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

PHD2 and EQ-g

  • Please log in to reply
14 replies to this topic

#1 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 10:32 AM

Hi Mates!

I noticed a quite strange behaviour when I use Hysteresis algorithm.

I have some deep peaks that I can't reproduce if I use the Lowpass one.

Of course, I can use LP instead of hysteresis but it should be nice to understand why it behaves like that.

 

Here in attachment my guiding log.

Please have a look at the 47 minutes and 58 seconds guiding run recorded at 20:38:28 yesterday night against the following ones with the LP.

By the way LP isn't perfect too, I have some peaks which I could reasonably assume that are PE, but it's far better than hysteresis.

 

Thanks for looking.

Bests.

 

Roberto

 

Attached Files


Edited by rob77, 16 August 2021 - 10:36 AM.


#2 Tapio

Tapio

    Aurora

  • -----
  • Posts: 4,817
  • Joined: 24 Sep 2006
  • Loc: Tampere, Finland

Posted 16 August 2021 - 10:50 AM

Have you read this ?
https://openphdguidi..._algorithms.htm

 

And have you tried Predictive PEC Guide Algorithm ?



#3 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 11:10 AM

And have you tried Predictive PEC Guide Algorithm ?

 

Nope. What I've tried yesterday has been to train PE via EQMOD, guiding using PHD2 hysteresis algorithm (this is why 47 minutes - it's around 5 EQ-r cycles).

The problem was that, as you can see plotting the graph, the errors do not look like periodical while - on the other hand - they seem so with Lowpass.

 

I haven't tested yet EQMOD PE training guiding with Lowpass to see if I can improve it. Wanna try tonight.

 

Bests



#4 rgsalinger

rgsalinger

    Cosmos

  • *****
  • Moderators
  • Posts: 9,536
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 16 August 2021 - 11:41 AM

"Increasing the hysteresis value willsmooth out thecorrectionsat therisk of being too slowto react to a
legitimatechangein direction."

 

The default value is 10 for hysterisis. You're using 65 for some reason. I think that's the problem - slow reversals if the initial error is large enough. 

 

Rgrds-ross



#5 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 12:03 PM

I'll try reducing it.

 

Thanks



#6 michael8554

michael8554

    Apollo

  • -----
  • Posts: 1,417
  • Joined: 19 Aug 2020
  • Loc: Wiltshire UK

Posted 16 August 2021 - 12:27 PM

Some very strange settings in your GuideLog.

 

You have a poor combination of very high Hysteresis settings and low Max Aggressions settings.

 

So PHD2 is really struggling to correct RA errors.

 

For instance, in your 20:38 Hysteresis session, there are some huge 25arcsec spikes, and PHD2 is sending tiny RA pulses for over 3 minutes before managing to correct.

 

Same in your Low Pass sessions.

 

So there are no worthwhile comparisons between the two modes, because your settings aren't giving either any chance of working correctly.

 

I would start over with a new Profile using default settings, make a new Calibration at Dec = 0, run Guide Asssistant, and accept its recommendations.

 

And your Guidecam focus could be improved, use the HFD reading in the Star Profile window.



#7 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 12:46 PM

Yes, I have to reset the parameters and run the guiding assistance for sure.

I did it few days ago, then I've started changing them.

What anyway needs to be understood IMHO is why |20|'' spikes are generated with hysteresis while aren't with LP (besides the fact that they are smaller in amplitude, the spikes that you can see while guiding in LP are potentially PE).

 

Thanks.

Cheers.

 
 

 

 


Edited by rob77, 16 August 2021 - 12:48 PM.


#8 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 12:53 PM

IMHO, analyzing the hysteresis session, the ones that I've highlighted in red are PEs.

The huge ones are something that I am not able to sorting out and that seems to not happen with LP.

 

immagine.png

 

 

Anyway I will retest tonight to rule out some external factors that could have generated those spikes during the hysteresis session and that disappeared in the next one with LP.

I live on a terrace on top of a 3 floor building so - maybe - during the hysteresis session some vibrations happened and didn't repeat later on.

 

 

Bests.


Edited by rob77, 16 August 2021 - 01:07 PM.


#9 Tapio

Tapio

    Aurora

  • -----
  • Posts: 4,817
  • Joined: 24 Sep 2006
  • Loc: Tampere, Finland

Posted 16 August 2021 - 01:14 PM

Nope. What I've tried yesterday has been to train PE via EQMOD, guiding using PHD2 hysteresis algorithm (this is why 47 minutes - it's around 5 EQ-r cycles).

The problem was that, as you can see plotting the graph, the errors do not look like periodical while - on the other hand - they seem so with Lowpass.

 

I haven't tested yet EQMOD PE training guiding with Lowpass to see if I can improve it. Wanna try tonight.

 

Bests

I'm not sure what you mean by 'training' here.

With Predictive PEC Guide Algorithm PHD is learning your mounts behaviour and learn to do corrections.

The other algorithms are just what they are.



#10 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 01:16 PM

EQMOD PEC training --> http://eq-mod.source...qmod_vs-pec.pdf

 

From page 8 on.

 

Instead of using PHD2 PEC algorithm, I tried to train the PE via EQMOD as explained in the mentioned PDF.

I'll do the other way round tonight, I'll try PHD2 PEC algorithm.


Edited by rob77, 16 August 2021 - 01:21 PM.


#11 michael8554

michael8554

    Apollo

  • -----
  • Posts: 1,417
  • Joined: 19 Aug 2020
  • Loc: Wiltshire UK

Posted 16 August 2021 - 02:52 PM

"why |20|'' spikes are generated with hysteresis while aren't with LP"

 

They weren't generated by the guiding, they occurred due to outside influences, and then PHD2 tried to guide them out.

 

And like I said, absolutely zero inferences can be made with those guide settings.



#12 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 16 August 2021 - 02:59 PM

 

They weren't generated by the guiding, they occurred due to outside influences, and then PHD2 tried to guide them out..

This is what I'll try to understand tonight.

I am substantially on the same page as you: these huge spikes must be due by some external factors.

 

But I need to clarify this point: 47 minutes with Hyst. algorithm, 2-3 huge spikes. 4 hours with LP, no huge spikes except PE.

It can be a coincidence but I need to reproduce it tonight.


Edited by rob77, 16 August 2021 - 03:00 PM.


#13 rgsalinger

rgsalinger

    Cosmos

  • *****
  • Moderators
  • Posts: 9,536
  • Joined: 19 Feb 2007
  • Loc: Carlsbad Ca

Posted 17 August 2021 - 12:13 PM

I disagree very slightly. I think that there is an outside influence here, but it might be guided out (become irrelevant) if the settings were not so odd. Once the OP has reset to the defaults, then we'll see what's what. Until that happens, it's hard to say if there's actually a problem or not. The acid test of guiding is not the graph, it's the output. 

 

I image remotely on two systems maybe 100 nights a year. The guide graphs are up on my screens most nights while I read, watch tv or just fiddle about. I can tell you that I see spikes all the time -- last time out it was windy - but they very rarely affect the results because any deviation is quickly guided out. 

 

Rgrds-Ross



#14 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 17 August 2021 - 12:35 PM

Yes, I agree. It very likely sounds something external influencing my guiding (I am trying to figuring out what it is. It could also be my building that has a kind of swinging, sometimes; I should test somewhere else).

 

Anyway, here it follows the log of 33 minutes guiding using PEC algorithm.

You can see there is a big spike nearly at the end. Overally is not so bad, IMHO. It could be better but my terrace is always under pretty annoying wind gusts.

 

Bests.

 

 

 

 

 

 

 

 
 

 

 

Attached Files


Edited by rob77, 17 August 2021 - 12:36 PM.


#15 rob77

rob77

    Vendor - FAST platesolver

  • -----
  • Vendors
  • topic starter
  • Posts: 1,000
  • Joined: 25 Feb 2012
  • Loc: Cuneo (Italy) / Belo Horizonte (Brazil)

Posted 18 August 2021 - 09:00 PM

As an update.

I am doing a session now using PEC algorithm and the guiding seems excellent (actually never reached such RMS values).

 

immagine.png

 

 

Thanks and cheers.

 

Roberto

 

 
 

 

 

 

 
 

 

 


Edited by rob77, 18 August 2021 - 09:00 PM.



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






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics