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

ASI 1600 bias vs flat darks...

  • Please log in to reply
37 replies to this topic

#1 Benni123456

Benni123456

    Viking 1

  • ****-
  • topic starter
  • Posts: 575
  • Joined: 20 Apr 2017
  • Loc: Munich, Germany

Posted 18 October 2020 - 11:52 AM

Hi, I see that it is sometimes mentioned one should not use bias frames for the asi 1600 but flat darks instead.

 

What I have noted is that when I use a very high number of bias files e.g. 200 or more, then the noise gets down.

Unfortunately, my flat box has a continuous dimmer, This means that I can not use a predefined setting since 0 would mean zero light and setting it to full would be too bright.

 

So I would have to take flats and flat darks after each observing session. Then, I only can take 25 flats and 25 flat darks, as I do not have much time in the morning out in the field...

 

I noticed that the 25 flat darks considerably increase my noise. The flats do so too.

 

I get less noise, if I just use a master dark.

 

 

I recently attempted to just use a dark frame and generate a noise free artificial flat by the pixinsight dynamical background extraction tool. The artificial flat led to considerably less noise levels. (to get such a flat one has to use a high density of the measurement spots and make sure the spots are not on stars. sometimes the density of the measurement spots has to be increased in some places.)

 

 

I still have to see whether the noise gets down even more, if i use darks, and a biasframe with very high image number 900 frames or so, and the artificial flat from pixinsight background extractor...

 

For now I just want to say that for me, the flats and flat darks did not really work....

And the bias must really have an extremely high number of files...

 

(Superbias did not work for me too)...

 

I just wanted to let others know this and ask whether you have made similar experiences....


Edited by Benni123456, 18 October 2020 - 12:02 PM.


#2 tkottary

tkottary

    Viking 1

  • *****
  • Posts: 532
  • Joined: 06 Dec 2015
  • Loc: SunnyVale ,CA

Posted 18 October 2020 - 11:54 AM

  0.2s bias works fine  for asi 1600



#3 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,516
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 18 October 2020 - 12:54 PM

a bias will only help if you can scale darks, since the 1600mm ampglow signal is nonlinear the darks are not scalable... 



#4 Benni123456

Benni123456

    Viking 1

  • ****-
  • topic starter
  • Posts: 575
  • Joined: 20 Apr 2017
  • Loc: Munich, Germany

Posted 18 October 2020 - 06:57 PM

a bias will only help if you can scale darks, since the 1600mm ampglow signal is nonlinear the darks are not scalable... 

Why?

I think pixinsight tests whether the scaled image has less SNR and if so it does not apply the scale....

It often does with bias...

 

 

Or am I wrong?


Edited by Benni123456, 18 October 2020 - 06:57 PM.


#5 Stelios

Stelios

    Voyager 1

  • *****
  • Moderators
  • Posts: 10,257
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 19 October 2020 - 03:37 AM

For those who *do* use dark flats: Do you match the exact duration of the flat (so if the flat is 0.2" you would take a 0.2" dark)--or do you match all exposures say 1 second and below with 1" dark flats?

 

It seems that most everyone has his own way to do dark flats. Fortunately, so far I am OK with just bias--but I plan to get an ASI294MM, and everyone says that camera needs dark flats. 

 

I would rather NOT shoot long duration (over 1-2") flats if I can help it. 



#6 Rasfahan

Rasfahan

    Sputnik

  • -----
  • Posts: 44
  • Joined: 12 May 2020
  • Loc: Hessen, Germany

Posted 19 October 2020 - 07:59 AM

I have an ASI1600MM. I did shoot „matching“ dark flats (i. e. 0.2“) at First, but since I started using sky flats to get rid of reflections this has become unworkable. So I took a lot of bias frames (1500), told Voyager to keep the sky flats between 0.2 and 1“ and now calibrate those with only the master bias. I recently tried a master flat dark of 50x0.5“ and it made no difference for integrations with up to 800 subs.

#7 gatsbyiv

gatsbyiv

    Viking 1

  • *****
  • Posts: 886
  • Joined: 29 Mar 2009
  • Loc: Doylestown, PA

Posted 19 October 2020 - 09:32 AM

As long as the sensor is cooled, there should be no meaningful difference between a bias and a short (~1") flat dark.  The only difference would be from thermal signal, which is negligible in that timeframe for a cooled sensor, and the step change in gain below about 0.1" if you're shooting really short bias frames.  As long as you're treating it like a flat dark (i.e. you're not using it to scale), you should be able to use a master bias as a master flat dark.  I use a master flat dark based on 50x1" exposures for all of my flats, which range from 0.2" to 8".


  • dan_hm likes this

#8 Seanem44

Seanem44

    Vanguard

  • *****
  • Posts: 2,259
  • Joined: 22 Sep 2011
  • Loc: Stafford, VA

Posted 19 October 2020 - 12:41 PM

There is no need for the BIAS with the ASI1600.  


Edited by Seanem44, 19 October 2020 - 12:41 PM.


#9 Stelios

Stelios

    Voyager 1

  • *****
  • Moderators
  • Posts: 10,257
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 19 October 2020 - 12:50 PM

As long as the sensor is cooled, there should be no meaningful difference between a bias and a short (~1") flat dark.  The only difference would be from thermal signal, which is negligible in that timeframe for a cooled sensor, and the step change in gain below about 0.1" if you're shooting really short bias frames.  As long as you're treating it like a flat dark (i.e. you're not using it to scale), you should be able to use a master bias as a master flat dark.  I use a master flat dark based on 50x1" exposures for all of my flats, which range from 0.2" to 8".

What does the "step change in gain" below 0.1" affect? What will go wrong if you *do* go below 0.1"? 

 

My current flats range from 0.001" to 1.3". I would have to change my acquisition significantly to bring the Lum all the way to  0.2" if that's needed.

 

So far I'm calibrating with master bias (taken at 0) and a master dark of 30" (calibrate and optimize both on), which works well with the cameras I have now--but probably won't work with the ASI294MM that I plan to buy. (I have the ASI1600MM-C, the predecessor to the ASI1600MM-Pro. I think whatever made flat darks a requirement started when they put out the -Pro version). 



#10 gatsbyiv

gatsbyiv

    Viking 1

  • *****
  • Posts: 886
  • Joined: 29 Mar 2009
  • Loc: Doylestown, PA

Posted 19 October 2020 - 06:20 PM

What does the "step change in gain" below 0.1" affect? What will go wrong if you *do* go below 0.1"? 

About a year ago, there was a thread documenting the behavior of exposures under about 0.1 second, showing that the amplifier kicks into a different mode for very short exposures.  There was a lot of technical detail proving this, but the actual size of the effect was small:  I think it was about one electron difference vs. a slightly longer exposure of 0.2s or greater.  This shouldn't have any meaningful impact on your images, but out of an abundance of caution, some people like to avoid taking dark flats (or flats) shorter than 0.2s.  

 

I only mentioned it so my answer was technically complete.  It's not anything that should trigger you to re-process your images. 


  • Stelios likes this

#11 Benni123456

Benni123456

    Viking 1

  • ****-
  • topic starter
  • Posts: 575
  • Joined: 20 Apr 2017
  • Loc: Munich, Germany

Posted 20 October 2020 - 04:59 PM

Actually, I noted that the flat duration makes a very large difference. When I make flats over 0.1 sec they have much less noise than the faster ones,,,,

It is paradoxial but important, especially since for the flats one has not too many images if one makes photos in the field....

 

I have now taken bias files 200 at 0.00032 sec and 800 at 0.1 sec.. They look totally different. The shorter ones have an entirely different large scale structure...

 

What I do not find ok is that I have not seen that zwo mentions this in the manual. When the sensor has such a behavior, why do they not say at which times this happens?


Edited by Benni123456, 20 October 2020 - 05:01 PM.


#12 csauer52

csauer52

    Messenger

  • *****
  • Posts: 432
  • Joined: 05 Jan 2011
  • Loc: Raleigh, NC

Posted 20 October 2020 - 05:35 PM

I have now taken bias files 200 at 0.00032 sec and 800 at 0.1 sec.. They look totally different. The shorter ones have an entirely different large scale structure...

Perhaps this is just my lack of awareness related to CMOS cameras as I still use CCD. Aren't bias frames simply supposed to measure the read noise of the camera therefore I was under the impression they should be zero duration images?

 

Is there something different with the architecture of CMOS cameras that requires taking bias frames of different lengths?


Edited by csauer52, 20 October 2020 - 06:21 PM.


#13 Benni123456

Benni123456

    Viking 1

  • ****-
  • topic starter
  • Posts: 575
  • Joined: 20 Apr 2017
  • Loc: Munich, Germany

Posted 20 October 2020 - 05:51 PM

Is there something different with the architecture of CMOS cameras that requires taking bias frames of different lengths?

Usually not. But ZWO does not make the sensor. Its from Panasonic i think....

When Panasonic engineers think it is nice to add an amplifier that starts with lower exposures, then there is not much that zwo can do....

 

But they could have documented that in their manual.... 

 

And I want to note that I really need many many biasframes to see the structures.... I now have stacked 800 or so....

 

What is strange:
When I take a superbias from 200 frames at 0.00032 sec, then the superbias shows some kind of hard edges that look defect.

like depicted in this thread:

https://www.cloudyni...gain-10-offset/

 

With 800 bias frames at 0.1 sec, the superbias would not show them for my camera, but rather dark "blobs", as it should be....

 

I do not know if a superbias would be better.

 

In the pixinsight forum, there is a thread that one would have to run it twice with different settings and invoke PixelMath

 

https://pixinsight.c...s-trouble.9411/

 

At that point, It would really be nice if the manufacturer could say something...

 

hm, no I guess the superbias would be very problematic. It would eliminate the fixed pattern noise and put it into larger structures....

 

I think it may probably be better if the stacking sees this noise?


Edited by Benni123456, 20 October 2020 - 06:38 PM.

  • csauer52 likes this

#14 dan_hm

dan_hm

    Viking 1

  • *****
  • Posts: 605
  • Joined: 31 Dec 2012
  • Loc: New Jersey

Posted 20 October 2020 - 05:51 PM

Perhaps this is just my lack of awareness related to CMOS cameras as I still use CCD. Aren't bias frames simply supposed to measure the read noise of the camera therefore I was under the impression they should be zero duration images.

 

Is there something different with the architecture of CMOS cameras that requires taking bias frames of different lengths?

The chip that the ASI1600 (at least the non-"pro" version) uses just behaves oddly when you take too short (or too long) of an exposure. That's the only reason bias -- in the traditional sense of taking the shortest possible exposure -- is not used. I don't believe this issue exists with any newer CMOS camera though someone correct me if I'm wrong.


  • csauer52 likes this

#15 Stelios

Stelios

    Voyager 1

  • *****
  • Moderators
  • Posts: 10,257
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 20 October 2020 - 06:12 PM

The chip that the ASI1600 (at least the non-"pro" version) uses just behaves oddly when you take too short (or too long) of an exposure. That's the only reason bias -- in the traditional sense of taking the shortest possible exposure -- is not used. I don't believe this issue exists with any newer CMOS camera though someone correct me if I'm wrong.

I always thought the opposite, which goes to show the confusion this lack of honesty from the manufacturer has caused. I have the ASI1600MM-C 3rd gen, the model before the -Pro. I have calibrated my (very short) flats with superbias *and* a short-ish (30") master dark with no issues, so I thought it was the -Pro differences (?) that introduced this issue. 



#16 gatsbyiv

gatsbyiv

    Viking 1

  • *****
  • Posts: 886
  • Joined: 29 Mar 2009
  • Loc: Doylestown, PA

Posted 20 October 2020 - 06:49 PM

I always thought the opposite, which goes to show the confusion this lack of honesty from the manufacturer has caused. I have the ASI1600MM-C 3rd gen, the model before the -Pro. I have calibrated my (very short) flats with superbias *and* a short-ish (30") master dark with no issues, so I thought it was the -Pro differences (?) that introduced this issue. 

I don't think there was any lack of honesty from ZWO here, or for that matter QHY or any of the other manufacturers that use the MN34230 sensor.  They probably didn't know, and it appears to not be well characterized by Panasonic either.  We are probably working at the extremes of the use cases they ever imagined for this sensor.  

 

Jon Rista nicely summarized why the weirdness at <0.2" probably doesn't matter here: The TL;DR is basically that the small variations are swamped by the shot noise of your flat frames.  

 

Also, I really don't recommend using the PixInsight Superbias process, if that's what you're referring to.  It was specifically designed for CCD architecture, as it averages along columns.  CMOS chips don't read out that way, as every photosite has its own amplifier.  


Edited by gatsbyiv, 20 October 2020 - 07:07 PM.


#17 Benni123456

Benni123456

    Viking 1

  • ****-
  • topic starter
  • Posts: 575
  • Joined: 20 Apr 2017
  • Loc: Munich, Germany

Posted 21 October 2020 - 02:41 AM

Jon Rista nicely summarized why the weirdness at <0.2" probably doesn't matter here: The TL;DR is basically that the small variations are swamped by the shot noise of your flat frames.  

For me I found that this noise matters.

Especially In your flats if they are also short..... When you use short flats and short bias, there is much more noise than it should be.

 

I have now run a test with my 800 frames 0.1sec bias, and no flats. There is much less noise than with the 0.00032 sec bias...

Flats with a slower speed than 0.1 sec also have much less noise than those at 0.00032 sec.

 

I would like it if zwo would ship these cameras with a biasframe for -15, -20, -25 °C at gains of 0, 139, 300 that is made out of a really large number of frames and a master dark at the same gains for  several exposures, say 1 sec, 10 sec 30 sec 60 sec 120 sec 240 sec, 840 sec that is also made out of a high number of frames, i.e. 1000 or so.

 

For a manufacturer it is easier to have a dark room with 20 cameras which are assembled to make a good dark frame library in one month...



#18 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 665
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 21 October 2020 - 09:41 AM

I have been using a 3rd generation ASI1600MM for several years and I have never used bias frames. Lights are calibrated using matched darks and flats are calibrated using matched darks.

Back in "the old days" when the filter passbands weren't matched as well as they are now we had different exposure times for each filter. Then scaling darks to match the different exposure times for each filter was almost a requirement. Bias frames were used to create scalable dark frames.

Is there a reason to use bias frames with the ASI1600?

Edited by Dean J., 21 October 2020 - 09:46 AM.


#19 gatsbyiv

gatsbyiv

    Viking 1

  • *****
  • Posts: 886
  • Joined: 29 Mar 2009
  • Loc: Doylestown, PA

Posted 21 October 2020 - 12:22 PM

Is there a reason to use bias frames with the ASI1600?

Nope.  It is not recommended to use bias frames with the ASI1600.  Dark flats, yes.  Bias, no. Most CMOS sensors don't deal well with scaling.  

 

But people sometimes use bias frames as dark flats, which can create some confusion around terminology.  (I.e., do you define it based on how you use it or how short the exposure is?)



#20 rwstanley

rwstanley

    Vostok 1

  • *****
  • Posts: 167
  • Joined: 04 Jun 2006
  • Loc: Illinois

Posted 21 October 2020 - 08:30 PM

I have the QHY 163M which is the same sensor as the ASI1600 and I had problems with bias frames and calibration also.  I found that the problem was that when I took 40 bias frames in a sequence, the rapid exposures would create a thermal signal that would affect the bias and make them inaccurate.  What I do now is alternate taking bias and dark frames in sequence so there is never a near continuous  readout of the chip and the bias frames work fine.  I use CCDAP and I just set up a bias and then a dark frame and loop it 40 times, works like a charm.

 

CS, Ron


  • jdupton likes this

#21 palaback

palaback

    Sputnik

  • -----
  • Posts: 31
  • Joined: 29 Mar 2009

Posted 22 October 2020 - 01:57 PM

This is a really useful thread, as I have been trying to perfect and simply my procedure for flats while there are continuous cloudy and smoky clouds here. So just for clarification it sounds like keeping flats from 0.5-10 seconds, and using dark flats of 1 second x 50 and no bias is the way to go. By using a neutral density film filter (0.6) and sheets of paper on my LED flat panel, and changing the 3 levels of dimming, I can keep in range of 1-5 seconds. But from what people have said here its still not clear to me if I need to worry about keeping my exposures in a tight range is that important, other than to make sure exposures are longer than about 0.5 seconds.  The main issue is for NB filters, they could go 10 seconds or more depending on what light intensity and diffusers I use. It sounds like calibrating 10 second ha flats for example with 1 second dark flats is ok.


Edited by palaback, 22 October 2020 - 02:00 PM.


#22 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 665
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 22 October 2020 - 02:41 PM

 The main issue is for NB filters, they could go 10 seconds or more depending on what light intensity and diffusers I use. It sounds like calibrating 10 second ha flats for example with 1 second dark flats is ok.

If you have to take 10 second flats to get sufficient signal, then if it were me,I would match them up with a set of 10 second darks.  That way you get rid of any thermal signal and amp glows that accumulated during those 10 seconds and you also eliminate the camera's bias as well.

 

I always try and match my lights and flats with darks of the same time and temperature.


Edited by Dean J., 22 October 2020 - 04:28 PM.


#23 magsterone

magsterone

    Ranger 4

  • -----
  • Posts: 366
  • Joined: 14 Mar 2006
  • Loc: Mobile, AL

Posted 22 October 2020 - 04:47 PM

So am I to understand that I should not create super bias in pixinsight? Just need to hear it again. I am slow at times. I have the asi1600mm pro. Thanks 



#24 rhart426

rhart426

    Vostok 1

  • -----
  • Posts: 150
  • Joined: 16 Sep 2017

Posted 22 October 2020 - 07:53 PM

So am I to understand that I should not create super bias in pixinsight? Just need to hear it again. I am slow at times. I have the asi1600mm pro. Thanks 

Yep, superbias bad. 

 

This thread has some really excellent information, thank you!

 

To sum it up; there are special cases and situations where you can make bias work, but using matched darks for your lights and flats, without bias, will always do the trick.



#25 gatsbyiv

gatsbyiv

    Viking 1

  • *****
  • Posts: 886
  • Joined: 29 Mar 2009
  • Loc: Doylestown, PA

Posted 22 October 2020 - 09:07 PM

If you have to take 10 second flats to get sufficient signal, then if it were me,I would match them up with a set of 10 second darks.  That way you get rid of any thermal signal and amp glows that accumulated during those 10 seconds and you also eliminate the camera's bias as well.

 

I always try and match my lights and flats with darks of the same time and temperature.

It's a good practice, but with the low dark current of these CMOS cameras, not completely necessary.  I measured 0.28 electrons difference in dark current at -15C between a 1- and 10-second flat dark.  Much smaller than the read noise.  Based on this, I use 1-second flat darks for all of my flats between 0.2 and 10 seconds.




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