Back to forums

TOPIC #1 BIAS VS DARKFLAT (ZWO 6200 MM PRO)

ZWO ASI6200MM Pro Russell Croman Astrophotography NoiseXTerminator Russell Croman Astrophotography BlurXTerminator Aries Productions Astro Pixel Processor (APP) Calibration
Brian DiazLinwood Ferguson
•24 replies•655 views
Brian Diaz avatar
Brian Diaz avatar
HI

Here is my first post about  the camera ZWO 6200 MM PRO

First  of all , carefully and without contaminating both Masters, apply the same value to both.


 TOPIC#1  (  BIAS VS DARKFLAT) TESTED 09/23/2004 

MASTERS DATA INFO :    MASTERS LIGHT HA 35 X 300S  
                                           780MM    F/6.5
                                           FILTER CHROMA 3NM 50MM
                                           6200 MM FULL FRAME MODE, 0C TEMP SENSOR ,GAIN100 ,OFFSET 50
                                           SKY QUALITY  SQM 16.55 , TEMP 84 F ,87 HUMIDITY ( MIAMI, FL USA )


BIAS INFO : MASTER BIAS  EXP 0.01/0 C /100 BIAS
DARKFLAT:  MASTER DARKFLAT  EXP 3S/0C/100 DARK FLATS
MASTERDARK : MASTER DARK EXP 300S/0C/100 DARKS
MASTERFLAT:   MASTER FLAT EXP 3S/0C/100 FLATS
 GAIN 100 OFFSET 50

SOFTWARE INFO:  ASTRO PIXEL PROCESSOR , PIXNSIGHT 


IMAGING PROCESSING INFO   (  FOR BOTH MASTERS )

 CROP ( SAME) 
 DBE : CORRECTION:  SUBTRACTION , SMOOTHING 0.200
 BLURXTERMINATOR  : SHARPEN STARS 0.25. SHPARPEN NONSTELAR 0.90
 NOISEXTERMINATOR : FACTOR 0.50
 STF :  TARGET BACK GROUND 0.15 . SHADOWS CLIPPING -3.1000
 END


RESULT ( DATA) 

SADR HA ( With master dark , flat and bias )

SNR :1.392E+06  61.44dbBackground Mean (normalized) = 0.14516760, (16-Bit) = 9513.55862008

Noise Evaluation: Ch |     noise     | count(%) | layers |0    | 1.813e-01 | 31.80       | 4         |



SADR HA ( With master dark , flat and Dark Flat )

SNR :1.723E+06  62db

BackgroundMean (normalized) = 0.14435297, (16-Bit) = 9460.17168219
 Ch | noise         | count(%) | layers |0    | 1.649e-01 | 25.99       | 4         |


 
IMAGING 


SADR HA ( With master dark , flat and bias )


      SADR HA ( With master dark , flat and Dark Flat )

       
CONCLUSION

I don't see a difference that can be detected by a human eye, much less that it affects or influences that final image

I would appreciate your comments 


Thank you

Clear Skies 
 
Brian
pfile avatar
pfile avatar
i would apply gradient correction only, then measure the noise. NoiseX  (and blurX) is potentially going to do something different to each image. the STF applied won't change anything in terms of the analysis but of course it changes how the image looks to the eye.

you can also LinearFit one gradient-corrected image to the other (assuming they have the same registration reference frame) and then divide one by the other to see if there is any hint of vignetting in the result. you might need to invert the result as most pixels will be very nearly 1.0 after that division.

having said all that i've only ever used bias frames to calibrate my flats with the 6200MM since i take sky flats and the exposures are all slightly different.
Helpful Insightful
Brian Diaz avatar
Brian Diaz avatar
thank you  for your comment

 the blur, noise , was apply after result , just for final imaging view
 
thank you
Linwood Ferguson avatar
Linwood Ferguson avatar
I did some testing a year or two ago, same camera.  I could not see any significant difference (I thought I could zoom way in and find something I thought was a difference but could easily be confirmation bias). 

Because I use a flat panel and can do fixed exposures, I use flat-darks, because it is more correct in theory.  in other words, I do not think it can ever produce results that are worse, and it MAY be better.
Brian Diaz avatar
Brian Diaz avatar
Hi

thank you so much , for you comment , linwood

i could not see any difference too

cs, Brian
Bill McLaughlin avatar
Bill McLaughlin avatar
Brian Diaz:
I don't see a difference that can be detected by a human eye,


IMHO, what else matters?  These are, after all, images for viewing  and not for science. I think sometimes that maybe too much in astroimaging comes at things from a "scientific-only" POV.  This is probably because the hobby attracts tech types by it's nature. I would certainty not argue that theory and the supporting science and math are not important - clearly they are very much so. But at some point one gets so far into the mathematical and theoretical weeds that one can fail to see the esthetic and/or practical garden.
Engaging
Linwood Ferguson avatar
Linwood Ferguson avatar
It may be worth explaining for those who stumble on this thread… 

Calibrating a flat with a dark is "correct" in the same sense as a light, in fact a flat is a form of light.

However, unlike most astro lights, a flat is taken at very short exposure, and so the difference in noise from a bias at 0.001 seconds (or whatever) and one at 1 second is pretty negligible, you might really see a difference in 0.001 and 300 seconds, but not 1 or 2 seconds. 

So the theory is if your flats are dynamic and have exposures which vary a lot, it's a pain to make darks at the same shutter speed, and using bias darks are a big win for a tiny (maybe invisible) loss. 

The same logic applies to those whose flats are at consistent exposure times, but just don't want to bother taking matching flat-darks for something that may not make a difference, just the logic is a bit less compelling.

And doing matching flat-darks if your flats are consistent is, in my mind, not much work for one less thing to worry about.
Helpful Insightful Respectful
Brian Diaz avatar
Brian Diaz avatar
hi, bill

thank you for comments

i agree , however there are something very  important there

it is CURIOSTY , "I don't have any special talents, but I am deeply curious."  Albert Einstein

cs, Brian
Linwood Ferguson avatar
Linwood Ferguson avatar
Bill McLaughlin:
Brian Diaz:
I don't see a difference that can be detected by a human eye,


IMHO, what else matters?  These are, after all, images for viewing  and not for science. I think sometimes that maybe too much in astroimaging comes at things from a "scientific-only" POV.  This is probably because the hobby attracts tech types by it's nature. I would certainty not argue that theory and the supporting science and math are not important - clearly they are very much so. But at some point one gets so far into the mathematical and theoretical weeds that one can fail to see the esthetic and/or practical garden.

As a counter-point, a lot of what we do with raw images and final stacks are specifically aimed at magnifying very, very small differences so they become visible.  Practically every image we take starts off with invisible data, brought out by subtracting noise and stacking signal and as we stack trying to deal with everything from pixel-to-pixel variations in the sensor to starlink trails. 

There is much we can't see that ends up making a difference upstream by final processing.  It's hard to know that this is for sure not one.   I THINK it is as well.  But...
Helpful Insightful Respectful Engaging
Steven Rosen avatar
Steven Rosen avatar
I have never used dark flats.  I don't think it's worth the  
extra work. 

I use a dark master library that I create in the Spring and Fall for NB and RGB at several temperatures (-5C, -10C, -15C and - 20C. and exposures ( 60, 90, 180, 240, and 300sec.). 

I take fresh flats and bias for each refocus  filter every night I shoot.

I'm very satisfied with the results.
Helpful Concise
Brian Diaz avatar
Brian Diaz avatar
hi, steven

thank you so much for your comments

 cs, Brian
Linwood Ferguson avatar
Linwood Ferguson avatar
Steven Rosen:
I take fresh flats and bias for each refocus  filter every night I shoot.
.

Why do you take biases nightly if darks are only every 6 months?
Luca Bartek avatar
Luca Bartek avatar
If you were to remove the "regular" darks from the equation as well, you'll find that they're also completely unnecessary, and I would argue you're better off not using them.
According to the ZWO website, "At a cooling temperature of 0 °C, the dark current noise is only 0.003e/s/pixel"
For your 300sec image that would be 300x0.003e/sec/pixel = 0.9e/sec/pixel
This is less than the read noise.
You risk introducing more noise with your dark than dark current you're removing.
Of course if you're comparing a masterbias made of 100 frames and a masterdark made of 100 frames, you'll get a similar result.
But you can easily do a masterbias of 1000 frames and you'll even out all randomness. Also, the camera supports captures of 0.00032sec, so why not get them as close to a true bias as possible?

See my reasoning in more detail in this video I made on this very topic: https://www.youtube.com/watch?v=EQGWr34Fu0c (start at 6:39)

Darks are not meant to remove noise. They're meant to remove dark signal, which you don't have.
Helpful Insightful Engaging
Linwood Ferguson avatar
Linwood Ferguson avatar
Now the argument to use NOTHING for a flat dark to me is much more interesting than using a bias vs matching exposure dark.   It's related to the argument never to use bias and dark together for a light (with CMOS when not scaling darks), the extra math operation does more harm than good (plus your dark - bias my go negative).
Luca Bartek avatar
Luca Bartek avatar
Linwood Ferguson:
Now the argument to use NOTHING for a flat dark to me is much more interesting than using a bias vs matching exposure dark.


this couldn't work because you have this baseline you're subtracting from your lights (be it bias or dark) and you have to subtract it from your flats as well as flat calibration is a division rather than a subtraction. With a great camera like the 6200mm you may not notice a visible difference but it is definitely there, and if you want to have a correctly calibrated image (e.g. for photometry) this is a must
Brian Diaz avatar
Brian Diaz avatar
thank you , luca

for you comments

that  was just a test about the zwo 6200 with the most important  eyes view 

 i have done  different test  with dark , without dark , with flat and dark flat and all version

i will unload a different    final imaging , one with dark flat and dark flat , flat and bias only and  flat and darkflat 

and is impossible for human eyes see the different  

nobody can see  the different and to be honesty that is the most import the final or not 
   personally i take more care when processing my imaging 

 CS, BRIAN
wsg avatar
wsg avatar
I haven't used darks in 2 years, just flats and dark flats entered into WBPP as bias.




scott
Brian Diaz avatar
Brian Diaz avatar
thank you so much for your comment , WSG

my next test is with out dark
only flat and dark flat and also  flat and bias

cs, Brian
starpixels avatar
starpixels avatar
I've noticed that the astro community has slowly trended toward the use of flats & dark flats (or biases) exclusively while skipping the use of darks. This of course coincides with the use of newer cmos cameras with their back illuminated chips. Something like the IMX571 has negligible dark current (at 0C and below). That being said, as soon as someone implies you don't need darks, you will surely hear from the other camp.  Some might say it's generational - old timers with PTSD from the early CCD cameras are sticking with what works for them. If you poke around online you might even find mathematical formulas that supposedly prove that you are literally spinning your wheels every time you add dark frames into the mix (again, with the newer cameras). I'm agnostic on the issue so I won't try to sell either idea here.  If you maintain a dark library, it probably won't matter much in terms of time expenditure. If you feel better taking a rigorous approach to calibration, then why not? Sometimes "best practices" are not about having a clear win, it's just something you do to cover the bases.
Well Written Insightful Respectful Engaging
Astro With RoRo avatar
Astro With RoRo avatar
Steven Rosen:
I have never used dark flats.  I don't think it's worth the  
extra work. 

I use a dark master library that I create in the Spring and Fall for NB and RGB at several temperatures (-5C, -10C, -15C and - 20C. and exposures ( 60, 90, 180, 240, and 300sec.). 

I take fresh flats and bias for each refocus  filter every night I shoot.

I'm very satisfied with the results.

That sounds like far more work than using dark flats. If you use the same exposure time for your flat frames each time then you can just built the dark flats when you build your dark library and keep them for the same time frame you keep your darks for. Using a flat panel or light box allows this to be done quickly and easily, even fully automated should you wish.
Helpful
David Payne avatar
David Payne avatar
Nice post Brian.  I agree with you, both images are great.   If you really want to know the difference, just subtract one image from the other (with renormalization).  From the statistics and level of detail in the subtraction image, you might get a technical answer, but I think any difference is small.  I am still believer in a dark/bias library that I apply to both flats and lights and only update 1/year, but this requires a cooled camera.

Cheers,
Dave
Well Written Respectful Concise Supportive
Brian Diaz avatar
Brian Diaz avatar
thank you so much, David

CS, Brian
Nick Grundy avatar
Nick Grundy avatar
I really can't tell the difference at all. (there's more trained eyes for that around here)

I should flats and darkflats with every period (1-2 weeks of clear skies it seems), but darks every 3-4 months, if that. I've never done bias frames as I wasn't sure they were needed and never got to it. 

If someone tells me adding bias frames to my calibration will improve quality, I definitely do it. I just haven't heard that before. 

Is it a holdover practice from ccd cameras? (I only have cmos)
Linwood Ferguson avatar
Linwood Ferguson avatar
Nick Grundy:
If someone tells me adding bias frames to my calibration will improve quality, I definitely do it. I just haven't heard that before.

I don't think anyone is suggesting ADDING bias to a cmos workflow that includes darks (whether for flats or lights). 

In CMOS days bias was used with darks to scale darks (maybe for other reasons).   The idea being that if you think conceptually of a light has having a (Signal + Bias + Dark), and the dark component was proportional to length of exposure, then you could calibrate a light with a different exposure by Light - Bias - Dark * (T / T0)  where T was the actual exposure and T0 was the exposure of the dark. 

Because there's so little difference in dark and bias this works very poorly for CMOS, at least in my experience, and I think it's generally accepted you should never use both bias and dark.  Someone please correct me if I'm wrong.
Helpful Insightful Respectful
Brian Diaz avatar
Brian Diaz avatar
the point is  that in the end there is no  different that it can be detected by the human eye, so it doesn't  matter 
wish variant you use no one will know

for me , i prefer to spend my time processing the final image in the end it is what counts and we see 100%