Ok… Well I am not sure about the star detection delta. That is one thing that did not change from 2.3 to 2.4… just the way in which we order them. I’ll take a look at some of your sample data.
Here is a run from tonight. Honestly I have not seen a curve that smooth before tonight. I included the log file in the dropbox folder. Nebulosity rejection is full weak and AF sample size is at 100.
Chris
One of the benefits of having star packs is that I can run the same exact data through both versions of SGPro and look for anomalies and unexpected deltas. I used the AF packs you sent over (thank you) and validated that there is no difference at all in terms of star detection (between 2.3 and 2.4). For AF pack 5 (originally captured in 2.4), both 2.3 and 2.4 found stars as:
13,15,15,22,37,35,29,39,14,19,15,15
AF Pack 2 (originally captured with 2.3) found stars in 2.3 and 2.4 as:
29,45,42,58,60,48,33
It appears as though the number of stars found and used for calculations is not version dependent, but is only differentiated by data.
One thing I might recommend… In your 30 sec exposures, some have significant trails. If you cannot avoid this during AF, you should either guide or reduce the exposure time. These frames introduce bad data into the algorithm and while SGPro can recover from a few of these, too many will come to a bad focus position.
Ken –
Thank you so much for looking into this and apologies if it ended up being a bit of a fools errand!
I’ll keep tinkering; again, many thanks. Last night/tonight I’ve been imaging with the Hyperstar at f/2, and AF seems much more stable.
Brian
It’s not a problem… and also, just as a note, the sample slider will have no effect for you unless you keep it very low… there are just not a lot of “good” stars. We reject stars that may look good to the eye, but will likely provide poor HFR data. You probably don’t need to use this slider at all. It is meant for people that have 20 or 30 good stars and then 70 parts noise and hot pixels that overwhelm the HFR calculation. In this case you would set the value to 30 and the other junk would get ignored.
An interesting thread - I use an 8" RC at it’s native 1600m and I am finding the auto focusing routine inaccurate compared to a bahtinov mask. I hope that the developers can work on the longer focal length focusing as it really can be a deal breaker I think.
Closing as duplicate.