Dark edge banding in Bias frames which disappeared over time

Discussion in 'STX and STXL Series Cameras' started by Niall MacNeill, Sep 18, 2021.

  1. Niall MacNeill

    Niall MacNeill Standard User

    Joined:
    Jun 23, 2017
    Messages:
    95
    I have an SBIG 16803.
    I calibrate my images with Bias and Dark frames as one would expect. The Darks and Bias frames and the Masters produced through integration, have always shown a dark left edge, as have my Lights. I have read some posts on this subject and seen this referred to as the DC recovery edge effect. What exactly is that? In any case it calibrates out......at least it used to calibrate out. Here are single Bias (0.1 sec, -20C) and Dark (600 sec, -20C) that I took in September 2020. Single Dark -20c_600sec_None_004_20200906_062214.jpg
    Single Bias_0-1 secs_0.1sec_None_003_20200729_225101.jpg
    This comparison of an uncalibrated image with a Bias/ Dark subtracted image, shows how effectively this dark edge calibrated out for images I took in late 2020 early 2021.

    Luminance image NGC 253_comparison uncalibrated image with Bias & Dark calibrated image.JPG

    I have always had a small amount of amp glow in the top left corner, but in recent months I noticed that a glow has appeared along the left hand edge of the images and is now quite pronounced. I thought this was worsening amp glow. It was getting so bad, that in the end, a month or so ago, I commissioned RBI pre-flash, which I knew to be a solution to amp glow. However, as I am sure you know, RBI pre-flash does generate more dark current noise which I have had to deal with. I went back to my Darks, recalling the dark left edge and I wondered it that had anything to do with it.
    I compared a recent uncalibrated Light of NGC 7293 and discovered that the light had no such dark edge and that the calibrated image had the glow.
    Luminance image NGC 7293_comparison uncalibrated image with Bias & Dark calibrated image.JPG

    It became clear that the Bias/ Dark subtraction, where I had the dark banding was under-subtracting the left edge relative to the rest of the image, resulting in the light band. This after all was not amp glow.

    I wondered where the dark banding was coming from so I subtracted the MasterBias from the MasterDark. The "calibrated" Dark showed no dark banding, so it is clear it was in the Bias.


    MASTERDARK_600sec_-20C_c.jpg

    I went back and reviewed my deep sky data from the time I created the calibration images in September 2020 and it is clear the dark edge in the images has been diminishing with time. Therefore the dark edge in the MasterBias has resulted in progressively incorrect calibration and an ever increasing bright band on the left side of the images.

    I should have renewed my Bias and Dark calibrations before a year had elapsed. I realised I had been slack with that, but I had presumed there would be some minor changes and miscalibration that would result. I am, however, staggered that the Bias frames could have changed so much, such that the DC edge effect has progressively disappeared.

    This then is a comparison between the MasterBias from 2020 with the one I have just completed.

    You can see the difference is massive. Can you give me to understand what has happened here and more particularly your recommendation on the frequency to redo my calibration frames. Is there anything else I need to know about this phenomenon?

    Comparison between Bias September 2020 and Septemebr 2021.JPG

    This is a DropBox link to all the original files you may need if you wish to check out my uncalibrated and calibrated images. The folders I have set up within I believe are self-explanatory.

    https://www.dropbox.com/sh/qjbcthvd1izbvkv/AADYlAdmJe4nstdVnnpJQGgFa?dl=0

    Thanks & regards,
    Niall MacNeill
    Wattle Flat,
    NSW,
    Australia
     
  2. Colin Haig

    Colin Haig Staff Member

    Joined:
    Oct 27, 2014
    Messages:
    7,379
    Location:
    Earth
    Hi Niall,
    The left edge effect is a function of the STX cameras's DC restoration circuit. The CCD chip receives various voltages to bias the substrate of the chip and to operate the amplifiers and logic.
    There is a small amount of time between when the voltage is changed and when it recovers to the normal operating voltage for imaging; @Doug can explain this better.
    That time difference between voltages results in a slight difference at that left edge.
    If you were to look at different manufacturer's implementations, you'd see small differences.

    As a standard practice, it's a good idea to update your bias and darks libraries periodically - perhaps 2-4 times per year.
    The CCD is an analog device, and camera behaviour will vary over time. If it didn't we'd just send you a "lifetime dark" and "lifetime bias" frame when we ship it ;-)

    We'll have a look at your files and let you know if there is anything of concern. It looks to be a few ADU, but the FITS will tell us.
     
  3. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    Please note that I can't read xisf format images, so I couldn't open your 2021 images. Also processed images, even stacked, changes things in unknown ways, so they are not useful in diagnosing camera issues. Please just supply a single raw bias frame in FITS format for each year.

    I agree with Colin about updating your calibration frames periodically. It will also help mitigate any changes in hot pixels due to accumulated cosmic ray damage.

    Technical stuff follows:

    There are all kinds of tradeoffs in the design of these cameras. The video comes out of the sensor with a high positive offset voltage - closer to the amplifier supply voltage 15V than ground. This has to be removed, obviously, before we can convert it to a number. This is called "DC Restoration." Also there's a big transition in the voltage of the output that happens during the vertical clocking, which affects DC restoration.

    The DC is blocked using a capacitor circuit, plus a resistor to pull the output towards ground. This ends up forming a time constant. If you make the time constant too long or too short you will see "tails" to the right of hot pixels and bright stars. If you make it "just right" the double-correlated sampler will remove any residual drift, but depending on the sensor, clock rate, and other factors you may get a "shape" to the bias on the left side of the sensor. That shape is generally very stable, but might vary over long periods of time. Thus the recommendation to reshoot your calibration frames periodically.

    If the shape changes quickly/suddenly there may be an issue with the camera; perhaps one of the internal power supplies went out of tolerance.
     
  4. Niall MacNeill

    Niall MacNeill Standard User

    Joined:
    Jun 23, 2017
    Messages:
    95
    Thanks Colin & Doug.
    Doug, I put two single FITS files into the 2021 folder. The Masters are in .XISF format as they were generated out of PixInsight, so I can only give you a single sub. There are FITS format images in all the other folders.
    I wouldn't say the change has been sudden, in that it has been happening progressively for a year, so hopefully I don't have an issue with the internal power supplies. If you look through the FITS format images I sent of the various objects I've been imaging, you can see that the dark edge in the uncalibrated images has been decreasing i.e. getting less dark. Of course, I didn't do new Bias frames in between, so I can only go by what I see in the Lights and there, the left edge of the calibrated images gets commensurately brighter.
    I just calibrated some recent Lights, with new Darks and Bias and aside from a bit of amp glow in the corner, they look perfect. I put the JPEG comparison in the folder. Here you see at left an uncalibrated image, in the middle calibrated (Bias and Darks only) with the recent 2021 data and at right the same image calibrated with the 2020 data. NGC 7293 comparison uncalibrated vs 2020 calibrated vs 2021 calibrated.JPG

    I don't think the darks have changed that much, but the Bias clearly has. Of course, redoing the Bias frames is very quick at less than a half an hour's time investment. The darks however, as you know take much longer. I generally like to do 35 exposures for the Darks, so for my 30 min images, is takes nearly 18 hours.
    I am still surprised at the extent of the change in the Bias frames, which suggests that the DC restoration process you described has changed such that the ' "shape" to the bias on the left side of the sensor' has been altered. Is that a cause for concern? As noted it still calibrates out perfectly.
    I am thinking that I will redo my Bias frames every couple of months and redo the Darks every 6 months.
    Thanks for shedding light on this issue.
    Niall
     
    Last edited: Sep 20, 2021
  5. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    Thanks for the individual FITS images. When investigating potential camera issues I need completely raw data. Any processing will have unknown effects. Even the slightest difference in brightness/contrast adjustment on a JPEG makes direct comparison utterly impossible.

    Looks like the shift was about 2 ADU from 2019 to 2020. It was considerably greater in 2021.

    I was looking at the firmware history on the STX. The only change potentially affecting clocking was back in 2016, when support was added for the 16801 sensor. It should not have impacted the 16803 in any way, but I can't rule that out entirely as I haven't looked at the changes in detail. But I should still ask whether you did a firmware update between the 2020 and 2021 images.

    There are no obvious defects in your images. There is no evidence of a change in trailing charge between 2019 and 2021, so it looks like the sensor is still clocking out normally. There's nothing obviously wrong with the camera.

    At this point I'd suggest just keeping an eye on it. Right now it seems to be working well, with appropriate calibration. If the change becomes more dramatic or more rapid then it may be time to send it in for a check.
     
  6. Niall MacNeill

    Niall MacNeill Standard User

    Joined:
    Jun 23, 2017
    Messages:
    95
    Thanks Doug, that seems sensible.

    I did commission a new Intel NUC to run the observatory in July. As I recall I downloaded the latest firmware for the camera, although I can't say whether it differed to that on my ASUS Laptop, which I was using previously.

    I will keep an eye on it. Hopefully, I don't need to send the camera to you as being located in Australia that would put me out of action for weeks I would imagine.

    Thanks for your response on this.

    Niall
     

Share This Page