Autofocus help


#1

I didn’t have “save auto focus packages” checked (will be sure to do that next time), but want to see if anyone has tips on how to best configure autofocus in the latest beta. This is with an SCT 11" Edge, OAG, DSLR, IDAS D1 filter. What I’m seeing: I’ll see the HFR look something like, from left to right:

4.0 / 3.5 / 3 / 4.5 / 5 / 6 / 5

First – I’m not sure why the HFR is getting worse then better at beginning of the run (right side of the numbers). I saw the same behavior before I put backlash compensation to accommodate for mirror shift, but that shouldn’t be the cause now. Not quite sure what is going on.

Second, and mainly: AF keeps expanding the range presumably to get a better V. The problem is, the stars quickly become donuts with an SCT so I have to keep the step size pretty low and settle for a shallow U. Once the range starts expanding, it seems to runaway a bit expanding the range several times. I hit cancel, and the focuser is returned to some location (the middle of the new expanded range?) but that’s way out of focus now.

Is it possible to disable the range expansion? Any other tips to get a solid run? Unfortunately clouds rolled in before I got a chance to do too much troubleshooting.


#2

i get this odd behaviour with my ‘none edge’ C11. I am still workjing on it. I find 9 points with a step size of 80 ish works but not consistently to produce a ‘curve of some kind’ with a minimum point. But it’s what I’d call ’ rough as old boots’.
It always has the second point higher than first - as per your data.

I’ll post again once I have experimented a bit more
Paul


#3

You need to control how much area you’re making the run on. I had the same problem… it was fixed by reducing the step size after I figured out where HFR 3 was on each side of the focus and controlling the focus around that point.

Basically, figure out where you’re in focus (mine is step 6300-ish). Then, I figured out that 37 steps 9x gets me HFR 3 on each side of the focus (roughly 360 steps total, I did this manually using the HFR tool and working my way out on each side of focus).

If I let it go too far out, I get the same behavior you’re seeing.


#4

This is true. When you get too far out, you see this behavior. Basically HFR of donuts (HFR lower), then, as you go in, donuts disappear and you get big fat stars, (HFR higher), then it gets lower from there.


#5

Thanks all for the tips – I’ll definitely do some manual tinkering next time I’m out.

Under what conditions does SGPro auto-expand the focus range (and can this be disabled)? I will try to dial in the range a bit more, but it seemed that small focus step sizes were creating a fairly flat “U” … I’m not sure if it was this or another reason the focus range was expanded, but the end result was that instead of taking the best position of the original curve, the AF routine went off into the weeds a bit expanding the range – however, once we are into “donut range” the HFRs are not reliable.

I’d end up cancelling the AF run as it was running for quite some time (10 minutes?) – this was with 20 second exposures. I’m not quite sure if it would have recovered had I not cancelled.


#6

It happens when it detects it hasn’t created a ‘V’. Honestly, the only time I see it now-a-days is if I haven’t re-centered my focus from a previous attempt. If you dial in your focus before you start the sequence, it will never happen.


#7

I definitely have to run some tests and tweak a bit (although, it’s been working great in 2.3), part of this may be what I was imaging (large galaxy so I need to look closer at making sure only stars were getting selected). However, the challenge here is if I increase step size, I get that dip at the far end as I’m too far outside focus, but otherwise a tighter V. Decrease it and I’m getting a softer U.

During a running sequence, my approach has been to focus quickly (fewer data points, typically 5) and frequently, primarily to adjust for temperature and altitude. So, I’m never really out of focus by more than a few steps.

Jared/Ken, please consider a feature request to simply disable the auto expanding focus capability. I think it’s a fantastic concept, but also seems more can go wrong in some equipment configurations.


#8

Here’s the thing, the amount you want to focus is fixed… Find whatever gets you from HFR 3 on each side, divide it by 9 and you’ll be golden.

I fought it during the initial beta too. Honestly, just figuring out your steps (which was Ken’s solution for me) solved the problem. I’d recommend trying that.


#9

Thanks Chris, appreciate the pointers.

When you say HFR 3, do you mean 3x the best HFR possible? Example: best focus with an HFR of 2 is at step 6000. Say the low end gives an HFR of 6 at step 5800, high end of HFR 6 at step 6200. 400 / 9 steps is 45. So, for 9 data points, use 45 step size.

Now, I’m speculating here and happy to admit to user error, but the challenge has been getting 3x HFR before donuts start appearing. IIRC, the best HFR I was seeing was around 2.8 and I couldn’t really get beyond 6 or so, so this is why I’ve got a fairly flat U.

Perhaps my exposure time is too short or too long … I use around 20 seconds, which seems reasonable for f/7.

Looks like I’m clouded in for awhile, but will tinker more as soon as I can.

At what point though does AF give up, and what happens?


#10

No, I mean an HFR value of 3. Whatever step position gets you there is the upper and lower end. It isn’t a big range (360 steps or so for me). Its the higher HFR values that cause autofocus to fail. It fails above a value above HFR 3.


#11

OK, now I’m confused.

I thought conventional wisdom was to multiply the best obtainable HFR by about 3 or so to get your ideal end points for the AF run to get a nice V. An absolute HFR of 3 would almost certainly mean a flat U for me, since I’m typically between and HFR of 2 and 3 when imaging at long focal lengths.

Wouldn’t HFR always be a relative value based on optics, camera, and seeing?


#12

Generally you want the max HFR to be 3X the lowest HFR. However that might not always be obtainable. You essentially want to go as large as you can before the AF starts to go “off the rails” in this case. You’ll probably want to leave a little padding in there as well. So if you find that you start to have problems around 3 HFR then you might want your max to be around 2.5HFR or so.

The 3X is only a guideline and it works well for scopes without a central obstruction. Scopes with a central obstruction will likely need to lower this amount.

Jared


#13

Makes sense Jared, thanks!

I’m not trying to beat a dead horse here – it’s no problem to lower the range to avoid limitations like those imposed by a central obstruction. The result is a flatter U than ideal, but I also don’t want AF to keep expanding the range trying to find a better fit and ultimately getting lost.

I don’t think there’s a “disable adaptive focus” option (consider a feature request?) so in the meantime, in the scenario where I’m stuck with a fairly flat U how can I best configure things to prevent AF from expanding the range?


#14

While I will not rule out an option to disable dynamic focusing, we will not be adding this option during the beta phase (as we are looking specifically to collect data around this feature).

Also. While not ideal, if AF fails to find curves it will recover fairly nicely when runaway focus is detected (usually between 15 and 19 attempts… (we may redefine what runaway focus means prior to release). I know this can be a time suck if you are using long exposure AF frames, but it will recover if left unattended (by using a weighted average of the lowest 3 points found during the runaway attempt). This will usually provide a fairly safe focus position and will continue the sequence.


#15

OK, good to know it will recover, but if a single run takes about 5 minutes, 15 to 19 attempts is too long considering I’m trying to run it 1-2 times per hour. I’d definitely love to see some user configurable gates here eventually.

Next time out, I’ll try to tweak as much as I can, collect as much data as possible and hopefully get things working well, but otherwise, I don’t really have a choice but to go back to 2.3.


#16

Sorry… not 15-19 attempts, I meant 15-19 exposures in a single attempt.


#17

Ah! That makes more sense. Now I really need to pay more attention – I must’ve been right on that cusp when I cancelled – it had been running for about 10 minutes.


#18

Sorry to resurrect an old thread. I have been using SGP with my three refractors without issue, (once I had sufficient exposure). I have just started with a 10" RC and my HFR, at best is about 2.5 (which by all accounts is pretty good) with 1x1 binning.

I was already aware of the issues concerning central obstruction and reduced my V-curve span. In my setup, the autofocusing is more miss than hit though. The HFR is folding back when it hits about 3.5. It is not robust enough to leave it unattended. I venture to ask, having not tried a combination of either Maxim or Focusmax with an RCT, is it possible for a HFR calculation algorithm able to read a de-focused SCT/RCT ? Is this something for the V2.5 SGP roadmap?


#19

Yes,

You can find 2.5 stuff here:

It is under something about “improved ideas for SCT focus”. It will be a while though.


#20

I do had the same experience with my Newton. autofocus is somehow difficult to master.
isn’t a “quick” turnaround possible for that kind of obstructed imagers…?


www.mainsequencesoftware.com