Hello! As I’ve posted previously, I’m getting a number of images that have the darkest pixels about 300 ADU below the darkeest pixels of the master dark in my SBIG ST402 camera with BVIC filters - 65000 full well. The images have darkest pixels about 800 and the master darks have darkest pixels about 1100 ADU.
In trying to solve this, I’ve exhausted most possibilities, and I am exploring camera specific electronics aging issues.
I am wondering how subtracting 400 ADU from my master darks would affect the quality of photometry. With simple subtraction, the darkest pixel in the master dark would be about 700 ADU - below the darkest pixel of the image.
If the photometry is still good, then there may not be a need to pursue expensive repair of an 18-year-old legacy camera (or to replace the entire imaging train!) at this time, and I would be able to pursue such if electronic drift/noise gets worse.
Thank you for your guidance and best regards.
Mike
Can you tell us more about these images? What are the exposure times of the darks and of the target images? How far apart in time did you take them? Did the temperature of the air or the camera change during the interval between darks and target images?
Thank you for your note. I’ve been making runs of RR CET SPV. B filter at 120 seconds. I rotate through I and V filters and then back to B. In my run last night, all were affected, though in previous runs, some I and V images were OK. There was no evidence of rapid temperature change that I could see, and the ST402 compensates fairly quickly. I made the new master darks about a week ago, making sure there were no light leaks - telescope capped and make during the night. The ST402 has an internal shutter as well to make dark frames
I use a master dark calibrated with a median combine of 25 dark frames at the same temperature and time. I use twilight flats at 1-second integrations, using a median normalized combine since ADUs vary between 5000 and 60000 - the linear range of my ST402 from previous testing.
I found this from Arne Hendon in the AAVSO forum from 08-January-2023. (There were a couple of threads dealing with negative pixels over the last few years.)
“I think the reason you are getting negative pixels is because you don’t have much sky background. Subtracting a dark frame with its noise then gives you positive and negative pixels, with a mean around zero. This is perfectly normal. Do your dark subtraction and flatfielding, and then apply an offset, in that order.”
When I try this in MPO PhotoRed Utilities, it does not appear to allow negative pixels before the offset is applied. It appears to show a wrap situation. I’ve included an image of RR CET processed through MPO Canopus before application of the offset. (The image ends in P). There are no negative pixels, and the histogram appears to show a wrap situation. It would look fairly Gaussian if the brightest pixels were brought to the lower part of the histogram.
I tried the same with AIP4WIN. It kept negative pixels, but even after the addition of an offset of 400 aftercalibration, the histogram does not appear Gaussian, so I am not sure if it is handling things well. I have attached that file.
I’ve asked Mr. Warner at MPO Canopus how the program handles negative pixels during calibration. The nice thing about MPO Canopus is that it handles batch calibration well, so I can calibrate dozens of images from a single SPV during a night’s run easily.
Are there any other programs that handle negative pixels and offset addition after calibration well? Perhaps PixInsight?
I am still curious how the photometry might be affected if the offset is subtracted from the master dark and then applied to the image, rather than applying the offset after dark and flat application.
Thank you for having a look at these images and best regards.
Mike
RR CET_LIGHT_B_120.00_-9.79_2026-01-09_0045_P.fits (767.8 KB)
RR CET Test AIP4WIN_0001_S2F.FIT (765 KB)
Mike
Just following up with this. I’d like guidance to see if a correction actually needs to be made when negative pixels are present of if I can just subtract a constant from each pixel of the master dark frame so that there are no resulting negative pixels in the calibrated image.
I may be mistaken, but it seems that a correction constant applied to each pixel of the master dark would cancel out as the Net flux is calculated, so the raw instrumental magnitude would be the same. However, I know that I may be mistaken. I'd like folks to tell me what I may be doing wrong so that I can find the correct method to calibrate my images when negative pixels are present after subtraction of the master dark frame.
For example, I imaged DY HER last night - a pulsating variable with a period of about 3.5 hours - and obtained about 300 images. Most of the images had darkest pixels darker than the darkest pixels of the master dark, resulting in negative pixels after calibration. As I reviewed past forum comments, it appears this can happen at dark sites, and, since my CCD camera is 25 years old, it may be more susceptible to this problem. (I have SBIG ST402. The offset cannot be adjusted. At 16 bits, images have up to about 62000 ADU count and its linearity goes from about 2000 ADUs to 60000 ADUs by testing)
One commentator suggested subtracting a constant from each pixel of the master dark image, calibrating, and then adding the constant back to each pixel of the calibrated image.
Last night, I had nearly 300 images of DY HER, so such a method would be very labor intensive since it has to be done by hand with each image in AIP4WIN. (I'm also trying Siril, though I've not figured out its pixel math yet. I'll look at astroPI soon as well.) And, I would hate to throw out 300 otherwise good images.
I'm looking at various other ways to handle this since MPO Canopus does not handle negative pixels. However, I'm wondering if a correction actually needs to be made other than subtracting a constant from each pixel of the master dark.
From what I can see, I'm wondering if the "true" instrumental flux calculation would be the same as the instrumental flux calculated if 400 ADUs are subtracted from pixel of the master dark in order to ensure that there are no negative pixels after calibration. If Z = Master dark value per pixel (average for demonstration purposes), then I believe the calibrated image flux (excluding flat correction for now) becomes
Net Flux F = Sum(objectFlux + skyFlux - Area*Z) - Area(average sky background/pixel - Z)
Then the Instrumental magnitude = -2.5log(Net Flux/Exposure time) + Constant, and the master dark values cancel out.
Since the calibrated object, sky, and Area all have the master dark pixel values subtracted from them, it seems to me that a constant subtracted from each pixel of the master dark would then cancel out in the Net Flux equation as well, resulting in the same flux, and, therefore, the same instrumental magnitude as a "true" calibrated image that does not have a constant subtracted from each pixel of the master dark. For example, using a correction of 400 ADUs subtracted from the master dark would yield
Net Flux F = Sum(objectFlux + skyFlux - Pixel Area(Z - 400ADU/Pixel)) - Pixel Area(average sky background/Pixel - Z - 400 ADU/Pixel)
That is my reasoning.
I would appreciate folks helping me by poking holes in my logic as appropriate (since my head hurts even with this simple exploration of image calibration equations!) so that I can find another method (as needed) to use with my images from the dark site when standard calibration results in negative pixels. Best regards.
Mike
One additional note. As noted above, Arne suggested, "Do your dark subtraction and flatfielding, and then apply an offset, in that order.”
I'm suggesting applying the offset to the master dark before subtraction since MPO Canopus cannot handle negative pixels (resulting in a wrap error) and AIP4WIN has to perform the calculation manually on each individual image, which becomes impractical with numerous images. I'm looking at other tools at present that can provide calibrated FITS images that can be used by MPO Canopus since it works very well with time series ensemble photometry with AAVSO integration
As I noted, I'm wondering if the new offset can be applied initially rather than at the end of the process. Best regards.
Mike
It is hard to tell but the first image appears to have a lot of “saturated” pixels. This is the same or similar to the problem you previously had it appears. The second images is a lot better. Probably because it has code to deal with the “hot” pixels.
Jim (DEY)
Hello! No new images. I believe the ones you referenced were the original ones in which I was trying to get a handle on strange behavior that turned out to be negative pixels
In my most recent post, I wanted some feedback about the validity of a technique I've used a couple of times - the effect of subtracting a small amount of ADUs from the master dark in order to prevent negative pixels. I am unsure of the effect that might have on magnitude error calculation.
I presume that calculated variable magnitudes would be unchanged since the same constant is introduced for any variables and comps in the image. From what I can tell, the amount of possible introduced error would be small. However, I'm out of my depth in this, so I would appreciate others' guidance. Best regards.
Mike
Hello! I think I’ve thought of a new work flow that might be appropriate, and I would appreciate input from those with more experience than me.
Would using Siril to calibrate images using the same master darks and flats that I have (as well as any updated masters in the future) but using 3 sigma hot and cold pixel cosmetic correction be appropriate?
I'd like to continue use my older CCD model (20-year-old ST402 with BVIC filters) under very dark skies so that I don't need to spend nearly $5000 to update to a newer CMOS camera with photometry filters that would likely be better suited for the dark site.
If I process the images with Siril using 3 sigma hot and cold pixel cosmetic correction, that seems to get rid of negative pixels. An additional benefit is that it gets rid of hot pixels that were not corrected with my dark frame and previous calibration program despite creating the master dark at the same temperature and integration time with nearly 80 raw darks. If folks remember, I had had problems identifying why I had a magnitude bump with ZTF J050738.71+301649.2. It turned out to be a hot pixel that came just close enough to the variable periodically to be included in the target annulus. The previous calibration program was not correcting for all the hot pixels for some reason even with a quality master dark that I had created.
I tried the Siril 3 sigma correction with the images of DY HER that I obtained last night, and the results seemed consistent with past results. Of note, the target images still had darkest pixels darker than the master dark by about 70 ADU - just enough to create negative pixels that threw off the reduction program.
So, would the 3 sigma hot and cold pixel correction introduce unacceptable errors/uncertainties into the reductions?
Thank you and best regards.
Mike
The RR Cet image you posted earlier in this thread looks like warm/hot pixels. No?