Autofocus problems

[old_eyes]

I have recently replaced my Equinox 80 ED with an Esprit 120. Both used a Lakeside focuser.

With the Equinox I was able to successfully autofocus in both NINA and Voyager. However, with the Esprit I am struggling. With Voyager I am unable to complete the V-Curve calibration. I'm also unable to use the Localfield mode that looks at the full frame. NINA uses a full frame autofocus routine, and that is working either.

The focus moves smoothly under manual computer control, and I can manually bring the scope to a successful focus using computer control. Once in focus, the scope holds that focus across the sky and right up to the zenith, and I can move away from focus and back again without seeing any backlash or slippage. So the focuser and the Lakeside controller seem to be working properly.

I have adjusted the step size to see if that makes any difference, but with no success.

I am sufficiently far north that we have lost astronomical dark completely, and I would not normally be imaging at this time of the year except that I'm trying to set up and test new equipment. I wondered whether the lighter background is reducing the contrast of the images and confusing the algorithms. Certainly, the error bars on the star size measurements seem to be much higher than I've experienced before.

That two completely separate pieces of software are failing to autofocus, previous successes, and the fact that I can focus 'manually' by directly controlling the stepper motor, suggest there is something about the data that is causing the autofocus algorithms to fail.

I would be grateful for any ideas I can try now, or should I just wait for the return of astronomical dark?

[old_eyes]

I have carried out a few more trials.

The V-Curve generated starting at the in-focus position looks like this:

It alsmost looks like the focuser is at its limit beyond about 9400, but the stepper range is from 0 - 1250. Yes the focuser is racked a long way out, but that is normal on the Esprit 120. I did two runs with backlash set to 50 or 100 steps and they look the same. There is no significant slipping of the focuser as I can move away from the focus point and back to the same position and hte same HFD.

Does anyone recognise this problem?

[wimvb]

Can you reduce the focuser range? Atm, best focus is at abt 9180. If you reduce the range to +- 200 steps from this value, you should be fine.

[old_eyes]

That makes sense, but I am not sure which adjustments I need to make in Voyager’s autofocus routine. More delving into the settings I think.

thanks for the suggestion Wim

[Starflyer]

The section on the right is typical of backlash. Set some backlash compensation, say 300 steps and test again.

[symmetal]

The individual focus points appear to be around 16 stepper motor steps apart. You only need around 9 or so focus points centred on the optimum focus position to dertermine optimum focus. The two lines through the V shape would then follow the V graph slopes and intersect at the correct focus position. This is as SGP used to do autofocus, although it now uses curve fitting to the V graph instead. You just needed to specify the stepper distance for each focus step and the total number of focus steps. In my case it was 25 stepper movements for each focus step and 9 focus steps in total. That meaned it moved the stepper out by 4 x 25 steps initially and then took 9 focus values moving in by 25 steps each time. You have to be in the ball park of correct focus before starting the routine or it may be unable to find focus.

From your graph you appear to have selected 16 stepper movements for each focus step and around 60 focus steps in total. This is causing the focuser to move out around 500 stepper steps when you initiate the focus routine which is way too high. I would make your focus setup similar to mine which also uses a lakeside focuser, 25 stepper movements per autofocus step and 9 autofocus steps in total. I assume Voyager has similar setup parameters to achieve this.

As Starflyer says horizontal focus steps at the right of the curve generally indicates backlash correction not being set high enough in the focuser setup, but that is indicating an awful lot of backlash which is unlikely. It's possibly the focus is so far out at that point that the focus routine can't determine a proper HFR value and just keeps stepping in until it starts responding to changing HFR.

Alan

[wimvb]

25 minutes ago, symmetal said:

It's possibly the focus is so far out at that point that the focus routine can't determine a proper HFR value and just keeps stepping in until it starts responding to changing HFR.

That would be my guess.

These focus routines are only reliable near best focus. Especially when you have a reflector (but not only then), the donut shape of out of focus stars throws off the algorithm that determines hfr. When the point spread function of the out of focus star gets larger, the algorithm can "lock on" a hot spot instead and return a ridiculously small hfr. That's why it is necessary to restrict the focus range as Alan wrote. Excellent write up, btw, Alan.

[old_eyes]

Thanks for the suggestions. I think that the v-curve routine stepping too far away from the focus point might indeed be the problem. I can check that out with manual focusing. What does the image actually look like at different focuser positions?

I am trying to make this work with Voyager, because in general I love the software. It really works for me.

I don't know whether it is possible to limit the range of the autofocus routine in the way you can with SGP.  It worked fine for me before with the Equinox 80, but the change in scope has thrown some curveballs.

Looking at how the HFD reacts to manual focusing, it looks like there may be some slip in the focuser. I understand it is not uncommon for Esprits to need a bit of adjustment, so my next step is to make sure there is sufficient tension and definitely no slippage before further attempts to get it to work.

[wimvb]

2 hours ago, old_eyes said:

What does the image actually look like at different focuser positions?

Stars will become fuzzier when out of focus. For a reflector they will become donuts, which autofocus routines generally have difficulties with. Depending on the speed of the scope, the V curve will be more or less steep. I would think that any autofocus routine has ways to adjust the focus range, because there is such a wide variety of  focusers, focus motors and scopes.

[old_eyes]

I have a refractor so the stars tend to become bigger and fainter until they disappear into the background. I think I will manually step through the ficus region and note what is actually happening to star shapes. The v-curve algorithm in Voyager uses a single star and jumps about all over the place generating the v-curve. It is really hard to follow by eye what the star shapes are doing. A manual walk through may help me diagnose problems.

[wimvb]

If you set the camera to continuous streaming, and slowly change the focus, you can see what is happening.

[wimvb]

@old_eyes: I assume you have seen this?

 

[old_eyes]

Thanks Wim. Yes I have. Followed all teh instructions but still ahving a problem. I think making sure everything is mechanically secure is the next check.

[old_eyes]

OK, so I think the problem is solved, and it has nothing to do with the software, just my own stupidity.

It was backlash, but not in the focuser itself.

There were comments on various forums that Esprit 120 focusers sometimes arrived from the manufacturer a little slack. Possibly due to vibration in transport loosening the adjustment screws. So I wanted to check that out.

The stepper motor drive shaft is connected to the focuser drive shaft with a coupling sleeve locked into place with two grub screws. I released the coupling at the stepper motor end so that I could move the focuser manually. It seemed to be tight but smooth. I thought it would be a good idea to check the focuser tension screws, and to do this I had to remove the stepper motor mounting plate to get at the screws. When I did so, the coupling sleeve slid off the focuser shaft without undoing the screw. It was not locked down!

Either I had never tightened it down, or I had not tightened it enough and it had worked loose. This meant that when the stepper motor changed direction the motor moved without the focuser shaft moving until the protruding end of the locking screw re-engaged with the flat on the focuser shaft.

Massive backlash issues!

I reassembled the stepper motor with the sleeve locking screws thoroughly tightened, and checked the ability to move the focuser accurately backwards and forwards using a vernier caliper. It all seemed good, and the next clear night I was able to get a successful initial v-curve in two separate trials. I then used robofocus in Voyager successfully in a short imaging run.

So, as is usual, operator error is the underlying cause of the problem.

We learn something each time we do something stupid, but it is a very slow process!

Thanks for all the suggestions and my apologies for worrying about what Voyager was doing, instead of the mistakes I had made.

[wimvb]

It's always a relief when things get a simple explanation and can easily be resolved.

Good luck with your gear.

[symmetal]

I've had the same grubscrew work loose on my lakeside focuser on my FLT98, the obvious sign being the first V curve step being near horizontal. Also it shows as significant free movement if you operate the dual speed focuser on the opposite end of the shaft to the lakeside fitting. On the Voyager setup it takes HFD 40 as its upper V curve limit which is what your first V curve was showing at the horizontal section, which may have caused it to be horizontal for so long, rather than just backlash which wouldn't have helped.

Rather than just manually setting a V-curve step size which you can then adjust as required to get the most useful V shape, Voyager sets the step size to a value that gives a HFD change of 1 which is why it needs an initial setup trial to determine it.

I assume it doesn't step so far out now on its initial focus move, and takes less than your indicated 60 focus steps to determine correct focus. 🙂

Glad it's all working now though. 😀

Alan

[tomato]

Glad you sorted it out. My NINA focus routines on my Esprit 150s are currently not as robust  as they normally are, but then I’m trying to use them in a twilight sky which I’m sure isn’t helping.

Incidentally I recently discovered NINA has a number of curve fitting functions for the focus routine, other than the simple V profile.

[old_eyes]

2 hours ago, symmetal said:

I've had the same grubscrew work loose on my lakeside focuser on my FLT98, the obvious sign being the first V curve step being near horizontal. Also it shows as significant free movement if you operate the dual speed focuser on the opposite end of the shaft to the lakeside fitting. On the Voyager setup it takes HFD 40 as its upper V curve limit which is what your first V curve was showing at the horizontal section, which may have caused it to be horizontal for so long, rather than just backlash which wouldn't have helped.

Rather than just manually setting a V-curve step size which you can then adjust as required to get the most useful V shape, Voyager sets the step size to a value that gives a HFD change of 1 which is why it needs an initial setup trial to determine it.

I assume it doesn't step so far out now on its initial focus move, and takes less than your indicated 60 focus steps to determine correct focus. 🙂

Glad it's all working now though. 😀

Alan

Well I have a good v-curve and the focuser worked when trie, so I hope I am OK. Time will tell 🙂

[old_eyes]

2 hours ago, tomato said:

Glad you sorted it out. My NINA focus routines on my Esprit 150s are currently not as robust  as they normally are, but then I’m trying to use them in a twilight sky which I’m sure isn’t helping.

Incidentally I recently discovered NINA has a number of curve fitting functions for the focus routine, other than the simple V profile.

I like the full frame autofocus routine on NINA. It is more tolerant of initial focus than Voyager, and I often use it to check where I am and get a good focus. NINA is better in real time, and Voyager better for automation.