Bias Time Dependence

Discussion in 'Legacy Models - Community Support' started by JoshuaFrechem, Oct 23, 2018.

  1. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    I have been using a calibration library from around June and started noticing I was getting bad correction in my stacks. So, I started making a new library and my bias frame structure has changed quite a bit. Attached are both 200 frame stacks using the same pixel rejection and stacking at the same temperature. The calculated read noise is about the same around 11e-. But, my darks have a measurable amount more noise using 100 subs per master dark. Is this normal or could it be indicative of something bad? I know it changes, but it seems to be worse than I would expect over a 6 month period.

    Thanks,
    Josh


    New Bias - Top
    Old Bias - Bottom
    newbias.JPG old bias.JPG
     
  2. Colin Haig

    Colin Haig Staff Member

    Joined:
    Oct 27, 2014
    Messages:
    7,378
    Location:
    Earth
    I'm just another customer, and thought I might help.
    More information would be needed - like which camera this is and what version of software, drivers and firmware produced the images.
    Can you provide the FITS images (or at least the FITS header) for both examples?
    Temperature, aging, power supply, how long the camera was running (internal temp), cooler temperature and power, are factors that may have an impact.
    If you updated firmware, that could also have an impact.
    It's sometimes helpful to think about "what has changed".
     
    JoshuaFrechem likes this.
  3. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    Hi Colin,

    STT-8300 at -20C. The first set was taken inside, I believe, at 22C while the second set was taken outside at around 13C. I have considered that but I don't see how, with CDS, that readout circuitry at such a low ambient temp and a 10C difference in temperature would change the bias structure so drastically.

    Power supply is a Mean Well 12V 6.67A supply from Mouser that was used for both data sets. Cooler power was probably around 50% for the first set and 25% for the second set. In the summer, my cooler runs a max of 55% and I keep the same cooling temp all year round for consistent calibration. The same firmware was on the camera for both. The second data set was with 6.18 while the first set was with CCDOps. Both processed in Pixinsight using the same rejection and stacking parameters. I can upload FITs later today.

    The biggest reason I am worried is because I have had MDL crash multiple times where I had to unplug the power to the camera with it in an operational state. I have been told it should be fine, but it still worries me. I expect it to change over time, but it seems more drastic than it should be, but I suppose one of the most important things is whether or not it corrects the data properly.

    Now that I own MDL, I can keep the software a constant from now on and I doubt many more firmware updates will be released for the STT. So, that likely won't change much either. But you are right, I definitely like the scientific approach you propose. Fortunately (maybe?) there was a large gap in time between the summer and now where I wasn't able to image much due to school. So the current run of calibration frames should be good for my recent data. All of these variables are why I really dislike not having a ROR anymore.
     
  4. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    Yes we would need FITS files to make any determination.

    I wouldn't worry about killing power to the camera with the cooler operating. With a delta-T of under 60 degrees, and given the significant thermal inertia of the CCD stack, the sensor will not experience thermal shock.
     
  5. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    Sorry for the delay. Cloudy weather and lots of work made me lose track of astro stuff. Here are the two Bias frames. I am trying to track down banding in my images, so I was going through my calibration frames and it reminded me to upload these. Let me know what you think!

    I can't find how to attach FITs here, if possible. So, they are at these links:
    https://drive.google.com/open?id=1-8xRMagJ4miIKAIvmq54udZesZBgQzsi
    https://drive.google.com/open?id=1Q9SohLQQq86l2JHjK_AG2XoSyFqhKJkv

    Thanks,
    Josh
     
  6. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    There's something going on there. Best guess is it's not flushing properly/completely. There's a chance it's a hardware problem, but it's most likely a configuration (EEPROM) or firmware problem. It's definitely not a sensor problem.

    First thing to do - download the latest SBIGDriverChecker (get the full package, don't just download driver updates - we've improved the driver checker itself). You can get that from:

    ftp://ftp.sbig.com/pub/SetupDriverChecker64.exe

    Use it to update your SBIG drivers. Then use it to check whether the camera firmware is up-to-date, and if not then update it.

    Next get the CCDOPS software and use it to take a new bias frame. I'd like to get a FITS uploaded.

    If that looks clean then we can try MaxIm DL again. If not then we might want to remote in to have a look at your camera's configuration settings and make sure they're okay.
     
  7. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    I already have CCDOPS installed and will give that a go tonight and get back to you. I will also get the new driver checker. I believe I have it installed already, but I will try again (It might be 6 months old if the new one was released more recently than that). I recently tried to update my drivers and only an SBIG driver was out of data. I tried to force it to reinstall the Firmware but it just said that it didn't need to be updated.

    Thanks,
    Josh
     
  8. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    I did another stack of 200 at -20C in CCDOPS in "High Resolution" mode. The frames downloaded much faster than in Maxim DL. Is this normal? I couldn't find a setting like in Maxim DL that controls the readout rate. The bias looks pretty different. Also, this bias has several hot pixels. Those hot pixels look smeared to the right like there is a CTE issue.


    https://drive.google.com/open?id=1FrpZ0w6ZGzAPxVB-50qWOzkXy_CKaXLA
     
  9. Colin Haig

    Colin Haig Staff Member

    Joined:
    Oct 27, 2014
    Messages:
    7,378
    Location:
    Earth
    I`m just another customer, and had a fast look at this, to see if I could help.
    All of the downloadable files were NOT produced wth MaxIm or CCDOps, they were processed with PixInsight
    You said you used CCDOPS - but didn`t for the XLA file -PixInsight again.
    In all cases, you averaged 200 frames, and the resulting data values are around 0.016 to 0.017, and a hot pixel shows as 0.020.
    However, somehow you are using PixInsight wrong, and the pixel values are so small that you must have normalized it too, making it useless to assess what is going on in your camera.

    Try again, but follow Doug`s suggestions.
    Do 1 bias frame, in CCDOPs, 1 time, no processing, be specific about what you did, and don`t withhold information.
    Good luck.
     
  10. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    Colin is right. We need unprocessed raw data from the camera.
     
  11. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    My apologies, I thought you wanted master frames since the first two were masters. But, as a sidenote, pixinsight scales the data between 0-1 by default and for the dynamic range of this and pretty much any camera this linear scaling is just changing the scales. So, I'm not sure what you mean by using pixinsight incorrectly. Pixinsight loads an integer image properly and converts it to a float only when doing some processing to it. The weird values for the master has to do with dealing with floating point numbers when processing. Since the first two bias frames were masters, I assumed you wanted another master. I didn't see anywhere that said a single frame.

    I'll upload 2 bias frames from 4 different nights. 2 each from the first two nights, 2 from ccdops, and 2 from last night in Maxim.
     
  12. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    The problem with processed data is I can't tell what the camera is doing and what you are doing. No matter what software you use.
     
  13. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    I understand you needing raw data. It just didn't click that is what you were asking for because I posted masters and was just thinking of masters because the change in structure is more *visually* apparent but not necessarily useful here. Would you like many bias frames from each night or just 1-2?
     
  14. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    If the change in bias is visible in one frame, then one is enough. If not then please post half a dozen. Please also post a "before" frame.
     
  15. Colin Haig

    Colin Haig Staff Member

    Joined:
    Oct 27, 2014
    Messages:
    7,378
    Location:
    Earth
    Joshua - my apologies to you, I had forgotten PixInsight default behaviour. Sorry,
    Colin
     
  16. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    Not a problem, Colin. I would rather be told if I am doing something incorrectly. You've been very helpful throughout my threads and others.

    I am working on getting several batches uploaded. I deleted my CCDOPS group so I am retaking them. The images appear almost instantly in CCDOPS vs about 2 seconds in Maxim DL. Is this because CCDOPS communicates more directly with the camera than Maxim DL or might that suggest I have something enabled that is slowing down the communication?

    The few things I like about CCDOPS is ambient temp reading and more information on the camera. Are these available in Maxim DL somewhere or is it considered superfluous information? Also, is there a low gain mode for the STT8300? The manual shows 0.38 and 2.3 throughout.

    The change is not very noticeable with only a few frames. I added 20 bias frames for each. You are more experienced, so you might see it in only a few frames. I added more just in case. I calibrate my images with darks and bias frames so I get masters of 200 so I feel confident I am not injecting a bunch of noise.
     
  17. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    Here are 3 sets of data. Unfortunately, in the last 6 months, while reorganizing data several times, all of my old calibration frames became corrupt and only the masters that were on a different HDD remain. With 20 images I can barely notice a difference between the MDL stacks but there is a noticeable difference between MDL stacks and CCDOPS. From my calculations, both give about a 10.5e- read noise, but the pattern difference gets me.

    I have been having strange horizontal bands in my OIII data (primarily, but also in other NB stacks. I haven't done LRGB in a while). But, it is there in my uncalibrated stacks too. It is only on the side of my OAG prism, so I am wondering if it is somehow related to that? But that is what started this whole bias frame issue for me. One of my old cameras had a CTE issue, so I sometimes fear the worse since I live in an area where condensation is often an issue.

    https://drive.google.com/open?id=1blfreAxk2Zk2LksrEpRo97IVscExzjDU
     
  18. Colin Haig

    Colin Haig Staff Member

    Joined:
    Oct 27, 2014
    Messages:
    7,378
    Location:
    Earth
    Doug - you'll have to play with this - I don't see any significant issues that look to be software related. I have a few comments below

    The CCDOps31st and MaxIm 30th frames are very similar. The MaxIm 20th are a bit higher across the board, but still seem pretty reasonable.

    Comparing individual bias frames, they have a range from about 957-1227.
    However, there are two (MaxIm 30, -0268, -0288) that are a bit different, and the temperature is a tiny bit different as well.
    Am wondering if the cooler current changed and this is giving the variation in the bias frames.

    Joshua - is the Mean Well supply a GST90A ?
    e.g.
    https://www.meanwell.com/webapp/product/search.aspx?prod=GST90A

    The specs on that supply aren't great - the +12V output can vary +/- 5% (which means 11.4V at its lowest).
    Am thinking you might have a borderline low-voltage situation.
    What wire gauge did you use?
    Why don't you use the original power supply that came with the camera?

    One minor thing - you probably already know - the screen stretch impacts what the images look like. Try using RANGE instead of Manual in MaxIm's Screen Stretch window.
     
  19. JoshuaFrechem

    JoshuaFrechem Cyanogen Customer

    Joined:
    Oct 1, 2018
    Messages:
    41
    Location:
    Virginia
    I chose the Mean Well because it stated 5% total regulation and 1% ripple just like the ICCInventus power supply shown on the SBIG site. I measured higher than 1% peak to peak ripple under no load for 400us, but the no load voltage regulation was well within the 5%. I'll try to see if I can monitor the supply voltage with a load but an 80W load might not be easy for me to test and I don't want to probe around my camera with it on. I haven't witnessed any yellow or red lights except for once and that was with a 5A supply, which is why I went to 6.67A

    I have tried using range and with even 20 subs it is hard to see. But with 100+, a structure becomes very apparent and the ccdops has horizontal banding like I might expect of a bias. The maxim dl bias has a shelf about half way up as can be shown in the master at the top. I stacked them in Maxim dl as well and saw the same result to verify it was pixinsight settings.

    I no longer think it is a sensor issue, which Doug also said, because it changes with the software. It's present in uncalibeated images. That makes me wonder why the bias isn't correcting it properly? Bias does appear to help the issue, but it doesn't fully correct it.

    Another note is that I can't figure out how to do a 0 exposure in ccdops. It defaults to .12s and I noticed the range is a few hundred higher than one of the maxim sets. I would expect the slightly longer exposure might be responsible.

    How much is the temperature varying? I watch my setup image most of the night and the most I've ever seen it vary is +/-.1C.

    The original power supply died and at the time, I didn't have the money to pay for the one from sbig. Sox I opted for one from a reputable supplier with the same specs. The power supply has 16gauge coming off and I have a 2.5ft extension made of 18gauge. From my calculation the voltage drop is 0.2V for the extension and .12V for the primary cable. I remember seeing Doug say they ship with 13.8V supplies now? I might go ahead and do that if that is recommended that way I know it isn't an issue.
     
    Last edited: Nov 1, 2018
  20. Doug

    Doug Staff Member

    Joined:
    Sep 25, 2014
    Messages:
    9,932
    CCDOPS appears to limit bias frames to the minimum light frame exposure time. MaxIm DL does not do that.

    I can see a fractional ADU change in the bias level about 3/5 of the way down the image. It amounts to 0.02 ADU. That's not much.

    Subtracting the two sets of bias frames almost eliminates the artifact, so clearly subtracting is working.

    You may see a very small shift in overall bias level with ambient temperature (not sensor temperature); some cameras exhibit this effect for complex reasons.

    I would recommend acquiring bias frames for every observing session, and see if that helps.

    If you see this phenomenon on processed light frames... I'm not sure the bias correction is responsible. The shift is much smaller than other noise sources. We may need to get a full set of calibration and light frame images to investigate further.
     

Share This Page