Hi Jasem,

i made a quick indoor-test, these logs cant be right?


with Prop 1.0

AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint 865 Measurement 850 Error 15 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 15.000000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I1AA4610> "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS2 Setpoint 69345 Measurement 69334 Error 11 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 11.000000 I: 0.000000 D: 0.000000 "

AZ-GTi Alt-Az WiFi : "[SCOPE] New Tracking Target AZ 0.158621° (913 microsteps) AL 12.039262° (69346 microsteps) "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint 913 Measurement 898 Error 15 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 15.000000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I1AA4610> "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS2 Setpoint 69346 Measurement 69347 Error -1 Rate 1941967896 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: -1.000000 I: 0.000000 D: -0.000000 "

AZ-GTi Alt-Az WiFi : "[SCOPE] New Tracking Target AZ 0.161405° (929 microsteps) AL 12.039272° (69346 microsteps) "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint 929 Measurement 914 Error 15 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 15.000000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I1AA4610> "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS2 Setpoint 69346 Measurement 69348 Error -2 Rate 1941967896 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: -2.000000 I: 0.000000 D: -0.000000 "


Logs with Prop 1.8

AZ-GTi Alt-Az WiFi : "[SCOPE] New Tracking Target AZ 359.719199° (2071982 microsteps) AL 12.041549° (69359 microsteps) "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint -1618 Measurement -1627 Error 9 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 16.200000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I106120F> "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS2 Setpoint 69359 Measurement 69370 Error -11 Rate 1941967896 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: -19.800000 I: 0.000000 D: -0.000000 "

AZ-GTi Alt-Az WiFi : "[SCOPE] New Tracking Target AZ 359.721986° (2071998 microsteps) AL 12.041537° (69359 microsteps) "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint -1602 Measurement -1607 Error 5 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 9.000000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I171201B> "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS2 Setpoint 69359 Measurement 69349 Error 10 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 18.000000 I: 0.000000 D: 0.000000

AZ-GTi Alt-Az WiFi : "[SCOPE] New Tracking Target AZ 359.724772° (2072014 microsteps) AL 12.041524° (69359 microsteps) "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint -1586 Measurement -1596 Error 10 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 18.000000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I138900D> "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking -> AXIS2 direction change. "

AZ-GTi Alt-Az WiFi : "[SCOPE] New Tracking Target AZ 359.727559° (2072030 microsteps) AL 12.041512° (69359 microsteps) "
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS1 Setpoint -1570 Measurement -1576 Error 6 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 10.800000 I: 0.000000 D: -0.000000 "
AZ-GTi Alt-Az WiFi : "[SCOPE] CMD <I1099B16> "
Image received. Mode: "Guide" Size: 2466240
AZ-GTi Alt-Az WiFi : "[DEBUG] AXIS2 Setpoint 69359 Measurement 69327 Error 32 Rate 1941967888 Freq -677309488 Dir ��{è�_��{��� "
AZ-GTi Alt-Az WiFi : "[DEBUG] Tracking AZ P: 57.600000 I: 0.000000 D: -0.000000 "

(But also good news -> Fixed Rate seems to work indoors -> now i really can slow down the speed -> i will test this evening under real sky with fixed rates)

Hi,

tested yesterday under real sky - unfortunately it didn't work. The drift was there and i could not overcome it with the prop settings.

Michael, can u confirm that too?

Regards

I built and tested the latest version last night indoors. Like you mentioned, there's an issue with the logs due to the data type definitions for tracking rate and potentially error. I didn't test outdoors, since I figured having the tracking rate and frequency unavailable would be working in the blind. Once we get the data type definitions corrected and we tune the PID such that we reduce the error to zero,and we still have drift issues, I have one or two more suggestions which may help.

Setting the tracking rate manually will be difficult to "guess" as you might have experienced last night, however if we get the PID portion right, it would at least get you in the ballpark and you could adjust by 1000 hz increments or so to dial in.

I tried again for two hours - no luck.

@jasem can you take a look at the broken logs, so that we can eventually try Michaels other ideas. That would be great!

I pushed a fork which Jasem can correct if necessary. I tested it, and it appears to have fixed the logs enough for us to continue testing.

Here are the changes:https://github.com/indilib/indi/pull/1874/files. To be incorporated in the master, it has to be approved in the github workflow. In the meantime, you can modify the .cpp file as I did on the link I provided and re-build the driver.

For testing, try starting with the proportional term of the PID at ~0.35 on the gain and moving up very slowly until you see the error essentially bounce from around 0 to 3 or 4 (or as low as you can get it). Then, slowly increase the integral term from 0.05 until you are near 0 error consistently. Finally, use the derivative term to get consistently zero. Be sure to wait, as it isn't instantaneous and the tracking improves with time. Here's a pretty decent explanation of PID "theory" if you aren't familiar with it: www.ctrlaltftc.com/the-pid-controller/tu...-of-a-pid-controller.

I will test this tomorrow if the skies are clear as I am about to hop off, and go to bed. We need to test this thoroughly before we move on.

Thank you. Since most users aren't expected to fine tune the PID parameters themselves, is there a way to start with sane-default and then perhaps automatically tune it to reduce the errors based on your observations?

No problem! Only a small contribution compared to the work you've put into this effort. In regard to your question, yes, I think we should be able to get to the point where the user doesn't have to touch these settings or very little as you mentioned by letting the program automatically tune the PID based on a default initial guess.

I am providing the settings I tuned with this morning which haven't been tested in the field but was able to get error down to zero. Logs are attached for reference. Mat, could you try these settings, and see if get similar results for error? My plan is to test these settings visually if skies are clear this evening.

Hi Michael,

thank you for your changes on the driver - ofcourse i tested your changes and routine already this morning indoors

I was able too to correct error down around 0, but had to go higher on Axis1 (1.8 - 0 - 0,05) but i will try your exact settings indoors and i will try on real sky as soon as possible. I noticed that is super important to wait after making a new settings to settle down in the logs, as you mentioned. in the past I was probably always much too fast.

I'm very curious what the results will be!

That's good news that you were able to get the error down, Mat.

I've attached some data analysis from the data I got this morning that I had submitted in a text file earlier. It's an excel file but I had to zip it.

When looking at the data, I noticed two things offhand (Note. I haven't fully analyzed the data yet). Firstly, it appears Axis 1 may be polling less than Axis 2 for the PID routine? 1 second versus 2 seconds. This was going to be one of my suggestions I mentioned a few days ago (adding back the adjustable polling until we figure out how often the PID should calculate the error and adjust the speed), however I wasn't expecting to see a different delta t for each axis. I was having a bit of a time with getting axis2 to settle down during the initial tuning of the PID which this may be a contributing factor.

secondly, it appears we may need to add some logic to exclude tracking factors which are less than 1 as this will produce a frequency greater than 16mhz. The program must have interpreted the excess speed as the maximum speed and set it accordingly. This may be a moot point, but something to look at later if we want the system response time to be quicker upfront.

If you have bandwidth, take a look at the data and see if you spot any other potential issues. Thanks

**UPDATE** ---> also note difference between the max/min frequency adjustments versus elapsed time between axis 1 & axis 2 even with throwing out the starting data points (Axis 1: 50000 hz difference versus Axis 2: 11,900,000 difference). Could go back to the polling frequency difference between the axes.

I tested the new version yesterday. I think it needs a lot of tests to finally confirm it, but it went very well for me.

It takes a long time to get close to the 0 Error value, I've always been way too fast here in the past. I startet with your values, but for Axis1 i had to change it:

1.8 - 0.01 - 0.05
0.75 - 0.01 - 0.13

(the tuning took me about 10-15 minutes)

In the end I was able to achieve that the error value was around zero (+-2). So it was possible for me (for the first time) to take a 60 second single frame (unguided):



(Raspberry Pi HQ Cam - single 60 Sek with Omegon 76-342)

After that i turned on guiding and it went well for over 7 minutes (than i stopped):



Polling was here at 500ms in the options, but i didnt saw a big difference.

I think that were the best results i ever hat with IND AZ-GTI(x) combo. I think we need to do a lot more testing to make sure it works reproducibly. But I'm confident and excited about the progress.

MAT, that is awsome image! I see that you are using HQ Camera for imaging? Are you imaging with Ekos? Which INDI driver are you using for HQ Camera?
Will this awsome work help with guiding with AZ-Gti EQMod connection too?