Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Why do you need bias frames if you do darks?

  • Please log in to reply
9 replies to this topic

#1 Ocelot

Ocelot

    Explorer 1

  • -----
  • topic starter
  • Posts: 52
  • Joined: 31 Mar 2016
  • Loc: Ottawa, Ontario

Posted 22 April 2019 - 07:29 AM

Hello!
I understand that bias frames will contain signal that the camera electronics generates. This signal is always present and needs to be subtracted from the light subs. However, since the bias signal is always present, it will be there in the dark frames as well, which are already subtracted from the light ones. Thus, if we subtract both dark and bias frames, will we not subtract the bias signal twice? I’m totally confused, can someone clarify, please?

#2 RedLionNJ

RedLionNJ

    Soyuz

  • *****
  • Posts: 3708
  • Joined: 29 Dec 2009
  • Loc: Red Lion, NJ, USA

Posted 22 April 2019 - 07:56 AM

You don't subtract separate bias if you're only using lights and darks.


  • Michael Covington likes this

#3 whwang

whwang

    Mercury-Atlas

  • *****
  • Posts: 2843
  • Joined: 20 Mar 2013

Posted 22 April 2019 - 08:10 AM

You will need bias for your flats, unless you take flat darks.
  • Michael Covington and bobzeq25 like this

#4 Ocelot

Ocelot

    Explorer 1

  • -----
  • topic starter
  • Posts: 52
  • Joined: 31 Mar 2016
  • Loc: Ottawa, Ontario

Posted 22 April 2019 - 08:28 AM

For flats, I guess, it makes sense. Thanks!
  • bobzeq25 likes this

#5 Kendahl

Kendahl

    Apollo

  • *****
  • Posts: 1439
  • Joined: 02 Feb 2013
  • Loc: Omaha, Nebraska

Posted 22 April 2019 - 07:50 PM

Here's a good way to think about Lights, Darks, Flats and Dark Flats. Bias frames are an approximation to Dark Flats.

 

Pre-processed Image = (Light - Dark) / (Flat - Dark Flat)

 

The Dark should match the Light in ISO (if you're using a DSLR), duration and temperature.

Flat - Dark Flat should be a Master Flat built out of many Flats and Dark Flats (or Bias Frames).


  • Michael Covington and Ocelot like this

#6 BQ Octantis

BQ Octantis

    Surveyor 1

  • *****
  • Posts: 1603
  • Joined: 29 Apr 2017
  • Loc: The Red Centre, NT, Australia

Posted 24 April 2019 - 02:47 AM

G'day Ocelot,

 

Your darks do indeed contain bias noise. The main reason to separate out the bias noise from the darks is that bias noise is always the same for a given ISO whereas dark noise changes with exposure duration and sensor temperature. As such, when you scale your darks to match the noise in your subs, if you don't subtract out the bias you end up erroneously scaling it as well—resulting in removing too much or too little bias noise from the sub. By separating the bias you can subtract it from the sub exactly; you can then scale the Master Dark (dark - bias) and subtract it from the sub separately. The formula for a calibrated sub thus becomes

 

Subcalibrated = (Subraw - Bias - (Darks - Bias)scaled ) / (Flats - FlatDarks)

 

You use FlatDarks to calibrate your flats instead of bias frames because FlatDarks contain noise above the bias noise (just like any darks—but they're only ~10x longer in duration than bias frames). However, you can use them unscaled (and thus with the bias included) because they're so short in duration (hundredths of a second) that their thermal noise isn't that far off from that of the flats.

 

Cheers,

BQ


Edited by BQ Octantis, 24 April 2019 - 07:41 AM.


#7 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23743
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 24 April 2019 - 05:03 PM

G'day Ocelot,

 

Your darks do indeed contain bias noise. The main reason to separate out the bias noise from the darks is that bias noise is always the same for a given ISO whereas dark noise changes with exposure duration and sensor temperature. As such, when you scale your darks to match the noise in your subs, if you don't subtract out the bias you end up erroneously scaling it as well—resulting in removing too much or too little bias noise from the sub. By separating the bias you can subtract it from the sub exactly; you can then scale the Master Dark (dark - bias) and subtract it from the sub separately. The formula for a calibrated sub thus becomes

 

Subcalibrated = (Subraw - Bias - (Darks - Bias)scaled ) / (Flats - FlatDarks)

 

You use FlatDarks to calibrate your flats instead of bias frames because FlatDarks contain noise above the bias noise (just like any darks—but they're only ~10x longer in duration than bias frames). However, you can use them unscaled (and thus with the bias included) because they're so short in duration (hundredths of a second) that their thermal noise isn't that far off from that of the flats.

 

Cheers,

BQ

It needs to be noted that dark scaling (particularly when it is automated via noise evaluation) does not work well with any camera that has amp glows. Further, scaling darks is not guaranteed to produce the best calibration in general, for any camera. If you do not do dark scaling, then bias frames are not necessary, or they may only be necessary to calibrate flats.



#8 BQ Octantis

BQ Octantis

    Surveyor 1

  • *****
  • Posts: 1603
  • Joined: 29 Apr 2017
  • Loc: The Red Centre, NT, Australia

Posted 27 April 2019 - 07:15 AM

It needs to be noted that dark scaling (particularly when it is automated via noise evaluation) does not work well with any camera that has amp glows. Further, scaling darks is not guaranteed to produce the best calibration in general, for any camera. If you do not do dark scaling, then bias frames are not necessary, or they may only be necessary to calibrate flats.

Too true, Jon. I would actually recommend bias frames over dark frames—indeed, it was the strange blue blotches in my shots of the Large Magellanic Cloud that led me to bias frames. It turns out that oversubtracting hot red pixels in darks that results in blue blotches. You don't get these with bias frames, and with dithering and sigma reject stacking, the hot pixels go away entirely.

 

BQ


Edited by BQ Octantis, 27 April 2019 - 07:18 AM.


#9 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23743
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 27 April 2019 - 10:41 AM

Too true, Jon. I would actually recommend bias frames over dark frames—indeed, it was the strange blue blotches in my shots of the Large Magellanic Cloud that led me to bias frames. It turns out that oversubtracting hot red pixels in darks that results in blue blotches. You don't get these with bias frames, and with dithering and sigma reject stacking, the hot pixels go away entirely.

 

BQ

If you have glows, you cannot use bias frames. Glow is a dark signal trait that is not present in bias frames, and can only be captured in dark frames. Further, bias frames won't capture any of the hot pixels or DFPN due to DSNU, so while you might avoid over-correction of some hot pixels (and if this is occurring, I would figure out why...you shouldn't be over-correcting hot pixels in just one channel...), you would be leaving behind plenty of DFPN to muck with the rest of your signal. 

 

Darks have their place. Biases have their place. Biases are not really an alternative to darks. Dark flats may be necessary depending on the camera. Dark scaling may work, if you have an exceptionally well behaved sensor with nothing but spatially random DFPN that grows at a very consistent rate, but not many cameras are that well behaved. 

 

For most people, the simplest approach to calibration is to subtract an unscaled well-matched master dark. This should correct the full DFPN profile across the sensor, and correct it well, with the vast majority of cameras. Flats may be able to use just a bias if the exposures are short, however if they are long enough that glows appear or for normal dark current to grow to a non-trivial offset, then flat darks would be necessary.



#10 the Elf

the Elf

    Surveyor 1

  • *****
  • Posts: 1758
  • Joined: 06 Sep 2017
  • Loc: Germany

Posted 27 April 2019 - 12:53 PM

A lot of good information given here. Not knowing how familiar the OP is with the topic at all I want to add some more information, although not exactly an answer to the original question.

Bias and dark frames contain information about the camera sensor and the sensor only. Flats contain information about the optics, namely the vignetting and of all the dust motes on different surfaces. As the sensor does not change much over time, bias frames can be reused for a long time and so can darks if the sensor temperture is known. Dust motes are always on the way, that is why you take flats every time. In my oppinion (and I got a lot of positive feedback for that statement) propper flats are the best means of making your processing easy. Think of a flat frame as almost as important as the light frame and invest in a decent flat field. I did some experiments with all the calibration frames, intentionally using darks with shorter/longer exposure time, higher or lower temperature, no darks, no flats, etc. I'm still working on the topic and plan to write an article for my website. For now I can tell (and theory predicted it) that the exact dark matching and the noise estimation based scaling is relevant for the dark regions of the image only. So with a bright object that covers most of the image like M42 or a globular cluster the impact of darks is comparably low. For dim objects they are more important. Flat darks are not needed if the exposure time for the flats is short, say fractions of a second. In that case calibrating the flats with bias only is fine. Given the little efford it is wise to take rather more than less flats. There is a difference between flats for mono cameras and OSC flats: in case of OSC you choose the exposure time so that the brightest of the three colors red, green and blue is well exposed but not too much. The level of the other two is determined by the flats color and often leads to very low signals in the red channel. One way to fight this is a large number of flats. I startet with 50-100 flats to fight the noise in the reds. Meanwhile I have an Aurora by Gerd Neumann. I did some testing and ended up with 3/8 orange Lee foil in the flat field to shift the blue color more to white. Now the exposure time is a bit longer but all channels are on the same level within a factor of less than two. Now I take only 30 flats and they are virtually free from noise. A lot of methods for flats are recommended but being an engineer in optical testing I find some of them bizare and strongly recommend neither to use T-shirts nor any laptop or tablet screens. Lambertian characteristic is the keyword that many people never heared before. You need a source with no structure, same output all over the area and lambertian characteristic, that defines the change of intensity by viewing angle. Laptop screens are awful in this point. With the EL foil all I do in processing is subtract a linear background (function level = 1) for the sky gradient. Period. Never had any trouble. In the US Spike-a-Flat seems to be equal to the European Aurora. A well known CN member recommends it and I would trust him. <smile>


  • kingjamez and Ocelot like this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics