SGPro Auto Focus Question

Hi Group,
Before my question, I’d also like to echo compliments to how well SGPro works. I was away for a bit over 3 weeks, and this last Saturday was my first night back. I upgraded to the latest SGPro and PHD2, created a new sequence for M101, and ran close to 8 hours successfully. Really nice software!

The only issue was a midstream problem with autofocus.

  1. Sequencing was in the following filter/event order: Ha (12), LUM (12), R (6), G (6), B (6).
  2. AF was at each filter change and after every 5 images.
  3. Temperature Compensation was at 113 steps per degree C.

LUM started around 11:24 pm (file 232444) and finished the last file (013831) around 1:49 am. From the start of LUM through 12:38 am (file 001227), focus was acceptable. Although file 001227 did suffer a DEC bump which made it unusable. But from files 001227 to the next (002318) there was an AF, and I believe a report of almost a 1 C temp drop. It looks like SGPro did what it was told to do, which was adjust the focus position by about 100 steps (from 29377 to 29473). However, this data appears to have been bogus and SGPro never recovered from the out of focus state through the remaining 7 images. The AF Packs 7 and 8 show the focus imbalance, where AF Pack 6 (focus going into LUM) shows acceptable.

File 013831 was the last (12th) of the LUM images and the 7th that was out of focus. File 014945 was the next file, and the first of the RED images. And it was back into acceptable focus.

With my setup, I expect to have to sacrifice a few images to DEC misbehavior, and maybe rarely to mis-focus. This session was unusual in that consecutive AF executions (AF Packs 7 and 8) failed to correct focus. Also interesting was the AF for the following filter (RED) had everything back where it should be.

Sorry for the long wind. Any thoughts on this?

It’s hard to say and it may just be a lack of comparison of “good focus run” data, but this stuff looks a bit murky… almost like thin clouds. Anyhow… regardless of this, SGPro is picking up LOTS of false stars (hot pixels, noise, etc). It is averaging 100 stars, when in reality you only have about 30 (maybe). That means that 70 star imposters have snuck in and overwhlemed HFR data presented by the real stars. AFPacks 7 and 8 are probably as a result of flying off the rails during the AFPack 6 run (bad starting pos). Running through AFPack 6, while limiting your stars to 20, yielded much better results.

Thanks for your reply Ken.
The sky was forecast for clear, but thin clouds may have passed over.

  1. Are you suggesting that I reduce the AF sample size from the default (100) to something closer to 20? Sorry, but I’m a little unclear on how you phrased that part of your response.

  2. I’ve been skeptical of how well the MicroTouch package actually represents the ambient temperature. I may turn the temperature compensate feature off for a while.


This will vary from target to target, but yes. You should run AF manually on a target before letting loose on a sequence.

The exact accuracy is not super important, just its ability to change over time. There was another request to add hysteresis for this output to help smooth this out.

Sorry to keep picking on this.

Your comment about doing an AF manually prior to sequencing…
In this case the LUM AF hiccup occurred only after 12 (10min) subs of Ha and 5 (10min) subs of LUM. At that point the last 7 subs of LUM lost focus. And then the AF recovered as it went into RED, GRN and BLU.
So, at the start, SGPro was satisfied (I think) with the initial AF HFR values.

But FWIW, this level of scrutiny (reviewing the AF HFR) is new for me so I will take more care going forward. Your suggestions here and some of the discussion in your and Jared’s presentation on the AI channel has given me more insight.