Autofocus with narrowband filters

I know this topic has been covered more times than I can count, and I’ve read most of them. What I am seeing is after running autofocus, I am getting better HFR by making manual adjustments myself.

I started by running autofocus with my 3nm Astrodon H-alpha filter. I took 8-second exposures and binning was 2x2. Even though the v-curve looks terrible, its obvious that the best position here is at the 30806 position.

So I take a sub and the HFR for the shot is pretty awful at 4.11.

So the very next thing I do is use the “out” button on the focus control panel to rack the focuser out by 100 steps. Now I take another sub and get much better HFR; this time 2.27.

What am I doing wrong here? Any suggestions on how I can get autofocus to give me better HFR?

I know there have been a great many people who have been using offsets to handle this, and I would love to try that, but first things first. I need to understand what is going wrong before I can get to the point of setting up offsets.

Thanks in advance for any suggestions you can provide.

Edit: here is my autofocus settings, in case anyone wants to see those. Let me know if I need to post my SGP logs here too, but I wasn’t sure if those would be useful for diagnosing this problem.

8 second exposures is almost certainly too small using a 3nm filter at 2x2 binning. For instance when I focused with my 3nm filters I would use 60 seconds at 4x4. You just don’t get much data in 8 seconds using those super narrow filters. This is generally why we recommend to setup your narrowband filters to focus with a lum filter and set up the filter offsets.


I am setup right now and I just did autofocus with your recommendations. I picked 4x4 and used 60 second exposures. I wish I had taken a screenshot of the v-curve, but it looked a little better in shape, and ended up recommending a focus of 30914. The HFR was 4.20 and 4.36. So I racked the focuser out by 100 steps again. I did this about half way through the exposure where I ended up getting an HFR of 3.4. The next shot, which was done fully at the new focal point, resulted in an HFR of 2.83.

So something still isn’t working.

I would also suggest you increase your step size by 50% or more.

Ok, I increased my step size from 125 to 190. Ran autofocus (again with 60 sec exposures) and got a better looking curve. Focal position was 30868.

Took a 2 minute exposure and got HFR of 5.3.

Racked the focuser out 100 steps and took another. Got an HFR of 3.58.

Racked it out again by another 100 steps and got an HFR of 2.55.

Maybe there is something wrong with my hardware? I have been reading other people’s posts about stuff being loose in the focuser. I’m using a Focuser Boss II on my 2.5" FT focuser.


In SGP, the HFR is reported in terms of pixels, not arc seconds, and is the average of many stars in the sub. When you take a “short” auto focus frame (ie, 10 - 15 seconds), the stars are small since the short exposure time does not cause the star sizes to grow. When you take a longer sub, the stars get bigger and that raises the HFR value. So the HFR of a downloaded sub will almost always be larger than the HFR reported by the auto focus routine. Binning will also affect the calculated HFR. A 2x2 binned sub will typically have an HFR half that of an unbinned sub. So if you auto focus at 2x2 you might get an HFR of 1.2. Then you shoot a longer, unbinned sub and see an HFR of 2.6.

Also, I see your min star size was 4. That could be too small depending on the focal length of the scope. The auto focus routine is largely empirical – try stuff to see what works. Going from 125 steps to 190 steps between data points made the V curve better. Increase the number of steps again; change the min star size; and finally, you should have an odd number of data points; try 9. You had 8 in your original screen shot.

Also, since you are always backing the focuser out to get a better HFR, this could be symptomatic of an incorrect back lash setting. If the current backlash setting is 0 then it sounds like it should be 100.


Thanks for the reply. I will do that as soon as I get my clear skies back.

I have always been a little unclear about the “min star size” setting. I took it to mean that SGP will consider anything smaller than 4px to be hot pixels and not try to use them in HFR calculations for the full frame. Is that wrong? Can you suggest what I should be using for this setting? You mentioned it is based on the focal length of the scope. Mine is 480mm and the scope is an f/6. My camera has 2.4 micron pixels and I have a pixel scale of 1.03 as/px. Since I must be misunderstanding the purpose of this setting, I have no idea what it should be.

I never realized there was a backlash setting! I just found it and set it for 100, so I am anxious to try this at my next opportunity.

Thank you for all the suggestions!


The min star size is related to the issue of not processing hot pixels as tiny stars. You are imaging with a rather short focal length (small star images) and using a camera with small pixels. You are kind of at the far end of the focusing continuum! When tweaking the auto focus, only change one parameter at a time – that is, don’t change the min star size and the number of steps simultaneously. As far as min star size start at 2 and slowly work up to 6 or even 8. Also don’t increase the number of steps between test runs by too much. You could over shoot the optimum setting without realizing it. If so, you are kind of down the rabbit hole after that. Always make note of any setting that does relatively well. If things go to hell, you have a known point of return.

One other comment about HFR values – they have no absolute meaning. They are just reference values that your configuration of hardware produces. The only thing you are going for is to get a minimum HFR value – that will always be the best focus point.

And finally, remember SGP assumes the image is reasonably well focused to start with. So, get a decent manual focus before even trying the auto focus routine.


You might also need to play with your backlash settings. If you have a lot of backlash it may not come back to the ideal HFR when it reverses direction. Although if it’s far off SGP will give you a warning that the HFR doesn’t match expectations.

I just wanted to report back and say that this solution PERFECTLY fixed my issues. Now I’m getting excellent V curves when focusing with clear, and then switching to my NB filters gives me excellent HFR, right on the money.

I was also pleased to notice that when I changed the step value for backlash, it would adjust the focuser the proper number of steps so it wasn’t necessary to rerun the focus operation. Bravo!!!

Thanks to all for your help, and thanks to Main Sequence for making such a great product!

1 Like