How To Take Bias Frames on a DSLR?

[oldpink]

if you have BYEOS it has an option to take darks, flats and dark flats

[Russe]

Can I "make" them in DSS?

[oldpink]

I have DSS installed but so far no data to process so not used it

[Xplode]

I have taken a look at the bias frames from my 6D and the noise is all random so bias frames = waste of time.

[Coco]

Interesting post.. I am experimenting myself with bias with & without.

[oldpink]

added my first Bias frames tonight
shot at the same ISO but at 1 /4000 with cap on
interesting results so far, off the bat my pictures are better but I was playing with ISO and white balance settings as well which improved things a lot

[dph1nm]

I have taken a look at the bias frames from my 6D and the noise is all random so bias frames = waste of time

The main purpose of a bias is to remove the pedestal signal, not noise. My 1000D has a mean bias level of 256 - this needs removing before you can use a flat field successfully. The dark obviously contains this pedestal as well, so you don't need both.

NigelM

[lensman57]

I'm with you on this one. I don't know where the idea that Canon Bias frames are "flatter than a ..." comes from.  They're certainly not and a clear fixed pattern emerges when you stack enough of them. I guess the magnitude of the pattern may not be huge on some models, but given the other limitations of DSLRs every little improvement helps the final image.  (See here: http://www.blackwaterskies.co.uk/2013/09/pixinsight-dslr-workflow-part-1-bias.html).

Where the jury is still out for me is where best to apply the master bias once you have made it.  I certainly subtract it from flats before stacking the master flat.  I am undecided as to whether to use it as part of the dark calibration process or not though.  I've tried a few different approaches so far and the differences are marginal so still on the fence at the moment.

Hi Ian,

I tend to agree with you, as for the Darks,  don't these contain the Bias already? Once the Darks are stacked I guess so do the contained Bias and the Master Dark is then subracted from the lights before division by the master Flat so after this stage all the Lights are Clean and flat fielded before stacking. I really don't know anything about Canon's Bias subtracting each exposure, that was news to me and I have not read anything in craig Stark's articles about this either.

A.G

[oldpink]

been meaning to ask this
I noticed my raws have this  added to the filename _00292stdev_ with different numbers is this to do with auto correction and will it effect sub's

[IanL]

Hi Ian,

I tend to agree with you, as for the Darks,  don't these contain the Bias already? Once the Darks are stacked I guess so do the contained Bias and the Master Dark is then subracted from the lights before division by the master Flat so after this stage all the Lights are Clean and flat fielded before stacking. I really don't know anything about Canon's Bias subtracting each exposure, that was news to me and I have not read anything in craig Stark's articles about this either.

A.G

Yes the darks do contain the bias, and I have been looking further in to this as part of my research on DSLR darks.  I have some way to go on it before it is ready for prime time, but I am now clearer about where to apply bias and why.  Bear in mind that this is specifically aimed as PixInsight processing (or other software that does dark frame scaling, so not DSS or similar). I'll quote from my as yet unfinished blog post:

It is important to note that in order for PI's dark frame optimisation to work to best effect, the master dark must have the master bias subtracted.

There is plenty of guidance out there that says exactly the opposite, i.e. don't bias subtract the master dark and don't bias subtract the light frames as by leaving the bias in the master dark you can subtract both bias and dark from the light frames in one operation. In general that is good advice, but when using software that performs dark scaling it does not apply.

The reason is that the optimisation routine tries to match the dark current using a single (linear) scaling factor which is applied to the whole dark frame. We start with our light frame and our master dark:

Light = Signal + Bias + Dark Current

Master Dark = Bias + ~Dark Current

Where Dark Current is the true dark current in the light, and ~Dark Current is the dark current in the master dark which may not match Dark Current due to temperature or exposure length differences (or indeed mysterious on-camera processing).

If we do not subtract the bias from both the light frames and the master dark, the dark calibration routine has to perform the following calculation:

Calibrated Light = Signal + Bias + Dark Current - Scaling x (Bias + ~Dark Current)

Multiplying the sum of Bias and ~Dark Current by a single scaling factor makes it harder (if not impossible) to subtract the correct amounts of both bias and dark current from the light frame.

If instead we subtract the master bias from the light frame and from the master dark before dark frame optimisation, the following calculation is performed:

Calibrated Light = Signal + Bias - Bias + Dark Current - Scaling x (~Dark Current + Bias - Bias)

Which simplifies to:

Calibrated Light = Signal + Dark Current - Scaling x ~Dark Current 

Hopefully you can see that we have subtracted the correct amount of bias as it nets to zero, whereas in the first case above we would be likely to subtract too little or too much bias depending on the dark scaling factor. Furthermore, the dark frame optimisation routine now has better starting conditions and more chance of finding the true value of Scaling.

In other words, if you do not subtract the bias the master dark will be a less good match for the light frame; either the optimisation routine will fail with a 'No correlation ...' error or it will succeed but with a less good match that might otherwise be possible. (This does not contradict Juan's [PixInsight Author] assertion that the routine is a purely numerical solution by the way, we are simply trying to load the dice in favour of a better outcome).

Note: Since the master bias is the most frequently applied calibration frame, it has the most opportunities to add noise to the final image. As we have seen previously, making a really large, clean master bias is trivially easy, whereas I have found that making an equivalently large master dark is a lot harder. Therefore subtracting the master bias and master dark separately is a practical time-saver that will result in a better final image due to the relatively larger/cleaner master bias frame.

So the upshot is that it is probably a good idea to subtract the bias separately since you can make a much larger bias stack in an hour or two, whereas to create a large master dark stack takes a lot longer (assuming you are trying to control the temperature of an uncooled DSLR - it took me weeks to collect a large library of reasonably closely matched darks).  Aside from this, if you are using dark scaling in your software, it is better to subtract the bias separately as you should get a better matched scaling factor, but otherwise you could just subtract the bias as part of the dark subtraction (assuming you don't care about the bias being a smaller stack as a consequence of doing so).

The other reason for using a separate bias is that you can just bias subtract your flats and not bother with darks at all, assuming the flat exposure is reasonably short (not much dark current will accumulate in a few seconds).

[IanL]

been meaning to ask this

I noticed my raws have this  added to the filename _00292stdev_ with different numbers is this to do with auto correction and will it effect sub's

I don't know what capture software you are using, but check the filename settings (most have options to include certain data items in the filename, like temperature, date, time, exposure length, ISO, etc.) At a guess that number is the standard deviation of the image, which should correlate to the amount of noise in the image.  Again borrowing from my upcoming blog post on the topic, here is a graph of the median standard deviations of sets of dark frames taken at different temperatures.  As you can see, the standard deviation increases as the temperature increases.  This is what you'd expect as more dark current accumulates at higher temperatures for a given exposure length, and therefore the dark current noise also increases:

The information in the file name won't affect any processing software; depending on the package you are using it will typically perform a weighted average of each of the images when stacking.  So 'better' images will be given a higher weighting and will contribute more to the pixels in the final image.  Weighting can be done in different ways, but simpler software may just look at the exposure lengths (from the EXIF data created by the camera) and weight longer exposures more (since they should contain more signal).  Better software will try to analyse the noise in the image (e.g. using the standard deviation or more sophisticated methods) or indeed perform the ideal process which is to estimate the signal to noise ratio of each image and favour the images with higher SNRs.

[lensman57]

Yes the darks do contain the bias, and I have been looking further in to this as part of my research on DSLR darks.  I have some way to go on it before it is ready for prime time, but I am now clearer about where to apply bias and why.  Bear in mind that this is specifically aimed as PixInsight processing (or other software that does dark frame scaling, so not DSS or similar). I'll quote from my as yet unfinished blog post:

It is important to note that in order for PI's dark frame optimisation to work to best effect, the master dark must have the master bias subtracted.

There is plenty of guidance out there that says exactly the opposite, i.e. don't bias subtract the master dark and don't bias subtract the light frames as by leaving the bias in the master dark you can subtract both bias and dark from the light frames in one operation. In general that is good advice, but when using software that performs dark scaling it does not apply.

The reason is that the optimisation routine tries to match the dark current using a single (linear) scaling factor which is applied to the whole dark frame. We start with our light frame and our master dark:

Light = Signal + Bias + Dark Current

Master Dark = Bias + ~Dark Current

Where Dark Current is the true dark current in the light, and ~Dark Current is the dark current in the master dark which may not match Dark Current due to temperature or exposure length differences (or indeed mysterious on-camera processing).

If we do not subtract the bias from both the light frames and the master dark, the dark calibration routine has to perform the following calculation:

Calibrated Light = Signal + Bias + Dark Current - Scaling x (Bias + ~Dark Current)

Multiplying the sum of Bias and ~Dark Current by a single scaling factor makes it harder (if not impossible) to subtract the correct amounts of both bias and dark current from the light frame.

If instead we subtract the master bias from the light frame and from the master dark before dark frame optimisation, the following calculation is performed:

Calibrated Light = Signal + Bias - Bias + Dark Current - Scaling x (~Dark Current + Bias - Bias)

Which simplifies to:

Calibrated Light = Signal + Dark Current - Scaling x ~Dark Current 

Hopefully you can see that we have subtracted the correct amount of bias as it nets to zero, whereas in the first case above we would be likely to subtract too little or too much bias depending on the dark scaling factor. Furthermore, the dark frame optimisation routine now has better starting conditions and more chance of finding the true value of Scaling.

In other words, if you do not subtract the bias the master dark will be a less good match for the light frame; either the optimisation routine will fail with a 'No correlation ...' error or it will succeed but with a less good match that might otherwise be possible. (This does not contradict Juan's [PixInsight Author] assertion that the routine is a purely numerical solution by the way, we are simply trying to load the dice in favour of a better outcome).

Note: Since the master bias is the most frequently applied calibration frame, it has the most opportunities to add noise to the final image. As we have seen previously, making a really large, clean master bias is trivially easy, whereas I have found that making an equivalently large master dark is a lot harder. Therefore subtracting the master bias and master dark separately is a practical time-saver that will result in a better final image due to the relatively larger/cleaner master bias frame.

So the upshot is that it is probably a good idea to subtract the bias separately since you can make a much larger bias stack in an hour or two, whereas to create a large master dark stack takes a lot longer (assuming you are trying to control the temperature of an uncooled DSLR - it took me weeks to collect a large library of reasonably closely matched darks).  Aside from this, if you are using dark scaling in your software, it is better to subtract the bias separately as you should get a better matched scaling factor, but otherwise you could just subtract the bias as part of the dark subtraction (assuming you don't care about the bias being a smaller stack as a consequence of doing so).

The other reason for using a separate bias is that you can just bias subtract your flats and not bother with darks at all, assuming the flat exposure is reasonably short (not much dark current will accumulate in a few seconds).

Hi Ian,

Many thanks, this is all very clear and yes as far as the CCD is concerened I don't bother with Darks, just around 100 Bias frames and then Flats with exposures around 0.05~0.2s depending. For DSLR I do use Darks.

A.G

[oldpink]

I use BYEOS and yes it puts the exposure, ISO and temp as well as the _stdev_ info, I'll see if I can find out what the software says about this