Tradheli Autotune RC2 - Alpha level testing

@bnsgeyer is the man to ask what best practice is for Heli, but I my attitude with any aircraft is to treat it like it is trying to destroy itself or me until I have tested and validated each part of the tune and dependent modes.

1 Like

Just a quick update @bnsgeyer , finished all balancing and tuning, still getting high VIBE in hoovering.
Looked for the difference between the two helis that work fine and this troubled one, found that this is the only one without shock absorbers installation under the Pixhawk, need to install one :slight_smile:

Hello Bill @bnsgeyer, Finished the new pixhawk shock absorber installation, and all the necessary alignments and tuning.
A big difference :slight_smile:
had some tail wagging, had to decrease yaw P and D rates gain, still can see some fluctuations on the PIDY target versus actual. but ot seen on the heli tail movement.
Is there anything to do to eliminate it ? can i proceed to auto tune ?

Attached a .bin file of very low hover with some ground touches.

https://drive.google.com/file/d/1wMgFjBGzzoLivDGEc7K-FWsIt9umM5ru/view?usp=sharing

@ZvikaF Glad to hear that you were able to quickly get back into the air. Vibes look acceptable and the IMUs are tracking each other. How high are you flying? Your altitude is barely showing anything. Looking at the vertical accelerations, they are ok. Did you make sure your blades are tracking well? I believe I am seeing a 1/rev or 2/rev in there. By the way donā€™t forget to add fast attitude and PID to your LOG_BITMASK settings. The attitudes seem to be tracking well enough to start the autotune. Yaw looks good enough as well to start autotune. Iā€™m surprised that your rate P gain is that low. Do you know the rate P gain value where it starts to oscillate?

Do another flight to look at althold and be sure it isnā€™t having issues again before doing the autotune. When you do start conducting autotune testing, Please be sure that you are flying in an area that is at least 50m by 50m. In the Max Gain, Rate P and Rate D testing, it will wonder a bit and you donā€™t have the ability to control it while it is testing like you do with the Angle P test. So if you make any inputs to the controls it will stop the test and give you control and start the test again once it sees the controls centered for 2 seconds. I would suggest starting with the yaw axis to see if it can tighten up the tune in that axis.

@bnsgeyer Thanks for the quick response, as i have some stock of spare parts for my fleet of 450, it was only a matter of dis / assembly time. this one needs a mast of 108 mm and I had only 111, so i had to extend too much the horns, and apparently, the new must is also slightly crooked, this might gives the 1-2/rev you see. tracking is fine, ordered the 108mm mast that will be changed on arrival.
Dew to a gusty day and tight area, hovering was at about 20-50 cm above ground, the high Z acc. readout seen, happened when slashed to the ground.
Having RPM telemetry on my radio (the newest Yappu LUA) helped a lot, the ESC seems to respond differently for every flight. this may explain the powered decent that had me soft crash the heli, during the last auto tune session a few months ago,.

Took the yaw P + D rates down to stop tail hunting, will check razing P influence.

Had fast attitude and PID bits on (65533) will add IMU_FAST as well, do you need IMU_RAW ?
changed the LOG_BITMASK to 425981 (skipped camera).

My autotune flights are done in a remote open area of few hundred meters, but the local pre checks is done in house at a tight placeā€¦

Thanks again for your effort and knowledge sharing, hope to bring some helpfull data on the auto tune.

:+1:

Thanks to you all for testing.

Interesting. I guess I didnā€™t remember seeing the PID message. No I donā€™t need Raw IMU. I will look again. For some reason the data looked like it was at a lower sample rate. As long as you have the default bits as well as fast attitude and PID, then I am happy.
Thanks!

1 Like

Hello Bill, @bnsgeyer

I tried the Autotune RC2 today with a 380 size heli and wasnā€™t too successful, only managed to do the VFF tune on pitch.

Used 4 batteries (14mins each) trying to do the ā€œMax gain, Rate D and Rate Pā€ part of the tune on pitch and found the 23 second sweep is just too long for small heliā€™s. Is there any way to break this up? Without some kind of control inputs or position hold/return in the 23 seconds window, our small heliā€™s just drift to fast or have gone to far see. Lol

@Steve_Mitchell Hi Steve. Glad you had a chance to try it. Yes I think the VFF tuning is difficult for the smaller heliā€™s. They donā€™t tend to reach a steady state rate and that causes issues for the tuning. You should know what a good value is for your heli. I would continue with the max gain, rate p and rate d tuning. If you need help determining the VFF then send me a log of it doing that test and I will give you what I think should be the right number. Post the logs in any case so I can see what it is doing.

As for the max gain, rate p and rate d tuning, I realized that it was pretty long and added velocity feedback but that must not be helping. One other reason that could be happening is due to the swash servos not being trimmed so that there is no I term being used when holding a hover on a calm day. But I still think the trim should have helped and kept it close to level.
Please post you logs so I can have a look. Thanks for testing.

Try the yaw axis. Maybe it will hold stationary long enough to get through the sweep.

@bnsgeyer Okay, Iā€™ll give it another go in a bigger area (200x200m). Also, will try the yaw axis first.
This is the VFF tune flight. I had no problems here. Was fast and easy to do.
https://drive.google.com/file/d/170IqFLDcIMlyhkxvtq4DAXd-uC76osr4/view?usp=sharing

This is one of many flights trying to get just the (2) max gain, rate p and rate d tune done and falling. It just drifts off too fast.
https://drive.google.com/file/d/1aco6PR519u3c-XRB08XPG77_J300wPta/view?usp=sharing

Heli specs:
Heli Size, 380
Number of main rotor blades, 2
Main Rotor Diameter in meters, 0.843m
Tail Rotor Diameter, 0.195m
Takeoff Weight,1.900kg
Rotor Speed (RPM) =2220

I was fortunate that I was able to complete the sweeps within a 10-15 meters. It was tricky as it tended to drift in one direction so I had to start it drifting very slowly in the other direction. It may have taken me 2 to 3 times but I was able to complete it. But I was working with a very small area (20m x 30m).

You can try FF but I have it set so that if it determines a value of VFF less than 0.025 (i think) that it will just set it to zero. Just be on the look out for that. for the Sweeps, I wouldnā€™t let it get going too fast. No more than 3 m/s. a slow drift is ok.

Glad to hear this. I will have a look. I thought this was where you had a lot of difficulty as well. Glad to hear it went better than I expected. I will take a look

1 Like

Testers @JoshW, @ZvikaF, @Steve_Mitchell, @Murdoch
Give me a day before you do anymore testing. I have an improvement that will help the aircraft maintain position better while it is conducting the max gain, rate p and rate d frequency sweeps. I would like you to use the newer version instead.

Bill

1 Like

Thanks Bill @bnsgeyer
Iā€™m on hold now anyway, waiting for the correct mast part before going out to try the auto tune

Testers @JoshW, @ZvikaF, @Steve_Mitchell, @Murdoch
I have completed my SITL testing of the latest changes. They work much better to help hold the vehicle position while doing the Max Gain, Rate P and Rate D tuning. I have provided a gain (AUTOTUNE_VELXY_P) that can be adjusted to help hold position. I defaulted it to 0.1. Try that first and then increase in 0.05 increments. I donā€™t have a good feel for how high you would really want to make this. 0.2 seemed to work well for my 600 heli.
Also I have put an angle protection on the autotune. It will kick you out of tuning and back to waiting for level if the vehicle roll or pitch attitudes exceed 30 deg. The might happen if you had 3 m/s or more velocity on the aircraft when it went into autotune. So be sure your ground speed is near zero when it starts the sweeps or dwells. If you see this more than once, then pull the log and post it so I can take a look.

I plan on making the binaries tomorrow night and should have them posted for you on the release page. I will probably append an ā€œrc-2.1ā€ to the name so you know they are the new versions.
Regards,
Bill

2 Likes

@Steve_Mitchell Had a chance to look at your data while building the firmware for each of the boards. I appears that my VFF tuning did not do a very good job. It is interesting to see that the steady actual rate is matching the steady desired rate however when you look at your other flight where you were tuning rate P and Rate D, it is clear that it isnā€™t near enough VFF. Here is one of the VFF tuning tests.

However looking at the initial gain for the frequency sweep, it is too low.


Here you can see the actual rate not matching very well with the desired rate. I would recommend initially setting ATC_RAT_PIT_VFF to 0.1. Right now that is a rough guess based on what your data looks like. Start testing with the Max Gain, Rate P, and rate D tuning. I want to see if first my improvements help you get through this tuning quicker and how the gain looks at low frequency.

Testers @JoshW, @ZvikaF, @Steve_Mitchell, @Murdoch
Iā€™ve updated the release page with the new firmware. This should work much better to help hold the vehicle position while doing the Max Gain, Rate P and Rate D tuning. I have provided a gain (AUTOTUNE_VELXY_P) that can be adjusted to help hold position. I defaulted it to 0.1. Try that first and then increase in 0.05 increments. I donā€™t have a good feel for how high you would really want to make this. Only increase it such that it minimizes drift during the Max gain, Rate P and Rate D tuning. Also I have put an angle protection on the autotune. It will kick you out of tuning and back to waiting for level if the vehicle roll or pitch attitudes exceed 30 deg.

2 Likes

@bnsgeyer Thanks for taking a look at my flight logs and adding hold position helper. Will test your recommended changes this weekend.

Looking at my params, I noticed ATC_RAT_PIT_I saved at 0.02873141. What part of the Autotune was this change made from the default 0.1 ?

When I do the yaw tune, Iā€™ll set ATC_ACCEL_Y_MAX to 54,000 instead of the defalut 27000, Is that right?

When it tunes VFF, it changes the I gain to 0.5 the VFF gain. You can set that to 0.05 for pitch since I am having you set VFF to 0.1

Yes. You could probably set it even higher. Before doing autotune. Set to 54000 and fly it with out autotune, then bump it up to 75000 and fly that without autotune. Then run autotune with that value. I think we have been overly conservative with this because we inherited the multirotor default and no one has questioned it.

1 Like

Well, almost no one :slight_smile:

Hi Bill and Tradheli people,
The ATC_ACCEL_Y_MAX as calculated by Leonards tuning guide for multirotors (and the spreadsheet and the MP plug-in) seems quite conservative and is probably just a safe starting point. Usually multirotors can easily use 2 or 3 times the calculated value, particularly for smaller quads. The default values based on prop size probably suit carrying a payload where acrobatics are not required. This can be left in place if you dont feel like tuning yaw for some reason - with multirotors we usually dont need a lot of yaw tuning, or even any, unless you see an issue or want to get everything just right. Pitch and roll tuning has the biggest effect of course.

The default value of 27000 suits multirotors with 10inch props, so maybe Tradheli could get a more suitable value loaded by default. Iā€™m not certain about how that would happen.