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

QHY183M & Flats = Frustration

  • Please log in to reply
94 replies to this topic

#51 kingjamez

kingjamez

    Vanguard

  • *****
  • Posts: 2055
  • Joined: 03 Oct 2006
  • Loc: Fairfax, VA

Posted 17 September 2019 - 12:31 PM

In looking more into this I found this: https://www.astropix...ting/#post-4533 .

 

I was amazed, it describes this issue to a T. I use APP for calibration and PI for everything else. Perhaps this is why I haven't come across this issue, APP fixes it without user interaction.

 

I want to be very clear that I'm no expert. However in thinking about how the calibration process works, I can't find a good reason to use a bias frame if you have matched darks and flat darks.

 

A regular, uncalibrated dark will contain all of the bias offsets, and it's subtracted from a light frame just like a bias is. Flats can be a good use for a bias, but if you are taking flat darks anyway then the bias is redundant.

 

Through all of my testing, I've yet to see a case where it makes sense to use a seperate bias ever, as long as you are taking flat darks. 

 

So even if your camera's bias wasn't over-correcting I don't know why it would be useful. 

 

I'd still like to know why your bias has a higher median ADU than your darks. So far all of my bias frames have a lower median ADU than all of my dark frames as they should.

 

One possibility is that when your master bias was made, the rejection filters were too aggressive causing the median to rise. No idea if that's right, but it's a possible explanation.

 

-Jim  



#52 terry59

terry59

    Cosmos

  • *****
  • topic starter
  • Posts: 9169
  • Joined: 18 Jul 2011
  • Loc: Colorado, USA

Posted 17 September 2019 - 03:48 PM

Thanks for all of the analysis Jim. This has been very informative. I read the link as verification of using a pedestal but it didn't provide enough specificity to determine how APP does it. The reason I am on this road is because I haven't gotten a good set of flats yet so dark flats haven't mattered. I'm going to try some sky flats next and see how that goes. 


  • kingjamez likes this

#53 Stelios

Stelios

    Cosmos

  • *****
  • Moderators
  • Posts: 7719
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 17 September 2019 - 04:09 PM

I have rerun everything using an uncalibrated master dark. Here are the two results (no bias and bias with a pedestal of 100) with only a STF applied. I'm curious to know what people think

I found the left image *sllllllightly* more contrasty. But it could well be the resolution limitations.



#54 bobzeq25

bobzeq25

    Hubble

  • *****
  • Posts: 16931
  • Joined: 27 Oct 2014

Posted 17 September 2019 - 04:18 PM

Just another data point.

 

With my 183C I use a Spike a Flat panel for flats, and bias instead of flat darks.  The two are often interchangable, although some cameras seem to need dark flats.  the 183 does not seem to me to be one of those.

 

The thing about using flats without either one of those is that the math just doesn't work right.  The reason is that the flats have very high ADU levels, and you divide the lights by them (rather than subtracting them, as with darks).  No bias or dark flat, and the math doesn't work correctly.  You do calibrate the flats with bias or flat darks before integrating them into a master flat.


Edited by bobzeq25, 17 September 2019 - 04:20 PM.


#55 terry59

terry59

    Cosmos

  • *****
  • topic starter
  • Posts: 9169
  • Joined: 18 Jul 2011
  • Loc: Colorado, USA

Posted 17 September 2019 - 04:37 PM

I found the left image *sllllllightly* more contrasty. But it could well be the resolution limitations.

The one on the left had pedestal added

 

smile.gif 



#56 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23722
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 17 September 2019 - 04:46 PM

Thanks Jon....I really like this camera. It is beating the tar out of the KAF8300. As to flats, it is taking many, many seconds to reach the 30k ADU level when using the Flatman and that pattern persists regardless of what I've tried so I think you are correct as to the cause because I don't see anything like that with the KAF8300 and its mechanical shutter.

Yeah, with the shutter you wouldn't have to worry about readout. Well, if you need tips for open aperture blue sky flats, let me know. They are pretty strait forward, and your exposures will usually be about 0.01-0.1 seconds.

 

I have been reusing my flats a long time. Well, at least, with the FSQ106 and the NC. The rotator has been nice since it allows me to get exactly the same angle (give or take a fraction of a degree) each night. I have been using biases lately to calibrate, since at a tenth of a second or less there is no amp glow to speak of. So hopefully, once you get your flats figured out...if you are able to keep the camera on the scope and not change it's orientation, you won't have to worry about it again for a while.



#57 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23722
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 17 September 2019 - 05:12 PM

OK, what's "short" and what's "longer" ? I take flats depending on filter at exposures from 0.01" to 1.2". We'll stipulate that 0.01" is short, but is 1.2" "longer"?

 

Also I keep hearing this about bias, but I have an ASI1600MM-C, which is a Panasonic "M" camera, correct? I have been using a superbias (100 bias frames) taken at shortest possible exposure with SGP for *years* now, with no ill effects (following the Warren Keller approach in Inside Pixinsight). I am tempted to take some dark flats just to compare and see if there's any discernible difference, but it's not going to be major for sure.

Fraction of a second is definitely "shorter." With these CMOS cameras, usually there are two types of exposure control. Camera controlled and driver controlled. Now at one point the cutoff was different for ZWO cameras, but I think these days both QHY and ZWO use 1 second as the cutoff point for camera-controlled exposures. Below, the camera controls the exposure (hardware timing is necessary to get exact sub-second exposure lengths, basically). Above 1 second, the driver controls exposure. Now right around 1 second or so, the actual time the sensor is exposed before readout may not be exactly 1 second, since there can be some latency between the driver issuing the command to end exposure, and the camera actually receiving it and ending exposure. Once you get up to a few seconds (what I would call "longer"), then this discrepancy doesn't really matter since it is usually just a fraction of a second anyway.

 

As for bias. The Panasonic M-series sensors SPECIFICALLY (and only them) have some issues with extremely short exposures. Under 0.2 seconds, there is an inconsistent signal that will fill the frame...but, it does not fill it the same each frame. So literal bias frames, 0 second exposures (the minimum exposure on the Panasonic M cameras is usually 32 microseconds, or 32 thousandths of a millisecond), will usually exhibit a changing gradient across the frame from one frame to the next. This variable gradient tends to result in a sort of radial gradient once all the frames are integrated, with a darkish spot in one area somewhere within the frame that brightens to the edges. The crazy thing is, this gradient where it gets brightest can actually be up to a few ADU brighter than the configured bias offset level itself. As you start to increase exposure from 0 to 0.2 seconds, the level of the bias will start to increase. So there is also inconsistency in the mean level of the biases for exposures below 0.2 seconds:

 

z3HkeXk.jpg

 

When you use 0.2 second or longer exposures, this odd variability in the frames disappears. The frames become consistent, totally flat, and the entire frame falls within the read noise RMS of the configured bias offset. You can see in the chart above how mean and median levels flatten out once you reach 0.2 seconds. From that point on, you will slowly start to see an increase in mean level due to dark current, but it is nothing like when exposures are under 0.2 seconds.

 

Note that this testing was done on an ASI1600MM-Cool, and done some time ago. The same kind of testing was also actually done by a couple other people, one guy actually did similar tests before I did and at the time I wasn't sure I believed his results, then I did my own testing and found the same problem. There have been tests with both QHY and ZWO cameras, and both manufacturers have recommended 0.3 second minimum exposures (I think QHY officially documented it in their camera manual, I don't recall if ZWO did, but I know ZWO recommends it on their forums).

 

NOW, I am not sure if these manufacturers recommend 0.3 second exposures for ALL CMOS cameras. If they do, it may just be a blanket precautionary thing. In my own testing and real-world experience with the ASI183, as well as my experience assisting at least a couple dozen other people with calibration struggles with the IMX183 cameras in general, I can say with solid confidence that the IMX183 does NOT have the same problem. It's exposures work well all the way down to minimum exposure ("zero seconds" with most acquisition programs). You can use bias frames for calibrating short flats if you wish with the IMX183, there is no specific short exposure limitation. You can use flats down to tiny fractions of a second if you need to.

 

The Panasonic M cameras are fine cameras...but they do need a little bit more care and attention to the calibration frames because of the bias instability issue. Just keep in mind, the Panasonic M issues are just for that sensor, not every CMOS sensor on the market. Also keep in mind, the Panasonic M sensor was really originally designed as a M4/3 mirrorless camera/video sensor, and was generally intended to be used for video capture with an Olympus camera. So it wasn't explicitly designed for our kind of ultra low light market. It is sadly the only M4/3 sized mono CMOS sensor on the market right now, so if you need fast downloads, or need high gain for short NB exposures, or anything like that...there really are no alternatives. The IMX183 sensor is just slightly over 1/2 the area of the Panasonic M sensor...HALF the area! ;P

 

I am hopeful that there will be a small pixel mono APS-C sensor on the market, preferably from Sony (as they really do seem to know how to make a good sensor), sometime soon here. In all the talk about the forthcoming IMX455 (full frame, 36x24mm, mono CMOS), I think there were a couple of posts about a mono Sony APS-C that might find its way into a QHY camera at some point here. So there is hope that we'll get a larger mono sensor some time within the next year here, for those who need a larger frame and don't want to deal with the calibration nuances of the Panasonic M.



#58 OldManSky

OldManSky

    Surveyor 1

  • *****
  • Posts: 1967
  • Joined: 03 Jan 2019
  • Loc: Valley Center, CA USA

Posted 17 September 2019 - 07:18 PM

Great information as always, Jon.

I'm hoping for an APS-C sized mono (Sony?) sensor as well, with 3.0 - 3.5 micron pixels.  That would make me really, really happy :)


  • Jon Rista likes this

#59 Stelios

Stelios

    Cosmos

  • *****
  • Moderators
  • Posts: 7719
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 17 September 2019 - 07:22 PM

Above 1 second, the driver controls exposure. Now right around 1 second or so, the actual time the sensor is exposed before readout may not be exactly 1 second, since there can be some latency between the driver issuing the command to end exposure, and the camera actually receiving it and ending exposure. Once you get up to a few seconds (what I would call "longer"), then this discrepancy doesn't really matter since it is usually just a fraction of a second anyway.

 

This would indicate to me that I would get a variation in the ADU count for exposures under a couple of a seconds with a given filter. I have *not* experienced this, so unless this would not affect the ADU counts (which vary by 10's of ADU's in a 64K space) perhaps my own copy of the camera is immune.

 

On the second issue, looking at the graph you show, the variation is when exposure lengths vary. But if you are taking bias at "0" (or 32 microseconds) wouldn't you always have a consistent value? Are you saying that this value would *not* be what the effect of the bias (shutter noise) would be at longer exposures? If so, I wonder if the calibration (as per WK, I uncheck "calibrate" for the bias and check *both* "calibrate" and "optimise" for the master dark) is corrected by PI by using the master dark instead of the bias. If not, where would the difference show? What would it affect? This is my stretched (auto-stretch) superbias.

 

superbias.JPG

 

Also, wouldn't having a (say) 1" long bias mean that the noise level would be the shutter noise *plus* 1 second of dark noise (and thus, not a true bias)?



#60 kingjamez

kingjamez

    Vanguard

  • *****
  • Posts: 2055
  • Joined: 03 Oct 2006
  • Loc: Fairfax, VA

Posted 17 September 2019 - 07:30 PM

Ideally the bias is representative of the “real” minimum pedestal always present in the images. Until you get above 0.2s (or so) on the 1600mm-c you don’t get to the real minimum.

So far in my testing even at veryshort exposures the 183mm(the subject of this thread) has a stable, accurate, median ADU which can be thought of as the accurate minimum.

-Jim
  • bobzeq25 likes this

#61 terry59

terry59

    Cosmos

  • *****
  • topic starter
  • Posts: 9169
  • Joined: 18 Jul 2011
  • Loc: Colorado, USA

Posted 17 September 2019 - 08:48 PM

Ideally the bias is representative of the “real” minimum pedestal always present in the images. Until you get above 0.2s (or so) on the 1600mm-c you don’t get to the real minimum.

So far in my testing even at veryshort exposures the 183mm(the subject of this thread) has a stable, accurate, median ADU which can be thought of as the accurate minimum.

-Jim

Jim....can you specify "very short"? QHY has minimum exposure at 5us



#62 kingjamez

kingjamez

    Vanguard

  • *****
  • Posts: 2055
  • Joined: 03 Oct 2006
  • Loc: Fairfax, VA

Posted 17 September 2019 - 09:03 PM

Jim....can you specify "very short"? QHY has minimum exposure at 5us

Not really. I set my bias' test at 0.00, 0.10, and 0.20 and got a spread of  0.4 median ADU. Any of those would be sufficient. 

 

I don't know what 0.00 equates to, my fits headers just say 0. 

 

Again, I don't use bias frames, I use flat darks. 

 

-Jim



#63 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23722
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 17 September 2019 - 09:04 PM

This would indicate to me that I would get a variation in the ADU count for exposures under a couple of a seconds with a given filter. I have *not* experienced this, so unless this would not affect the ADU counts (which vary by 10's of ADU's in a 64K space) perhaps my own copy of the camera is immune.

 

On the second issue, looking at the graph you show, the variation is when exposure lengths vary. But if you are taking bias at "0" (or 32 microseconds) wouldn't you always have a consistent value? Are you saying that this value would *not* be what the effect of the bias (shutter noise) would be at longer exposures? If so, I wonder if the calibration (as per WK, I uncheck "calibrate" for the bias and check *both* "calibrate" and "optimise" for the master dark) is corrected by PI by using the master dark instead of the bias. If not, where would the difference show? What would it affect? This is my stretched (auto-stretch) superbias.

 

attachicon.gif superbias.JPG

 

Also, wouldn't having a (say) 1" long bias mean that the noise level would be the shutter noise *plus* 1 second of dark noise (and thus, not a true bias)?

Possibly. There is also the rolling shutter effect, which can affect exposures around that length as well. This introduces a slight gradient, which I forgot to mention. My guess is the gradient issue trumps the timing issue. The timing issue would be fairly small though, and with exposures to ~50% saturation, eh, the timing variation is just going to get lost in the noise anyway, really. The timing issue around 1 second would affect your lights, really, not the flats.

 

Regarding the variation under 0.2 seconds. It is not just variation with exposure length. You'll have to try it. The FIELD changes as well...with exposures under 0.2 seconds there is a variable gradient in bias signal across the field. So they are basically useless, and they actually introduce that gradient into anything you calibrate with them. So it is not just a variation in mean level with exposure, it is a variation on offset within each pixel across the field. This is tough to show without creating a video of it, which is why I haven't done it. You just need to run a simple experiment, though...grab a set of 0 second biases, then flip through them in PI's Blink tool. You will see what I am talking about.

 

Also...do NOT, not not not ever, use SuperBias!! :p Superbias destroys, no it utterly wipes out, ALL microstructure in the bias pattern. SuperBias, which I used to use myself, is IMO a really bad idea and can lead to worse quality integrations, especially with CMOS cameras and shorter exposures. You want to use a real bias, integrated from a good number of frames, to ensure that you are properly correcting for the REAL bias pattern. This pattern is quite structured...includes vertical banding as well as some smaller structures at a smaller scale across pixels (usually little diagonal structures). If you are using a bias, don't use SuperBias. It's useless.

 

This is what you should ultimately get out of a properly constructed master bias...note the pattern at 256 subs...this is the bias pattern that bias frames calibrate for. Technically speaking, it is in the darks as well, if you use just darks. The problem with the ASI1600 is with exposures under 0.2 seconds, the field is not flat, and there is this additional variation, this non-uniform dynamic gradient across the frame that varies the bias level. This gradient, being random and non-fixed, thus cannot be calibrated out...in fact, calibration with bad biases ADDS the gradient to everything you calibrate, even though at the same time you are removing the bias pattern seen below.

 

IsN9fKA.gif


Edited by Jon Rista, 17 September 2019 - 09:09 PM.

  • OldManSky likes this

#64 rockstarbill

rockstarbill

    Fly Me to the Moon

  • *****
  • Posts: 6320
  • Joined: 16 Jul 2013
  • Loc: Snohomish, WA

Posted 17 September 2019 - 09:45 PM

This would indicate to me that I would get a variation in the ADU count for exposures under a couple of a seconds with a given filter. I have *not* experienced this, so unless this would not affect the ADU counts (which vary by 10's of ADU's in a 64K space) perhaps my own copy of the camera is immune.

 

On the second issue, looking at the graph you show, the variation is when exposure lengths vary. But if you are taking bias at "0" (or 32 microseconds) wouldn't you always have a consistent value? Are you saying that this value would *not* be what the effect of the bias (shutter noise) would be at longer exposures? If so, I wonder if the calibration (as per WK, I uncheck "calibrate" for the bias and check *both* "calibrate" and "optimise" for the master dark) is corrected by PI by using the master dark instead of the bias. If not, where would the difference show? What would it affect? This is my stretched (auto-stretch) superbias.

 

 

Also, wouldn't having a (say) 1" long bias mean that the noise level would be the shutter noise *plus* 1 second of dark noise (and thus, not a true bias)?

I would not use a Superbias at all. Considering the CMOS cameras can download frames in moments, getting a good 256 stack wouldnt take much time at all. That is likely overkill, but its what I do since I take Bias and Darks on nights like tonight, where it is pouring down rain. 



#65 NorthField

NorthField

    Viking 1

  • *****
  • Posts: 933
  • Joined: 01 Jun 2017
  • Loc: SW Missouri

Posted 17 September 2019 - 10:22 PM

Clear as a bell here 😎

( I’ll say that again November 21st 🙄 )

#66 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23722
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 17 September 2019 - 10:27 PM

Also, wouldn't having a (say) 1" long bias mean that the noise level would be the shutter noise *plus* 1 second of dark noise (and thus, not a true bias)?

Yes, it would be a dark. A flat dark, if they were taken to calibrate flats.



#67 terry59

terry59

    Cosmos

  • *****
  • topic starter
  • Posts: 9169
  • Joined: 18 Jul 2011
  • Loc: Colorado, USA

Posted 18 September 2019 - 12:57 PM

Yeah, with the shutter you wouldn't have to worry about readout. Well, if you need tips for open aperture blue sky flats, let me know. They are pretty strait forward, and your exposures will usually be about 0.01-0.1 seconds.

 

I have been reusing my flats a long time. Well, at least, with the FSQ106 and the NC. The rotator has been nice since it allows me to get exactly the same angle (give or take a fraction of a degree) each night. I have been using biases lately to calibrate, since at a tenth of a second or less there is no amp glow to speak of. So hopefully, once you get your flats figured out...if you are able to keep the camera on the scope and not change it's orientation, you won't have to worry about it again for a while.

I collected some sky flats this morning....an interesting experience. With this Ha filter the sky was well through civil twilight before the ADU values broke 20k. This one at ~40k may represent what I need. I'm going to shoot a set of these and see

Attached Thumbnails

  • Capture.JPG

  • Jon Rista likes this

#68 kingjamez

kingjamez

    Vanguard

  • *****
  • Posts: 2055
  • Joined: 03 Oct 2006
  • Loc: Fairfax, VA

Posted 18 September 2019 - 01:07 PM

I did the same thing this morning after reading John's suggestion. I did diffuse it with some white fabric, but I was surprised at how bright the sky had to be before I got reasonable exposure times.

 

I waited until the sun had mostly come over the horizon so I could do a run without the sky changing brightness quickly. I eventually got a good run at around 1 second exposure with the sun mostly fully risen.

 

Not sure if I like this method as it seems pretty variable. My LED panel is always the same, but more hassle to setup.

 

-Jim



#69 bobzeq25

bobzeq25

    Hubble

  • *****
  • Posts: 16931
  • Joined: 27 Oct 2014

Posted 18 September 2019 - 04:28 PM

I did the same thing this morning after reading John's suggestion. I did diffuse it with some white fabric, but I was surprised at how bright the sky had to be before I got reasonable exposure times.

 

I waited until the sun had mostly come over the horizon so I could do a run without the sky changing brightness quickly. I eventually got a good run at around 1 second exposure with the sun mostly fully risen.

 

Not sure if I like this method as it seems pretty variable. My LED panel is always the same, but more hassle to setup.

 

-Jim

My Spike a Flat is easier to setup than sky flats. Park the telescope vertically, lay the panel on top.  My t-shirts have retired.  <smile>

 

You do pay for that convenience.


Edited by bobzeq25, 18 September 2019 - 04:29 PM.

  • Stelios likes this

#70 OldManSky

OldManSky

    Surveyor 1

  • *****
  • Posts: 1967
  • Joined: 03 Jan 2019
  • Loc: Valley Center, CA USA

Posted 18 September 2019 - 04:28 PM

terry, that looks like a Valentine's day flat.  Or the Heart nebula :)

 

Do you see a similar pattern with other filters, or just with Ha?

Curious...



#71 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23722
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 18 September 2019 - 04:46 PM

I did the same thing this morning after reading John's suggestion. I did diffuse it with some white fabric, but I was surprised at how bright the sky had to be before I got reasonable exposure times.

 

I waited until the sun had mostly come over the horizon so I could do a run without the sky changing brightness quickly. I eventually got a good run at around 1 second exposure with the sun mostly fully risen.

 

Not sure if I like this method as it seems pretty variable. My LED panel is always the same, but more hassle to setup.

 

-Jim

When I talk about blue sky flats, that is with the sun high in the sky. I'm not talking about twilight sky flats with a diffuser. I'm talking no diffuser, open aperture, empty blue sky, DURING THE FULL ON DAYTIME. Your exposures will be sub-second, and you'll easily be able to get 30-40k ADU. Not only that, with exposures around 0.1 down to 0.01 seconds, you can yank down 36 flat frames per channel in a matter of a couple minutes. 

 

IMO, there is nothing easier. In fact, I've programmed my setup, which I have no automated flat panel for, to take flats like this against the north pole some time after teh sun has risen, so there can be no stars in the flats. I don't need to worry about a diffuser, nor a flat panel, or anything like that. Just empty blue sky. 


Edited by Jon Rista, 18 September 2019 - 04:47 PM.

  • kingjamez and bobzeq25 like this

#72 Stelios

Stelios

    Cosmos

  • *****
  • Moderators
  • Posts: 7719
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 18 September 2019 - 05:30 PM

terry, that looks like a Valentine's day flat.  Or the Heart nebula smile.gif

 

Do you see a similar pattern with other filters, or just with Ha?

Curious...

No, it's Pluto at 2,000,000 power!


  • OldManSky likes this

#73 Stelios

Stelios

    Cosmos

  • *****
  • Moderators
  • Posts: 7719
  • Joined: 04 Oct 2003
  • Loc: West Hills, CA

Posted 18 September 2019 - 05:39 PM

I would not use a Superbias at all. Considering the CMOS cameras can download frames in moments, getting a good 256 stack wouldnt take much time at all. That is likely overkill, but its what I do since I take Bias and Darks on nights like tonight, where it is pouring down rain. 

I have the utmost respect for you and Jon Rista, but I don't understand why Warren Keller would recommend it (rave about it, really) in his book?

 

And then there's this post by Juan (PI creator).

 

Still, perhaps I'll create a masterbias of 500 0.2" frames and compare--although it won't happen today.



#74 bobzeq25

bobzeq25

    Hubble

  • *****
  • Posts: 16931
  • Joined: 27 Oct 2014

Posted 18 September 2019 - 06:07 PM

I have the utmost respect for you and Jon Rista, but I don't understand why Warren Keller would recommend it (rave about it, really) in his book?

 

And then there's this post by Juan (PI creator).

 

Still, perhaps I'll create a masterbias of 500 0.2" frames and compare--although it won't happen today.

This may be one key to understanding the difference of opinion.  In the Juan Conjero post you cite.

 

"If you just have 20 or 50 bias frames, a superbias will improve the performance of our dark scaling algorithm significantly."

 

But what if you don't use PI's dark scaling?  I believe that with CMOS (at least some cameras), you absolutely should not scale darks?  I don't scale them with my 183.

 

Not to mention that some CMOS need dark flats, instead of bias, which renders the discussion moot.

 

The dispute is international.

 

https://www.tommasor...o-superbias-no/

 

This may be another battle where the big boys disagree for small reasons, and it doesn't matter much for us small fry.  <grin>  I'd guess if we looked at two images we've done, same workflow, we'd see little if any noticeable difference.  Our other issues would dominate.

 

Also, from the PI forum thread:

 

"when I tried the Superbias I did not see any difference, that’s because I use 100 bias subs which is producing a reasonably good quality bias anyway."


Edited by bobzeq25, 18 September 2019 - 06:30 PM.

  • Stelios likes this

#75 jdupton

jdupton

    Surveyor 1

  • *****
  • Posts: 1950
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 18 September 2019 - 06:11 PM

Stelios,

 

   This is my rudimentary understanding of the SuperPixel Process. It could well be way off base.

 

   If I recall the internal descriptions of the SuperBias process correctly, it was based on observations of the way CCD chips work. In a CCD, all the pixels in a column are shifted down (or up) to the readout row. Because of this, each pixel's data traverses through other pixels in that column. This process can result in a somewhat predictable pattern of the data in that column. The purpose of the SuperBias process is to do an analysis of the data using a multiscale process to remove extra noise which makes the Bias frame have many of the characteristics of a stack of many more frames.

 

   In a CMOS camera, each pixel is read out (more or less) independently of other pixels in the same column. (They can share the same A/D but do not get shifted through other pixels as they are read and converted.) Because the readout is different, the assumptions made by the SuperBias process may not apply. It is not a case of SuperBias suddenly being wrong; it's just that what it does may not be appropriate for CMOS sensors given its origins and goals.

 

 

John


Edited by jdupton, 18 September 2019 - 06:36 PM.

  • OldManSky likes 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