19 May 2020 - APP 1.080 will soon be released with full Fujifilm RAF support, so that will include SuperCCD & X-Trans camera's 🙂 !
2019 September: Astro Pixel Processor and iTelescope.net celebrate a new Partnership!
Analyse Stars fails
I hope you are well.
Tonight I experimented with 4x4 binning of RGB using my Atik 314E. All was going well until "Analyse Stars". I canceled it midway through the Reds. Approximately 90% failed but the other 10% reported finding only 5 or 6 stars.
I've attached a screenshot that I took of Ekos during the capture of the Reds. As you can see there are plenty of stars.
(Hmmm, the forum is telling me that the maximum file size is 30 bytes! I tried deleting some existing files I had but it made no difference. It still says 30 bytes. I'll try again tomorrow.)
Please let me know if I can help. Tomorrow night I'll try 3x3. I've had success with 2x2.
Well, that was a monumental disappointment. 3x3 binning got one step farther than 4x4 but it did fail at Registration. As you recall 4x4 failed in Analyse Stars due to an inability to find stars in 95% of the frames. Tonight, 3x3 made it past Analyse Stars but I was unpleasantly surprised that it could only find about 25 stars in a rich star field. I didn't give up so I tried increasing the exposure 50% but that yielded the same number of stars. So it seems that APP is very, very particular about what constitutes a star. Thankfully, I can do 2x2 binning for RGB because 1x1 would be out of the question.
(The forum still reports that the maximum file size of an attachment is 30 bytes. No progress there?)
Been playing with APP some more this morning. Also going back to my M27 in LRGB using 2x2. Many more stars detected with 2x2. Fewer stars with 3x3. Hardly any with 4x4.
I think what is happening is this: APP needs to distinguish between legitimate stars and hot pixels. Hot pixels are single pixels. Stars on the other hand have a width and height that spans many pixels for bright stars and maybe only one or two pixels for faint stars. APP's star detection algorithm basically says: let's ignore "stars" that are only one or two pixels wide, let's just concentrate on the brighter stars in the image. With 4x4 binning almost all stars are reduced to just one or two pixels width so as a result 95% of my frames don't have sufficient number of stars. With 3x3 binning there are more stars detected so 60% of my frames are good. Finally with 2x2 binning there are enough stars detected to proceed on to registration with 100% success.
3x3 binning is close to "working" but 4x4 is an utter failure. I think it boils down to the fact that my focal length is too short for the sensor's pixel size. The CCD Suitability Calculator says that I am slightly under-sampling. If my FL was a little longer or if my pixels were a little smaller then 3x3 would work 100% successful.
The odd thing is that Ekos' plate solving feature defaults to 3x3 binning in order to keep the star count low which speeds up the solving process. Perhaps this shortcut can't be tolerated in image processing in APP and other post-processing software.
Did you calibrate your images with a master dark to get rid of the hot pixels before proceeding to Analyse Stars?
Yes I did. Darks, Flats, Bias, it's all there.
I was able to create a composite LRGB image by salvaging what I could of the 3x3 RGB. I was hoping for 2 hours total integration time but I was only able to achieve 26 minutes. (I would attach it but the forum still says: "Max file size 30B"). My wife likes it so I guess it's good enough for me too.
Thanks for your reply.
I'll give 3x3 another try. This time I'll increase the exposure of LRGB by 50%. That should make puny stars into fatties and thus satisfy APP. The only downside is that I am running with a camera having a relatively shallow full well depth of 13ke-. It's always a struggle getting good SNR per frame without saturating stars. Would anyone like to buy me a 50-100ke- CCD? 🤣
Well no such luck. I followed my plan of increasing exposure by 50% but unfortunately I still get registration failures using 3x3 binning.
So the lesson learned is this: pay attention to the "CCD Suitability Calculator" found here:
If you discover that you are slightly under-sampling at bin1, like me, then the most you can hope for is bin2. On the other hand if you are in the "green" then give bin3 a try. Do a 1-hour test LRGB of a bright DSO using bin3 to see if APP likes it. Finally, in the extreme case, if you are over-sampling then bin2 and bin3 are great options.
So is this issue now "solved" for you or is it worth going into deeper..