First Light with new AC4040FS1

Attached is a photo of my first light images with my new AC4040FS1,which I took last night. I included a single 5 minute HA sub and a quick integration of 3 hours of the HA data. I used the High Stack Pro readout mode, and I hope the darker banding comes though so you can see what I'm dealing with. I'm hoping someone can give me some feedback on what direction to look for solutions. The image was taken with my CDK14, inside my dome, and there was no moon present at the time.

 

@Merrel Miller

Hi Merrel.

The Diffraction staff would need to see the raw images to make a diagnosis, the attached snapshots are not sufficient to make a proper diagnosis.

Can you reply and attach a couple of raw unprocessed images straight from the camera in .fits format, if the images are too large and exceed the forum size limits then upload to a free hosting service, Dropbox or GoogleDrive etc and post a share link.

Note that Diffraction staff won’t look at files that are in PixInsight .xisf format, only .fits format.

FWIW, we run an old, non-commercial, CCD camera that exhibits the same quadrant structure. On that camera it’s due to the way that the sensor was designed with four independent read out amplifiers that read out the four quadrants of the sensor individually and small differences between the amplifiers shows as shadowing where the quadrants meet. The effect is most noticeable when the sensor is not cooled so I should ask did you have cooling active?

On our old camera, calibrating with flat frames correctly removes the quadrant structure, are you using flats in the calibration process?

When calibrating CMOS cameras in PixInsight you will achieve a better calibration result if you do not use bias frames, instead use darks, flats and matching flat-darks (dark frames that exactly match the flats exposure time) and when combining a stack of low noise CMOS subs in PixInsight you should experiment with noise evaluation enabled and disabled.
The noise evaluation algorithm in PixInsight does not compute accurate noise weighting values when read noise is so low in these new CMOS sensors, especially with narrow band data, which leads to over or under compensated flats calibration results.

The best place to discuss processing in PixInsight though is on their own forum where users with much more experience can help.

With no personal experience of these newer CMOS cameras I can’t think of anything else to suggest ATM.

HTH

William.

 

Thanks William. I can send the raw fits files if needed, but will wait until I'm requested to do so as they are ~32MB. I did have the camera cooled to -15 degree C but probably can go lower. I suspected that my master flat was not as good as it should have been. I did create and use master bias frames and sound like I'll need to learn to more about using dark flats. I read some of the posts from Doug and he indicated this type of sensor was quite different for my STT8300. Lot's to learn

I appreciate you taking the time to respond. Merrel

 

Merrel,

HighStackPro. Great idea but I experienced it it locking in CMOS chip patterns that I can't get rid of in dark subtraction. I spent months struggling with this last year and I solved it and my sanity by exclusively imaging in High mode. The AC4040 CMOS chip is very sensitive and I have yet to need any subs longer than 300s. Bright objects are good in HSP because of the gain in dynamic range, but nebula and galaxies do just fine in High mode at full gain, we well as narrowband. And if I have a very bright object (I recently imaged NGC 40 the Bowtie Nebula) I mix long and short exposures in Pixinsight to recover the dynamic range.

I have found the following works. Others may have developed different procedures, but this is my best to date with a CDK20, AC4040, AO-X.

1) Run the cooler at a temperature that keeps it below 50% power. I see the cooler circuit feeding back noise to the chip. I find -5C acceptable and -10 preferable and can achieve this except in the Phoenix summer when I need a little cooling fluid assist. The Chip in High mode gets you into a very low noise range at -5C and -10G is great. Anything else looks to be overkill. The key for me is less than 50% power to the cooler.
2) Take darks at the end of a run every night if possible or at least in similar ambient conditions to the light subs. I take 15-20 darks every night at the end of the run and match them to the lights taken that night. It's a pain, but otherwise I see CMOS chip patterns creeping in if the darks and the lights were taken at significantly different cooler power. If timing doesn't work to take darks every night try to take darks in same average ambient conditions as the lights so the cooler power is close to that used when taking the lights. I avoid dark libraries with the AC4040.
3) Ghosting of bright stars is a real issue for the AC4040. Throw out early darks that may have ghost images. I find it takes about 20-30 minutes for ghosts and amp glow to subside in the darks if I take them immediately after finishing an imaging session. Dithering and stacking has eliminated most ghost artifacts in my light images, but you need to really watch things when you have bright features constantly illuminating the chip. I did an experiment to take short darks a bunch of times after a light exposure with bright stars. The ghosts were not eliminated by a number of reads of the chip...it was a time factor after the shutter was closed.

You should be able to get good calibrations with darks alone that eliminate the CMOS patterns. Doug is correct that flats can also reintroduce the chip pattern to the image is the flat darks are not well matched. I take sky-flats and find that as long as the flat is taken to an ADU range of 1800-2200, then a few tenths of a second difference in the flat-darks does not produce any pattern that carried over into a calibrated image. For reference I use ACP for the sky-flats and their algorithm targets constant ADUs by slightly increasing the exposure as the sky dims. I calibrate with a flat-dark that is normally within 0.1-0.3 sec of the flat exposures for 6 sky flat subs.

I hope this give you some things to try to get things going quickly.

Greg

 

Greg, Thanks for the reply. I came to the same conclusion about High mode. While I haven't had a lot of time to image due to travel and weather since I purchased the AC4040, I have produced some good images. I literally was going to write you as I've been impressed with your images on Astrobin. Your procedures seem sound and I'll give them a try. If you are willing, I'd like to write you from time to time via email (mine is merrel@reagan.com) for some coaching on techniques that will improve my imaging. Again, thanks for your time and response. Merrel

 

Merrel,

I sent you an email so we can have direct communication. Thanks for the kind words on my Astrobin pictures. I am still hit or miss on making really good images ftom the AC4040 but each time I process, I learn a little more about my technique as well as the system.

 

Hi Greg and Merrel, I've been reading your posts, and hope to interact as well as I have issues. I'm just getting the AC4040 running, so I have no experience imaging with it yet. I think I'll avoid HDR and the post-acquisition combination process for now, and either use High mode or HighStackPro. Your comments about a dark library and cooler power % are interesting. I'll experiment with it as well. By the way, I'm not super clear about when one would use the Low readout mode? Have you experimented with that? The description says "The low-gain output provides the full dynamic range of the sensor, but loses the low-end signal." I'm not sure what is meant by the low-end signal reference.

Best regards,
Mark

 

Thanks Colin, I wasn't sure exactly what that meant. So High readout mode gets the faint detail but lacks some dynamic range. HighStackPro gets the faint details and gets the dynamic range by internal stacking of sufficient subexposures to end up with a 16-bit image (if the settings permit 16 subs), and doing a lot of subexposures just in High also gets dynamic range after post acquisition stacking of the 12-bit images or by mixing short and long as Greg noted.
I'll experiment with HighStackPro and High and the associated darks to see what I get, including whether I can avoid the chip patterns Greg mentioned in HighStackPro.
Best,
Mark

 

Thanks Doug. One last question that is somewhat off topic here, but related to calibrating the various images. I understand that if I use HighStackPro, I'll need dark frames in that readout mode as well. For flat calibration of HighStackPro images, do the flat frames need to be read out in HighStackPro as well? I ask because the flat frame exposures will many times if not always be shorter than would permit 16 subexposures to be made. That may not matter, and best practice may be to just do HighStackPro flats too (and flat darks using HSP). Thanks,
Mark

 

That is puzzling if taken literally. Maybe I'm taking it too literally. But I cannot possibly take flat darks (1-15 sec exposure) and light darks (150-900 sec exposure) in the same readout mode and subexposure settings if the minimum subexposure is 1 sec.

Isn't this correct?: The readout modes and subexposures really only have to match for flat darks and flat lights, and separately the readout modes and subexposures must match between for light darks and light (science) images. That would be tractable, though exposure scaling would help since I take a lot of different exposures in one night. Thanks for guidance.

 

Eric, you are correct.

Flat-darks must match the readout mode and exposure time of flat-lights. As long as the acquisition mode of the light and dark have to match.

Darks must match the readout mode and exposure time of lights. Flats are a calibrated image if taken correctly and therefore dimensionless when used to calibrate lights.

That said, I'm finding that flats can carry over the chip pattern to the final calibrated light. I'm doing some experiments to see if I can reduce this effect. A user of an FLI camera with a GSENSE4040 chip (same or different from the AC4040, I don't know but I suspect very similar in characteristics) was advised to take flats at a much lower ADU count that is used for CCD cameras. 10-20% of full well instead of 50% full well for CCDs. I have a suspicion for why this is true but until I confirm, I'll leave that as an unproven theory.

BTW, I also find that darks should be taken at similar stable cooler power settings as the lights were taken. I typically run the cooler so only 30-50% power are required. If I push the cooler "hard" I find residual chip patterns in my images, even with dithering.

For reference I am exclusively imaging so I really get worked up when I see residual patterns in my calibrated stacks. While faint, these compete with image details and limit what I can get out of a data set. I have uploaded a camera screenshot of the latest issues I'm wrestling with. Taken with flats acquired to 1800 ADU so I hope that the new lead I have will correct this. The vertical line in the image is after 20 pixel dithering for 65 subs of 300s, calibrated and stacked. the variation in the left and right of the line is 10 ADU. Given the OIII halo around M97 is only 30 ADU above background, this is a significant portion of detail that I'll lose if I can't calibrate out that residual chip pattern line.

Hope this helps,

Greg

 

If true, this makes CMOS a non-starter for me. I will absolutely not spend half of each night taking darks. My photometric programs take many different exposures in a single night (each target has a computed exposure time in each filter). CCDs performed this fine for years, but CMOSs apparently can't keep up. Great.

Whatever CMOS's advantages, so far the Great CCD-to-CMOS migration seems to have served mostly to increase sales of ibuprofen.

 

Hi Eric, I'm about to start to use the AC4040 with my photometric work (time series photometry for exoplanet follow-up for TESS). In the past the photometry I did for that type of work with CCDs usually involved some pre-planning of exposure time for the targets planned for the observing session (30 sec to a few minutes normally), and I did a set of darks at the same exposure duration before the run to create a master dark frame for processing real-time (rather than scaling longer darks). We were using AstroImageJ (and I will be doing that again) which has a data flow that presents and plots during the imaging session (in part so we can abort if we start to see an eclipsing binary and avoid wasting a night on it). What durations are you normally using? Probably a different type of photometry so there may be an issue for you. I am going to do some experimentation to see whether I can in fact use a scaled dark for this type of work. I'll let you know if I do and have results worth sharing.
Best,
Mark

 

Eric, I don’t actually spend half the night doing darks and I have built a dark library for each readout mode and acquisition time I use, for flats and lights. And unless I’m trying an experiment, I rarely take darks at the end of a run. I will say that when I was acquiring darks each night, it led me to the observations that there is an issue with Residual Bulk Image with this chip, and knowing that, it leads to different acquisition strategies.

I won’t try to convince you of CMOS advantage since I have taken a lot of “Ibuprofen” in last year. But a key advantage is likely chip availability in the future. I hope we do learn to use this sensor because the noise characteristics (except the fixed patterns) are outstanding.

Greg