×

INDI Library v2.0.6 is Released (02 Feb 2024)

Bi-monthly release with minor bug fixes and improvements

Internal Guider Changes--please read if you use it

  • Posts: 1208
  • Thank you received: 559
Summary
I have made some user-interface and algorithmic changes to the internal guider in the experimental software (3.5.5 beta).   If you have tweaked parameters in the internal guider version, may need to re-tweak some parameters (specifically control gain/aggressiveness & min move) when you switch to the new version of the guider.

Details

The most notable change to the user interface relates to the guider controls. 

Controls, how it is in 3.5.4:
In the current release, 3.5.4, and previous versions, on the Guider Page there is a sub-tab below the image display called "Control". This tab allows you to enter values for Proportional Gain, Integral Gain, Maximum Pulse, and Minimum Pulse for both the RA and DEC axes. Note that the values entered for the gains were numbers defaulting to 133 and 0 (where 133 approximated correcting half the guide error with each correction). Values for the Min/Max pulse parameters were in milliseconds--e.g. if you set min to 20, then no correction pulse would be sent that was less than 20ms, with the assumption that such small corrections probably were dwarfed by seeing noise.

Controls, how it has changed in 3.5.5beta
The Control sub-tab was removed, and the guider's Options menu (the menu you get by clicking Options on the bottom right of the Guide Page) was reorganized. Now the Guide Options menu looks as below. 

 

In the new Options Menu, the  Guide tab now has controls related to what was in the old Controls sub-tab. What was called Proportional Gain is now called Aggressiveness, and instead of the previous numeric scale with numbers like 133, the new parameter values should be between 0.0 and 1.0. Instead of assuming a generic response rate for your mount, this aggressiveness control now uses the calibrated mount values to decide how long a pulse is sent to correct the guide deviation. Aggressiveness=1.0 would tell the guider to try to fully correct the guide deviation each guide interval. A full correction is probably too aggressive, and would likely cause some oscillation. It likely better to use an aggressiveness of 0.6-0.8 but you can experiment with this.  Similarly, Integral gain uses a 0.0-1.0 "gain" to control how much the average of previous errors are weighted into the computation of the guide pulse. I would recommend starting with 0.0 for integral gain (i.e., not using it), but again feel free to experiment.  The min and max movement parameters are now in units of arc-seconds, instead of milliseconds of pulse. For instance. min_error = 0.2 means don't try to correct a guide deviation smaller than 0.2 arc-seconds. Using a value of 0.0 is probably not ideal as you may be "chasing the seeing".

Because the above units changed and now depend on the calibration of your mount, it wasn't possible to simply transfer over your old parameter values to this new version.

Dithering Controls:
The dithering controls were moved to a new tab, but weren't changed.

GPG Controls
GPG control gain and min move parameters were removed. GPG now uses the values in the Guide tab (above) for these value.

Calibration Controls
The only change to calibration is the addition of a "Max Move in pixels" parameter--it limits about how many pixels the guide star can be moved during the calibration process.

Summary
  • Change in name of Proportional Gain to Aggressiveness, and change to a 0-1.0 scale.
  • Aggressiveness now uses calibrated mount values to compute its pulses (both the direction and the milliseconds of pulse required to move the guide star one arc-second).
  • The Min- and Max-move parameters change from units of milliseconds of pulse to units of arc-seconds.
  • The Control sub-tab moved inside the guide Options menu.
  • The Dither controls moved to it's own tab in guide Options menu.
As always, please feedback any issues or suggestions you may have.

Hy
2 years 7 months ago #74923
Attachments:

Please Log in or Create an account to join the conversation.

  • Posts: 1009
  • Thank you received: 133
Great work Hy! Will try ASAP.

Just one question: I assume that means I have to re-do the guider calibration after the change, so that all values are properly adjusted and propagated? (I so far still use my calibration from May.... ;) )
2 years 7 months ago #74926

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Sounds great! I'd love to give this a try and give you some feedback. I'm using 3.5.4 stable downloaded from the website. How would I go about getting this version of the guider (sorry if this is a very basic question, but if you don't know, then you don't know!). Also, will I be able to swap back and forward between the guider versions?

If I understand correctly what you've changed... in order to get the same setup in the new version as in 3.5.4 (as a starting point from which to play around) I would need to do the following for each axis:
1. Get the 3.5.4 Proportional Gain, call it PG (in my case I leave it on the default 133).
2. Do a calibration and get the 1arcsec pulse length, call it PL.
3. Set the Aggressiveness to PG/PL.

From here I can try adjusting the Aggressiveness and see what happens. How does this take effect, does the guider respond immediately to parameter changes or do I need to restart it?

What's the algorithm for Integral Gain, i.e. how many previous values are averaged and is it a straight average or weighted in some way, e.g. more recent weighted higher? If GPG is used with Integral Gain how do they interact, or is it one or other but not both at the same time?

For calibration, the PHD2 folks recommend doing a calibration at a specific telescope orientation and then leaving it alone until the setup is changed. Presumably it would be the same here, but how can I force a calibration (PHD2 SHIFT+Start) when I want to and then leave it alone until I specifically want to do another one (values persisted).

You are probably already thinking about this but PHD2 offers suggestions for parameter settings based on the calibration. I think this would be very useful for the internal guider.
2 years 7 months ago #74929

Please Log in or Create an account to join the conversation.

  • Posts: 985
  • Thank you received: 160
"but how can I force a calibration (PHD2 SHIFT+Start) when I want to and then leave it alone until I specifically want to do another one (values persisted)."

Just click at the "clear calibration data" icon. Ekos will then re-calibrate once you hit the "Guide" button.

 

If you have "Store and reuse guide calibration when possible" in the calibration settings active

 

Ekos will continually use the latest calibration until you click "clear calibration data" again.
Last edit: 2 years 7 months ago by Alfred.
2 years 7 months ago #74942
Attachments:

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
Great thanks Alfred, good to know!
2 years 7 months ago #74948

Please Log in or Create an account to join the conversation.

  • Posts: 1208
  • Thank you received: 559
@Pit: Looking forward to your feedback. You need to re-do your calibration so long as you haven't done any of the normal things that might cause you to need to re-do it, e.g. rotate the guide camera sensor, change the weight balance, ...

@John, Great questions, see below:
    How would I go about getting this version of the guider
You're asking a more general question, which is "how do I get to use the KStars pre-release experimental/beta software?" Unless you compile the software yourself, the experimental releases are only available on Linux systems using the techniques described in www.indilib.org/download.html. If you build from source, you can look at the KStars README page at invent.kde.org/education/kstars#building
Keep in mind that running experimental software always has risk that it will be less reliable than released software. Also note that you'd not only get the new internal guider changes, but also any other changes that have been made since the last release. I, and I'm sure many other experienced forum users, will be happy to help futher if these web pages aren't clear enough, but I'd suggest either started a new forum topic on getting the latest release, or, probably better, finding the last such topic and continuing it.

  Also, will I be able to swap back and forward between the guider versions?
I do, but if this is new to you, it may not be straight-forward. Jasem's instructions show you how to do that. If you build from source, you can just run the new binary without installing it. If you are unsure, you should definitely fully back up your system, so that you can restore.

  If I understand correctly what you've changed....(procedure for finding the aggressiveness)
Yes, that makes sense. You don't need to re-calibrate, you could look at your current calibration--it's show in the Calibration tag in the guider options. You might want to skip that procedure and just use the default aggressiveness and see how you like it before making changes.

  Does the guider respond immediately to parameter changes or do I need to restart it?
It should respond immediately to changes in Aggressiveness and min/max moves.

  What's the algorithm for Integral Gain, i.e. how many previous values are averaged and is it a straight average or weighted in some way, e.g. more recent weighted higher? If GPG is used with Integral Gain how do they interact, or is it one or other but not both at the same time?
First, I did not modify nor experiment with integral gain as part of this change. In fact, I don't use it. I welcome someone looking further into and figuring out if it is helpful and how to configure it. I have looked at the code extensively, as part of this, and understand what is done. Integral gain looks at up the previous 50 guider iterations (upto, because there may be fewer available--e.g. on the 25th iteration it looks at the past 24).
It takes a straight, unweighted average of the guide deviation, e.g. it may wind up computing that the mean historical guide deviation is -0.5 arc-seconds. It then weights in that error with the current guiding measurement, using the aggressiveness (for the current measurement) and the integral gain weighting the historical measurement.  GPG does not use integral gain. It has its own hardwired mechanism. (Good point, I should grey it out when GPG is enabled.)

  You are probably already thinking about this but PHD2 offers suggestions for parameter settings based on the calibration. I think this would be very useful for the internal guider.
Agreed, a wizard for computing the best parameters would be great. I don't have any concrete news on this.

  For calibration, the PHD2 folks recommend doing a calibration at a specific telescope orientation and then leaving it alone until the setup is changed. Presumably it would be the same here
Yes, the PHD2 recommendations are good, and in general their advice applies to the Ekos internal guider as well.
 
Last edit: 2 years 6 months ago by Hy Murveit.
2 years 7 months ago #74955

Please Log in or Create an account to join the conversation.

  • Posts: 593
  • Thank you received: 276
@Hy thanks for the very detailed reply. I'll take a look at the links you included. I am Mac based so looks like I'll need to build Kstars myself. I may disappear down a rabbit hole for some time... I'll take your advice and check out the forum for helpful threads.
2 years 7 months ago #74960

Please Log in or Create an account to join the conversation.

  • Posts: 239
  • Thank you received: 38
I have had the best guiding I have ever had using this new version!  Thanks so much!

 
2 years 6 months ago #74985
Attachments:

Please Log in or Create an account to join the conversation.

  • Posts: 333
  • Thank you received: 23
Please, there is a way to manually change the guiding star without stopping the shoot?
Also, when we use the scheduler, the guide star is autoselected, but many times it select a star that is not good, then I ave to stop the scheduler and make all the stuffs manually
2 years 6 months ago #75102

Please Log in or Create an account to join the conversation.

  • Posts: 992
  • Thank you received: 155
Working well!

 


  
2 years 6 months ago #75119
Attachments:

Please Log in or Create an account to join the conversation.

  • Posts: 992
  • Thank you received: 155
Well - I spoke too soon. As shown above, I tried the latest version last night and the first set of 20 exposures guided perfectly (RMS of 0.3"). But after that - once I moved to a new target I was never able to get it to work. They made changes to the internal auto guider and whatever they did - it is now not usable for me. It took me an hour just to calibrate it well - and even then the guiding was never usable. I used up 4 hours of a perfect night trying. I then tried a few unguided exposures and I was able to get 180 second exposures at 2,000mm focal length - so the issue is with the guider, not the mount. So until this is fixed I am limited to 180 second shots, unguided.
PS: I always use 8 second guide exposures because anything less is not possible with an iOptron EC2 mount.
2 years 6 months ago #75150

Please Log in or Create an account to join the conversation.

  • Posts: 1208
  • Thank you received: 559
Please send me a problematic log, including a "debug for guide" log setting.

- There shouldn't be any changes to calibration, that should work as before. So, if that is the problem, then it's not the changes to control, and likely the older version would have done the same thing.

- Where were you pointing? In general guide doesn't work well very close to the pole.

- Do you think the issue was a bad choice of guide star? That wouldn't have changed, but could be your issue.

- Does it work well with the older software? Ideally you'd have the new and old binaries and be able to see if something in the new code affected things.

- What guide algorithm and what star detection algorithm are you using?

These questions apply to guiding, but from your note, I'm thinking that getting calibration to work better will solve things:

- is the problem with the guide star or the control

- did you adjust your aggressiveness (if it was over-reacting turn it down, if it was under-reacting, turn it up)

- is the problem in RA or DEC?

- 8s exposures is very unusual, but I trust your judgement on that.

Hy
2 years 6 months ago #75152

Please Log in or Create an account to join the conversation.

Time to create page: 1.735 seconds