biases - library versus per session

[troypiggo]

Since I've had a temperature controlled CCD, I've been using dark libraries.  And more recently, my dark libraries have been simplified by taking 1800s darks and scaling.  I'd read somewhere the darks for scaling should be somewhere around 3 to 5 times the longest subs you take.

Anyway, my question is about biases now.  Was just thinking about the pros and cons of a bias library versus taking them every time I go out.  I know they take but a fraction of the time darks do, most of the waiting time is the download time.  But if I can save some time at the end of a session, and if it's good enough for darks, why not?

I guess I'm curious if biases should be consistent enough over time?  Probably plan on taking new dark and bias (if applicable) libraries every 6 months or so.

Ooh, another quick couple of questions.  My dark libraries I use 20 shots.  I understand it's preferable/recommended to use at least the same number of shots for biases, so I usually adopt 20 biases.  I know the benefit vs number tapers off and at some point you just have to draw a line and say 20/30 whatever is enough.  I guess if I'm using libraries I could push more to say 30?  What do you do?

Read about superbiases here.  I'm on a Mac and I think the process/script was for Win only?  Should I be taking 50 biases or is that getting silly?  Can this superbias feature be used on a Mac?

Obviously flats will be taken per session to suit the optics/filters/binning etc used for each session.

[Carlos Milovic]

You may combine bias from a library with fresh ones, and thus, keep them at a large number, without making so many of them at the same time (and, you may discard the oldest). I wouldn't use less than 50 bias.

Superbias may be helpfull if you clearly see vertical patterns in your master flat, and differences between pixels in the same column are smaller than differences between columns. Then, the superbias would emulate a master bias done with thousands of frames, with just a few dozens (say, for example, 32).

The module right now is compiled for windows and linux, 64bits. Juan already has the code and documentation, so it will a matter of time to have it distributed and compiled for all other OS.

[troypiggo]

Wow!  50!  Ok.  A kind of a rotating library makes sense.  Thanks.

Maybe my 20 darks isn't enough as well?  If I increased this, do you have a recommended shelf life for them?  6 months too long?

[pfile]

how many frames do you have to use before you have a 'superbias' master? 100? 1000?

just curious. i probably do not have that many bias files laying around. i will wait for the SuperBias module to be published...

[Carlos Milovic]

To make the SuperBias process to work you just need a master of a few dozens... 20 is good enough. It is supposed to be a model of a master bias done with thousands of images. For example, for my Canon 300D, I made a superbias (from 20 frames) whose maximum differences from the master bias were under 2 ADUs. With a master done with 400 frames, the differences went under 1 ADU.

[astropixel]

And more recently, my dark libraries have been simplified by taking 1800s darks and scaling.  I'd read somewhere the darks for scaling should be somewhere around 3 to 5 times the longest subs you take.

That occurred to me the other day, but for DSLR's I think the noise would be prohibitive. A cooled CCD strategy, perhaps?

[Andres.Pozo]

Wow!  50!  Ok.  A kind of a rotating library makes sense.  Thanks.

Maybe my 20 darks isn't enough as well?  If I increased this, do you have a recommended shelf life for them?  6 months too long?

For my camera, with a Kodak KAI-04022 sensor, 50 bias is in the low side. I usually use 100.
This is an example of stacking between 10 and 80 bias:

[troypiggo]

And more recently, my dark libraries have been simplified by taking 1800s darks and scaling.  I'd read somewhere the darks for scaling should be somewhere around 3 to 5 times the longest subs you take.

That occurred to me the other day, but for DSLR's I think the noise would be prohibitive. A cooled CCD strategy, perhaps?

Yes, I'm using a cooled CCD (QSI583ws).  Not sure how transferable the theory would be to DSLRs, but can't imagine it'd be that different.  I know the noise would be much higher, but it's just pixels and numbers?

[troypiggo]

...
For my camera, with a Kodak KAI-04022 sensor, 50 bias is in the low side. I usually use 100.
This is an example of stacking between 10 and 80 bias:

Thanks for the visual!  That makes it crystal clear for me.  Overnight I took 100 biases and compared the integration to one of 20 biases and the difference is certainly noticeable.  That'll be my standard for now on.

[pfile]

And more recently, my dark libraries have been simplified by taking 1800s darks and scaling.  I'd read somewhere the darks for scaling should be somewhere around 3 to 5 times the longest subs you take.

That occurred to me the other day, but for DSLR's I think the noise would be prohibitive. A cooled CCD strategy, perhaps?

i have done 1800s darks with my 50d, but i had cooled the sensor to about 10C during these exposures. i've used these 1800s darks to calibrate 1800s and 1200s lights successfully.

[astropixel]

How many would normally be required? And is that a better method than combining a number of shorter darks?

[pfile]

well to be clear, i made a master dark from 20 or 30 1800s darks. i just let the camera run overnight after my target has gone behind the house...

[Philippe B.]

Wow!  50!  Ok.  A kind of a rotating library makes sense.  Thanks.

Maybe my 20 darks isn't enough as well?  If I increased this, do you have a recommended shelf life for them?  6 months too long?

For my camera, with a Kodak KAI-04022 sensor, 50 bias is in the low side. I usually use 100.
This is an example of stacking between 10 and 80 bias:

Same for me
Now, for my QSI540, it is at least 200 bias and around 50-75 darks. Mostly refreshed during nights after shooting.
At the end, there is a difference on the final image. Less defect and higher SNR.
Tested superbias but result was not better.