#1  
Old 20-09-2015, 09:12 AM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Auto focus question

I've been playing with auto focus settings in SGP trying to improve the reliability of the auto focus routine. The variables are: exposure length, binning, filter selection, number of points on the V-curve and number of steps between points. I've been trying various combinations of these and comparing the V-curve results with the focus point as indicated with a Bahtinov mask.

With the SN10/QSI683, the calculated critical focus zone is around 40um which is about 10 steps on the Moonlite focuser. What I'm finding is the V-curve is quite flat around the best focus point, especially if I use 9 points of 10 steps each, so the auto focus routine finds it difficult to achieve repeatable results. If I increase the number of steps between points to 23 I get better definition of the V-curve because the stars are a long way out of focus at each end of the curve, however this means each point on the curve is 92um apart which is more than double the critical focus zone width.

I've also found it is better to have more points on the curve as it is quite common for one or more points to be outliers causing the curve fitting to be thrown out. The down side is it takes longer to run the focus routine as you add more points.

SNR clearly plays a big part in calculating the focus metric (HFR in SGP) and I've found I'm getting better results with 2x2 binned 5 sec exposures than 1x1 binned 10 sec ones. The impact of SNR is quite noticeable if I use CCD Inspector to calculate FWHM of the autofocus images and compare this to HFR determined by SGP. I get quite good correlation between the two metrics when the SNR is better.

In calculating the critical focus zone I notice this changes with filter (wavelength) - 51um for red vs 37um for blue. I'm now wondering if there is any advantage using the red or blue filter rather than the Lum to perform the focus routine and use filter offsets pre-determined with the Bahtinov mask when changing to a different filter to image.

Has anyone played with using the RGB filters for focusing? Are there other auto focus routines that would work with SGP?
Reply With Quote
  #2  
Old 20-09-2015, 09:45 AM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Hi Peter,

To my knowledge, using filter offsets will be faster but possibly less reliable.

I only use narrowband filters (3nm), and have found that 20-second exposures in autofocus routine work very well, combined with 7 data points, and about 20 steps for the focuser (also Moonlite, telescope is f/5.6). Such settings give a fairly sharp V-curve pretty much each time. More steps are counterproductive -in my case as the stars then are way out of focus and HFR values are initially not changing too much resulting in a fairly horizontal trend (not good for finding optimal focus).

Hope it helps a bit
Reply With Quote
  #3  
Old 20-09-2015, 09:57 AM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Thanks for the info Slawomir.

I believe your system has a CFZ of around 78um so I can see why 20 steps works well for you. I think the f5.6 system produces a more defined V than my f4 optics. Are you binning for autofocus?

Why would filter offsets be less reliable?

There must be a lot of people imaging with f4 or faster systems - I wonder how are they focusing?
Reply With Quote
  #4  
Old 20-09-2015, 10:17 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by peter_4059 View Post
There must be a lot of people imaging with f4 or faster systems - I wonder how are they focusing?
Strongmanmike uses an Atlas on his AG12. It has a tiny step size (105,000 steps moves the total range of about 11mm from memory.) The Atlas appears to be the focuser of choice on fast RH scopes as well.

Sorry, I can't help otherwise. I don't have experience at less than f/4.9.

Cheers,
Rick.
Reply With Quote
  #5  
Old 20-09-2015, 10:30 AM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Rick - sounds like you think it is a mechanical issue? I've had the dial gauge on the focuser and stepped it in and out and get very reproducible results.

I suspect it is the HFR calculation or variation in HFR due to SNR, guiding error or something else. I've seen reference to flat v-curves from fast optical systems in the Focusmax manual. I also notice the Focusmax V-curves recommend 30-40 points and aim to get a large variation in HFR over the curve.
Reply With Quote
  #6  
Old 20-09-2015, 10:32 AM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Hi Peter,

Yes, I think theoretical CFZ for my system is between 70-100 um, but in practice it seems smaller than that. I am not binning for autofocus.

I do not use offsets for filters as I feel there may be more variables involved in obtaining good focus than just temperature, so I prefer refocusing on the night when on the target and with the filter in place rather than using offsets.

I think you can also use half-steps with Moonlite ,if needed.

With half-steps you have 0.00008" of focuser movement per half step (2 microns per half step) - should be sufficient for f4 system
Reply With Quote
  #7  
Old 20-09-2015, 10:40 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by peter_4059 View Post
Rick - sounds like you think it is a mechanical issue? I've had the dial gauge on the focuser and stepped it in and out and get very reproducible results.

I suspect it is the HFR calculation or variation in HFR due to SNR, guiding error or something else. I've seen reference to flat v-curves from fast optical systems in the Focusmax manual. I also notice the Focusmax V-curves recommend 30-40 points and aim to get a large variation in HFR over the curve.
I haven't thought about it deeply, Peter, because I've never had much trouble with automated focus. A couple of things I've noticed are that I generally get better looking V curves on faster/shorter FL systems (even the same system with a big reducer) and that just being in the CFZ is not good enough. I've always assumed that the first effect is due to the greater impact of seeing at longer FL.

Have you thought about playing with an eval copy of FocusMax 4? (or one of the earlier free versions.) Might be an interesting experiment.

Cheers,
Rick.
Reply With Quote
  #8  
Old 20-09-2015, 10:42 AM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,877
I also use an Atlas focuser on my Riccardi Honders.

Its quite easy to use and is very precise. At present I am not using Focusmax or other software but about to implement one.

One thing that does concern me about automated focusing is the fact that all electronic focusers I have used including the Atlas (the most precise) seem to focus best moving out towards the focus point rather than in.

So you manually focus, you get a spike in the sharpness graph (I use CCDsoft at the moment and about to stop using that as well in favour of Sky X). Then if you go past it its not a matter of reversing the same number of steps. It will be further away. I suppose this is some friction or fighting gravity.

I wonder how that impacts on V curves.

Anyway, focusing is a breeze with the Atlas and precise focus is easy to achieve and obvious. The image also looks a certain way.

Roland Christen focuses using a single star and I think probably in the centre of the image. Use subframing to get a good look at the stars and you can compare them when the new image downloads to see if they get smaller.

I would rough focus on 2x2 binning and then use 1x1 if I wanted a tad more accuracy.

This is at F3.8.

Another factor is unless your camera is perfectly square to the scope then the whole image will not come to focus at the same focus distance.
You can see the top or left or right side being sharp but the other side not as sharp. That is your camera not square or it is flexing slightly.

In that case, square up your camera and pack out the offending sides or if still a slight error a compromise focus position will be required. An automated focus routine must do some average of the image or just uses one star so it would be safest to pick one in the middle where these flexes revolve around and is really the midpoint.

Roland also focuses in such a way to allow a bit of change as the temp shifts.

How much it shifts with temperature is something you'd have to log with accurate temp info.

The Atlas has a temp sensor in it but its inside the focuser and I find its slow to respond to ambient and is always showing it as warmer than the ambient. I tried to work out how much the focus shifts with 1C and it was a bit hard. My best guess is about 50 steps out as it cools (AP RHA). That's not much.

I focus at the beginning of the session and then again after an hour or so and it usually refines a little. Luckily the RHA is pretty temp stable focus-wise but there is a change. Not like an FSQ which shifts a lot.

Focus can also be hard in bad seeing. Comparing 2 images becomes hard as one has better seeing than the other so even though you are improving focus it can look worse. So if you think you are on track to sharper focus but the next image looks worse, wait a second and take another to check.

I use Astrodon LRGB Gen 11 filters which promote they are parfocal. Its very handy. I do find that they are in fact parfocal so I focus using Luminance and that seems to be correct for RGB Ha O111 and S11. Very convenient! Baaders promoted that they were parfocal and they were way off. That was a few years ago so I don't know if they have fixed that. Also they promoted they were 1:1:1 colour combine and they were way off! Gee honesty in advertising!

Hope this helps.

As I say I am about to use FocusMax Pro so I will be posting questions soon myself! I am overdue to using it. But I am confident my manual focusing is super accurate and that's the Atlas. Its a beautiful piece of astro gear. Super expensive but amazing.

Some scopes hold their focus better than others. My CDK is all carbon fibre and it does not shift focus that much but there is a minor change with temp.
The old RCOS carbon fibre promoted focus and forget and same focus night after night and that was somewhat true. Perhaps not entirely.

Greg.
Reply With Quote
  #9  
Old 20-09-2015, 11:00 AM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Quote:
Originally Posted by RickS View Post
A couple of things I've noticed are that I generally get better looking V curves on faster/shorter FL systems (even the same system with a big reducer) and that just being in the CFZ is not good enough. I've always assumed that the first effect is due to the greater impact of seeing at longer FL.
I'm finding I get better curves with the Paracorr (increases the fl by 15%) compared to the MPCC however I'm keen to see if I can get the MPCC to work with autofocus.

Quote:
Originally Posted by RickS View Post
Have you thought about playing with an eval copy of FocusMax 4? (or one of the earlier free versions.) Might be an interesting experiment.
I'm under the impression focusmax does not have camera control built in?

"FocusMax works seamlessly with MaxImDL v5 and V6, CCDSoft V5, and TheSkyX Professional Camera Add On. (More platforms coming.)"
Reply With Quote
  #10  
Old 20-09-2015, 11:03 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Quote:
Originally Posted by peter_4059 View Post
I'm under the impression focusmax does not have camera control built in?

"FocusMax works seamlessly with MaxImDL v5 and V6, CCDSoft V5, and TheSkyX Professional Camera Add On. (More platforms coming.)"
Ah, yes. That would be a problem if you don't have one of those packages.
Reply With Quote
  #11  
Old 20-09-2015, 11:05 AM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Greg - thanks for the info. I can get great focus with the Bahtinov mask however I'm interested to hear if anyone is auto-focusing with a fast (f4) system. The more I read, the more convinced I am that what I'm seeing is what I should expect. The issue is SGP only performs a single focus run whereas Focusmax allows parameters to be averaged over multiple configuration runs up-front.

Last edited by peter_4059; 20-09-2015 at 12:20 PM.
Reply With Quote
  #12  
Old 20-09-2015, 11:19 AM
DavidTrap's Avatar
DavidTrap (David)
Really just a beginner

DavidTrap is offline
 
Join Date: Jun 2009
Location: Brisbane
Posts: 3,032
Surprised about needing longer exposures.

What Mag star are you aiming at? I centre up a mag 5-6 star and with luminance, my exposure times are ~1sec or shorter with Bin 1x1. I've not gone much longer for fear of saturating the pixels. This is an f7, 130mm refractor. I might have to try longer exposures and have a good look at the histogram.

DT
Reply With Quote
  #13  
Old 20-09-2015, 11:22 AM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Hi David - what software are you using to focus? SGP autofocus doesn't slew to a bright star - it calculates the average HFR of the field that is being imaged.
Reply With Quote
  #14  
Old 20-09-2015, 11:31 AM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Quote:
Originally Posted by DavidTrap View Post
Surprised about needing longer exposures.

What Mag star are you aiming at? I centre up a mag 5-6 star and with luminance, my exposure times are ~1sec or shorter with Bin 1x1. I've not gone much longer for fear of saturating the pixels. This is an f7, 130mm refractor. I might have to try longer exposures and have a good look at the histogram.

DT
Hi David,

With SGP you stay on the target (do not slew to a bright star) and that's why longer exposures are often needed, in particular with narrowband. SGP also looks at the entire image and calculates HFR for many stars (bright and dim) and uses that for calculating optimal focus. That's why I think you do not need many focus runs with SGP, as in each focus-step it uses an average for hundreds of stars.

A few times I rerun autofocus routine to check if SGP calculates the same focus, and it always has been to within one or two focuser steps.
Reply With Quote
  #15  
Old 20-09-2015, 02:14 PM
Peter.M's Avatar
Peter.M
Registered User

Peter.M is offline
 
Join Date: Sep 2011
Location: Adelaide
Posts: 947
I got in contact with the developers of SGP about this issue when i first started setting it up. The problem you are having i think is because your step sizes are too small. Try setting it to 20 or 30 between steps. The software does not choose the focus point from the minimum HFR of a single point, it calculates the intercept of the straight lines made from each half of the V. It is not necessary to actually sample the correct focus point in your V curve. As soon as i increased the step size I got reliable focus every time.
Reply With Quote
  #16  
Old 20-09-2015, 03:08 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
I have an f4 with a Moonlite and use SGP. After playing around a lot, finally settled on 9 steps of 5 pulses, smart focus disabled and darks enabled (essential). Usually use 5 second exposures for lum/RGB and 10 for NB, but occasionally may need to double exposures if the star field is not very strong - all 1x1 and focused through the current filter. Also get more reliable results by setting the number of stars to 15 with the above settings (it works better if it does not have to try to focus on dim noisy stars). I refocus every 3 frames in poor seeing and every 5 frames in good - usually use 3-5 minute subs, so this is about 1/4 to 1/2 hour between refocus runs.

These settings work OK for my system in seeing up to about 4 arcsec FWHM - if it is worse than that, the focus may sometimes drift off through outliers, but I do not image much if the seeing is that bad. If need be, increasing the exposure time can generally get things back under control in really bad seeing and, as a general comment, seeing variability is always the issue if there is a problem - never mechanical precision.

overall, the focus routines in SGP are not quite perfect, but they still work really well in good seeing and do as good a job as I could - and I can be asleep when they do.

maybe also try without the MPCC. these probably introduce significant aberrations, which may possibly increase the width of your CFZ

Last edited by Shiraz; 20-09-2015 at 03:22 PM.
Reply With Quote
  #17  
Old 20-09-2015, 03:16 PM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Thanks for the info Peter and Ray. Interesting Peter is doing quite large increments whereas Ray uses quite small ones. On my system 20 steps would be 80 micron which is double the theoretical CFZ whereas 5 steps would be half the CFZ. I'd love to have a look at your SGP logs if you have one handy that includes an autofocus run.
Reply With Quote
  #18  
Old 20-09-2015, 05:04 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
FWIW, here is a plot from last night's log showing an event where refocus did something useful - started at 151 and ended at 144 - as good as I could have done. Maybe I could increase the step size a bit in conditions like this, but it still worked well.
Attached Thumbnails
Click for full-size image (sgp.jpg)
45.7 KB60 views

Last edited by Shiraz; 20-09-2015 at 05:16 PM.
Reply With Quote
  #19  
Old 20-09-2015, 06:10 PM
Slawomir's Avatar
Slawomir (Suavi)
Registered User

Slawomir is offline
 
Join Date: Sep 2014
Location: North Queensland
Posts: 3,240
Hi Ray,

It looks to me that you should double your step size.

Just one thing with autofocus in SGP- you need to fairly close to a good focus for the routine to work well.
Reply With Quote
  #20  
Old 20-09-2015, 07:39 PM
peter_4059's Avatar
peter_4059 (Peter)
Big Scopes are Cool

peter_4059 is offline
 
Join Date: Jun 2007
Location: SE Tasmania
Posts: 4,532
Ray,

Thanks for sharing. Here's a few of my better v-curves. Although the curve looks ok and correlates quite well with CCD Inspector FWHM, the real test for me is the look of the stars and the Bahtinov mask results.

Peter
Attached Thumbnails
Click for full-size image (FocusPlots1.JPG)
120.1 KB57 views
Click for full-size image (FocusPlots2.JPG)
118.4 KB46 views
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 05:52 AM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement