Predictive PEC: beta3

[Edgar Klenske]

Dear all,

thank you all for your time and your valuable feedback on beta2! Just in time for the weekend, I have finished the work on beta3 and I'm looking forward to see your results!

Changes (relative to beta2)

• smooth transition between learning phase and predictive guiding
  
• improved period length identification (+bugfix)
  
• speedup in the prediction process

Installing (and Going Back)

Andy produced installers for Windows and OSX that you can use as if you would install any other version. This beta release is based on the current PHD2 2.6.3. It shows "-GP_beta3" in the title bar, so that it is easy for you to see which version you are using.

Since some of the standard parameters have changed, please reset the parameters of the "Predictive PEC" algorithm by clicking on "Reset". Please enter your mount's period length (or a rough guess).

Please note that you'll have to switch back to your usual guide algorithm before reverting to the standard version of PHD2, otherwise you won't have any guiding at all.

The Algorithms

This beta release adds a new "Periodic PEC" guiding algorithm that adapts to the periodic error or your telescope and increases performance by using predictions. The "Periodic PEC" algorithm should work out of the box as drop-in replacement for any algorithm you're currently using for the RA axis. It even continues guiding based on predictions when the guide star is temporarily lost or the guiding is in pausing state (for refocusing, for example). Note, however, that slewing during pausing is likely to degrade performance, since it changes the worm phase.

This release also includes a new "Predictive Drift" guiding algorithm for the Dec axis. This guider isn't finished, yet, and shouldn't be used for testing so far.

Testing and Reporting Problems

We are interested in performance and usability, but of course also in bugs that may reveal themselves during testing. Therefore, please report any problems you encounter.

Note that the "Predictive PEC" algorithm needs to "see" about 2 periods of the periodic error until it reaches its full performance. Therefore, each test should cover at least 3 period lengths.

Depending on how much time you can spend on testing our new feature, please go through the following steps and send us your guide logs afterwards. You can use pausing and dithering as usual, you can even try to do regular exposures during testing. Just be aware that the performance might be worse than usual, potentially ruining your exposures. ;)

1) Switch your RA guiding algorithm to "Predictive PEC" and let us know if you like it.
2) Compare your usual algorithm to the "Predictive PEC" algorithm, report RMS errors.

If you know your mount's worm period (even only roughly), please enter this in the settings page before you start, but leave "auto" enabled so that the guider can adjust this number later.

Please report your mount type, the period length of the periodic error (if you know it) and send us your guide log (ekle...@tue.mpg.de). If anything breaks, please also send us your debug log, too!

Thank you for testing!

Known Issues

• The "Predictive Drift" guiding algorithm for Dec might not work at all. Please don't use it in this release since it could degrade the performance of RA guiding.
  

Further Information

If you are interested in the details of the algorithm or implementation, you can have a look at the scientific article (open access) or the developer documentation. If you have any questions, please don't hesitate to post them here or ask me directly via eMail (ekle...@tue.mpg.de).

Thank you for reading and testing!

All the best,
Edgar.

[Brian Valente]

Hi Edgar

great! hopefully it will be clear tonight for more testing.

should we turn on auto this time?

any other tips on how to adjust our use?

Brian

[Brian Valente]

Edgar I have one more question

 

If we know the periodic error for our mount, is there really any value to selecting the “auto” option for determining the PEC?

 

It seems like it could introduce more uncertainty and potential for error?

 

 

 

Thanks

 

Brian

 

 

Brian Valente

Brianvalentephotography.com

[Bryan]

Edgar

Can we tinker with the two parameters 'on the fly' or should we stop guiding, if we change one or the other?

Also,  I'm showing my ancient control theory here.  Is control gain roughly equivalent to Proportional in a PID controller?    Is the prediction gain equivalent to Integral or Derivative gain or neither?

Thanks!

Bryan

[Edgar Klenske]

@Brian, you're right. If you know your period length exactly, you're adding uncertainty when you're using the "auto" setting. So, for best performance, turn off auto and enter your period length.

However, I'd be interested in seeing whether the "auto" setting works, too. So, ideally, try it once with Hysteresis, once with PPEC (exact period) and once with PPEC (auto), if you have enough time to your disposal. We want this algorithm to work for all users in a "PHD" manner. Therefore it is important to try this out, too, so that we see potential bugs. :)

[Edgar Klenske]

@Bryan, if you want to play with the settings to find out what's best for your mount, you can of course do this on the fly. Let us know what you found out, then! For best comparability it will be of course be interesting to see a full run with constant parameters, since that's the usual case.

You're absolutely correct, the "control gain" is equivalent to the Proportional part in a PID controller. The prediction gain, however, is not present in a PID (which is a pure feedback controller). The prediction gain is part of a feed-forward path in this controller, where we try to compensate the disturbance (from the gear) in a lookahead fashion. If you're interested in the details, I have the links to the paper and documentation in the first post, and I'm happy to answer questions!

[Terry Fishlock]

Anyone know the period length for an Atlas mount?

Terry

[steve]

El 04/03/2017 a las 21:32, Terry Fishlock escribió:
> Anyone know the period length for an Atlas mount?

My eq6 is 479s. I believe the eqg is the same. HTH.

[Andy Galasso]

I added a page to the PHD2 Wiki. Let's use that to collect and share worm period information.

https://github.com/OpenPHDGuiding/phd2/wiki/Mount-Worm-Period-Info

Andy

[Vince]

Celestron CGEM worm period is 478.689.

Vince

[Andy Galasso]

On Sun, Mar 5, 2017 at 3:24 PM, Vince <vprof...@gmail.com> wrote:

Celestron CGEM worm period is 478.689.

Thanks, wiki page updated.

[djkr...@gmail.com]

Ioptron IEQ45PRO period is 336.6 seconds  (256 teeth in RA worm gear)

Dave

On Saturday, March 4, 2017 at 5:11:14 PM UTC-5, Andy Galasso wrote:

[Andy Galasso]

On Sun, Mar 5, 2017 at 3:53 PM, <djkr...@gmail.com> wrote:

Ioptron IEQ45PRO period is 336.6 seconds  (256 teeth in RA worm gear)

added, thanks

[Bryan]

When PHD2 beta3 starts, I get a registry error.  Clicking OK exits the error box and PHD2 seems to run OK.  This error occurs everytime I start beta3.  It was not seen in beta or beta2.

Screenshot attached

Bryan

[Vince]

I get the same error.  Using Windows 10.

Vince

[Brian Valente]

Mine installed and running just fine under windows 10

Brian

Sent from my iPad

--
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.
For more options, visit https://groups.google.com/d/optout.

[Edgar Klenske]

Hej Bryan and Vince,

thanks for pointing out this error! I think I could fix the bug right away. Andy updated the installers (thanks!).

Would you please download the new installer (same link) and see if it works?

Best,
Edgar.

[Vince]

New installer downloaded and installed, I get no error now, thanks Edgar.

Vince

[Brian Valente]

Edgar

this is a few days old but i did another run with Beta 3 and everything looked great. 

https://www.dropbox.com/s/py1bj5ze4ax2uro/PHD2_GuideLog_2017-03-03_194858.txt?dl=0

i played with the settings a bit, but I think you are correct that the default settings you suggested seem to be best for my setup. increasing aggressiveness seemed to create more guiding instability

I'm planning on moving to an ONAG soon, i wonder how that will impact PPEC guiding if at all

Brian

[Bryan]

Edgar

I also get no error now.

THANKS!

Bryan

[mar pal]

Hi Edgar, Thanks for working on this interesting algorithm.

How would this work with dithering?  I'd imagine it would cause some issues?

Martin

[Andy Galasso]

Hi Martin,

The new algo handles dithering and can also handle pausing guiding for focusing.

If you are interested in the details of how it does it, Edgar fill you in. But having tested it myself, I can confirm that dithering and pausing works great.

Andy

[Brian Valente]

Yes, dithering and pausing works just fine with PPEC algorithm, I can confirm that as well on my Losmandy G11

 

 

 

Thanks

 

Brian

 

 

Brian Valente

Brianvalentephotography.com

 

--

[dz8...@gmail.com]

Andy,  

For your database, the Paramount MyT has a worm period of 269.26 seconds.

[Andy Galasso]

Thanks Doug, table updated.

[rwstan...@gmail.com]

I tried Beta 2 and ran into some problems when clouds came in, guiding became very erratic after that.  Is this situation addressed in the algorithm?  If a cloud or a period of clouds passes over will that affect the predictive guiding that follows when the clouds pass?

Thanks,
Ron Stanley

[Brian Valente]

Ron

 

Do you have the guide log you can post?

 

I’m guessing the answer is no, it doesn’t account for that.

 

Thinking about this, I wonder how many cycles of learning there is on the algorithm: does it refine continuously, or after a few periods, it has enough and stops learning?

 

It might be a good option to be able to set this as a variable in the user interface, for situations like you describe. If we anticipate a cloudy night, maybe set learning cycles to 4, and for clear skies leave it at 0 (which means continuous)?

 

 

Thanks

 

Brian

 

 

Brian Valente

Brianvalentephotography.com

 

From: open-phd...@googlegroups.com [mailto:open-phd...@googlegroups.com] On Behalf Of rwstan...@gmail.com
Sent: Friday, March 10, 2017 12:30 PM
To: Open PHD Guiding <open-phd...@googlegroups.com>
Subject: Re: [open-phd-guiding] Re: Predictive PEC: beta3

 

I tried Beta 2 and ran into some problems when clouds came in, guiding became very erratic after that.  Is this situation addressed in the algorithm?  If a cloud or a period of clouds passes over will that affect the predictive guiding that follows when the clouds pass?

Thanks,
Ron Stanley

--

[Andy Galasso]

Ron,

Yes, the new algorithm handles periods where guiding is interrupted, like for passing clouds. We would need to see your guide log and debug to assess what happened with the erratic guiding.

Andy

​

[Brian Valente]

Ron was your guiding interrupted, or just crapped out due to the clouds?

 

Thanks

 

Brian

 

 

Brian Valente

Brianvalentephotography.com

 

From: open-phd...@googlegroups.com [mailto:open-phd...@googlegroups.com] On Behalf Of Andy Galasso
Sent: Friday, March 10, 2017 1:29 PM
To: Open PHD Guiding <open-phd...@googlegroups.com>
Subject: Re: [open-phd-guiding] Re: Predictive PEC: beta3

 

Ron,

 

Yes, the new algorithm handles periods where guiding is interrupted, like for passing clouds. We would need to see your guide log and debug to assess what happened with the erratic guiding.

 

Andy

​

--

[Tyler]

Edgar,
This is a really interesting concept.  Thanks for you work developing it and I look forward to testing once I get my system back up and running....

Question/thought:  It seems like this could really benefit folks who have mounts with non-integer gear ratios since those harmonics can't easily be correct by normal PEC.  I have a CGEM with the 8/3 ratio which fortunately isn't too bad on my particular mount but other folks are sometimes not as lucky.... If the user were to provide your software with the worm gear period and also acknowledge the presence of a non/integer gear ratio, I believe you could start reducing most of the harmonics after the first couple of worm gear cycles and then update the predictive PEC algorithm to so that it attempts to reduce the 8/3 harmonic once 6+ worm cycles have been completed....

Just wanted to relay the idea but recognize you're probably more focused on getting the more generalized capability working for the larger community- which I totally understand....

Thanks,

Tyler

[hnau]

Here's my guide log from last night utilizing PPEC on beta 3 on my AZ-EQ6 GT.  I struggled with thin wispy clouds all evening, so I ended up calling it quits after a couple runs of ~6 gear cycles.  Guiding was not bad, however the 0.55-0.6'' in RA RMS I observed was a little worse than what I normally see with hysteresis (less than .5'' is typical and down closer to .2'' on great nights) .  That could have easily been my seeing conditions.  Also, after the first run I realized that despite entering 479 as the period length, I forgot to uncheck auto.  When I went back in the settings mid way through the night I saw the value had jumped up to ~482, at which point I unchecked auto and reset the value to 479.  Overall though, I think PPEC is looking promising.  Thanks for your hard work on implementing this new algorithm.  Please let me know if you have any questions or if there's anything else I can do.  The next two nights' forecasts look quite a bit better (then again, yesterday's said 0 clouds... sigh), so I will give it another shot, hopefully with much longer runs.

[rwstan...@gmail.com]

Hi Brian,

I lost the guide star for around 20 minutes, after that the guiding became very erratic.  To be fair I had quite a bit of Dec backlash which I have since tried to correct.  The guide log is attached.  I was using Beta 1 and not Beta 2 as I said earlier.  The section in question is Section 7 after that I switched to Hysteresis.

Thanks,
Ron

[Edgar Klenske]

@Martin, the new algorithm collects data about the periodic error and then uses it to predict the gear error in advance. You're right, dithering and pausing would in principle disturb the algorithm, but it is possible to do the guiding based on the prediction only during settling and pausing phases. This way the algorithm copes well with dithering and pausing in most cases, as long as the telescope isn't slewed to a different target position during the pause. If you want to know the mathematical details, you can have a look at the documentation (see link in my first post).

Best,
Edgar.

[Edgar Klenske]

@Tyler, thanks for your thoughts on this!

The problem with the 8/3 mounts at the moment is that my algorithm forgets the data after the guiding is stopped. This is because the algorithm gets much more complex if the data stems from different runs, since the data needs to be stored persistently and new data needs to be aligned to the old data robustly and correctly. Therefore, my algorithm can only work well for very long runs on the 8/3 mounts. Anyway, you can enter the correct (non-integer) worm period and let the algorithm do its job. It starts with hysteresis and blends over to PPEC over two period lengths. This way it should be safe to use the algorithm, however, it might take a while until you see the potential benefits of the PPEC algorithm.

Best,
Edgar.

[Vince]

Hi Edgar, when you say PPEC starts with hysteresis and blends over to PPEC, does it use the parameters I have set in the Hysteresis algorithm?

Also, PPEC's default minimum move is 0.01 pixels.  Is this the recommended setting for all setups?  My guiding pixel scale is .56 acrsecs per pixel.

Thanks,

Vince

[Edgar Klenske]

@Vince,

no, right now it uses the default Hysteresis parameters. Using the parameters of another guider would be hard to understand and configure. Either we use defaults, the same values as for PPEC, or we need to add extra hysteresis parameters.

Minimum move is implemented exactly as for Hysteresis. Just use a value that's good for Hysteresis for your setup. I came up with a rather small value because it helps RMS performance. However, this could cause sawtooth problems and therefore I'll set it to something higher with the next release.

Best,
Edgar.

[Vince]

Ok, thanks Edgar.

Vince

[Brian Valente]

Hi Edgar

just checking in. Any updates on a beta 4 or progress with predictive DEC?

Cheers

Brian

[Diego Colonnello]

Last night i tested this beta version, predictive pec is awesome, in a mount with an RMS  of 1.2 it came down to 0.7 with the use of a dark red filter on the guide camera, and down to 0.42 with the Predictive Pec......... it is simply amazing.

[bw_msgboard]

Hi Diego, thanks for reporting your results.  Could you tell us a bit more about the mount – what kind of mount is it and had you already applied a periodic error correction to it?  What is the image scale you’re using for guiding (arc-sec/px) and what do you think is going on with the use of the red filter?

 

Thanks.

Bruce

 

---

--

[Diego Colonnello]

I have an old EQ6 pro that i have fixed and tuned, belt mod and all the wiring passing inside the mount to avoid any dragging, i am using a red filter and it gives me a nice steady star to guide with the OAG, so far i think that PHD2 has done the biggest difference on my imaging....... please keep the good work.

[Brian Valente]

Andy

worm periods

Losmandy Titan 318.1262611 sec

Losmandy G11T 318.1262611 sec

[Andy Galasso]

got it, thanks, wiki has been updated

[Suresh Mohan Neelmegh]

Hi Andy,

 where can i enter the worm period ? im not able to find that.

Sorry

Suresh

[Andy Galasso]

Hi Suresh,

The worm period setting is in the Advanced Settings window (the brain), Algorithms tab, under Predictive PEC. If you enter the known worm period you can un-check the Auto check-box.

Since you are posting in this thread for beta 3, are you using the Beta 3 version still? Everyone using the beta should upgrade to 2.6.3dev2 or newer to get the latest improvements and fixes, the beta builds are now obsolete.

Andy

[Donghun Kim]

Hello

Can PPEC help correcting constant RA drifts due to factors such as atmosphere refraction and imperfect RA tracking speed?

Thank you!

Donghun

[Brian Valente]

Donghum

 

PPEC can help correct for errors due to periodic error of your mount. It does not itself correct for atmosphere refraction or RA tracking speed.

 

Those are hopefully corrected up as part of the normal guiding process. As I understand it PPEC also uses hysteresis as part of its RA guide plan, so you aren’t giving any of that up, just adding the periodic error correction.

 

The way I think about it, it kind of eliminates the need to program PEC into your mount.

 

 

 

Thanks

 

Brian

 

 

Brian Valente

Brianvalentephotography.com

 

From: open-phd...@googlegroups.com [mailto:open-phd...@googlegroups.com] On Behalf Of Donghun Kim
Sent: Saturday, April 1, 2017 11:25 AM
To: Open PHD Guiding <open-phd...@googlegroups.com>
Subject: [open-phd-guiding] Re: Predictive PEC: beta3

 

Hello

--

[Andy Galasso]

Donghun,

Yes, the algorithm does model and correct for constant RA drift in addition to periodic error components. There's a nice diagram of the model on page 4 of the algorithm description document available in the PHD2 wiki here..

Andy

[bw_msgboard]

Hi Donghun.  As Andy said, the PPEC algorithm will likely detect these drifts, but so will the other guide algorithms.  Steady drift is generally the easiest thing to guide out, so it’s rarely a problem for anyone.  Are you measuring uncorrected drift by analyzing your guide logs?  If so, you should probably post a log so we can take a look.  But if you’re reaching this conclusion on the basis of elongated stars in your images, you probably have a different problem entirely.

 

Good luck,

Bruce

 

---

From: open-phd...@googlegroups.com [mailto:open-phd...@googlegroups.com] On Behalf Of Donghun Kim
Sent: Saturday, April 01, 2017 11:25 AM
To: Open PHD Guiding
Subject: [open-phd-guiding] Re: Predictive PEC: beta3

 

Hello

--

[bw_msgboard]

From: open-phd...@googlegroups.com [mailto:open-phd...@googlegroups.com] On Behalf Of Brian Valente
Sent: Saturday, April 01, 2017 11:29 AM
To: 'Donghun Kim'; 'Open PHD Guiding'
Subject: RE: [open-phd-guiding] Re: Predictive PEC: beta3

 

Donghum

 

PPEC can help correct for errors due to periodic error of your mount. It does not itself correct for atmosphere refraction or RA tracking speed.

 

Those are hopefully corrected up as part of the normal guiding process. As I understand it PPEC also uses hysteresis as part of its RA guide plan, so you aren’t giving any of that up, just adding the periodic error correction.

 

The way I think about it, it kind of eliminates the need to program PEC into your mount.

 

Hi Brian.  This might prove to be true, but I would be surprised and I don’t think we have any data yet to support such an ambitious claim.  If your mount has permanent periodic error correction that works well, I think you’d be well advised to use it.  Then PPEC can try to handle the residual errors that can’t be fully eliminated by the periodic error correction in the mount.  Measuring the baseline periodic error is usually done with guiding disabled and can easily take as long as 20 minutes to collect enough data for the statistics to hold up.  Why would you want to effectively do this every time you started imaging?  But as I said, that’s just my expectation based on the theory.  If you have the time, maybe you could run some careful back-to-back tests with PEC enabled and disabled to see what happens.  We’d definitely like to see the results if you do that.

 

Have fun,

Bruce

 

 

 

Thanks

 

Brian

 

 

Brian Valente

Brianvalentephotography.com

 

From: open-phd...@googlegroups.com [mailto:open-phd...@googlegroups.com] On Behalf Of Donghun Kim
Sent: Saturday, April 1, 2017 11:25 AM
To: Open PHD Guiding <open-phd...@googlegroups.com>
Subject: [open-phd-guiding] Re: Predictive PEC: beta3

 

Hello

 

Can PPEC help correcting constant RA drifts due to factors such as atmosphere refraction and imperfect RA tracking speed?

 

Thank you!

Donghun

--
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.
For more options, visit https://groups.google.com/d/optout.

[mahaffm]

I tried out the Predictive PEC: beta3 last night and got very good results. I saw what looked like a 30% improvement in my guiding. My setup is: CGE-Pro, Celestron HD11 - 2800 FL, OAG. My typical error I was seeing was approximately 0.35".in total Guiding error (RA and DEC).

Mark

[Andy Galasso]

Mark,

Excellent! Glad to hear you had such great results and thanks for reporting back.

FYI, the beta is now obsolete and the Predictive PEC code has been incorporated into PHD2 proper, so you should probably upgrade to the most recent development build to continue using PPEC.

Andy

[mahaffm]

Hi Andy,

I use a QHY5III174 mono version with my Celestron OAG. I see that there is a Beta version for native support for QHY cameras. I would like to test it our but was wondering if all of the Predictive PEC (PPEC) is included in with the QHY beta. We have limited clear nights so I'd like to continue to use the PPEC while testing out the native support for the QHY camera.

Thanks,
Mark

[Vince]

Mark's suggestion would be helpful for me too.  I would like to use the latest QHY dev build with the latest PPEC version.

Vince

[Andy Galasso]

Mark, Vince,

The latest PHD2 dev build has both Predictive PEC and the new QHY camera support.

Andy

[Vince]

Cool, thanks Andy.

Vince

[mahaffm]

Perfect Thanks Andy,

I'll give the updated Beta a try on the next clear night out.

Mark

[mahaffm]

Hi Andy,

I downloaded the latest dev build and was able to test out the beta update using my QHY5III174 Mono camera using the QHY Camera setting in PHD2 and did not have any issues. I rebuilt the Bad-pixel Map as well as the Dark Library without any issues. Since it is currently cloudy out I was not able to do any testing under the stars but was able to loop the camera for about 15 minutes with various exposure times and gains without any issues. I went as low a 5 for the gain and as high as 95. Looping exposures from 1 to 10 seconds without any issues.

Mark

[Andy Galasso]

Excellent, thanks for the update.

[Richard Cardoe]

Installed v2.6.3dev3, Win10, no problems during installation.

Equipment: CEM60, GSO 8"RC, 0.75x reduced to 1218mm focal length, guided with SX OAG and Lodestar x2, for a guiding pixel scale of 1.46 arcsec/px.

Guide rate was 0.5x in RA and 0.8x in DEC, actually I calibrated PHD at 0.8x in RA but later reduced it, it seems more effective sometimes for controlling RA than reducing settings in PHD, I don't know how much that might effect the predictive PEC algorithm or reading the logs, but worth noting.

My CEM60 has a periodic error of 11.5 arcsec pk-pk and is fairly new to me so I'm still working on getting it dialled in, I've also been struggling to get guiding working reliably, my RA is consistently 50-100% worse RMS than DEC, and so I'm still getting eggy stars at 1218mm focal length even at 5mins.

I suspect the periodic error may be a factor, hence my testing of this promising new guide algorithm.

Typical values from a guiding session:

Hystoresis algorithm RA and DEC

RA 0.57" (0.39px) RMS

DEC 0.34" (0.23px) RMS

Total 0.66" (0.45px) RMS

RA Peak 2.2"

DEC Peak 1.55"

With Periodic PEC algorithm RA and Hystorsis DEC:

RA 0.44" (0.30px) RMS

DEC 0.37" (0x25px) RMS

Total 0.58" (0.39px) RMS

RA Peak 2.49" (1.71px) 

DEC Peak 2.01" (138px)

It seems some improvement in RA and a slight reduction in DEC, though conditions were not great it was windy.

I make it approx 23% reduction in RA RMS and approx 8% increase in DEC RMS, actually bringing those values closer together reduced the eggyness of my stars.

Enough promise to keep testing with it for sure. I would be very interested in any suggestions for improving my settings.

Other notes, the CEM60 has a worm period of 299.18 seconds (as does the CEM60-EC) these could be added to the wiki page. When I let the auto period calculator run it came up with 294.807 over 218 samples and 307.440s over 1057 samples. Perhaps something to tweak there, I settled for just entering my period manually.

PHD2_GuideLog_2017-04-12_012136.txt is the log with predictive PEC enabled, there are several runs in there. Log segments 4 and 13 are with auto worm period calculation running, there others are fixed. The final segment 15 is 233 mins showing just over 46 worm cycles.

For comparison PHD2_GuideLog_2017-04-08_021717.txt is my more typical guiding with the CEM60. Struggling guiding to be honest, you can see a clear repeating wave in RA during which looks very much like my periodic error, segment 8 shows a long run. There are long periods in RA where it seems like it just isn't responding and then suddenly switches from east to west or vice versa and then does the same again.

Hopefully the logs are still useful.

Great work on PHD2 everyone.

Thanks,

Richard

[bw_msgboard]

Hi Richard.  Thanks for the very clear report, this makes it easier to follow what’s going on.  We’re still trying to get a feel ourselves for how this algorithm works on different set-ups, but I’ve made some comments below:

 

---

From: 'Richard Cardoe' via Open PHD Guiding [mailto:open-phd...@googlegroups.com]

Sent: Thursday, April 13, 2017 5:04 AM
To: Open PHD Guiding

Subject: Re: [open-phd-guiding] Re: Predictive PEC: beta3

Installed v2.6.3dev3, Win10, no problems during installation.

 

Equipment: CEM60, GSO 8"RC, 0.75x reduced to 1218mm focal length, guided with SX OAG and Lodestar x2, for a guiding pixel scale of 1.46 arcsec/px.

Guide rate was 0.5x in RA and 0.8x in DEC, actually I calibrated PHD at 0.8x in RA but later reduced it, it seems more effective sometimes for controlling RA than reducing settings in PHD, I don't know how much that might effect the predictive PEC algorithm or reading the logs, but worth noting.

 

This really isn’t a good idea and it definitely makes it harder to interpret what PPEC is doing.  You should leave the mount guide speed settings at whatever values were used for calibration.  If you think you see over-correction, you should reduce the aggressiveness factors – that’s why we have them. <g>  Based on what you did, I’d expect to see some under-correction in RA, and that’s what I see, especially in the 4hr run.  

 

My CEM60 has a periodic error of 11.5 arcsec pk-pk and is fairly new to me so I'm still working on getting it dialled in, I've also been struggling to get guiding working reliably, my RA is consistently 50-100% worse RMS than DEC, and so I'm still getting eggy stars at 1218mm focal length even at 5mins.

I suspect the periodic error may be a factor, hence my testing of this promising new guide algorithm.

 

Have you applied conventional period error correction to the mount?  Is the 11.5 a-s pk-pk number before you applied PEC?  Or is the 4/8 log done with no PEC at all, just conventional guiding?

 

Typical values from a guiding session:

Hystoresis algorithm RA and DEC

RA 0.57" (0.39px) RMS

DEC 0.34" (0.23px) RMS

Total 0.66" (0.45px) RMS

RA Peak 2.2"

DEC Peak 1.55"

 

With Periodic PEC algorithm RA and Hystorsis DEC:

RA 0.44" (0.30px) RMS

DEC 0.37" (0x25px) RMS

Total 0.58" (0.39px) RMS

RA Peak 2.49" (1.71px) 

DEC Peak 2.01" (138px)

 

It seems some improvement in RA and a slight reduction in DEC, though conditions were not great it was windy.

I make it approx 23% reduction in RA RMS and approx 8% increase in DEC RMS, actually bringing those values closer together reduced the eggyness of my stars.

 

Enough promise to keep testing with it for sure. I would be very interested in any suggestions for improving my settings.

 

I think the long run on 4/13 shows a definite improvement in RA with a nice reduction in the obviously-repetitive error we see on 4/8 – but maybe there’s more to be had.  If you go back to normal procedures and don’t fudge the mount guide speed and you still see some under-correction in RA, try boosting the ‘predictive weight’ by a small amount.   That’s just a suggestion on my part, I don’t have enough experience with the algorithm yet to be confident about it.

 

Other notes, the CEM60 has a worm period of 299.18 seconds (as does the CEM60-EC) these could be added to the wiki page. When I let the auto period calculator run it came up with 294.807 over 218 samples and 307.440s over 1057 samples. Perhaps something to tweak there, I settled for just entering my period manually.

 

With auto-adjust enabled, this period will be tuned to match the stronger components of the residual periodic error that’s actually measured.  For now, we think it’s going to be best to let the algorithm fine-tune the value after giving it a good starting point.  But if you know from your own long-term analysis that there’s a specific error frequency you’re trying to attenuate, then you might want to specify that value and disable auto-adjust.  We think most users will not be in that situation though.

 

PHD2_GuideLog_2017-04-12_012136.txt is the log with predictive PEC enabled, there are several runs in there. Log segments 4 and 13 are with auto worm period calculation running, there others are fixed. The final segment 15 is 233 mins showing just over 46 worm cycles.

 

I’ll tell you, I don’t see much to be unhappy about here.  With a total RMS of 0.57 a-s, you were probably guiding at the limit of the seeing, particularly if you had windy conditions.  This run should have produced nicely round stars, and if it didn’t, I think the problem lies elsewhere, not in guiding. For example, optical aberrations like collimation error can produce oblate stars.  The more you try to drive the total guiding error down below these levels, the more you’re dependent on seeing and the harder it will be.   Speaking from personal experience, when I get guiding results like these using an OAG, out-of-round stars are simply never a problem using 10-min exposures.

 

Good luck,

Bruce

 

For comparison PHD2_GuideLog_2017-04-08_021717.txt is my more typical guiding with the CEM60. Struggling guiding to be honest, you can see a clear repeating wave in RA during which looks very much like my periodic error, segment 8 shows a long run. There are long periods in RA where it seems like it just isn't responding and then suddenly switches from east to west or vice versa and then does the same again.

 

Hopefully the logs are still useful.

 

Great work on PHD2 everyone.

 

Thanks,

Richard

 

--

[Richard Cardoe]

Hi Bruce,

Thanks for the comprehensive reply.

I understand how it wasn't a good idea to fiddle with the guide rate behind PHD's back, that's why I thought it best to mention it. I'll make sure to not do so in future testing :)

It was a hangover from my HEQ5 where a lower RA rate than DEC really helped stabilise things, but I wouldn't normally change it after calibration.

11.5" a-s pk-pk is the native periodic error, measured with PemPro. have attempted conventional PPEC using PemPro, but without any significant success, the best I've had is reducing the 11.5" pk-pk native error to 10.5", within the margin of error that I probably wasn't doing some right at all and actually having no effect.

Both of my the 4/8 and 4/13 logs are without PPEC enabled, just guiding.

I too hope there is more to be had from this good start with this algorithm, I noted the under-correction myself and did increase the 'predictive weight' slightly to 65%, I was being conservative, I'll test next time with it at that level and without lying to PHD about my RA guide rate :) Then perhaps tweak higher in small increments.

Regarding auto-tune of the worm period, I'll test with manually entering 299.18 and enabling the auto-tune.

I do have one particular question, it was my understanding that as long as the total RMS error was under the pixel scale of the imaging setup then you don't see errors in non-round stars, i.e my total 0.57" as/px against the 0.91" as/px resolution of my OTA/Camera combination. Is this right? Even with the stark difference in RA and DEC RMS I have in my setup? If my understanding is correct then that might point to other causes of non-roundness you suggest, such as the notorious fickleness of collimation the GSO RC or other aberration from reducer spacing and so forth.

I have experimented with my ED80 refractor at 510mm focal length (different reducer setup), it too doesn't show completely round stars at 10mins and I'd expect it to with a resolution of 2.13 as/px.

More testing required, when the skies cooperate :)

Thanks again for you insight.

Richard

[Jay L]

Edgar,

Thanks for putting this together. I am looking forward to trying it out. A question: I had done some guidelog analysis of my system (https://youtu.be/Xxbc-aUIqzo) and identified several mechanical harmonics (not just the fundamental). Does, or will, your algorithm "learn" what harmonics are present and develop a multi-harmonic predictive response for the mount?

Jay

[Andy Galasso]

Hi Jay,

Great job on that video, I think your analysis is spot-on.

To answer your question about harmonics, that is correct, the algorithm does deduce the harmonics and apply predictive corrections. You may be interested in this document describing the algorithm, available on the PHD2 Wiki.

Andy

[Andy Galasso]

This topic is now locked for replies.

The Predictive PEC Beta 3 is obsolete, and the Predictive PEC algorithm is now included in PHD2.

Feel free to start a new discussion thread if you have any questions or issues regarding Predictive PEC.

Andy