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

Not RBI on CMOS but this looks similar

  • Please log in to reply
62 replies to this topic

#1 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 08:11 PM

Here is an interesting test that I encourage all imx294 owners to try at home.

I setup the QHY294 on a lens then pointed the lens at the blinds in my home office. This provided a nice contrasty reference image. 

The curious thing that I observed is that if I left the lens uncovered, then capped it and took a 4 second shot right away, the 4second dark showed a ghost image of the blinds. The subsequent darks were always OK with no residual image.

Being familiar with RBI on my CCD sensors I tried the same trick at different temperatures from ambient +/- 22c to -20c. The temperature didn't seem to affect it. It always happened, even at 22c.

 

Here are the reference shot and the ghost image. What is interesting is that all I had to do to create a ghost image was to uncover the lens without taking a frame.

 

What tipped me off on this was the fact that when doing plate solving, the first image after a slew showed faint trails of bright stars. I though it was the mount settling time so I redid the experiment waiting a couple of seconds for the Mach1 to settle but that didn't change the trails. 

 

Anyone want to try this?
I don't think that in practice this really impacts the images (maybe flat frames?) but it's interesting nonetheless.

Attached Thumbnails

  • RBI light.jpg
  • RBI.jpg

  • james7ca and numchuck like this

#2 OldManSky

OldManSky

    Mercury-Atlas

  • *****
  • Posts: 2,799
  • Joined: 03 Jan 2019
  • Loc: Valley Center, CA USA

Posted 28 May 2019 - 08:20 PM

That is interesting.

Incomplete clearing of the wells before a new exposure (seemingly only affecting nearly-saturated pixels)?

 

That would also seem to explain the star trails...


  • TieDyeAstronomer likes this

#3 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 08:23 PM

Right, that would be like RBI, except I didn't take an exposure prior to the dark. All I had to do to create the ghost image was uncover the lens.



#4 OldManSky

OldManSky

    Mercury-Atlas

  • *****
  • Posts: 2,799
  • Joined: 03 Jan 2019
  • Loc: Valley Center, CA USA

Posted 28 May 2019 - 08:34 PM

Right, that would be like RBI, except I didn't take an exposure prior to the dark. All I had to do to create the ghost image was uncover the lens.

Yeah, but...with an "electronic shutter" the light is always hitting the pixels and creating electrons (holes) in the wells.  Whether you're taking an exposure or not.

What's supposed to happen when you start an exposure is the wells get cleared, and start accumulating during the exposure time from zero.

Except...not zero.

I'm going to have to try this on my 183MM, although I haven't seen the star streaks you mentioned during platesolves...



#5 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 08:36 PM

No need to try with the 183. I just did and it's fine. It doesn't do it.

I thought the wells only collected electrons when energized, no?


  • OldManSky likes this

#6 jdupton

jdupton

    Vanguard

  • *****
  • Posts: 2,214
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 28 May 2019 - 08:44 PM

Andy,

 

   That is a very interesting observation. I will have to find a way to try it with my camera. (I don't have a lens adapter so would need to use a scope.)

 

   You may have stumbled onto something I suspected but had no way to reliably demonstrate. I found in my testing of my camera, that doing one frame after another sometimes gave me inconsistent results. I also found and subsequently described here in another thread on the IMX294 that if I inserted "dummy" short (0 or 1 second) exposures between all the frames I wanted to keep, the inconsistencies appeared to go away. 

 

   I had described to another user that I viewed it as something akin to an RBI-like effect but was shot down as not being possible. (As you say, not RBI but very RBI-like.) You have real data that will help me duplicate my experiment in a more controlled manner. While I don't think it is really RBI as is found with deeply cooled CCDs, I think it may be a real effect and have something to do with a form of auto-dark-current compensation happening on the chip.

 

   Thanks for posting this very interesting finding!

 

 

John


Edited by jdupton, 28 May 2019 - 08:47 PM.

  • numchuck likes this

#7 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,085
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 May 2019 - 08:44 PM

Hmm, that is rather strange. I wonder if other people's 294s do that... The sensor reset that occurs before a new exposure should have cleared that. With a BSI sensor, RBI should be impossible since charge is read and cleared through the entire substrate, so there are no real charge traps that could accumulate charge between exposures. For an FSI CMOS sensor...I guess there is nothing to prevent potential charge traps in the substrate... I am not sure what the 294 is...FSI or BSI.

 

I would also be curious to see if the IMX183 does it. I've not seen that on my ASI183 before, but I haven't really done a test quite like that. 



#8 DmitriNet

DmitriNet

    Vostok 1

  • -----
  • Posts: 144
  • Joined: 13 May 2017

Posted 28 May 2019 - 08:47 PM

That's is CMOS-specific RBI.



#9 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 08:47 PM

John, I remember your observations with respect to inserting dummy frames. That's the first thing I thought of when I saw this phenomenon. 

 

Jon, my QHY183c does not show this behavior. The dark frames are always perfect. The 294 is a BSI sensor. 


  • jdupton likes this

#10 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,085
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 May 2019 - 09:00 PM

John, I remember your observations with respect to inserting dummy frames. That's the first thing I thought of when I saw this phenomenon. 

 

Jon, my QHY183c does not show this behavior. The dark frames are always perfect. The 294 is a BSI sensor. 

Hmm, is the 294 global shutter? I wonder if it is, maybe some charge is getting stuck during readout...

 

EDIT:

 

It's a Pregious, so it is global shutter. I wonder if that has something to do with it. IIRC the IMX183 is an Exmor R, which is a different kind of pixel architecture. 


Edited by Jon Rista, 28 May 2019 - 09:03 PM.

  • OldManSky likes this

#11 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 09:22 PM

Do you think that this may contribute to the miscalibration issues that some people are reporting.

#12 jdupton

jdupton

    Vanguard

  • *****
  • Posts: 2,214
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 28 May 2019 - 09:31 PM

Andy,

 

   It might contribute to a very small miscalibration error from Dark Frames due to a slightly higher than normal ADU (the residual plus the normal dark current). I think it may be much more of an issue with Flat Frame calibration. With short exposures, the extra counts in both the Flat Frame and the Flat-Dark can throw off the calibration of a Flat Frame to a greater extent.

 

   It might be hard to do a controlled experiment to determine the extent of any effects. Bad Light Frame calibration from a "distorted" Master Flat may be tough to quantify. In any case, it may be wise to be aware of the effect when shooting Flats and Flat-Darks.

 

 

John



#13 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 09:38 PM

Yes I was referring specifically to flats. I wonder if I should take flats interspersed with zero duration darks and with the flat panel turned off during those exposures. Darks aren’t an issue since there is no light hitting the sensor and whatever charges are left, they are flushed out after the first dark. It would be prudent to discard the first dark frame I guess.

#14 jdupton

jdupton

    Vanguard

  • *****
  • Posts: 2,214
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 28 May 2019 - 10:01 PM

Andy,

 

   That sounds like a reasonable approach. 

 

   I am not sure that you need to turn the flat light source off during the short dummy exposures. You could try it both ways to see if it makes a significant difference. I'd the interested in your results. For Flat-Dark Frames, I suspect that the exposures are short enough already that dummy exposures between each doesn't make much of a difference assuming you take all the Flat-Dark Frames in one session. It certainly wouldn't hurt to use dummy frames in between, though.

 

   One note on Library Dark Frames for calibration of target exposures -- that is exactly where I first noticed the inconsistencies in my testing. I was shooting Dark Frames of 0 to 960 seconds duration to characterize my camera and had mixed up the order of exposures. I found that most of the short duration frames had a slightly higher mean ADU when they followed a long duration dark frame than if they followed another short duration frame.

 

   That is when I happened onto the idea of inserting dummy short exposures between all longer frames (regardless of type). The short frames seemed to help "flush the small residual values" following long (>30 seconds) exposures. I started inserting the dummy frames for all sequences (lights, flats, and darks) and the inconsistencies seemed to go away.

 

   Based on that, I would recommend even putting the dummy short frames between long dark frames on the hunch that dark current accumulation from a prior exposure may affect the next frame in a manner similar to the light-generated electrons in the wells as you observed. One or more short frames seems to flush the system giving better consistency between Light Frames or long exposure Dark Frames. At least that is what I thought was happening based on my experiments.

 

 

John



#15 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 10:14 PM

So i don’t think that there are trapped charges that aren’t completely flushed out. This would not explain the star trails in the first exposure after a slew. Right? I think those are fresh electrons collected after the last exposure and before the next one. My hunch is that unless you turn off the light source the camera will keep accumulating photons (electrons) between frames that then are added to he next frame.

Edited by andysea, 28 May 2019 - 10:15 PM.


#16 jdupton

jdupton

    Vanguard

  • *****
  • Posts: 2,214
  • Joined: 21 Nov 2010
  • Loc: Central Texas, USA

Posted 28 May 2019 - 10:20 PM

Andy,

 

   You make a good point about the star trails. Your experiment should be able to show any differences if you run it both turning the light off between frames versus leaving it on and inserting short frames between. I should be able to duplicate that experiment in the day or two so we can compare notes on the results.

 

 

John


  • andysea likes this

#17 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,085
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 May 2019 - 10:55 PM

Yes I was referring specifically to flats. I wonder if I should take flats interspersed with zero duration darks and with the flat panel turned off during those exposures. Darks aren’t an issue since there is no light hitting the sensor and whatever charges are left, they are flushed out after the first dark. It would be prudent to discard the first dark frame I guess.

If a short exposure between flats clears the residual, then I would say that sounds like an effective remedy. I would contact QHY or ZWO though, and see if they can look into it. These guys usually get the sensors with the companion FPGA, if one is an option...it may be that they can fiddle with the FPGA to correct the issue, maybe do a double-reset automatically on readout or something. Worth a try, at least, to make the camera easier to use, as this is a bit of a tedious problem...



#18 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,085
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 May 2019 - 11:00 PM

So i don’t think that there are trapped charges that aren’t completely flushed out. This would not explain the star trails in the first exposure after a slew. Right? I think those are fresh electrons collected after the last exposure and before the next one. My hunch is that unless you turn off the light source the camera will keep accumulating photons (electrons) between frames that then are added to he next frame.

An electronic shutter camera is in effect always exposing. A reset is issued at the START of a new exposure to clear out the wells before the exposure starts. With an electronic CMOS shutter with per-column readout this reset usually goes very quickly, tiny fraction of a second. Once an exposure is completed then a rolling readout/reset is performed where each row is read, and after being read is reset (although this exact mechanism may depend on the sensor design, some sensors may do a full sensor reset after readout, and of course a global shutter is different as well). 

 

Once readout is done...well, the sensor is "exposed", uncovered and literally open to incoming light...there is no shutter to block light. And the photodiodes are sensitive to light...so, they pick up photons, convert em to electrons...all the time. If you do a slew, and the aperture is uncovered, then the sensor is going to be exposing while you slew. The trails in the stars that occur while you expose during a slew are going to occur no matter what. Then, once you finally center your object after the slew and kick off a new exposure, THAT exposure will reset the sensor on its initiation. I the reset does not fully clear the wells of charge, as seems to be the case with the ASI294...then yes, you would see remnants of the brighter star trails that occurred during your slew. Because technically speaking...the slew WAS your last exposure. wink.gif


Edited by Jon Rista, 28 May 2019 - 11:04 PM.

  • andysea likes this

#19 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 11:08 PM

Perfect thank you Jon! That explains it. I started a ticket with QHY to make them aware of this issue. I don't know if the ZWO variant has the same behavior. Unfortunately I no longer have mine.



#20 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 24,085
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 28 May 2019 - 11:12 PM

Thinking about this more...a lot of global shutter designs use CCDs (charge coupling, really) for the "backing memory" during readout. So, the main pixel well would be charge-coupled to another well to shift the "live" charge out of the current exposure, obscured behind the main pixel (or perhaps in a masked area). We know that CCDs can bloom. I wonder...since the IMX294 is a Pregius and thus global shutter...if perhaps bright pixels are effectively "blooming" into a backing memory via the charge coupling. I also then wonder if during a reset...are only the main pixel wells reset? Could that leave a ghost image in the frame memory where charge bloomed (leaked in) from the main wells during the previous exposure? That might explain why another readout clears the ghost...


Edited by Jon Rista, 28 May 2019 - 11:14 PM.


#21 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 28 May 2019 - 11:29 PM

That would be a logical explanation. It seems strange that only the main wells are reset. Unless a full reset would not be irrelevant for the main application that this sensor was designed for.



#22 HH_ASTRO

HH_ASTRO

    Sputnik

  • -----
  • Posts: 49
  • Joined: 17 Oct 2017

Posted 29 May 2019 - 10:42 AM

Looks like an insufficient power supply, I would check your power adapter.

#23 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 29 May 2019 - 12:53 PM

I use a 10a 13.8v pyramid supply.

#24 andysea

andysea

    Soyuz

  • *****
  • topic starter
  • Posts: 3,665
  • Joined: 03 Sep 2010
  • Loc: Seattle, WA

Posted 29 May 2019 - 03:05 PM

So in order to rule out USB power issues I redid the test using startech powered USB hub and no external power supply, cooler was off of course.

Here is the result.

 

Attached Thumbnails

  • RBI light2.jpg
  • RBI2.jpg


#25 freestar8n

freestar8n

    Vendor - MetaGuide

  • *****
  • Vendors
  • Posts: 9,304
  • Joined: 12 Oct 2007

Posted 29 May 2019 - 06:08 PM

In my look at documents on the 294 I don't see indication it is global shutter - but it sure has a lot of stuff going on in those pixels.  Can someone point to a reference that it is global shutter?

 

It's definitely geared toward high speed video - including a very specialized HDR mode.  In that sense the main application modes may not care about this effect - even though it would matter for deep sky.

 

And does it really have each color in a 2x2 pattern?  I didn't realize that.

 

I think this is a good discovery by Andy and it may need special handling - just like RBI.  At least if any artifacts show up.  It's good if there is no temperature dependence.

 

Frank




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