Manual Focusing, HFR star button, large sensor, not accurate


Hello SGP Team and Ken welcome back! I just wanted to bring up some issues I’ve experienced since the new HFR changes. I have seen inaccurate HFR representations. I am not sure if this is a result of resizing the image to help speed up the HFR calcuation with wide sensors on slower computers.

Here is a comparison of some data where the focus is gradually declining then corrected.

According to SGP it appears the focus is fine statistically, however when comparing in another program the FHWM is declining. When I am out in the field it appears to me that everything is fine because the HFR number was declining or around the same level, however in reality it is actually getting worse.

It’s also quite difficult to see if you are out of focus sometimes with the bayered matrix (in my opinion). Especially on tablets out in the field, on a 24" monitor a little easier.

I have shared frames 1->9 if you find this helpful:

Thank you again for your excellent program and any time have on this issue.



Yes, the bicubic resample of the image seems to have a much more destructive effect on images with the bayer matrix intact.

When I disable that, I get this:

  • Frame 1: 2.83 (Initial focus)
  • Frame 7: 3.15

Not sure exactly how to handle this… the HFR calculation is simply way more computationally expensive. I’ll try to think about it some…


I thought about this some… Since “good” sequence images should not have donut stars (the primary driver for HFR refactor), I am thinking about using the older, less precise, but WAY faster routine for star analysis… Maybe. That routine had a nebulosity rejection parameter that is no available (would be too confusing to add HFR parameters that affect different things). SO… I will put a little time into seeing if there is a generally acceptable set of parameters to feed the old routine (or an application level setting to use the faster one).


Ken, I really appreciate your time on this. I love to be able to trust those HFR numbers coming in.

I’ll keep those frames available for download for any tests as it shows a slow gradual focus loss which is a typical scenario with temperature drop or focuser drift.

Thanks again!


Hi Ken, Just wondering if this was still possible? Thank You.


I not really sure. The two settings for this AF method (Nebulosity rejection and max stars) are no longer adjustable so I am afraid of introducing a very frustrating image history option. I suppose we can try it. I have also been looking at ways to possibly make the current method faster, but I am running into a series of brick walls there…


I wonder if you should consider doing a simple debayer (eg bilinear) on OSC (one shot camera) images before other processing. If I blow up an OSC image to 200% it is pretty apparent that there is noticeable pixelation that may be interfering with star detection. Stars have a checkerboard pattern which I suspect is because the green pixels are more sensitive. Perhaps consider a modified debayer to reduce the image to gray scale. Just interpolating the green pixels might do the trick. The result could be much better for star detection.


DesertSky, correct me if I’m wrong, but most (including my qhy10) OSC cameras when binned are outputting mono images. And when u frame and focus wouldn’t u be binning 2x2 or more anyways to help speed up the process? I guess I’ve never looked zoomed into my 2x2 in SGP to see if the matrix is still there. After reading this thread I’m wondering if my focus is always a bit out of focus now since I’m using an OSC in SGP.

After writing this I’m also realising that we OSC users still need accurate HFR stats in 1x1 , otherwise watching image history for focus data is ineffective.


I believe that is correct when binning.