Predictive PEC - Predictive Gain and Reactive Gain

[endless_sky]

Hello,

I am trying to optimize certain parameters, especially for RA guiding (I can get good enough DEC guiding), since it's all over the place.

I am using a Sky-Watcher NEQ6 mount and I get a total RMS error anywhere between 0.6-0.7 arc-sec to 0.8-0.9 arc-sec, even a few minutes later, just by slewing to a different (close-by) target.

Here is a log file: https://openphdguiding.org/logs/dl/PHD2_logs_9YQ5.zip

Guiding session #1: 0.7 arc-sec (discarding the guiding assistant part)

Guiding session #2: 0.9 arc-sec (discarding the guiding assistant part)

I know my mount is capable of doing better, as I have seen streaks of 0.5-0.6 arc-sec for 30-40 minutes at a time. So I think it's just a matter of finding the correct settings and enable it to do that for the whole duration of the imaging session.

Now the question: I understand while using Predictive PEC algorithm for RA, I have two components to the corrections being applied: a predictive correction and a reactive correction. The amount of those are defined as percentages, in the appropriate fields (predictive gain and reactive gain).

What I would like to know is if there is a way of knowing, when a correction is applied in RA, how much of it was due to the prediction and how much due to the reaction?

It would be nice, for example, if the correction bars that are displayed in the guiding graph would be solid for the component due to the prediction and clear (as they are now) for the component due to the reaction. This would be extremely useful in fine tuning the predictive gain and reactive gain, in case of over or under-corrections being applied.

Right now I don't know which one to touch, in case my guiding is over-correcting or under-correcting, since I don't know which of the two issued the correction (the predictive or the reactive?)

If this can be done by analyzing the data in Excel or with the Log Viewer, can somebody point me in the right direction on how to split the components of the corrections so I can adjust those settings? Of course, doing it like this would mean seeing the outcome in the next session, after analyzing the data - it would still be very nice to be able to see it at a glance by taking a look at the correction bars, during the guiding session and apply adjustments as needed.

Thank you for your time and input!

Matteo

[bw_msgboard]

Hi Matteo.  I think you’re off on a tangent here, diddling around with these parameters isn’t going to handle your RA guiding limitations – the limitations are in the mount.  You’ve got something like a 15-25 arc-sec uncorrected periodic error, seen here:

 

 

 

 

This is happening at the native worm period of around 480 seconds, and the PPEC algorithm is doing a pretty impressive job of controlling it.  But you also have the unwelcome and well-known 120 sec tracking error that is probably limiting your RA guiding at this point.  If you can improve the long-period periodic error via periodic error correction software, you can then force the PPEC algorithm to focus on the 120-sec oscillation.  Other users of this mount have been somewhat successful in doing that.  It requires setting the periodic length parameter to 120 sec and turning off the auto-adjust option.  

 

Getting back to the reactive and predictive aggressiveness controls, these rarely need attention and they aren’t going to help with your problem.  When you think you see a change in the guiding after fiddling with them, you are probably getting fooled by seeing (which can change on short time scales) or mechanical variances in the RA drive system.  Trying to show these things in the UI would be a real mess and not useful – they are vector sums, so they can’t be represented by a visual overlay approach.  If you want to examine performance at this level of detail, you’ll have to extract the information from the debug log files – you can do text-searches on ‘PPEC rslt’.  But I think you’d be better off taking a different approach.

 

Good luck,

Bruce

 

---

--
You received this message because you are subscribed to the Google Groups "Open PHD Guiding" group.
To unsubscribe from this group and stop receiving emails from it, send an email to open-phd-guidi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/6db9cf30-3f73-407f-ac4c-a2590cfbef23n%40googlegroups.com.

[Steve]

Hi

Make certain that the mount is mechanically perfect with new bearings then cleaned and re-greased.

Move the motor driven gear a tooth at a time relative to the idler gear until the 122s peak minimises. Careful though with the peak at 10s which is also a function of the position of the gears mentioned. IOW, a compromise.

Set the PPEC period fixed at 122s.

The only way I've found which works.

HTH.

--

[Matteo Gaetan]

Thank you Bruce and Steve for your responses.

I'll take a look at the debug log, I thought all the guiding commands were in the guide log. I'll give it a look just for curiosity.

Unfortunately, my mount, being an old version, doesn't support permanent PEC. All I could do was, when I was using EQMOD, train PEC and play it back through EQMOD, provided I always parked the mount before switching it off. But my Windows laptop died and now I am using KStars/EKOS on a Raspberry Pi. These don't have the same semi-permanent PEC capabilities that EQMOD had. In fact, they only have a predictive one just like PHD2.

So I don't have the option of running PEC with another software and using a different period for the PHD2 predictive one.

Or could I run two PHD2 at once? Just kidding... Or maybe it can really be done? Worth to ask, I suppose.

Unless there is a software I don't know of, that allows to play a PEC recorded curve, while PHD2 works on a different period.

Any ideas?

Also, thanks Steve for the suggestions about making the mount work better. Can you point me out to a website/tutorial for disassembling and correcting these gear meshes? I would very much appreciate it!

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAB7Jd2Kyw0Ap0zYoFcBMV2jOspa1q96HHKPadAf8tq2SeP5AEg%40mail.gmail.com.

[Brian Valente]

Hi Matteo

>>>>Unless there is a software I don't know of, that allows to play a PEC recorded curve, while PHD2 works on a different period.

That's pretty much how mount-based PEC and PHD's PPEC work today, on every mount

The mount-based PEC programs the mount to reduce periodic error, it's a one time fixed correction. 

PPEC is dynamic - it measures and estimates periodicity of errors, and then applies them proactively. 

If you are really interested in seeing the difference, there are a couple of approaches you could take:

1. do some guiding, then evaluate the first worm period vs. later worm periods. During the first worm period, the PPEC algorithm doesn't apply any predictive values, so it's primarily hysteresis (i.e., reactive). a few worm cycles later, it's tuned up it's predictive algorithm and should show improved guiding

2. you can do some guiding where one session is predictive only ( predictive weight 100%, reactive weight 0%) and then a separate session where it's reactive only (predictive 0%, reactive 100%). If you do them back to back with as identical conditions as possible (same area of sky, same side of pier, same seeing conditions, etc.) you might be able to see some differences there

This is in theory only, there are many other variables, i don't know how illuminating it would be, but some ideas there for you 

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAOaJL4zDL0xcyPC5HxfF9YcRg86DGrYBRvccBiyqFx0HA2S5Uw%40mail.gmail.com.

--

Brian 

Brian Valente

portfolio brianvalentephotography.com 

[Matteo Gaetan]

Thank you, Brian.

I have been using predictive PEC since I have started autoguiding - about ten sessions ago... Not long, I admit.

But I like the results. Like you said, I tried hysteresis only, and it gave me worse results.

Unfortunately, for the reasons I stated above, I cannot use PEC directly at the mount and take advantage of the predictive PEC of PHD2 at the same time. I understand that together they would produce better results. Even more so if they could target different periods.

Since the 120s period is so common with NEQ6 - EQ6R Pros, would it be possible for PHD2 to register more than one prediction and use one for the 120s period and the other one for the 480s (summing or subtracting the necessary corrections as needed)?

Another alternative would be to have a software that can play a recorded PEC over a few periods (creating the curve should be possible using PHD2 and PEMPro, for example) alongside with PHD2 doing the predictive part on another chosen period.

The problem is I don't know if such software, outside of EQMOD (which doesn't work under Linux/Raspberry), exists.

Other than acting via software, I guess the only possible approach is doing some mechanical tuning as suggested by Steve - but in order to do this, I really need to find some guidelines on how to open the mount and where to act.

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAJa45i7XgVz7Aakw_wU5qB8jufwa%2BMORqQZWx%3Dy0C-JCWWm8Zg%40mail.gmail.com.

[Brian Valente]

>>> Since the 120s period is so common with NEQ6 - EQ6R Pros, would it be possible for PHD2 to register more than one prediction and use one for the 120s period and the other one for the 480s (summing or subtracting the necessary corrections as needed)?

no, it doesn't really work that way. It evaluates and creates a predictive periodic error correction for whatever it sees as periodic,including calculating the period if you enable that (this is a rough explanation, but there aren't multiple cycles or periods that you specifically target).

If your primary constraint is the 120seconds, you can set your period for 120 seconds and disable the auto adjust period. 

that way you can specifically target that period. Larger periods like 480s are usually easier to guide out using regular hysteresis.

I had a similar situation with a mount that I used mount-based PEC to eliminate the primary (and longer) worm period, and then used PPEC targeting a 41 second period. It was quite effective. This is slightly different from your situation because you can't use PEC but i still think it's a strategy worth trying if the 120sec is your main concern

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAOaJL4xafoF1hdrPR%3DCum4xRnuz20QYdr1D2Q7Y02P2fCckAqg%40mail.gmail.com.

[endless_sky]

Thank you, Brian - understood.

Well, in order to target the most offending period, I just collected some data using Frequency Analysis on the raw RA data from some guiding sessions within the past 6-7 times I went out imaging.

Here are the results (peak to peak, in arc-sec) for the periods 10s, 120s, and 480s (roughly)

10s: 0.3 0.3 0.3 0.3 0.3 0.1 0.1 0.5 0.1 0.5 0.5 0.5 0.4 0.2 - average = 0.3

120s: 1.2 0.7 1.1 1.0 0.8 0.8 0.8 1.4 2.2 2 1.7 1.8 1.7 1.8 - average = 1.4

480s: 10 6 8 8.5 8 8 8.4 12.5 18 14 15 16 14 15 - average = 11.5

As you can see, in some of those sessions the 480s period peak to peak was actually quite good (8-10 arc-sec) and some of those where 3 hour long, uninterrupted guiding sessions. Then it started going towards the 12-15 arc-sec. And I didn't change anything to the mount or the gear used.

Anyway, could you tell me if this is the right approach (using the averages):

0.3 arc-sec peak to peak for 10s, means 0.03 arc-sec/sec

1.4 arc-sec peak to peak for 120s, means 0.012 arc-sec/sec

11.5 arc-sec peak to peak for 480s, means 0.024 arc-sec/sec

It would seem the most offending one is the 10s period, followed by the 480s, while the 120s has little to no contribution (it's 2-2.5 times smaller than the other two). Am I understanding this right?

If I am, I should actually tackle the 10s period, or keep at the 480s.

If, however we take the highest offending numbers for each:

10s: 0.5/10 = 0.05 arc-sec/sec

120s: 1.8/120 = 0.015 arc-sec/sec

480s: 18/480 = 0.0375 arc-sec/sec

The 10s still seems to be the worst, followed by the 480s. Should I tackle the 10s or it's so high frequency that it's contribution would likely not even matter? I use 2 to 3s guiding exposures, anyway, so it would only span 5 or 3 frames. But, other than that, I don't really see a benefit going after the 120s period, since it seems the smaller of the three. It's worth a try, though. Also because 480s is a multiple of 120s, so maybe the longer period would also benefit from minimizing the shorter one.

Thanks again and sorry for the lengthy post, but I like playing with numbers and making things better, if I can.

Matteo

[Steve]

Hi

http://www.astro-baby.com/EQ6%20rebuild%20guide/EQ6%20Strip%20Down%20Home.htm

When I used the acronym PPEC, I was referring to the Predictive version which is available in phd2. 

I got nowhere with the old vspec playback pec. PHD2's PPEC just works and is the only way with the 122 second frequency.

Cheers

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAOaJL4zDL0xcyPC5HxfF9YcRg86DGrYBRvccBiyqFx0HA2S5Uw%40mail.gmail.com.

[Matteo Gaetan]

Thank you very much, Steve.

For the link and the tips.

So, there's hope, even without software PEC. Unfortunately, checking clear sky app, I won't have another suitable night for at least a week...

But, as soon as I'll have an opportunity, the plan is to test with the 120s period. I'll do some guiding with the 480s, at the beginning, just to have something to compare it to, then I'll switch and see if I get an improvement.

I'll report here my findings.

Since I don't have the luxury of leaving the auto-adapt period on (otherwise PHD2 will resort back to the 480s, correct?), any chance you can tell me what the 120s depend on so I can calculate the exact number to put? Or is 120 good enough?

Thanks again!

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAB7Jd2KHg-v%3DSJTrq1h44U6Q00JpyACLnP_SdsLex9QD7bMPWg%40mail.gmail.com.

[Matteo Gaetan]

I meant 122s, sorry.

[Steve]

Fix the RA period in PHD2 to 122s. Read, 'fix'. 

Cheers

To view this discussion on the web visit https://groups.google.com/d/msgid/open-phd-guiding/CAOaJL4wkdh_9aohve9bmA_m61OqDV%2Bb1grr68iBn5DvY%3D%3DJ8CA%40mail.gmail.com.

[endless_sky]

So, it appears my mount (or PHD2) just doesn't like the California Nebula...

Thursday night (the 12th) was clear enough for me to go out and do some tests - I also just received a few spacers and I needed to test for backfocus optimal distances, anyway.

So, I ended up trying some guiding, too. In a matter of 15 minutes, I went for my best ever guiding (0.66") with a 480s peak-to-peak PE of 25", to my worst ever (1.23"), still with a 480s peak-to-peak PE of 20+".

What changed? I slewed from M31 to the California Nebula. I then tried 122s for the fixed period and got 1.03". Went back again to 480s (self adjusting) period, and got 1.04".

PE peak-to-peak is around 20" for all these cases. See attached guide log file.

I know the guiding sessions are short, as clouds were rolling in and I didn't have much time, but the results are pretty consistant with my previous sessions. IC 1805, M33 - only two subjects beside the California - I shot so far with auto-guiding, were all between 0.7-0.8" total RMS, with some excursions into the 0.5-0.6". But every time I imaged the California, I couldn't go below 0.8-0.9".

I thought it was probably caused by the worm gear meshing with a different part of the RA gear, but the peak-to-peak PE errors of around 20" in pretty much each guiding session seem to state otherwise. If I had better portions of the RA gear, those peaks should get better in certain areas and worse in others.

Anyway, changing from 480s to 122s didn't seem to make much difference in this particular scenario, but I sure will try again and for a longer session - weather permitting.

I also bought an amperometer and I plan to check current absorbtion while the RA and DEC axes are (independently) spinning, with the mount as is now. Then I'll try tightening the gears as explained by Astro Baby tutorial, until I can minimize every play the mount can possibly have, while continuing checking the current and making sure I don't go overboard (to avoid binding / motor stalling). Then I'll see if my guiding improves.

Anyway, my current image scale is 2.08 arc-sec/pixel, so I should still be good while guiding below or around 1" total RMS. However, I like it much, much better when I can get it closer to 0.5!

Guide log: https://openphdguiding.org/logs/dl/PHD2_logs_38gw.zip

Thanks for looking,

Matteo