Darks and Bias frames

[ngwillym]

I seem to be getting conflicting views on whether it is necessary to take both dark and bias frames to calibrate lights.

Some web sites say yes, some say no - because the bias in included in the dark.

In the past I've only ever taken lights, darks and flats

What is the consensus here, and does it vary depending on the type of imaging device - DSLR, CCD, webcam etc

 

thanks in advance

Neil

[craigg]

The book 'Making Every Photon Count' suggests subtracting the bias from the darks because the processing subtracts darks from the subs, but uses some sort of average/divide function (something like that, doing this from memory).  So if you subtract bias from the darks, then process the subs using bias subtracted darks and the bias, you get the best results.

Thats my current understanding, but like you say there's a lot of conflicting information.  Invested to see what other (more experienced) folks reply with.

[Qwiati]

I am not more experienced, but I agree with craigg and the book. I am in favour of using BIAS master frame or even Super BIAS especially that it takes very short time to take many of them.

From Pixinsight pre-processing tutorial:

It is also very important to have a bias library, because thermal noise must be bias-subtracted: only thermal noise must be rescaled in the master dark to match the thermal noise in the light frame.

[vlaiv]

If I'm correct, bias frames introduce more noise, I was also under impression that bias frames are useful, but when looking at the math it seems that they do more harm than good.

Both dark and bias frames are used to remove unwanted signal, not noise. Both dark and bias frames introduce more noise when applied. From my consideration, I've come to conclusion that it is best to use only darks. Maybe there are cases where using both bias and darks give better results (specific ratio of dark current noise to read noise, or something like that), but that would be beyond my understanding / knowledge.

Here is my reasoning:

- when stacking every component of noise reduces by square root of number of frames stacked (for average stacking methods), signal levels remain the same.

- when subtracting or adding frames, signals add/subtract but noise is calculated by following formula: square_root(square(noise_a)+square(noise_b)) (this is kind of pythagoras theorem for noise which makes perfect sense, since two random signals (noise) ought to be orthogonal to each other, much like two independent vectors)

So when calibrating dark frames with master bias frame before stacking you increase noise level in each one of dark frames. Stack of uncalibrated dark frames will have lower level of noise than stack of calibrated dark frames, for same stack size.

Then you take light frame and calibrate it with lower noise master dark - you add certain amount of noise. If you calibrate it first with master bias (add some noise) and then with master dark that has more noise - you add even more noise. It is obvious that darks only calibration will introduce less noise whilst still removing both read signal and dark current signal from light frame.

 

[MARS1960]

3 hours ago, ngwillym said:

I seem to be getting conflicting views on whether it is necessary to take both dark and bias frames to calibrate lights.

Some web sites say yes, some say no - because the bias in included in the dark.

In the past I've only ever taken lights, darks and flats

What is the consensus here, and does it vary depending on the type of imaging device - DSLR, CCD, webcam etc

 

thanks in advance

Neil

I'm pretty sure if using a DSLR with a CMOS sensor you dont need to take bias frames, i read it on here somewhere.

As for CCD, i think it is the general consensus that if you bother with bias frames then only apply them to your lights via the darks.

[wornish]

Very informative link here.  Modern DSLR sensors don't need darks and flats.

http://www.clarkvision.com/articles/astrophotography.image.processing/

[Physicist13]

that article does only refer to taking images through standard camera lenses for which the lens info is known (ie distortions ets..so the camera automatically takes them out). His discussion on bias and dark is useful though! you need to know where your noise sources come from and if they are really random or not. "background" fitting routines can do the job of a "flat" to a pretty good extent.

P

 

[andrew s]

The reason you get different advice is because there are different situations.

1) If your dark's are exposed for the same time as the flats then you don't need bias frame since they are simply subtracted.

2) If your darks are exposed for a different length of time than the flats then you need bias frames as the darks will be scaled to the same time as the flats and both the flats and the darks must be bias subtracted. This is automatically done in most software.

The same applies to image frames.

I always use a master bias frame as I then use master darks exposed for the same or longer time than my image frames for both flat and image frames.

One thing to be aware of is that if you take a rapid succession of bias frame or short darks or flats the temperature regulation of the camera may not be able to maintain the temperature accurately so I programme a delay long enough to allow the temperature to settle.

 

Regards Andrew

[dph1nm]

If you are subtracting darks from your lights you do not need a bias, as the bias is, indeed, included in the dark. BUT you must somehow subtract the bias from any flats you take, or they will not work correctly. You can do this either with a bias, or with what is quaintly called a flat-dark (or is it dark-flat, I never remember!). i.e. a dark frame with an exposure time the same as the flats. Personally I would stick to using a bias.

The main purpose of bias subtraction is to remove the artificial signal manufacturers add to prevent negative numbers occurring in your raw images. Different manufacturers have a different approach to this, but certainly Canon DSLRs do require this subtraction.

NigelM

[ngwillym]

Some useful expalnations here, thanks everyone.

 

Until now - I've only being doing short exposure imaging - either lunar/planetary or DSO with integrating video camera with exposures typically of less than 8 seconds - and until now - I always took my darks at the same exposure as the lights and haven't done a lot with flats - so now I know that for the best result I need to change my methods