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

How does one mitigate the RBI in their 16803 sensor?

  • Please log in to reply
58 replies to this topic

#26 swalker

swalker

    Associate Editor - Sky & Telescope

  • *****
  • Vendors
  • Posts: 1,637
  • Joined: 22 Jan 2007
  • Loc: 42.9225°N, 71.2242°W

Posted 19 January 2018 - 08:43 AM

In my experience using two FLI PL16803 CCD cameras for the MDW Sky Survey (www.mdwskysurvey.org) we use RBI pre-flash protocol on our darks, flats, and light exposures. it is really the only way to ensure complete removal of residual stars, particularly since we image more than a dozen fields per night per scope.


  • dvalid and SeymoreStars like this

#27 SeymoreStars

SeymoreStars

    Mercury-Atlas

  • *****
  • topic starter
  • Posts: 2,555
  • Joined: 08 May 2014
  • Loc: Pennsyltucky

Posted 19 January 2018 - 09:40 AM

In my experience using two FLI PL16803 CCD cameras for the MDW Sky Survey (www.mdwskysurvey.org) we use RBI pre-flash protocol on our darks, flats, and light exposures. it is really the only way to ensure complete removal of residual stars, particularly since we image more than a dozen fields per night per scope.

Thank you Sean. May I ask what camera control software you use?



#28 swalker

swalker

    Associate Editor - Sky & Telescope

  • *****
  • Vendors
  • Posts: 1,637
  • Joined: 22 Jan 2007
  • Loc: 42.9225°N, 71.2242°W

Posted 19 January 2018 - 10:25 AM

 

In my experience using two FLI PL16803 CCD cameras for the MDW Sky Survey (www.mdwskysurvey.org) we use RBI pre-flash protocol on our darks, flats, and light exposures. it is really the only way to ensure complete removal of residual stars, particularly since we image more than a dozen fields per night per scope.

Thank you Sean. May I ask what camera control software you use?

 

ACP Expert (which incorporates Maxim DL) http://acpx.dc3.com/


Edited by swalker, 19 January 2018 - 10:28 AM.

  • SeymoreStars likes this

#29 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 19 January 2018 - 11:24 AM

Sean,

That looks like a very interesting project!

John



#30 SeymoreStars

SeymoreStars

    Mercury-Atlas

  • *****
  • topic starter
  • Posts: 2,555
  • Joined: 08 May 2014
  • Loc: Pennsyltucky

Posted 19 January 2018 - 11:50 AM

In my experience using two FLI PL16803 CCD cameras for the MDW Sky Survey (www.mdwskysurvey.org) we use RBI pre-flash protocol on our darks, flats, and light exposures. it is really the only way to ensure complete removal of residual stars, particularly since we image more than a dozen fields per night per scope.

Sean I should lend you my NV device with Ha filter you can just look up and see it.

20171104 202117

  • swalker likes this

#31 BenKolt

BenKolt

    Vanguard

  • *****
  • Posts: 2,158
  • Joined: 13 Mar 2013

Posted 19 January 2018 - 12:36 PM

Thank you Jeff and John for all the references.

 

I think the first thing I ought to do is evaluate just how much if any of an RBI problem I actually have with my new FLI ML16200.  It's quite new, arriving just last month, so this time of year I'm not getting much chance at all to use it outside.  I've had only two sessions with it, where I set cooling to -30C and -40C.

 

I apologize if this has been covered before or if I will discover all this in the links, but could you suggest a good sequence for evaluating RBI?  For the foreseeable future I'll have to do it indoors, but I can achieve -30C cooling (maybe more) with this camera.  Is there a recommended number of flats I could take (with a light panel) followed by darks?  How many?  What exposure times?  Illuminate the histogram to halfway, 2/3rds? Etc.?

 

Thank you very much.

 

I also apologize for asking about the 16200 here as it's not the same camera as the OP.  I can take my conversation to a new thread if you prefer ...

 

Ben


Edited by BenKolt, 19 January 2018 - 12:41 PM.


#32 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 19 January 2018 - 02:04 PM

Ben,

The first (and easiest) thing you can do is to take some dark data.  Use the same exposure and temperature that you plan to use for your lights. You will be able to see the RBI trap pattern inherent in the sensor and you can subtract two darks (then divide by sqrt(2)) to evaluate the dark noise.  

 

To see RBI effects, I think that you'll probably have to get out under the stars.  Set your dithering distance to be a large value...say 15 pixels and shoot a field of bright stars with a long (ish) exposure (15-10 minutes will certainly work.)  Then examine your subs for ghost star images near the bright stars.  You can also just shoot the moon and then shoot a dark field.  That will very clearly demonstrate RBI effects; but, that's a pretty extreme example.

 

John



#33 gezak22

gezak22

    Gemini

  • *****
  • Posts: 3,275
  • Joined: 15 Aug 2004
  • Loc: On far side of moon. Send help.

Posted 19 January 2018 - 03:17 PM

 

I've had absolutely no issues in dealing with RBI in a 6303. It showed strong RBI, but it was so reproducible that it calibrated out just fine.

I was not aware that RBI can be removed via calibration. How did you do that? Did you use NIR preflash before calibration?

 

It's my understanding that RBI is caused by residual image (AKA ghost image) from previous sub exposure being shown on next sub exposure. So how is RBI calibrated out? If you didn't do any dithering, then RBI image may be completely be covered by the next sub exposure and calibration may work.

 

Peter

 

Maxim DL had an option for RBI preflash, and it removed the artifact that the previous image left behind on the sensor.

 

Edit: The preflash imparted another artifact (see here), and it was that artifact that was calibrated out wit normal darks and flats.


Edited by gezak22, 19 January 2018 - 03:20 PM.


#34 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 19 January 2018 - 03:26 PM

Geza,

Yes, that's what pre-flash does and that's why you have to build new dark and bias libraries to properly calibrate any image taken with RBI clearing flash.  As I understand it, that pattern is due to internal stress in the bulk silicon.  Mechanical stress in the bulk material affects the amount of charge that can accumulate in the traps so it shows up as a pattern in your darks, lights, and bias data.

John


  • swalker likes this

#35 BenKolt

BenKolt

    Vanguard

  • *****
  • Posts: 2,158
  • Joined: 13 Mar 2013

Posted 19 January 2018 - 05:11 PM

Ben,

The first (and easiest) thing you can do is to take some dark data.  Use the same exposure and temperature that you plan to use for your lights. You will be able to see the RBI trap pattern inherent in the sensor and you can subtract two darks (then divide by sqrt(2)) to evaluate the dark noise.  

 

To see RBI effects, I think that you'll probably have to get out under the stars.  Set your dithering distance to be a large value...say 15 pixels and shoot a field of bright stars with a long (ish) exposure (15-10 minutes will certainly work.)  Then examine your subs for ghost star images near the bright stars.  You can also just shoot the moon and then shoot a dark field.  That will very clearly demonstrate RBI effects; but, that's a pretty extreme example.

 

John

Thanks, John.

 

Since I won't be able to get under the stars any time soon (without getting both the camera and me drenched), would I be able to get meaningful results by taking some flats followed by darks?  I may try that anyway in addition to your multiple dark data suggestion.

 

Let's say that I "blast" with a series of flat frames and then immediately measure a series of darks over time.  Could I see RBI if it's present, what might it look like, etc.?  Any suggestions on dark exposure duration?

 

I'm going to play around with this over the weekend, but if you or anybody else has some suggestions on how best to do this with just flat - dark combinations, please let me know.

 

Thanks again!

 

Ben



#36 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,673
  • Joined: 12 Oct 2007

Posted 19 January 2018 - 05:16 PM

In my experience using two FLI PL16803 CCD cameras for the MDW Sky Survey (www.mdwskysurvey.org) we use RBI pre-flash protocol on our darks, flats, and light exposures. it is really the only way to ensure complete removal of residual stars, particularly since we image more than a dozen fields per night per scope.


That makes sense for consistent and reliable results all over the sky where sometimes you have bright stars and sometimes not. You can run as cool as you want and dither.

But at the same time for certain objects you may have incurred more noise with pre flash as opposed to running a bit warmer. So the question of which is best is still open for certain situations - but if you just want it to work and not get ghosts, pre flash makes sense.

Frank

#37 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 19 January 2018 - 05:38 PM

 

Ben,

The first (and easiest) thing you can do is to take some dark data.  Use the same exposure and temperature that you plan to use for your lights. You will be able to see the RBI trap pattern inherent in the sensor and you can subtract two darks (then divide by sqrt(2)) to evaluate the dark noise.  

 

To see RBI effects, I think that you'll probably have to get out under the stars.  Set your dithering distance to be a large value...say 15 pixels and shoot a field of bright stars with a long (ish) exposure (15-10 minutes will certainly work.)  Then examine your subs for ghost star images near the bright stars.  You can also just shoot the moon and then shoot a dark field.  That will very clearly demonstrate RBI effects; but, that's a pretty extreme example.

 

John

Thanks, John.

 

Since I won't be able to get under the stars any time soon (without getting both the camera and me drenched), would I be able to get meaningful results by taking some flats followed by darks?  I may try that anyway in addition to your multiple dark data suggestion.

 

Let's say that I "blast" with a series of flat frames and then immediately measure a series of darks over time.  Could I see RBI if it's present, what might it look like, etc.?  Any suggestions on dark exposure duration?

 

I'm going to play around with this over the weekend, but if you or anybody else has some suggestions on how best to do this with just flat - dark combinations, please let me know.

 

Thanks again!

 

Ben

 

Ben,

Yes you could do that.  If you can shadow half of the flat field, it will make it easy to see the RBI effects with a single sub.   My guess is that you'll only need to take a single flat exposed to 2/3 of the well depth to clearly see RBI effects.  I use 1200s at -40C for darks but I'm sure that you should easily see RBI with a 300s exposure.

John



#38 BenKolt

BenKolt

    Vanguard

  • *****
  • Posts: 2,158
  • Joined: 13 Mar 2013

Posted 19 January 2018 - 06:53 PM

Ben,

Yes you could do that.  If you can shadow half of the flat field, it will make it easy to see the RBI effects with a single sub.   My guess is that you'll only need to take a single flat exposed to 2/3 of the well depth to clearly see RBI effects.  I use 1200s at -40C for darks but I'm sure that you should easily see RBI with a 300s exposure.

John

 

That's a good idea.  I'll see if I can set that up this weekend.

 

Ben



#39 BenKolt

BenKolt

    Vanguard

  • *****
  • Posts: 2,158
  • Joined: 13 Mar 2013

Posted 20 January 2018 - 02:33 AM

Well, well, well!

 

I did an (indoor) RBI measurement test this evening with my FLI ML16200, and I do believe I am seeing the effects of RBI.  I was able to cool down to -35C but no lower since I was indoors.  The temperature held steady the whole time.  I took a series of flats with the intensity dialed to bring the histogram to peak around 2/3rds of the full range.  I then immediately took 3 20-min dark frames.  I called them, in sequence from start to finish, D1, D2, D3.

 

First, here are the histograms (with no stretching applied) taken from HistogramTransformation in PI and put together to show the shifting of the peak in time.  Am I correct that the shift to the left indicates RBI?  I note that the shift from D1 to D2 is noticeably larger than the one from D2 to D3.  Again, D1 was started immediately after the flats, D2 20-min after the flats, D3 40-min after the flats.  If I do something like this again I'll run another dark or two just to complete the measurement.

 

Dark Frame Histograms.jpg

 

This next figure shows the difference histograms (with stretching applied) of the dark frames.  The top one is (D1 - D2) / sqrt(2) (as John suggested above), the second one (D1 - D3) / sqrt(2), the bottom one (D2 - D3) / sqrt(2).  The red lines indicate the approximate location of the peaks of the histogram distributions.  The difference of D1 - D3, is slightly larger than D1 - D2, which makes sense because the previous showed the shift to the left in time.  The difference D2 - D3 distribution peaks nearly at 0, again indicating that D2 and D3 are closer to being the same.

 

Dark Frame Difference Histograms.jpg

 

I can see some large structure pattern noise in the D1 - D2 and D1 - D3 frames, however the pattern is not discernible in the D2 - D3 frame.  I found these hard to render and show anything interesting as JPG files here, so I'll just skip them.

 

Finally, here are some stats (normalized) from the various images.  These show the same results in numbers:

 

RBI Stat Table.jpg

 

Today I have made contact with the SGP developers, and they say that they will expand the FLI camera setting options in the software to include RBI mitigation control.  In the near future I hope to be able to test the effectiveness of these settings.

 

Let me know if there are any other results of this test or other tests that may be of interest.

 

Best Regards,

Ben


Edited by BenKolt, 20 January 2018 - 02:39 AM.


#40 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 20 January 2018 - 02:25 PM

Ben,

Yes, I believe that the shift in the peak  is due to RBI leakage.   If you compute un-normalized histograms, you should also see the peak value decrease over time.  

 

My suggestion to subtract darks to determine the noise requires that you use two identical data sets.  Subtracting two data sets at -40C with no RBI clearing and dividing by sqrt(2) will show the rms noise without RBI clearing.  Then subtracting two data sets at -40C with RBI clearing and dividing by sort(2) to get the rms noise at -40C with RBI clearing.  Subtracting two darks with differing amounts of leakage current will show the change but you cannot use the difference to compute the rms noise.

 

John



#41 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,673
  • Joined: 12 Oct 2007

Posted 20 January 2018 - 06:24 PM

RBI and pre-flash have come up many times in CN over about the last ten years - and I haven't seen a clear demonstration to show that pre-flash and max cooling is always the way to go - despite that often being the conclusion.  As an obvious example - if you have no rbi and you pre-flash - you have definitely made things worse since you increased the noise with no benefit.  And for a given exposure time and temperature - there will be some point where running warmer without pre-flash will introduce the same noise as running cooler with pre-flash - which means that at any temperature in between, you would be better running warmer and not doing pre-flash.

 

But for scientific work with zero tolerance for artifact - pre-flash is pretty much a requirement.  For aesthetic imaging, however, it would depend on the scene and your tolerance for any artifact that might appear - and your willingness to remove it with manual post-processing.

 

As for ways to measure it - here is a good summary by Richard Crisp:

 

https://www.techbrie.../articles/10873

 

Note that as far as he is concerned, you just want the noise introduced by pre-flash to be comparable to read noise - and he is fairly adamant pre-flash should always be used.  But if it is the same as read noise, you have effectively increased read noise by sqrt(2) - and there may be less total noise if you run warmer and don't use pre-flash.  So to me it depends on the exact values of the noise terms, plus how prone the scene is to artifact - and your tolerance for their impact and/or manual removal.

 

Frank



#42 swalker

swalker

    Associate Editor - Sky & Telescope

  • *****
  • Vendors
  • Posts: 1,637
  • Joined: 22 Jan 2007
  • Loc: 42.9225°N, 71.2242°W

Posted 20 January 2018 - 06:34 PM

 

In my experience using two FLI PL16803 CCD cameras for the MDW Sky Survey (www.mdwskysurvey.org) we use RBI pre-flash protocol on our darks, flats, and light exposures. it is really the only way to ensure complete removal of residual stars, particularly since we image more than a dozen fields per night per scope.


That makes sense for consistent and reliable results all over the sky where sometimes you have bright stars and sometimes not. You can run as cool as you want and dither.

But at the same time for certain objects you may have incurred more noise with pre flash as opposed to running a bit warmer. So the question of which is best is still open for certain situations - but if you just want it to work and not get ghosts, pre flash makes sense.

Frank

 

Yes, we took that into account. Part of the reason we shoot 12x20 minutes per frame and dither. Compromise is necessary in a project of this scope. The results are still extremely deep from the dark site at New Mexico Skies.


Edited by swalker, 20 January 2018 - 06:36 PM.


#43 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,673
  • Joined: 12 Oct 2007

Posted 20 January 2018 - 07:24 PM

Yes, we took that into account. Part of the reason we shoot 12x20 minutes per frame and dither. Compromise is necessary in a project of this scope. The results are still extremely deep from the dark site at New Mexico Skies.


As a concrete example - if in your case the pre-flash noise in those frames is equal to read noise, then your effective read noise is sqrt(2)r.  And that means that if you disabled pre-flash, you would have achieved the same snr with 24x10m frames - and your exposures wouldn't need to be as long, and you would have dithered more - etc.  Or you could have left it at 12x20 and achieved higher SNR - ignoring rbi artifact if any.

 

And if you are sky background limited and read noise isn't an issue, you have less need for extreme cooling in the first place and can run warmer to avoid artifact.

 

Anyway - I think the benefits of pre-flash involve a lot of factors and the best compromise for aesthetic imaging will depend on many things.  

 

Frank



#44 swalker

swalker

    Associate Editor - Sky & Telescope

  • *****
  • Vendors
  • Posts: 1,637
  • Joined: 22 Jan 2007
  • Loc: 42.9225°N, 71.2242°W

Posted 20 January 2018 - 09:17 PM

 

Yes, we took that into account. Part of the reason we shoot 12x20 minutes per frame and dither. Compromise is necessary in a project of this scope. The results are still extremely deep from the dark site at New Mexico Skies.


As a concrete example - if in your case the pre-flash noise in those frames is equal to read noise, then your effective read noise is sqrt(2)r.  And that means that if you disabled pre-flash, you would have achieved the same snr with 24x10m frames - and your exposures wouldn't need to be as long, and you would have dithered more - etc.  Or you could have left it at 12x20 and achieved higher SNR - ignoring rbi artifact if any.

 

And if you are sky background limited and read noise isn't an issue, you have less need for extreme cooling in the first place and can run warmer to avoid artifact.

 

Anyway - I think the benefits of pre-flash involve a lot of factors and the best compromise for aesthetic imaging will depend on many things.  

 

Frank

 

Without pre-flash, there is variable amp glow in the corner of all the 16803 cameras we have tested (we also tried out an SBIG STX 16803), rendering each of them inadequate for the project without use of the RBI protocol. Our exposures still needed to be 20 minutes for two reasons. The first is since we are imaging through a 3-nanometer HA filters, we have excellent skies, and 20-minute subs barely the sky background limit.  And then there's also the matter of storage space. The project is expected to generate more than 5 terabytes of data. Your suggestion doubles that amount, because all files from the camera are 32-megabytes each. As I said, there are compromises in such a large project. We've already accumulated 1.4 terabytes of data in 19 months of activity (which is more than 1,000 completed fields plus calibration frames)


  • psandelle likes this

#45 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,673
  • Joined: 12 Oct 2007

Posted 20 January 2018 - 09:34 PM

Hi Sean-

 

I have no doubt your choices are optimal for your goals - I was really mapping your situation to aesthetic imaging - which is what most of the CN discussions are about.  For scientific work - I think pre-flash is a requirement and there is no other option - if rbi has any chance of introducing artifact.

 

But for aesthetic imaging in a backyard with intermittent clouds - where you are spending many nights on a single object that has no faint stars - there is more motivation to address the noise terms optimally while keeping exposures as short as possible.  You know that people concerned about pre-flash are worried about sensor noise - because they feel a need to go very cold in the first place.

 

But your point about non-repeatable amp glow changes everything - and it says that regardless of rbi - pre-flash is needed for consistent exposures.  If that's the case then pre-flash is the way to go - period.  So for people considering whether or not to use pre-flash - if you find the amp glow does not calibrate away well - I would try pre-flash and don't worry about its associated noise.  It may be just what you gotta do.

 

Frank


  • swalker likes this

#46 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 20 January 2018 - 10:34 PM

Noise isn't the only consideration.  By eliminating amp-glow (which is also an RBI artifact,) RBI clearing makes mosaics much more manageable.

 

John


  • swalker likes this

#47 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,673
  • Joined: 12 Oct 2007

Posted 20 January 2018 - 11:33 PM

Amp glow by itself isn't a problem as long as it is consistent.  But if it is variable - then it acts as a noise term and cannot be removed by calibration.

 

So yes - anything that introduces inconsistent exposures that prevents calibration from working well is a factor.  And if it is present, then you don't need to study or measure the noise introduced by pre-flash - because you have to use it anyway.

 

Even if you didn't have bright stars that left residual images you could have this other problem.  So this adds another question to the whole decision to use pre-flash.  If your dark current is low in the first place or you are sky background limited -  and if you find amp glow calibrates away well - then you may be in a situation where you are better off not using pre-flash.  But if you see amp glow after calibration - and pre-flash results in cleaner frames after calibration - then use pre-flash.

 

What if the pre-flash itself is variable?

 

I think lot of this stuff just has to be tried so see what works best for a given camera and imaging objective.

 

Frank



#48 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 21 January 2018 - 12:57 AM

Corner amp-glow in the 16803 always appears to be variable.  As the trap charge leaks down, the amp-glow diminishes with each exposure.  You either have to toss out a few initial frames or put up with gradients in the stacked data, which will vary depending on how many of the initial frames are included.  Amp-glow in the initial frames can be reduced by taking 50-100 bias frames before the sensor is cooled.  This trick works pretty well but it still doesn't make the glow perfectly repeatable because it relies on trying to empty the charge traps while the sensor is still warm.  Properly implemented, pre-flash completely fills the traps so that amp-glow is totally eliminated.  In order to be done correctly, the pre-flash sequence has to completely fill the bulk charge traps.  I personally use a cycle of 5, 0.5 second flashes for each exposure to fill the traps.

 

Mosaic fitting is made a lot easier if you start without the gradients that amp-glow introduces along the edges and in the corner(s).  For mosaic work, I would always take the trade-off against noise with pre-flash for a better fit.

 

John


  • swalker likes this

#49 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,673
  • Joined: 12 Oct 2007

Posted 21 January 2018 - 01:28 AM

Corner amp-glow in the 16803 always appears to be variable.  As the trap charge leaks down, the amp-glow diminishes with each exposure.  You either have to toss out a few initial frames or put up with gradients in the stacked data, which will vary depending on how many of the initial frames are included.  Amp-glow in the initial frames can be reduced by taking 50-100 bias frames before the sensor is cooled.  This trick works pretty well but it still doesn't make the glow perfectly repeatable because it relies on trying to empty the charge traps while the sensor is still warm.  Properly implemented, pre-flash completely fills the traps so that amp-glow is totally eliminated.  In order to be done correctly, the pre-flash sequence has to completely fill the bulk charge traps.  I personally use a cycle of 5, 0.5 second flashes for each exposure to fill the traps.

 

Mosaic fitting is made a lot easier if you start without the gradients that amp-glow introduces along the edges and in the corner(s).  For mosaic work, I would always take the trade-off against noise with pre-flash for a better fit.

 

John

That's all fine and means the amp glow is variable and you can't calibrate unless you use pre-flash - which somehow makes the frames more consistent.  But I think that is a separate issue from RBI.  And if you have any kind of calibration problem that affects mosaics - I would say it is also a problem for a single image.

 

There are threads on problems with amp glow in cmos cameras - but I don't think it's a problem because it is consistent and calibrates away well.  (As long as the gain doesn't change for some reason).

 

So I consider this issue of amp glow variability to be completely separate from the removal of residual bulk image - which is a residual image from the previous exposure that contaminates the next exposure.

 

There are cameras with pronounced amp glow - but it is repeatable so it isn't a problem.

 

And there are ccd's that can be cooled well and don't show rbi - so you don't need to worry about pre-flash - and the camera may not even have pre-flash.

 

Then there are known cameras prone to rbi - but amp glow isn't an issue or it is repeatable.  So it is not clear if pre-flash has benefit.

 

Finally there are cameras with rbi and amp glow that is variable.  And for them you probably need to use pre-flash.

 

But I think a lot of people with cameras prone to rbi and possibly variable amp glow aren't using pre-flash.  So I guess they should all start using it right away?  Have their prior images been hurt much?

Frank



#50 jhayes_tucson

jhayes_tucson

    Fly Me to the Moon

  • *****
  • Posts: 7,388
  • Joined: 26 Aug 2012
  • Loc: Bend, OR

Posted 21 January 2018 - 02:13 AM

Amp-glow is an RBI artifact.  It's caused by NIR emission from the column amps that travels into the bulk material partially filling the bulk charge traps.  That's why it is eliminated with pre-flash--the traps are completely filled.   In this case we are talking about the 16803 and I believe that the majority of amp-glow occurs when the sensor is first powered up.  Other sensors may have amp-glow that is constant from frame to frame making it possible to calibrate out; however, in the 16803, it is variable from sub to sub because of the rate at which it leaks down during each read-cycle of the sensor.  The problem can be propagated into the stack if frames that contain amp-glow are included in the stack.  Furthermore, dithering will cause a small variation in the gradient between stacks. Corner-glow will also propagate to the opposite corners in the stacked data if some of the data was taken after a meridian flip.  None of this is a huge issue for a single exposure but it's a pain for mosaics.  These effects become pretty obvious when you work with a lot of data taken with a 16803 sensor.

 

John




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