CMOS calibration files and routines

[kirkster501]

Hi all, I read that with CMOS your darks should exactly match the lights in terms of length, with no scaling of the master dark??

Do you also need to have dark-flats  (to be clear, darks that match the length of the flats?) ??

And in some places I read you should use BIAS and others that they are not needed....

I have recently acquired a QHY268C OSC CMOS and getting my head around the differences from CCD.  Since it is OSC I at least only need one set of flats and flat-darks and not per filter.

Thanks

[Adam J]

9 minutes ago, kirkster501 said:

Hi all, I read that with CMOS your darks should exactly match the lights in terms of length, with no scaling of the master dark??

Do you also need to have dark-flats  (to be clear, darks that match the length of the flats?) ??

And in some places I read you should use BIAS and others that they are not needed....

I have recently acquired a QHY268C OSC CMOS and getting my head around the differences from CCD.  Since it is OSC I at least only need one set of flats and flat-darks and not per filter.

Thanks

Depends on the exact sensor and how it performs its internal calibration.

Its a new camera and sensor I dont know what the correct answer is, however what I do know is this:

1) Dark Flats always work with CMOS.

2) BIAS doesnt always work with CMOS.

Hence, unless you want toexperement with the sensor yourself and determine if BIAS is viable then you will need to use Dark Flats until someone else determines otherwise.

Adam

[TerryMcK]

It also depends on what you use to calibrate. For instance AstroPixelProcessor barfs if you don't have bias files but does continue after you acknowledge the message. DeepSkyStacker just gets on with it.

On my OSC's I use lights, darks, flats and sometimes bias although it can sometimes (but not always) have a negative effect and the amp glow can still show. Removing the bias frames helps. I've never used dark flats or flat darks (whatever terminology it is!).

I haven't got my head around PixInsight calibration processes yet as it seems to be complicated. I'm sure once you have done it a few times it comes send nature but my number 1 calibration tool of choice is without doubt APP.

 

I also use APP for my two mono cameras (although the 1600MM is out on long term loan). Again using exactly the same lights, darks, flats (maybe bias) but lights and flats are per filter of course.

[andrew s]

If it has amp glow don't use scaled darks as amp glow Is nonlinear with exposure.

Regards Andrew 

[vlaiv]

26 minutes ago, kirkster501 said:

Hi all, I read that with CMOS your darks should exactly match the lights in terms of length, with no scaling of the master dark??

Yes, this is correct.

It is in part because of the design choices for CMOS sensors. Some CMOS sensors distinguish between short and long exposures. Anything up to 1 second of exposure is done "on sensor" - meaning that sensor electronics keeps time and decides when the exposure is finished. This is needed for high FPS counts that CMOS are capable of.

Longer exposures are in fact "bulb" exposures as far as sensor is concerned - driver (or external electronics) is the one giving commands - start exposure / stop exposure, and here sensor does not keep time.

There is also difference in how sensor does internal cleaning / calibration and so on with these two types of exposures.

For example, I just used my ASI178mcc to do some wide field shots. Look at these stats:

These are darks having same exact settings - except for exposure length. One dark is for "filler" subs (over exposed star cores) and second dark is regular 60s exposure dark.

How on Earth can 1s dark have higher mean value than 60s sub? Dark sub consists of bias signal and dark current signal. Bias signal should be the same and dark signal raises with time. Logic dictates that 60s sub has higher mean value than 1s sub due to accumulated dark current.

However, what really happened is that 1s exposure is "short" exposure and is being managed by sensor while 60s exposure is long exposure and is being managed by driver. Bias level (although I had offset set the same on both) is different for some reason and it is bias level that is higher in short exposure.

This means that you can't record bias for long exposure as bias will be like 0.001s long and it will always be managed by sensor.

For this reason, there is no dark scaling in conventional sense.

What you could do is to take 60s darks and 120s darks and extract bias from those with a bit of math and then use again bit of math to do dark scaling. I have not done this to check if it would work on CMOS (it might not for some reason) but in theory it should work. On the other hand - I just use darks of matching duration and have no need for bias.

36 minutes ago, kirkster501 said:

Do you also need to have dark-flats  (to be clear, darks that match the length of the flats?) ??

And in some places I read you should use BIAS and others that they are not needed....

There is a problem with this question, and the problem is the word "NEED". You don't need flat darks but you want to shoot them because it is the proper way to do calibration.

Flat needs to be pure light signal and it is not unless you subtract both bias and dark current signal (depending on your flat exposure). With nicely behaved sensors you can in principle use only bias as flat exposures tend to be very short and not a lot of dark current accumulates to make any difference but as we have just seen above - CMOS are tricky in that respect, you must first make sure your bias is actually bias and is usable.

If you want to do proper calibration - just use flat darks. They are of the same duration as flats so very easy to shoot. You can also create set of them and reuse them - just make sure you have all the parameters the same.

 

[kirkster501]

Ok guys many thanks.

I am taking set of BIAS and Darks at my chosen Gain/Offset/Temperature/Exposure.  I will play around with this when I get some real data.

[Uranium235]

Belt n' braces approach is best for CMOS ive found (well the ASI178 anyway).

Darks, Dark flats, Bias, Flats

I even match the setpoint cooling for all calibration frames (yes, for flats as well so you get a correct ADU measurement as hot pixels skew the readings)

[DaveS]

It might not be applicable to the latest CMOS cameras, but I've found using Bias as well as Dark frames messed up the calibration, leading to imperfect removal of the amp-glow. I *think* it's because the Dark frame already contains the Bias signal, but I may be wrong. This was with the ASI 1600 amd 183 MMC cameras.

I ended up using Dark, Flat and Dark Flat frames.

YMMV.

[kirkster501]

The chip used in the ASi2600 and the QHY268C does not have amp glow.  I created master flats and BIAS just playing around yesterday and they are extremely clean.