SGPro autofocusing help.

[tooth_dr]

Any tips of fine tuning it?  I have followed the SGPro instructions and set up my autofocus routine according to that.

I've ran the focusing routine a few times tonight, and it never seems to find what I would visually determine as the best point.  Just now the computer settled on position 2415 based on the V-curve, however I could clearly see my diffraction spikes werent merged.  I subsequently ended up moving it to 2422 and the spikes merged.  My settings are 30 steps and 9 points, have tried upping it to 42 also (this gives me a 4-6x HFR increase over focus).  When I watch the best fit curve, it doesnt always seem to be the best fit if that makes sense, in that the lowest point according to SGPro isnt what I would say should be the lowest point.  I use 8s exposures, 2x2 binning.

Any help very gratefully welcomed.

 

[Starflyer]

Have you got any screenshots of your V curves, you can tell a lot by the shapes of a few focus curves.

I'd look at 3 - 4 HFR, any more gives you a V curve with a sharp point at the bottom, this makes it hard for SGP to determine best focus, you want a slightly rounded bottom.

 

[blinky]

Just a note - your VCurve might not look like the above, that screenshot is from the previous version, where the best focus was the crossover points of the 2 lines.  Now SGPro uses a best fit curve and best focus is the lowest point on the curve. 

The other thing that can mess things up is high cloud, often I cant even see it but when I try and focus my focus points are not great meaning the best fit curve is not totally accurate either. It does all work though, just takes a bit of tinkering to get it right!

[inFINNity Deck]

Hi tooth_dr.

The current version of SGPro uses a second order (parabolic) fit on the data points, so it is best to choose settings that make the curve look like a parabola and thus less like a V.

Nicolàs

[tooth_dr]

2 hours ago, Starflyer said:

Have you got any screenshots of your V curves, you can tell a lot by the shapes of a few focus curves.

I'd look at 3 - 4 HFR, any more gives you a V curve with a sharp point at the bottom, this makes it hard for SGP to determine best focus, you want a slightly rounded bottom.

 

Hi Starflyer - thanks for the reply.  30 steps gets me an increased HFR of x 4.  I've been using that, but changed it last night to see if it worked any differently - actually to see if the sharp curved helped!

2 hours ago, blinky said:

Just a note - your VCurve might not look like the above, that screenshot is from the previous version, where the best focus was the crossover points of the 2 lines.  Now SGPro uses a best fit curve and best focus is the lowest point on the curve. 

The other thing that can mess things up is high cloud, often I cant even see it but when I try and focus my focus points are not great meaning the best fit curve is not totally accurate either. It does all work though, just takes a bit of tinkering to get it right!

Hi blinky - I actually have a second autofocusing unit running simulataneously on another scope, using the DeepskyDad AF3 system.  It is operating a much slower F ratio but appears to be accurate.  I set them both up the same way.  The problem with the focusing has been there for a few months so I think conditions arent a contributing factor.  I dont have an example of the focusing from last night, but have attached one from a few weeks ago on the other scope below.  Just to reiterate this isnt the focuser/scope that is causing me the bother.

  

 

 

1 hour ago, inFINNity Deck said:

Hi tooth_dr.

The current version of SGPro uses a second order (parabolic) fit on the data points, so it is best to choose settings that make the curve look like a parabola and thus less like a V.

Nicolàs

Thanks Nicolas - I note that you are using 7 points and the difference between the in focus HFR and the end points is only 0.5.  What setting are you using?

[inFINNity Deck]

26 minutes ago, tooth_dr said:

Thanks Nicolas - I note that you are using 7 points and the difference between the in focus HFR and the end points is only 0.5.  What setting are you using?

Hi Adam,

that screen shot actually isn't mine, but of one of the other guys together with whom we built this new routine (I was responsible for the automatic outlier-detection part). At the moment I have no screen dumps of my own set-up in action, weather permitting I will post one later this week. Currently I am using 11 points in the focus routine and, once centred, the HFR values lie between approximately 2.4 and 7-8. Here is one of my focus-runs from the testing stage, processed in a stand-alone application:

The step-size of 15 is for my focuser, which uses a home-made 1:16 geared planetary drive on a standard SkyWatcher focuser (the OTA is an Esprit 150ED). So for the original 1:11 drive the step-size would need to be around 10.

Nicolàs

Edited December 22, 2020 by inFINNity Deck

[tooth_dr]

1 minute ago, inFINNity Deck said:

Hi tooth_dr,

that screen shot actually isn't mine, but of one of the other guys together with whom we built this new routine (I was responsible for the automatic outlier-detection part). At the moment I have no screen dumps of my own set-up in action, weather permitting I will post one later this week. Currently I am using 11 points in the focus routine and, once centred, the HFR values lie between approximately 2.4 and 8. Here is one of my focus-runs from the testing stage, processed in a stand-alone application:

The step-size of 15 is for my focuser, which uses a home-made 1:16 geared planetary drive on a standard SkyWatcher focuser (the OTA is an Esprit 150ED). So for the original 1:11 drive the step-size would need to be around 10.

Nicolàs

Very interesting!

Would you have any input into the value to use for min star detection?  I wonder why it's not going to the point of best focus?

[inFINNity Deck]

Hi Adam,

I am using 6px at 1x1 binning on my ZWO ASI1600MM Cool Pro. Exposures are 7 seconds and are always taken in luminance.

Nicolàs

[tooth_dr]

10 minutes ago, inFINNity Deck said:

Hi Adam,

I am using 6px at 1x1 binning on my ZWO ASI1600MM Cool Pro. Exposures are 7 seconds and are always taken in luminance.

Nicolàs

Ok, thanks.  I'm using 2x2 binning, 8 seconds on luminance and 15 narrowband.  Perhaps this is too long for the scope 180mm/F2.8

[Laurin Dave]

32 minutes ago, tooth_dr said:

Ok, thanks.  I'm using 2x2 binning, 8 seconds on luminance and 15 narrowband.  Perhaps this is too long for the scope 180mm/F2.8

Hi Adam I use 2x2 with 3/15 at f7 and 3/10 at f5.5 so I would try reducing the exposure for your faster system .. (although I’d thought SGPro rejects saturated stars)  and also reduce the step size, you could increase the number of points just in case..  also do you know the relationship between your critical focus zone and step size in micron ..  I’d imagine there’s some guidance on this somewhere in the ether..  

Dave

[tooth_dr]

18 minutes ago, Laurin Dave said:

Hi Adam I use 2x2 with 3/15 at f7 and 3/10 at f5.5 so I would try reducing the exposure for your faster system .. (although I’d thought SGPro rejects saturated stars)  and also reduce the step size, you could increase the number of points just in case..  also do you know the relationship between your critical focus zone and step size in micron ..  I’d imagine there’s some guidance on this somewhere in the ether..  

Dave

Thanks Dave.  I did have an issue in the beginning with my Hitec Astro Mount Hub Pro and Moonlite.  The CFZ of the Epsilon is 17um, and the focuser originally went in steps of 10um.  I sent the MHP back to HA and they updated the firmware to accommodate 2um steps.

I will try shortening the exposure time and increasing the steps.  I've noticed that sometimes it selects two stars as one star - if there are a couple of stars close together it puts a circle around both of these.

[Starflyer]

Are your problems with the Moonlite?  Have you tested it for slippage with a digital caliper?  Measure at one point, move out one or two thousand steps, move it back to the original point and remeasure?  It needs to return to the same point and do so repeatably.

They're beautiful looking focusers but not great for imaging, I've had two and gave up on both for autofocus.

[tooth_dr]

1 hour ago, Starflyer said:

Are your problems with the Moonlite?  Have you tested it for slippage with a digital caliper?  Measure at one point, move out one or two thousand steps, move it back to the original point and remeasure?  It needs to return to the same point and do so repeatably.

They're beautiful looking focusers but not great for imaging, I've had two and gave up on both for autofocus.

The focuser is just back from Moonlite following an upgrade and I would have no reason to suspect any issues.  In fact the whole thing, stepper motor and all, was setup by Ron, I just unpacked it and fitted it to the tube! I’m well below the weight limit that it has been set at so I don’t suspect slippage, but I can do a check tomorrow.  It does give a good v curve though, I regret not taking a photo of the screen last night.  It just seems to lack precision when selecting the best fit points, and ultimately the exact focus position.  Perhaps my exposures are too long.  I really need a clear spell to check these suggestions. 

[tooth_dr]

It's forecasting clear skies tonight unbelievably - so my plan is to reduce the exposure time and try this again.

Using SGP - i went through the instructions (in focus then move to 4 x HFR) - the calculations give a 30 step size.  Will I just stick with this?  If I make it 11 points, does that mean the total excursion is the same, but points of the graph are just closer together?  This has got to be a benefit?

 

Thanks in advance

Adam.

Edited December 23, 2020 by tooth_dr

[inFINNity Deck]

13 minutes ago, tooth_dr said:

It's forecasting clear skies tonight unbelievably - so my plan is to reduce the exposure time and try this again.

Using SGP - i went through the instructions (in focus then move to 4 x HFR) - the calculations give a 30 step size.  Will I just stick with this?  If I make it 11 points, does that mean the total excursion is the same, but points of the graph are just closer together?  This has got to be a benefit?

 

Thanks in advance

Adam.

Hi Adam,

the 11 points will be separated by 30 steps each. The advantage of more points is that SGP has more data to fit the curve through and thus to correctly detect outliers. Downside of the 11 points is of course that it takes more time to finish the routine (11 images are taken). The instructions to calculate the step-size provide a good starting point. After the first run you will see whether you get a V- or parabolic-curve, but you should aim at the latter as that will match the parabolic fitting better. So if a V-curve is shown, it is better to slightly reduce the step-size and try again. From a physics point-of-view the HFR has a hyperbolic response (so all optics show a V-curve when far enough out of focus), the near-focus part of which can be approximated by a parabola. Implementing a best fit method is easier when using a second-order best fit instead of a hyperbolic fit and that was the reason why that model was implemented.

Nicolàs

[newbie alert]

I don't use sgp but I measured the backlash on my moonlite yesterday.. to be honest I didn't think I had any but measuring it proved that I did.. I think Ray did a u tube video on backlash compensation using sgp( simplified astro)

[tooth_dr]

22 minutes ago, newbie alert said:

I don't use sgp but I measured the backlash on my moonlite yesterday.. to be honest I didn't think I had any but measuring it proved that I did.. I think Ray did a u tube video on backlash compensation using sgp( simplified astro)

Thanks, I have my backlash set to 50 and 'IN'.  I may increase it though as I dont think having too much matters.  I'm not sure it is a backlash issue though

[tooth_dr]

28 minutes ago, inFINNity Deck said:

Hi Adam,

the 11 points will be separated by 30 steps each. The advantage of more points is that SGP has more data to fit the curve through and thus to correctly detect outliers. Downside of the 11 points is of course that it takes more time to finish the routine (11 images are taken). The instructions to calculate the step-size provide a good starting point. After the first run you will see whether you get a V- or parabolic-curve, but you should aim at the latter as that will match the parabolic fitting better. So if a V-curve is shown, it is better to slightly reduce the step-size and try again. From a physics point-of-view the HFR has a hyperbolic response (so all optics show a V-curve when far enough out of focus), the near-focus part of which can be approximated by a parabola. Implementing a best fit method is easier when using a second-order best fit instead of a hyperbolic fit and that was the reason why that model was implemented.

Nicolàs

Thanks Nicolàs, this is all very clear and I understand what you are saying.  30 step size currently gives me somewhere between a curve and V shape - much like my post above.  I'll try 11 data points later along with a slightly smaller step size.  I dont mind if it takes 5 minutes longer, as at F2.8 it is very finickity, so a few minutes getting exact focus is not an issue.  I appreciate the help.

Adam

 

 

[newbie alert]

18 minutes ago, tooth_dr said:

Thanks, I have my backlash set to 50 and 'IN'.  I may increase it though as I dont think having too much matters.  I'm not sure it is a backlash issue though

To be honest I've not tested it yet and looking at the forecast it's not going to be anytime soon 

If you have high cloud that's enough for the plot on the chart to be slightly out, so I always check with a mask now..

When it works it's super fantastic