Happy Friday Folks,

Here's another Ekos Focus video, this time on Focusing using FWHM

Happy focusing!

Hi John,

Another very interesting video. Thank you for this !


I have one small suggestion: the sound volume of the recording is rather quiet. Maybe it's my laptop, but I have to set both the slider in Youtube and the system volume to maximum. So it may be useful to turn the audio recording up a little for a next video.

Anyhow, thank you again for your beautiful work !

I actually recommend everyone to watch John's videos. They contain an incredible amount of information, allowing you to understand the focusing process much better and achieve fantastic results.

Thx for the feedback, I'll try to make the sound better next time.

I'm already looking forward to the next video!

Enjoy the weekend!

Hi John,

I was actually wondering how the focuser module handles backlash.

During a new focus attempt, the focuser first takes a large number of steps in one direction and then returns in smaller steps to the other direction, taking and analysing a light every time. In the first step back, the backlash will be eliminated and this is often clearly visible on the parabola/hyperbola as a 'bad' measurement.

After finding the "perfect focus point" after all iterations, the focusses moves towards that point.

But I would expect the focuser to first pass that point a bit and then return to the "perfect focus point", this to eliminate the backlash during the final and very important step of the focussing process. Is that the case, or are other smart techniques being used?

So lets say you measure focuser backlash and its 100.

If you set AF Overscan > 100, e.g. to 150, then on each outward movement, be it the first move of an Autofocus, last move of an autofocus, filter offset movement, adaptive focus movement, etc etc, Focus will add 150 steps to whatever the desired move was, for example an outward move of 57 steps, so a total move of 207 steps. When the focuser reports the move complete, Focus will command a move inward of 150 steps. So the end result is a move to the desired position be always ending from outward to inward.

So on an Autofocus run, there are 2 "outward" steps: the first and last ones.

If you set the Backlash field (instead of AF Overscan) then the driver will adjust the focuser movements. Depends on how the driver is coded as to exactly how it works. Problem of course is that if backlash is 99 and you set 100 then you'll be over compensated by 1 tick and if backlash is actually 101 then you'll be under compensated. This is why I usually recommend AF Overscan.

I have thought about trying to calculate backlash automatically without needing the user to specify a value. Its possible by moving out a large step, taking a frame, calculating the HFR, moving in a few ticks and repeating until the HFR changes significantly. The number of focuser movement steps before the HFR moves "significantly" would be the backlash. When I tried this though I didn't find it to be very reliable because its difficult to know if a change in HFR is due to backlash being taken up, or just a random fluctuation in HFR, so I didn't pursue it.

Very clear explanation! Thanks for that!

Till now I did use an overscan of 1,000, but considering the Esatto focuser takes 731,000 steps over a travel of 25 mm, that is obviously far too little.

I increased the overscan setting to 10,000 and just took a series of 180 sec lights of the Pacman nebula. Quickly calibrated & integrated these lights, followed by a minimal post-processing. The result is clearly that the stars are much sharper than before including more detail in the nebula itself.


You made my day! (euhhh night)

Hi John,

After watching your interesting videos, the focusing process became much clearer to me and after adjusting the overscan setting my photos sharpened noticeably. In Pixinight, the measured FWHM has gone from an average of 4.7 to an average of just under 4.1 and less.

These are the settings I'm using now:




The large number of steps has to do with the Esatto focuser, which has a (theoretical) resolution of 0.04 microns. Of course, the mechanics do not achieve this in practice. I think the backlash is even larger than the 10000 steps I have now set as AF Overscan, but 10000 is the highest value I can enter.


So 2 question raised:

1. Would it be possible to increase the upper limit of the AF Overscan input?
2. Makes it sense to enter 10000 in the Driver Backlash field & 10000 in the AF Overcan field? Are they added while performing a focus routine?


Any feedback or advice is greatly appreciated!

Hi Fitchie,

To answer your questions:

1. Yes - no problem. Question is what to set it to?
2. This should work. Other folks have done this successfully with other focusers. I had a quick look at the Esatto driver and couldn't see any obvious reason why it would be a problem.

The driver backlash is limited in the driver to 10000.The Esatto is supposed to be low backlash and the driver has a limit of 10000 backlash steps. I know other folks are successfully using the Esatto with Ekos so are you sure you have >10000 steps of backlash? Are you sure for instance that the Esatto isn't slipping?

A check you could try is to take an autofocus run like this one...


How are the first datapoints on the right? I circled points 1, 2, and 3. If there was backlash these would be "flatter" because the inward steps are still taking up backlash and not moving the focuser so the HFR will not be decreasing.

If you then hover the mouse over the yellow circle you get the tooltip box containing the position and HFR, in this case 2.18. Now look at the HFR box under the v-curve (highlighted). This is the HFR of the last frame taken, in this case Focus takes a frame when its moved to the solution position. In a perfect world these 2 numbers will match (2.18). Its just luck that they exactly match. Random noise means they usually differ a bit. But the point is that last frame (taken at focus) should be near the minimum HFR on the graph. If its not, then that's an indicator of possible backlash.

If you set backlask=0 and AF Overscan=0 and try the above you'll get an indication of how much backlash you have. If you post up a picture of that it would give me an indication of how much to increase the max AF Overscan by.

Hi John,


The HFR for the yellow circle is on average double the last HFR.
Please see attached screen dumps for several Driver Backlash & AF Overscan settings...


#1
Driver Backlash = 0 & AF Overscan = 0




#2
Driver Backlash = 0 & AF Overscan = 0




#3
Driver Backlash = 0 & AF Overscan = 10000




#4
Driver Backlash = 0 & AF Overscan = 10000




#5
Driver Backlash = 10000 & AF Overscan = 10000




#6
Driver Backlash = 10000 & AF Overscan = 10000

Thanks for running the test and sending the screenshots FItchie. I was assuming you were using HFR but FWHM works just as well here - same principle. I guess given the title of this thread using FWHM makes a lot of sense!

Backlash = 0; AF Overscan = 0. Looks like there is a small amount of uncorrected backlash as the yellow dot is above the curve.

Backlash = 0; AF Overscan = 10000. Looks to me like you have no noticeable backlash with those 2 runs.

This implies your backlash is <10000.

Looking at run 1. Points 1 and 2 of imply some backlash (although star detection issues can produce the same effect). Maybe around have a step size or 2,500.
Run 2 implies no real backlash.

I would say from these curves your backlash is < 10000 so setting AF Overscan to 10000 should work fine. You could try tuning backlash down a bit. If say your backlash is 2500 then have Overscan set to 10000 works fine but you're moving the focuser a lot further than you need and wasting a (tiny) bit of time.

I don't see any benefit to allowing larger AF Overscan values from these graphs.

Hi John !


Thanks for your info. The more I read about it, the more interesting it becomes. I never actually thought that there was so much involved in focusing. What you are doing in the Aberration Inspector (back focus & tilting calculation) is absolutely awesome !

That being said, I had indeed understood that the value in the yellow circle refers to HFR, but it is the value of the chosen measurement technique, in my case FWHM. And then the graphs and numbers do make sense. Perhaps a suggestion: replace VAL in the tooltip box in FWHM or HFR, depending on the chosen technique, then it will be immediately clear.

As mentioned before, my stars are noticeably sharper and there is more detail in my photos since I adjusted the backlash/overscan setting, but I still think I should get better results. EKOS indicates that my FWHM is somewhere around 3.1 - 3.6, when I measure the lights in PixInsight I get noticeably different (noticeably less good) values.


FWHM:




Eccentricity:




How is the FWHM calculated in EKOS?
Does this happen on unstretched images and is the subframe scale of the camera taken into account or is it only at pixel level?