Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Filter offsets and focus variance

Astrophotography
  • Please log in to reply
12 replies to this topic

#1 klaussius

klaussius

    Apollo

  • *****
  • topic starter
  • Posts: 1,418
  • Joined: 02 Apr 2019
  • Loc: Buenos Aires

Posted 17 September 2021 - 12:14 PM

I have a question better suited for the more experienced folks over here I think.

 

I'm trying to define my workflow wrt filter offsets. I've mostly been getting by with no filter offsets because all my filters have been parfocal and my optics reasonably CA-free (using a newtonian), but since I will soon start doing some imaging with a refractor I thought I should start thinking about filter offsets.

 

I get all the theory and how to do them, what I'm not sure about is how to deal with seeing and other sources of variability when "training" those filter offsets.

 

If I want the best possible focus, I would want to compute very precise filter offsets. The way I'm planning to do so, is by doing several AF runs on each filter and taking the median focus position on each filter. Since filter offsets should never change for a particular system, I can spend the time to do it precisely once. That's all fine and dandy there.

 

Now I'm left with a set of offsets that sometimes are very small (my filters are, after all, parfocal), so what's the best course of action here? The process above reduces variation by external factors, but it won't entirely eliminate it.

 

If I make my focuser shift very small amounts between filters needlessly, I'm thinking I'll be subject to focuser drift due to backlash and slippage. I don't usually see much effect from any of those effects under normal imaging conditions, but if I switch filters and apply filter offsets often I bet that could change.

 

Should I ignore very small filter offsets? How do you define how small an offset is too small an offset? I've seen people compute the CFZ in microns, but I have no idea how to translate that into focuser steps.

 

On a related side question, how do you deal with AF variability? I've noticed successive runs tend to vary the resulting focusing position, understandably, do you just let it vary randomly when doing AF runs periodically or do you have some sort of procedure to weed out outliers?

 

Has anyone given these issues any thought?

 

PS: I should add, for context, that I'm writing my own software, which is why I ask these kinds of questions. I'm trying to define how the AF module should behave. I'm not using NINA or any other common software, though I've checked how they do things for inspiration of course.


Edited by klaussius, 17 September 2021 - 12:19 PM.


#2 ryanha

ryanha

    Viking 1

  • *****
  • Posts: 934
  • Joined: 05 Aug 2020

Posted 18 September 2021 - 12:17 PM

I have thought about these but don't have much by way of experimental data.

 

For focuser drift I think if you take a good clear night and in the middle of the night (once temps and darkness stabalizes) aim high in the sky and do 10 AF runs in a row, that will give you a sense of your variability through experiment.  So you can see what your standard devs are and then see how many std devs actually change your focus metrics (HFR, etc.)

 

For backlash, you can use an overshoot technique or just measure the backlash if it is non-uniform by direction.

 

When you do your "one night" of getting actual AF values, you might rotate through them to minimize nightly seeing drift (e.g. RGBRGBRGB vs RRRGGGBBB).

 

Lastly, once you have a set of offset computed, you might periodically do an AF run after filter change/offset application to get a measurement of your delta and then track whether that is outside of your std dev of errors.  (like you could start by doing this every 10 runs, and then back it off as you build confidence).

 

Either way, publish your results for sure!!

 

As an aside, if you are writing the software yourself, I have thought it would be interesting to do a machine learning based approach where you also take into account things like HFR delta frame by frame, and temperature change, and time since last AF).  Wouldn't that be cool?  Then you could eventually train the system to decide when to AF and what the offset should be.

 

 

--Ryan


  • klaussius likes this

#3 TXDigiSLR

TXDigiSLR

    Mariner 2

  • -----
  • Posts: 265
  • Joined: 02 Jan 2020
  • Loc: Dallas, TX

Posted 19 September 2021 - 12:50 AM

I use NINA and ALL of this is easily automated.   The new plug in for Filter Offsets runs through everything and does all the computations.  I set the sequencer to autofocus on every filter change.  I found with the filter offsets all set up, my autofocus routine runs quicker than before and I have a bit more imaging time.

 

I do notice that my AF numbers are NEVER identical from night to night.  Since everything is automated, as long as its close to start, everything goes quickly.



#4 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,332
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 19 September 2021 - 07:39 AM

The critical part in filter offsets (IMO) is backlash compensation.  If you haven't tackled that yet, start there.   As mentioned, the safest bet (again, IMO) is to make all moves end moving inwards, i.e., overshoot for outward moves, and then come back.

 

With a refractor, your filter offsets will be larger, as they are not defined by filters being parfocal, but by the focus curve of the optics.  And even good ones do have that... 

 

For measuring I'd pick a full moon night with reasonable seeing [u]and constant temperature[/b].  Then do several runs going through all filters, so the comparative averages are over the same ambient conditions.

 

As for how accurate - check your focus curve close to the focus, again best with good seeing.  That should show you how many focus units deviation are acceptable.  E.g., for my Sharpstar I can 'see' 10-15 units - focus shift for RGB is in the ±55 range.  The Stowaway isn't much different.

 

I personally refocus on filter change anyhow, as I take my data in large® blocks.  I still use filter offsets though, as the routine I use works better/faster when starting close to focus.


  • klaussius likes this

#5 klaussius

klaussius

    Apollo

  • *****
  • topic starter
  • Posts: 1,418
  • Joined: 02 Apr 2019
  • Loc: Buenos Aires

Posted 19 September 2021 - 12:38 PM

For measuring I'd pick a full moon night with reasonable seeing [u]and constant temperature[/b].  Then do several runs going through all filters, so the comparative averages are over the same ambient conditions.

I did just that with my newtonian, which is the one I use the most, and the one I know the most, so I can judge data quality from experience.
 

I did 3 runs on each filter with 7 points, 15s exposures. The whole routine took about 1h 45m for all 7 filters shocked.gif

 

I got the following offset data:

{
  "filter_offset[L]": 0.0,
  "filter_offset[R]": -21.0,
  "filter_offset[G]": -5.0,
  "filter_offset[B]": 19.0,
  "filter_offset[S]": -23.0,
  "filter_offset[H]": -21.0,
  "filter_offset[O]": 18.0,
  "filter_offset_dev[L]": 0.4714045207910317,
  "filter_offset_dev[R]": 3.091206165165235,
  "filter_offset_dev[G]": 1.247219128924647,
  "filter_offset_dev[B]": 12.552113589175153,
  "filter_offset_dev[S]": 33.24989557210001,
  "filter_offset_dev[H]": 1.632993161855452,
  "filter_offset_dev[O]": 4.4969125210773475,
  "filter_offset_dev": 17.261072871585956
}

Clearly I had a few bad runs that threw the std dev way off. Unavoidable stuff since I shoot from my roof, which is shared, someone I think came and turned on all the lights because when I got there to check things up they were on. bangbang.gif

 

It seems that even the newtonian has some CA in the system, which is interesting.

 

Anyway, I fudged the numbers manually to fix that anomaly, and they seem to have worked really well.

 

I could not measure any backlash whatsoever. I will have to be careful with my refractor which may have some backlash.

 

I will do as suggested, make sure to overshoot when applying the filter offsets. That's easy enough to do.

 

As for how accurate - check your focus curve close to the focus, again best with good seeing.  That should show you how many focus units deviation are acceptable.  E.g., for my Sharpstar I can 'see' 10-15 units - focus shift for RGB is in the ±55 range.  The Stowaway isn't much different.

 

Sadly, I did not store the focus curves for all the runs. I should have.

 

But judging from manual tests, I can see a minimal change in FWHM and contrast numbers with moves as small as 10 steps. The change is tiny, but it's consistently there. The FWHM curve is rather flat for a longer span, but I know if I'm not right in the middle of the focus zone I get curvature/coma near the edges. The edges seem to be way more sensitive to focus position than the center of the frame.

 

Right after the above I did a capture on NGC 55. Not the best target for a full moon night, but it let me test those filter offsets.

 

Even though when I tested it I could see no backlash or focus slip, and focus seems to be highly reproducible in this setup (I can park the focuser and go back to the exact focus spot quite precisely), it seems it did shift across the night. I don't know whether it truly shifted or it's just the target going lower in the atmosphere (I started capturing near zenith, and ended at about 40deg alt):

 

fwhm_L.png

 

That's FWHM in pixels of L subs, which are shorter subs so less affected by guiding, wind and whatnot. And there was some wind.

 

I did not refocus along the session to keep the test simple. Normally I would. Perhaps next time I will, to verify whether that FWHM increase is coming from focus or the atmosphere.

 

Average FWHM across all channels was similar enough:

 

  • L: 3.39504762021
  • R: 3.26168274124
  • G: 3.43056085689
  • B: 3.42531326975
  • H: 3.50410785882


#6 Wilsil

Wilsil

    Mariner 2

  • -----
  • Posts: 248
  • Joined: 09 Jan 2021
  • Loc: Perth, WA AUSTRALIA

Posted 20 September 2021 - 08:29 PM

Yesterday I had my scope back with a new adapter for my Esatto.

As it is full moon and clear skies, it was a good time for testing.

So I tried to focus in L first which worked.

Instead of the focuser sitting halfway at 7.5mm is was at 5.64mm. Not too bad.

Then I tried Ha, but couldn't get anything.

What I am reading here is that Ha is moved inward compared to L, is that correct?



#7 klaussius

klaussius

    Apollo

  • *****
  • topic starter
  • Posts: 1,418
  • Joined: 02 Apr 2019
  • Loc: Buenos Aires

Posted 20 September 2021 - 08:38 PM

Yesterday I had my scope back with a new adapter for my Esatto.

As it is full moon and clear skies, it was a good time for testing.

So I tried to focus in L first which worked.

Instead of the focuser sitting halfway at 7.5mm is was at 5.64mm. Not too bad.

Then I tried Ha, but couldn't get anything.

What I am reading here is that Ha is moved inward compared to L, is that correct?

Yes, but:

  1. It's a very small amount
  2. It's because the LRGB filters are of a slightly different thickness compared to the SHO filters. Barely 0.1mm difference, hardly noticeable except to the AF routine.

Not sure which are your filters, but when I had a different set of filters with larger thickness differences, LRGB and SHO focused quite a ways apart. About 200+ steps away.

 

Back then, I was using a ZWO LRGB 1.25" set and Antlia 3nm SHO set.

 

Now I'm using the same antlia filters with Optolong LRGB filters, which are 2mm thick, much closer to the Antlia set.

 

So it depends on which filters and which filter thickness you're using.



#8 Wilsil

Wilsil

    Mariner 2

  • -----
  • Posts: 248
  • Joined: 09 Jan 2021
  • Loc: Perth, WA AUSTRALIA

Posted 20 September 2021 - 08:57 PM

Thanks klaussius.

I have Optolong LRGB, Antlia 3nm Ha and borrowed Orange O3 and S2.

My focuser step size was calculated as being 6000-7000, so I used 6000.

The curve for Lum wasn't too bad.



#9 klaussius

klaussius

    Apollo

  • *****
  • topic starter
  • Posts: 1,418
  • Joined: 02 Apr 2019
  • Loc: Buenos Aires

Posted 20 September 2021 - 09:06 PM

Well, bear in mind that Ha needs longer exposures, which is why I was using 15s exposures. For L it's overkill, but SHO needs such long exposures to get accurate readings.



#10 Wilsil

Wilsil

    Mariner 2

  • -----
  • Posts: 248
  • Joined: 09 Jan 2021
  • Loc: Perth, WA AUSTRALIA

Posted 20 September 2021 - 09:15 PM

Agree, I used 5 seconds for Lum and initially 30 seconds for Ha.

But even 120 seconds didn't show anything. Something is not right.



#11 klaussius

klaussius

    Apollo

  • *****
  • topic starter
  • Posts: 1,418
  • Joined: 02 Apr 2019
  • Loc: Buenos Aires

Posted 20 September 2021 - 09:22 PM

Agree, I used 5 seconds for Lum and initially 30 seconds for Ha.

But even 120 seconds didn't show anything. Something is not right.

 

Clearly. 30s should be more than enough, and your filters are similar to mine so they should focus rather close to each other. At least LRGBH.



#12 Wilsil

Wilsil

    Mariner 2

  • -----
  • Posts: 248
  • Joined: 09 Jan 2021
  • Loc: Perth, WA AUSTRALIA

Posted 20 September 2021 - 09:25 PM

That's why I think something is wrong with the Ha filter.



#13 Wilsil

Wilsil

    Mariner 2

  • -----
  • Posts: 248
  • Joined: 09 Jan 2021
  • Loc: Perth, WA AUSTRALIA

Posted 21 September 2021 - 09:18 PM

Happy to report that all went well yesterday.

I used the Dark Customs plug-in in NINA and ran the offset calculation sequence on a bright star near zenith, past meridian.

Focused roughly in Lum and started the sequence.

Every filter was focusing, even Ha this time. No idea what I did wrong the day before.

 

Attached Thumbnails

  • received_1223007618163859.jpeg

  • klaussius likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics





Also tagged with one or more of these keywords: Astrophotography



Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics