Hello,

I own and regularly use for more than ten years a SteelTrack Baader with Robofocus, attached either to a C9Edge (mine) or C14Edge (of a local club). Up to last year, I used a Sbig ST10 without Ekos. My focus routines were the one of Equinox Image and/or a private software developed by a friend, with them I could achieve sharp focus (fwhm below 1.5'' were not rare at 3950mm at Montagne de Lure, France) and, more important for the following, the procedures give repeatable results. The method looks very similar to Linear 1 pass, except that the V-curve is sampled by outwards moves, the solution being reached by an inward move. The Baader has an engraved scale for visual checking ranging from 0 to 30mm, I calibrated the Robofocus so that one 30mm = 3000 ticks. Hence the CFZ is 8 ticks.

Recently I bought a new camera (Zwo2600mm) and connected it to a Stellarmate Plus with Ekos 3.6.8 running on it. I am really impressed by the ergonomy of Ekos and the quality of the underlying algorithms. I made a test session few days ago with the C9, on targets such as M36 or M35 in order to have many (up to 100) stars to focus on.

My idea was to make repeated linear 1pass routines with the luminance filter (I mean with no time interval between routines) in order to check if I get consistant results. I made it either with backlash=30 or 15 (the actual backlash is certainly lower than that) and oversan=0, or backlash=0 and overscan = 40 or 15. It is noteworthy that the backlash is only "outwards", ie is applied only after inwards move. I found no way to enter an inwards backlash, anyway overscan mimics that.

In all cases, the results were puzzling. I could achieve repeatable focus (I mean, with a spread less than the CFZ with no trend) up to seven times, but otherwise I observe a frank decreasing trend. For instance, I got 1283, 1283, 1276, 1272, 1257, 1245, 1242, 1232, 1213, 1207, 1211, 1210, 1211, 1210, 1204, 1184, 1169, 1156. A visual check on the engraved scale yields a value about 1300 (more than 1 mm shift would have been noticed), which is about the right value to comply with the EdgeHd back focus. I never experienced variations that high over a whole night, at most it was a 30 ticks shift because of temperature change. So Ekos delivers variable results, but the Baader remains more or less at the same position.
The most intriguing for me is the erratic behavior: it can be stable for few focus routines, then a decreasing trend (I have not observed an increasing trend). The V-curve always seems to be correct (R2 above 0.98) and the hfr at solution is consistent from a run to another, up to the unavoidable seeing variations, at least points on the V-curve off to 15 ticks or more from the solution are clearly worse. I tried to change the polling (Default is 1000 ms, I changed to 10 sec and 100ms), with no change. I also tried to focus on only one star, I also stopped and relaunched the Indi servers... no change.

An idea ? I attach the log.

Again, thanks for that so well designed tool.
Philippe

Hi Filliatre,

I'n not familiar with the Robofocuser but here are some thoughts:
1. With backlash you typically need to set an exact value. If you set it to 15 and the real world backlash is 10 or 20 then each time backlash comes into play you will either overcompensate or undercompensate. In each case the focuser position and counter will diverge. The effect of multiple autofocus runs would be what you observe - a trending focus point.
2. With AF Overscan if the real world backlash is greater than the AF Overscan value you'll get a similar effect to 1.
3. If something mechanical is slipping in the electronic focuser / telescope setup then again you'll get this effect.

The other thing to bear in mind is that if you move the focuser manually (i.e. other than with Autofocus) then the same comments as above apply.

The trend does appear to be more than you'd expect for temperature variation, so something else is causing this effect. You say that it appears that the focus algorithm is finding correct focus (within tolerance) which is good. I had a look in the robofocuser driver and it looks like backlash is handled by the device (rather than in the driver). So perhaps it would be good to understand more about how Robofocuser handles backlash. Do you have a manual on it?

I suppose another test would be to measure backlash and use that value and then try AF Overscan > measured value.

Thank you very much for your answer John.

With my previous focusing routines, I did not need to know the exact value of the backlash, I just have to overestimate it. Indeed, since they draw an outwards V-curve, and since the backlash correction affects only inwards movements, it acts exactly as the overscan for Linear 1-pass. Since I could make repeatable focus for years with the same hardware, I am confident on the fact that the actual backlash is lower than 15 (which corresponds to 0.15mm).

Now to answer your thoughts:

1) I fully agree, indeed before I started the procedure I was pretty sure to find a trend by setting a overestimated value of backlash. Your videos explains the point. What is puzzling is with that settings I was able to have repeatable focus (ie ekos finds the same solution) up to seven times in a row.
2) I used overscan = 40, then 15. IMO the actual backlash is lower than that, but it is worth trying in a next session.
3) I have checked for any loose screw in the Baader Crayford or in Robofocus… up to now I have not observed so huge drifts with this hardware, in more than ten years. Indeed, few years ago I made hundreds of inwards/outwards move, with the OT pointing at zenith and an heavy Sbig camera attached, the slip is about 1 tick per in/out move, no more. Of course it is worth checking again. But if the slip is greater than the CFZ, the HFR at ekos solution would be poorer that the one of the points around the minimum of the V-curve, I did not observed that.

Here is the manual ( robofocus.com/wp-content/uploads/2021/09...us-User-Manual-1.pdf . It is not very verbose about how the backlash is handled. BTW I did not find a way to select in Indi the direction of the backlash, while it is possible in the original driver.

Again, the most annoying thing is the erratic behaviour, sometimes it is stable, sometimes there is trend. A systematic error would have been easier to chase. As a result, the utility to measure the focus offsets between filters is messed up.

Thanks,
Philippe

Hi Philippe,

Thanks for the extra information and manual. Seems like the focuser backlash just adds the backlash value on In or Out as you add.

I agree its puzzling and of course being unpredictable is difficult to find. I'm not sure I can offer much help unfortunately other than to suggest retrying and see if you can somehow spot a pattern that gives a clue as to the underlying problem.

Thanks again, John.

The manual says that the backlash can be added in or on, and it was the case in the previous software I used. I did not find a way to set the backlash "in" with Ekos/Indi, it is always "out" (ie the correction is only applied after inwards moves). In itself it is strange.
Also, I did not get unpredictable behavior with previous software and the same hardware. Of course, perhaps I just failed to notice, one can never be sure... but in ten years or regular and fruitful sessions I am quite confident. That is why I think that somewhere (in the indi driver, in the linear 1pass...) I poorly chosed a parameter value. But which one ? It seems (emphasize on "seems") that the ability of linear 1pass to find the right solution is not affected. Since the quest is quite time consuming, can we at least have a shortlist of suspects?

Regards,
Philippe

Hi Philippe,

So for the backlash in / out there is no current functionality to support this in Ekos. I imagine it could be added to the indi driver and set there. I suggest you raise a change request for the Indi folks to see if they can add it.

I had a look at the log you sent which includes a print out of the parameters. Everything looks fine to me. The only "unusual" parameter you are using is Average Over = 3. Most folks leave this at 1. There's no particular reason this could be problematic but you could try setting to 1.

Other than that I can't really add anything above my comments to your initial post which would be the reasons I'm aware of that could cause the symptoms you've observed.

Thanks again John,

I forgot to say, the Robofocus is connected to the StellarMate through a serial-USB adapter. I dunno whether this can play, as I always use it this way without problems before. Indeed I wonder if the Robofocus makes all the steps that Ekos requires it to do, or if some can be "lost" randomly somewhere.

Perhaps other users of Robofocus have an idea? Or perhaps people with other focusing devices have experienced something similar?

Regards,
Philippe

I did check for tell-tale signs of the focuser not behaving itself, but I couldn't spot any in the log. Usually when this happens the actual focus will be "off" which isn't the case here - still its possible, of course.

You could try posting in the StellarMate forums - there's quite an active one on Discord. Also, if Robofocus have some groups (I don't know I'm afraid) someone might know something.