Autofocus issue


#1

I was imaging M51 last night and was struggling with the autofocus routine. It was ending up over a fairly large range for my setup (c9.25, .63 reducer, external focuser, ST2000XM). It never really would do a good v-curve and after stopping my sequence and working on the autofocus settings, I notice SGP would occasionally select the core of companion galaxy of M51 (NGC5195) and that would create a bad datapoint on the curve. This would happen a few times during the focus run. So the result would be sawtooth curve more or less. So I ended up disconnecting from SGP and using FocusMax via MaximDL on a nearby star.

I guess my question is how do I tweak the “nebula” slider bar to avoid SGP from using the galaxy core. I went from one end of the slide to the other and I could not get a reliable v-curve. I have followed the instructions on how to determine the step size and all that with my dial indicator. When galaxies and nebula do not cause problems, it works great. I kind of wish we could manually pick which stars in the FOV to base focus on as a override. Is that possible even? If it helps, I focus thru my Lum filter, 2x2, 5sec, dark frame enabled, 11 steps.

Thanks for any ideas/help.


#2

SGP uses 300 stars to compute focus metrics. So if it was just a few stars in the core or a few data points in the core then that’s likely not the issue.

By what you’re describing it sounds like more of an issue with seeing. I would start by increasing your step size to get a more dramatic V curve and see how that goes. To be completely honest I use about 2X my calculated value for step size and in bad seeing I increase it even more. I have found this works better for my setup than the calculated value. Your mileage may vary, but this works well for me.

Thanks,
Jared


#3

That is good to know Jared, thanks. My computed step size is 36 and typically I have it set to 160-200 in SGP with 9-15 steps. Looking at last nights image history, the HFR is averaging around 2.6. Not sure how that ranks with “bad”. The CSC did indicate poor seeing. Tonight is suppose to be clear again, so I will give it another try. I actually let it run and when I got up in the morning, I had a PHD2 guide failure, so the sequence aborted, but that is another thing to look into.

Is there a diminishing return on how large that step value can be set too? Obviously everyone’s situation is different and depending on seeing, the setting needs to be tweaked too. Will 2.4 be less finicky with regards to this?


#4

[quote=“pmumbower, post:3, topic:230”]
Looking at last nights image history, the HFR is averaging around 2.6. Not sure how that ranks with “bad”. The CSC did indicate poor seeing. [/quote]

That seems pretty low. Generally my sharply focused frames are around 0.7-0.8.

Yes, at some point the size will be too large and you’ll likely end up not being able to definitively tell where focus is given the samples. If you start seeing HFRs that are out of the single digits then your step size is likely too high. I always aim for my high HFR value on the graph to be about 3x my low. So if a focus frame is 0.8 then I shoot for the high of 2.4. I don’t get super worried about this…these are just guidelines. I would say that you probably want to keep your high HFR in the single digits though…probably under 5 HFR.

2.4 will not really change much in this regard to what the current 2.3 is for determining focus. Currently 2.3 uses the same best fit algorithm that 2.4 will be using. The main advantage in 2.4 is that it will be able to correct focus when you’re more out of focus where 2.3 would just leave you poorly focused and continue. I’m saying this assuming you’re using the latest 2.3 release.

Thanks,
Jared


#5

I think my better nights, which are rare the past 6 months in Michigan, have been in the 1.4-1.8 range. I image at 1480mm, which works to be around 1.05". So maybe I am pushing it with my location.

I always start the night by focusing with my Bahtinov mask. So maybe if I do take a 2x2 image(my focus bin) and calculate the HFR on that, I will know what range I need to be in. Have you found much difference with focusing on different bin modes? Sorry for all the questions, I am learning a lot about this feature and love it when it works good for me.


#6

That is good know Jarred, I have always used my calculated step size and get a bad curve a higher percentage of the time, normally I would run AF 2 or 3 times before I get a acceptable curve. I will try increasing the step size on the next clear night and see how that performs.

Thanks

Trevor


#7

Generally binning is better for AF. The more pixels a star occupies, the
more accurate the HFR measurement. Go with 2x2 or 3x3 at least.
pmumbowerhttp://mandrillapp.com/track/click.php?u=30191084&id=7f48a04f90304bdbbab8d548e2a0becf&url=http%3A%2F%2F191.234.49.38%2Fusers%2Fpmumbower&url_id=4e62e4b9dd6925cc4ee8f9e8bdb086c7fa397d6b
May 22

I think my better nights, which are rare the past 6 months in Michigan,
have been in the 1.4-1.8 range. I image at 1480mm, which works to be around
1.05". So maybe I am pushing it with my location.

I always start the night by focusing with my Bahtinov mask. So maybe if I
do take a 2x2 image(my focus bin) and calculate the HFR on that, I will
know what range I need to be in. Have you found much difference with
focusing on different bin modes? Sorry for all the questions, I am learning
a lot about this feature and love it when it works good for me.


#8

Quick question while tweaking my AF settings tonight: What do the various color boxes mean? I assume green is good and red is bad. Is that correct?


#9

That is correct. Green means that the point contributed to the slope calculation. Red means it was rejected. The lowest point of the V is also rejected as it could incorrectly skew the results.

Thanks,
Jared


#10

Ok that is very good to know. Since I am imaging at 1600mm with my ST2000, the fov is not huge. So like with my current project of m51, even with 2x2 binning on and 10sec exp thru the lum (dark sub on too), I do not get a lot of stars. Actually watching it AF several runs will tweaking the setting, I would say there were usually 3-8 green boxes per AF frame. So I am guessing that is where my AF issues are coming from.

Besides going with even longer exposures, is there anything else I can do? Kind of wish we had the option to invoke the “focus target” before the AF run begins so I get get over to a better FOV of AF. Especially while doing spring time galaxies when star are few. Pretty sure once I get working on summer time stuff, I will have plenty of stars.


#11

The SGPro AF routine is not guaranteed to work in all areas of the sky (for
exactly reasons you mention). For targets with low star counts, you can
add patterns that automatically pause the sequence for manual focus (just
like the patterns for AF), move the scope to a better region, run AF (or
use a bahtinov mask), then right click your target, click center on target
and continue the sequence. Someday, we may automate this process… I’ll
place it on the backlog. You can determine if you can trust SGPro to focus
for you by running AF on your target manually prior to starting the
sequence.


www.mainsequencesoftware.com