Single Frame S/N Analysis of WBPP Registration

Cygnusx1 said:
Believe me I'm always open to suggestions. As for using dark frame optimization I refer to https://www.cloudynights.com/topic/801195-dark-scaling-with-latest-cmos-imx455imx571/ (which can be as clear as mud) which has some good points. I only restarted using darks with the 1600 as my 2600 is in the shop. I am curious about the use of Offset effect on Hot pixels. I just noticed that on some randomly chosen IC405 frames Hot pixels were 2-3 times higher for narrowband than wideband. For my 2600 the gain was 100 and the Offset 50. Maybe use a different Offset for the two bands? I am rerunning my full set of IC405 data in WBPP. I had CC threshold set at 3 sigma.
As for the recurring "focus" possible issue, maybe, but it still doesn't explain why I have no problem at all doing a full processing load of LRGB frames with any of my data in WBPP. Just narrowband (hence my S/N question).
I really do have many many Tb's of data taken over the last 6 months that exhibit the narrowband registration issue, which is why I resort to the kluge of running SA outside WBPP. Focus has not been an issue (for the 1600 a definite maybe). I do get that WBPP is basically a wrapper to run the processes, I just can't nail down the SA parameter differences in and out of WBPP.
BTW all this is really good stuff! I love this kind of analysis and I appreciate all the inputs!
Click to expand...
It would be interesting to know what practical PI limits exist when working with massively oversampled images like these (i.e. 20 pixel FWHM). Simply scaling the images down by 2X (which doesn't result in any loss of useful information) allows them to align just fine with the default settings.
 
The resolution of my system with the C14 EdgeHD, a .7 reducer, and an ASI2600 (or 1600) with 3.76 micron pixels is .28"/pixels so yeah that is pretty tight. The 20 pixel FWHM is fairly large (6"). More typical is 7-10 FWHM. If I was sure all this was a focus issue that could be resolved with a better focuser (Moonlite?) I'd get one real quick. That brings me back to understanding PI performance knobs.
 
Cygnusx1 said:
The resolution of my system with the C14 EdgeHD, a .7 reducer, and an ASI2600 (or 1600) with 3.76 micron pixels is .28"/pixels so yeah that is pretty tight. The 20 pixel FWHM is fairly large (6"). More typical is 7-10 FWHM. If I was sure all this was a focus issue that could be resolved with a better focuser (Moonlite?) I'd get one real quick. That brings me back to understanding PI performance knobs.
Click to expand...
I do think the evidence points to PI simply not recognizing such large blobs as stars, given that both scaling the images down or using deconvolution to make the stars smaller result in workable images. Or maybe a serious enough focus error moves the profile too far from an ideal Moffat or Gaussian to result in classification as a star (and not a planetary nebula!)
 
Last edited:
But they’re the same ‘blobs’ in the wideband and RGB frames and WBPP has no problems at all. Even at 20 px FWHM that’s around 6 arcsec and I’m usually down a factor of two from that.
 
Cygnusx1 said:
But they’re the same ‘blobs’ in the wideband and RGB frames and WBPP has no problems at all. Even at 20 px FWHM that’s around 6 arcsec and I’m usually down a factor of two from that.
Click to expand...
I doubt they're the same. The RGB stars may have a much higher S/N, so the PSF measurement code may be more robust. Or there are simply more stars above some signal threshold used by the PSF measurement code.
 
Hence my original question about single frame S/N analysis. Also, understanding the PSF measurement model.
 
OK, answering to your post #20. I have a ASI2600MC and use this camera either with a dual band filter or without filter. My conditions are otherwise different from your's: I use a refractor (fl 530 mm, /f 5.0), gain 100, offset 50, exposure time 5 min, for all images. My experience with this equipment and settings are:
I dither each frame. I capture dark frames and don't use dark frame optimization. I need an output pedestal (100 DN) only for the filtered frames. CosmeticCorrection is used with great care in order not to damage the frames by removing faint stars, that means: a hot sigma of 13 is applied (about 300 pixels are corrected per frame). Probably you are considering the many warm pixels of the ASI2600 as hot pixels, but these are better removed by dithering and rejection in image integration.

With these settings, I never ever had a problem with star alignment.

Offset is not relevant in this connection. An offset of 50 is plenty for the ASI2600. If the raw subframes don't show clipping in the low range, there is no need to change this value.

Your larger fl and /f have to effects: less stars in the field of view and less signal. Maybe that an exposure time of 240 s is not sufficient in case of narrowband filters (I use 300 s at /f 5.0, gain 100 with the dual narrowband filter). Gain 100 is definitely appropriate for the ASi2600 plus narrowband filters.

Certainly, wrong application of CosmeticCorrection can do much damage. You should observe the number of corrected pixels per frame.

I don't claim that your problem is caused by poor focus - this is only one possibility, and you are the one who knows better if it applies or not.

And finally, another possibility is poor local seeing conditions at your site.

Bernd
 
Very good Bernd. I was just going to upload a log file of 20 hours of data processed entirely in WBPP. All narrow were 300 sec and RGB 240sec. I reduced the sigma to 2 in CC just for the heck of it and it registered about 75% of the narrows which might be a record high. No idea what caused it. Looking at the log file I found a number of narrowband FWHM's <=1px (a number I assume correlated to the number of failed registrations). I am not at all proficient in interpreting WBPP log files, but am learning.

logs - Google Drive

drive.google.com drive.google.com
 
I've been rereading "Guide to Preprocessing of Raw Data with PixInsight" and saw it recommended that Bias Frames taken for the ASI1600MM camera (Panasonic MN 34230 sensor) should be >.3 sec. Is this critical for this camera?
 
I've been able to put together a process that enables me to register all target frames in WBPP (in batches). Maybe I missed something, but I have not seen anything that uses this approach. Doesn't mean it's not out there, I just haven't seen it. What I found was I could register narrowband Ha,S2 and O3 frames as a group only (no LRGB). I checked registration of these frames on 4 targets that had previously completely failed. Worked on all of them. I just separated them from LRGB frames. I then aligned the LRGB batch with the Ha,S2 and O3 batch. In WBPP I used the Auto Mode for reference frame selection. I have attached some details of the WBPP cases and parametrics fyi. I have some thoughts why it might work, but I'm certainly not sure. Any "insights" would be welcome.
1713501718080.png
 
You must log in or register to reply here.
Back
Top