Autofocus failures

Hello again. I have a celstron C14 SCT and 0.8x reducer with a Starizona feather/ microtouch. I use autofocus at the beginning of my sessions to get a good focus. I focus manually first so the autofocus routine has something good to work with. I’m getting about a 1:5 success: failure ratio. Sometimes I get a nice u or v curve and a 90% + confidence indication and other times it repeatedly tries to focus, can’t get a good result and returns to the original focus point. Most of the time past midnight I’m not in attendance so can’t intervene. The successes are usually at the start of an imaging session, when I click ‘run’ shortly after I platesolve in a random area of sky. The failures are when I’m not in attendance - I’ve set the system to autofocus every 90 minutes, so the failures usually have a galaxy or two in the image, not sure if that affects the algorithm.

Just for an idea of scale, if I image M81 with this setup it more or less fills the frame. I’m mainly trying these days to image some of the Arp galaxies and these are usually much smaller, perhaps taking up 20% ish of the FOV (for the larger Arps).

I’m using 7 points and 125 steps per point which gets me donut stars for the first point. I also crop the image by about 15% to avoid any odd shaped stars at the FOV edge.

Any suggestions on how to improve things would be wlecome.



If you are getting donut stars at the extremies of your AF curve then I would reduce the step size. Measuring the HFR/HFD of donut stars is a problem.

Posting your log file would also help as it has all the data.

Thanks, it’s just donuts at the start of the autofocus routine, not at the finish point. I’ve empirically tested with different step sizes but step size doesn’t seem to affect the result. I’ll post the log this evening.


here’s the session log file - many thanks for your time with this.


here’s an example of what I get at the start of a session. Always gets there with one attempt if I ensure reasonable focus before the AF starts.



Hi Paul,

Looking at the AF curve you posted above and the one below which I extracted from your log using SGP Log Viewer, it seems to me that SGP is trying to do the right thing but is having trouble with your data.

The chart I’ve uploaded is for single AF run from the log, but this has repeated 4 times before managing to acquire a set of seven data points that the curve fit routine deemed to have passed the quality threshold.

Out of the five AF runs contained in your log fie only two managed to pass the quality threshold at the first pass, one required two passes, and two required 4 passes. I can fully understand your frustration.

Looking at the collection of curves it seems that they are all moreorless aligned around the same best focus position which suggests to me that the focusser is accurately repositioning the camera at each pass. So the issue would seem to be to be why during each pass do there seem to be one or two sufficiently aberant data points to bring the fit quality below the threshold?

Looking at the curve I pulled from your log something seems to be happening on the left-hand side of the curve. Not sure what but maybe your stars have started to donut and this has caused problems for SGP’s star detection and measurement routines. If this suggestion is correct I would suggest reducing your step size so that during AF runs the maximum HFR values you obtain do not go above c5.0 as this seems to be the threshold for the poor detections.

As you are focusing out to in the left hand issues woud not seem to be backlash related but I would suggest maybe increasing (doubling) your backlash settings to see if this helps at all with initial measurements at the right hand side.

I should also ask if the AF results you obtained on this night reasonably typical of what you get, or where the conditions unusually gusty or hazy, that might explain the poor AF data sets?

I sse you are using SGP version In v4, has introduced a new star detection and measurement option using ASTAP. I am a fan of this option. Do you have SGP configured to save your AF image packs? If so then if you would send me a link I would like to take a look and see if upgrading to v4 might be a benefit to you. Or do so when you have another set of poor results!

To conclude, here is the graphic for the second very poor AF run. This shows the signs of uncorrected backlash at the right but backlash fails to explain the left-hand side problems.

thanks for this analysis Mike, much appreciated. Yes this AF data is typical of what happens in my imaging sessions. I live in South Wales and in the last few years the winters have been consistently cloudy here, so not much chance to get data and experiment. The session the log file relates to was very clear sky. I have reduced the step size by one fifth and will try that next session - in fact I had made this change before posting the AF image, but mist took over about one hour later.

I think I have the backlash set around one thousand steps on the focuser. Perhaps I’ll try to measure the backlash. I have the coarse setting set at 100 steps and the fine at 30. So I could approach focus with coarse adjustments going in until I get good point stars then reverse direction and count the number of steps going out until focus starts to change. It would be rough but I think I could get it +/- 100 steps with care. I’d have to disable backlash compensation temporarily to do this. I’ll give it a a try.

Re V4, I’ll try that too. I read the posts regarding SGP star detection vs ASTAP with interest. Anything that helps to improve operational effectiveness is very welcome these days as there’s so little imaging time.

I don’t currently save AF packs, but I’ll start doing that. I’ll also look for the SGP log viewer, I use the PHD2 viewer for PHD logs and it’s very useful.
Thanks again,

Hi Paul,


SGP Log Viewer is here: SGP Log Viewer … Updated 4-7-2021 – Ceti Alpha 5

SGP AF Log Viewer is here: SGP AutoFocus LogViewer download |

They have a somewhat different approach but I find them both useful.

A couple of further things I think you might try are:

a) Try increasing your AF exposure times. This should boost SNR and maybe give more reliable star metrics
b) Do you have a mount settling time set (Control Panel/Telescope). Moving a camera back and forth during focusing has to cause some vibration which in turn adversely affect HFR measurements. Don’t go too high as it is applied after esach focuser adjustment so will mount up considerably.

If it’s any consolation weather in SW Surrey has been rather poor too.


thanks Mike that’s very helpful. re weather, yes I’m in touch with Greg Parker at the New Forest Observatory - he gets much the same.

The Download Link on the Ceti Alpha 5 page for SGP Log does not work, it takes you 3rd party subscribe page… not a down load. The Download Link Calls which is invalid.

Bummer really wanted this little App.

yes it looks like the link’s been hijacked…Anyone have a copy of the program which could be accessed? Or a link repair?


I think this is what you want. I’ve also alerted the developer that there is a problem.

Thank You very much!

fyi: The download page for SGP Log Viewer is fixed.

many thanks for that - I really should start working out my AF runs better and try and do offsets, rather than AF all the time.

Have you tried going BIN3 with 8 to 10 sec exposure, it works for me.

I have a 14 as well, IMHO I would reduce your step size for one thing. If your having donut’s, that’s to big an adjustment. Also did you also select “disable smart focus”. I also do 1x1 binning instead of 2x2. Works better for me.


1 Like

thanks Kip,

Yes smart focus is disabled in my settings. I have reduced the step size from 125 down to 80 and that’s reduced the donut effect. I tried 1x1 binning last night, but wow what a tming difference! Each autofocus point took about 2 - 3 minutes !! a nine point run took circa 18 minutes to complete. I think something is wrong there, so I swapped back to 2x2 binning.
thanks for input

Do you also have “fast download” set as well. How did you come up with the step size?


yes I’ve always had fast download selected Kip. Re the step size, I originally followed the guidance in the SGP help file and then to try to improve things I started experimenting. I always get ‘lop sided’ focus curves with more points on the right (as viewed on screen) than on the left. Last night it was a real mess and the curve looked like a random mass of points - you could’ve done a least squares fit and got a straight line!

I’ll have to start saving the focus runs and provide some of those in the hope that they may provide some insight into what’s happening.