were many hot, and warm, pixels of course, but underneath those, there was some small variation in the background level. I looked at many small regions and the minimum background count in each region; it varied from 278 counts to 306 counts. If I assumed that most of the statistical variation would have been smoothed out by stacking 20 darks, then some of this difference in 'background' counts could represent the real pixel to pixel difference in sensitivity and dark current. Then it's just a question of whether that is small enough to ignore, or to smooth out by dithering, having first used a bad pixel map to correct the gross variations.  And would the benefit of subtracting out this remaining small variation be outweighed by the cost of some noise added by the dark subtraction operation?  Definitely seems worth trying a comparison.

Adrian