Auto Focuser not Determining Correct Value

OS: Windows 10 X64

Auto Focuser not determining correct value

Auto focuser seems to be detecting the wrong value, with the log link below, it recommended a value of 7477 whereas 7503 gives a lower FWHM value and better eccentricity according to PixInsight Sub Frame Selector

I have Smart Focus disabled, but enabling it makes no difference

Run Auto Sequence

Heyo, You have either:

  • Mistakenly downloaded and overwrote your install of 3.1 with 3.0
  • Mistakenly captured with an older version of SGPro or
  • Attached the wrong logs

Ah, Dropbox wasn’t showing me all the files, so I gave the wrong one, here’s the correct one

That is the auto focus log, please attach the main log.

SGPro is functioning as designed here. That said, we may consider increasing the minimum fit quality… It’s not terrible, but it is at 84% and may have improved if recentered and run again. The current behavior allows anything better than 80% to proceed.

@jmacon Thoughts?

Of note, is that this run did trigger a warning that the HFR of the final position did not jive with where it should have been. Currently we only emit a warning, but there might be a lever to pull here… like:

AF is good if (Fit Quality >= 80% AND No HFR warning)

Several things going on here:

  1. Of your 4 successful focus runs, only this first one gave a value of 7503. The other three gave values close to the 7477:
    run #1: 7503 84% SII
    run #2: 7473 86% SII
    run #3: 7468 91% SII
    run #4: 7477 96% RED

  2. clearly you don’t want to be focusing with NB filters such as SII. Your RED filter has given a much better value. You had requested RED be used but there is a bug there. When that is fixed and you do your focusing with RED, you will get much better results. Run #1 gave an inferior result because it was a poor run with poor quality, exactly what we would expect. A rerun should have occurred but the default rerun threshold is set at 80% in SGP. Almost everyone should set it to at least 90% and a large majority of users should set it to 95%. Which is where I have it.

  3. you will likely get even better results focusing with L and/or need a lower exposure time.

  4. I think your backlash is set too low. All of your focus runs look this this one with the far right data point, which is the first image taken, showing the classic lower HFR value than would be expected because the backlash was not fully taken up, resulting in better focus at that focuser position. I suggest you double your backlash setting and see if that data point does not come in much higher. Where it is, unusually low, makes for a poorer quadratic fit. The routine eliminates one bad data point, but this one is not bad enough to get eliminated, thus giving you a poorer fit, and skewing the best fit a little to the right of what it would give if that data point was in optimum position, and making the Quality measure worse.

@Ken a better default quality threshold may be 90%, which is where we started. My best guess after looking at 3000+ quadratic focus plots from about 20 different users, is that 90% of our users will have most of their focus runs giving them a Quality better than 98%. So I have mine set at 95%. It is working perfectly. Those folks who need something less than 90% either have challenging equipment, or challenging targets such as globular clusters at very long focal lengths. @STAstro should have his set at least to 90%, and maybe higher using the L filter.

Just for anyone reading along, this was just a misunderstanding, no defect present here.

OK. I am going to set the default to 90% and then have folks that need it lower the min fit (instead of the other way around).

This certainly seems the case. I have had mine set to use L on both scopes, and SGP has always used L.

I just checked in the Sesto Senso software and the backlash is set to 0, so double of that is 0 :smiley:

From the 4 Runs, yes the first one gave a value of 7503 and the results in FWHM and Eccentricity were what I expected, it was subsequent runs where it didn’t quite work, and telling SGPro to go back to 7503 fixed that, so not sure if backlash is the issue, and I am not sure how to apply backlash to the focuser options as it is something I;ve never done before, so any guidance would be much appreciated.

I actually focus on Red at 2x2 Binning to get to focus, and I have un-selected the Filters per filter so that Red is always the chosen one

How to you mean it didn’t quite work? Are those runs that are not part of the log you posted? The runs you posted in the log all gave exactly the results we would expect. Is there a reason you are focusing with R?

What I meant there is that it reported a value that wasn’t correct for imaging with, it should have been back at 7503, as I manually set it back to 7503 to avoid wasting images.

Yes, I do not use a Luminance filter (As I never image in Luminance), so the next filter in line is Red, with Narrowband the SNR is a lot lower so since I Know all my filters are Par Focal it’s much better to Focus in Red at 2x2 Binning with a 5s Exposure right?

You are basing this on the sub frame selector in PI. It clearly does not agree with the HFR values produced by SGP, which for these 4 runs, are very much in agreement that something around 7473 is your best focus, at this particular time and temperature that you performed these 4 focus runs. We of course, are forced to use what SGP gives us since PI is not integrated into SGP.

However, on how many nights and with how many different focus runs have you compared the FWHM values from PI with HFR values produced by SGP? How do you know which one is better to use? It is also a good idea to get all your focus parameters refined and working well before giving early indication much weight.

As for focusing with R, since your filters are par-focal, that probably will work well for you. But I might just mention that if you have not done an extensive all night session just focusing on all filters, or something comparable over many nights, then comparing the temperature vs position plots to determine filter offsets, you may be surprised at the result. I imaged for a couple of years with my Astrodon filters and thought they were par-focal. After doing this process, I discovered they were not. In fact the NB filters differed a great deal from the RGB filters. I rarely image the L on my AG12, but I always focus with L because it gives a much stronger signal than the R and therefore produces much better focus curves and better focus for all the filters.

I have looked visually at the images, for example last night, some of the images that were rejected by pixinsight based on a focuser value of 7473 were actually less in focus than frames at 7503 (For a 5 minute gap)

I have done many “Full Nights” imaging, but this has only recently started to happen to me, so somtehing has changed and it’s not the hardware as I am running the same rig I have been for the past few months.

A good idea would be for you to image 5 or 6 nights where all your focus runs are with the R filter so they produce good focus curves, then post a link to all the main SGP LOG files so I can analyze a decent size set of them. It’s hard to conclude anything definitive with only one good focus run to look at.

where can i set the threshold? for focus % ?

Details here:

Beta 361 Running Great - FOCUS SUGGESTIONS

Same thing tonight, first focus routine recommended 7556 which was spot on, PixInsight did not reject a single frame, then 100 minutes later an auto focus routine ran and it recommended 7503 which PixInsight did not like, so I manually changed it back to 7556 and PixInsight is happy again

FWHM of first focus according to PixInsight is Circa 4.5
FWHM of second focus according to PixInsight is 5.468
FWHM of manually changing it back to 7556 is back to Circa 4.5

So maybe there is some backlash, again the warnings were there

Warning = HFR validation frame warning

I think I may have found the problem, and it may not be software related, I need to confirm :wink: