V4 dithering not working

Questions, bug reports, requests for enhancements, etc.
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: V4 dithering not working

Post by iMustBcrazy »

yes, 128 is a very interesting number! ;-)

Note that there is two issues here. One is not dithering properly. Do you see any evidence of that? The other is guiding which in general is quite good but yes I've notice it go off for inexplicable reasons.

I'm going to try PHD2 tonight to see if I get the same issues.
User avatar
Peter Polakovic
Posts: 2696
Joined: Thu Apr 21, 2016 9:38 am

Re: V4 dithering not working

Post by Peter Polakovic »

Dithering looks good, at least in the log. But sometimes it just gets lost. Maybe try multistar guiding. If it is really a lost star, it may help.
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: V4 dithering not working

Post by iMustBcrazy »

Peter Polakovic wrote: Mon Aug 02, 2021 7:47 pm Dithering looks good, at least in the log. But sometimes it just gets lost. Maybe try multistar guiding. If it is really a lost star, it may help.
Yes I’m always using multi star guiding with v4. That was my main interest in v4. It’s quite resilient to clouds 🙂 as well as being more accurate.
rumen
Posts: 133
Joined: Wed Feb 22, 2017 7:31 pm

Re: V4 dithering not working

Post by rumen »

By the way I had no time to look at the log. But if you use larger dithering value than the selection radius it may easily get lost.

By the way are you using some 8bit color camera for guiding?
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: V4 dithering not working

Post by iMustBcrazy »

rumen wrote: Tue Aug 03, 2021 2:00 am By the way I had no time to look at the log. But if you use larger dithering value than the selection radius it may easily get lost.

By the way are you using some 8bit color camera for guiding?
Hmmmm... I use 1 px and detection radius is 8.

Its a 16 bit mono, ASI120mm-s
User avatar
Peter Polakovic
Posts: 2696
Joined: Thu Apr 21, 2016 9:38 am

Re: V4 dithering not working

Post by Peter Polakovic »

No, this is not the case. The log shows dithering was less than 1px and was nearly settled down when it skipped 128 pixels. Such drift can't be really detected, so it is definitely bug, but I have no idea how it happened.
rumen
Posts: 133
Joined: Wed Feb 22, 2017 7:31 pm

Re: V4 dithering not working

Post by rumen »

No it is not nasty bug... It's a wrong setup :)
This is not "multistar", this is donuts guiding. See the log:

02:52:12.280270 indigo_server: Donuts: threshold = 1466.748, max = 8816.000, edge_clipping = 0px
02:52:12.340651 indigo_server: indigo_agent_guider[capture_raw_frame:535]: Drift 0.1296px (0.07073, -0.1086)
02:52:12.340737 indigo_server: indigo_guider_reponse(): P = 1.0000, I = 0.5000, response = -0.0649, drift = 0.0649, avg_drift = 0.0000
02:52:12.340768 indigo_server: indigo_guider_reponse(): P = 1.0000, I = 0.5000, response = -0.1121, drift = 0.1121, avg_drift = 0.0000
02:52:12.340810 indigo_server: INDIGO Bus: property change request

The thing is that this is something to be expected from donuts in some cases, it is even stated in the documentation. Along with that, edge clipping is set to 0, (i think the default is 8). This makes it more prone to this error as the edge of the frame may interfere. So please set the edge clipping to something more reasonable. Another situation in which this may happen is overly populated star field, in this case you should use rather aggressive amount of edge clipping and use only half or even a quarter of the frame. As with the multi star, more doesn't necessarily mean better, there is a sweet spot for each setup and conditions. Another source of such error is too small SNR, To work somehow, the algorithm requires SNR of >10 and to work reliably the SNR should be > 20, in your log I see SNR ~15 at times, there may be even less did not look too deep. So this may be case here.

Also there is another problem that can lead to this with Donuts. There should not be any bright stars on the edge comming in and out of the frame this may lead to such problems. This is why in Ain I show the clipping border (have no idea how it is visualized in AI).

So the bottom line: I see 2 problems which hit the limitations of this algorithm:
1. Border clipping setup - increase the border clipping 0 is not ok.
2. Bad SNR, this can be fixed by the edge clipping too, just use a smaller sub frame.

And if this does not work use another algorithm.
Attachments
Screenshot from 2021-08-03 10-46-36.png
Screenshot from 2021-08-03 10-46-36.png (882.61 KiB) Viewed 137 times
User avatar
Peter Polakovic
Posts: 2696
Joined: Thu Apr 21, 2016 9:38 am

Re: V4 dithering not working

Post by Peter Polakovic »

Yes, Rumen is right. I missed it in the log, but you used DONUTS, not SELECTION (e.g. multistar) mode

04:12:05.403948 indigo_server: INDIGO Bus: property definition
04:12:05.403974 indigo_server: 'Guider Agent'.'AGENT_GUIDER_DETECTION_MODE' SWITCH rw Ok 2.0 0 (null) { // OneOfMany
04:12:05.403997 indigo_server: 'DONUTS' = On // Donuts
04:12:05.404021 indigo_server: 'SELECTION' = Off // Selection
04:12:05.404044 indigo_server: 'CENTROID' = Off // Centroid
04:12:05.404067 indigo_server: }

You have to select these two buttons in toolbar to use SELECTION with MULTISTAR:

Screenshot 2021-08-03 at 12.55.05.jpg
Screenshot 2021-08-03 at 12.55.05.jpg (42.75 KiB) Viewed 134 times
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: V4 dithering not working

Post by iMustBcrazy »

Oh my mistake Peter. I thought when said "multistar" that you meant DONUTS. I didn't know there was a multistar method.

The edge clipping is 0 but that is the default in AG. I wouldn't have known what to do with that so I didn't change it. I'll change it to 8.

I'll show you the results from last night in the next reply...
iMustBcrazy
Posts: 338
Joined: Sun Jun 02, 2019 3:52 pm

Re: V4 dithering not working

Post by iMustBcrazy »

I started last nights session with PHD2 - I love their graphs. I wish you guys had better star profile/HFR/FWHM graphs. I can hardly see the ones in AI.

Note that I did NOT polar align. I worked from my balcony that faces south, so polar alignment was approximate. Also, I used ST-4 guiding - very reliable!

Here's the calibration, guiding assistant and guiding graph from phdlogview - looks pretty good - calibration and dithering worked fine. I was not able to get my imaging camera to work at same time as guiding camera so I could not actually get a stack to see the dithering result.

I would guess my PEC is about +/-25" as I seem to have captures only half of a PEC cycle with the assistant.
Screen Shot 2021-08-03 at 8.09.09 AM.jpeg
Screen Shot 2021-08-03 at 8.09.09 AM.jpeg (225.15 KiB) Viewed 132 times
Screen Shot 2021-08-03 at 8.09.16 AM.jpeg
Screen Shot 2021-08-03 at 8.09.16 AM.jpeg (301.25 KiB) Viewed 132 times
Screen Shot 2021-08-03 at 8.09.20 AM.jpeg
Screen Shot 2021-08-03 at 8.09.20 AM.jpeg (354.92 KiB) Viewed 132 times
Stats from guiding assistant:
Screen Shot 2021-08-02 at 10.50.54 PM.jpg
Screen Shot 2021-08-02 at 10.50.54 PM.jpg (482.4 KiB) Viewed 132 times
I'll put AG results in the next reply...
Post Reply