SGP Autofocus "Gull Wing" on left of graph

[kirkster501]

Hi all, Got an infuriating issue.  I'm on the way for a perfect "V" autofocus and then the very last sample will suddenly go down wrecking the autofocus and forcing it to rerun.  Often requires three complete AF runs and sometimes abandons it.  

Any thoughts as to why this is happening please?  I've messed with my step size, number of samples, backlash etc

Sometimes it works perfectly.  When it has a wobble It is always the last sample to the left

This one just about worked.  Is there a way to set the tolerance in SGP?   I am moving to NINA shortly so I am sure SGHP is cursing me...!

Thanks, Steve

Edited October 14, 2020 by kirkster501

[tomato]

Maybe I’m being too pragmatic, but I quite often get less than a perfect focus profile from SGP and in those instances I estimate the best setting from the data and go with that. In your example that would be 3870.

I’m still trying to complete a 12 panel mosaic, and when SGP runs the autofocus three times in search of the perfect fit, something has got to give.

[symmetal]

Your curve is only showing 7 steps. I use 9 steps which is the default i believe. It's more likely to ignore an odd sample.

Also, your step size as shown is too large. You need a couple of samples down near the optimum setting for best curve fit. At your oprimum HFR of around 1.3 shown your first sample on the top right should have an HFR around 3 or 4. The curve should be more of a U shape rather than a sharp V. A step size of around 20 should be better.

Backlash setting is fine. If it wasn't the first line segment of the graph (top right) would be less steep than the second segment.

The exposure is important in getting sufficient stars bright enough to measure and not too clipped. Too far from optimum exposure will give an uneven graph. Also bin the auto exposures at least 2x2. The exposures I use at 2x2 binning are 2 secs for L, 5 secs for R, G & B, and 25 secs for NB.

The only time mine ever does a focus rerun by itself is when the initial focus setting is too far off that it doesn't go through a minimum after around 7 steps. It then does a large focuser shift and tries again.

Alan

[kirkster501]

So you feel I am wandering too far from focus when the focus routine starts?  Ok guys thanks, I will try much a smaller step size next time out.

I have indeed always used nine steps, 2x2 binning and about 4s exposures and changed it to seven steps only that night to experiment.  I often get a "focus failed, assuming focus is outwards" (or words to that effect) when it runs the second (and third) times.  It is intermittent, flip a coin if it works or not.  When it does work I can get a perfect "V" but when it fails the left hand sample (occasionally but less so the right hand sample) -  like downturned wings.

I just never feel I can leave the rig to do its thing and every autofocus needs to be supervised, as things stand currently.

Edited October 16, 2020 by kirkster501

[symmetal]

The other thing is maybe the focus friction screws to remove any slop in the drawtube are possibly too tight and the focus tube friction when it's near the top left of the curve is a bit too much for the stepper motor. This would cause the final check focus to be possibly reported as being outside the predicted optimum setting though. I don't know what autofocus system you have, but the lakeside ones have significant gearing, so can usually cope with fairly stiff focusers.

Alan

[kirkster501]

On 16/10/2020 at 15:21, symmetal said:

The other thing is maybe the focus friction screws to remove any slop in the drawtube are possibly too tight and the focus tube friction when it's near the top left of the curve is a bit too much for the stepper motor. This would cause the final check focus to be possibly reported as being outside the predicted optimum setting though. I don't know what autofocus system you have, but the lakeside ones have significant gearing, so can usually cope with fairly stiff focusers.

Alan

Thank you for your thoughts Alan.  It is indeed a Lakeside focuser on my Takahashi FSQ85.  The whole focus run in the focus plot above on the top post is 3705>4035 out of a total of 7300 complete range of the focuser.  So I am only using a fraction of the focuser's potential stepping.  Indeed, when I make the step size smaller, it'll be less than that even.   I can't think there'd be so much of a delta difference in the "stiction" of the focuser in so small a total range of movement???  However, for sure, I will investigate this if dropping the step size does not address the problem.   

Your theory does hold water though.....  I did indeed have to tighten this focuser a tad a couple of years ago (before I stared to use AF) because it slipped when pointing to the zenith even under a moderate imaging load.  Hmm, (thought to myself) should I loosen it off a tad.....?   I will try the smaller step size first and fiddle with the exposures a bit.  Maybe even a slight lubrication of the focuser as well.

It would be good if in SGP (or NiNA) you could adjust what you thought was an acceptable result in an AF run.  I notice that anything below 90% causes a rerun.  Would be good if that could be a configurable and tunable option for this sort of situation.

[symmetal]

Do you find that the focuser value for what it believes is optimum focus tends to move in one direction. In other words each time you focus with the same filter the focuser position result tends to be a little less than previous optimum focuser position values, and that this trend continues from previous sessions. This indicates that the focuser is sometimes not moving as far as it should.

The gull shaped final step indicates that for that final step, the focuser only travelled around half as much as it was commanded, either due to too much friction or possibly, but unlikely, slipping under gravity as the stepper motor wasn't powerful enough to hold it. There is no feed back from the motor to the control software so the focuser position values are just counts of the pulses sent to the stepper. It has no idea as to whether the focuser actually moved as commanded.

If the focuser didn't move as far as was thought on the last step, then when the focuser moves outwards to the supposed optimum position for the final validation step it actually ends up further out than it 'thinks' it is. The focuser thinks it is at position 3860 but is actually at the physical position corresponding to say 3880, so the validation HFR is worse than it should be, hence the poor quality value.

From the graph the optimum HFR should be around 1.3 but your validation is most likely significantly higher than that.

I find that my validation HFRs are very close to the optimum HFR, and often slightly better, and the quality is 99 to 100% each time, so haven't experienced re-runs.

With a smaller step size you will find that the fitted curve matches the lower values better as it likes to see a shallower bottom V shape rather than a sharp one.

Alan

Edited October 19, 2020 by symmetal

[AngryDonkey]

Have you tried reversing the focusing direction i.e. focus out instead of in?

[Starflyer]

As mentioned, a smaller step size should sort this out. Your V should have more of a rounded bottom than a sharp point.