Copter 3.3r8 Wandering Yaw Tricopter

ArduCopter Copter 3.3
1

Hi all,

I have an interesting situation with a tricopter that had otherwise been flying perfectly well in Stabilize, Loiter, Drift, etc. with 3.2.1.

Since updating to 3.3, I had attempted an autotune that I think didn’t work well due to wind and my fairly constantly having to roll/pitch to keep it from getting too close to obstacles.

Now, with 3.3r8 I was unable to fly in stabilize and it was awfully twitchy. Yaw was wandering, and not consistently. So I reset everything to defaults, redid ESC, RC, Accel, and Compass calibration to try the new autotune in 3.3.

The wandering yaw persists. In this test flight, I just took off in the back yard and attempted to keep it at a steady hover. Right away yaw was wandering as if the compass was creeping one way, then I would adjust and get it nose out, then it would wander again one way or another. So even in Stabilize, I couldn’t keep it steadily nose out.

Does this Yaw/DesYaw show anything of interest? You can see it spinning, and this is me just trying to keep it straight.

[attachment=0]Screen Shot 2015-07-28 at 7.56.24 PM.png[/attachment]

It seemed an awful lot like the compass just being way off, but this exact craft was just flying perfectly and compass calibration has always gone just fine. Interestingly, just going between Flight Data and re-doing Accel calibration, there are a few times I would go into Flight Data and see the horizon as if the tricopter was completely upside down. I saw the “Compass Variation” error a few times, and one time a power cycle on the Pixhawk ‘resolved’ it, which seemed strange to me.

I am hoping this jumps out at someone and you can educate me on what it up, thanks!

2

here is the tlog as well in case it is helpful. In the beginning you can see me fiddling with repeating accel (and maybe mag) calibration, followed by crazy “Bad AHRS” and the HUD horizon flipping. Toward the end is the flight, and it appears to faithfully show what I experienced. What it felt like in flight both in Stabilize and Loiter was that it lost its reference point for yaw rotation so I had to constantly adjust the yaw stick to keep it straight vs. where it wanted to go at the time.

3

I just tested it again, only in Stabilize, and it truly is random. I would adjust rudder to compensate, then get it under control, then it would randomly spin somewhere else, requiring a completely different compensation. What does this suggest?? Noise in the servo signal? Why all of a sudden on a craft that had been flying fine for weeks?

Just in case it is meaningful, one change I made was moving the power module to the inside of the frame instead of hanging off the back, and mounted the FrSky voltage sensor to the underside. They’re not really “closer” to the Pixhawk or the GPS/Compass mast, but those changes and this weirdness occurred at the same time.

Compass seems perfectly fine to me.

4

I tested several more times, re-doing calibrations, going back to factory defaults, etc. to no avail.

Finally I reverted to 3.2.1 and touched nothing else, didn’t even redo compass calibration just to keep the changes between tests to a minimum, and it is back to the smooth flier I remember.

No more wandering yaw in stabilize, and solid Loiter even in today’s wind.

Can someone shed some light on the differences in Stabilize for 3.3 that could cause this? It truly felt like it was driving the servo randomly. Is the servo no longer on Channel 7 for the 3.3 version? It did “sort of” work.

I thought it would be interesting to look at RC_IN channel 4 (rudder) and RC_OUT channel 7 (yaw servo). It seems strange to me as it doesn’t directly track the stick input like other logs I’ve seen. What is 3.3 doing to operate the yaw servo in Stabilize?

[attachment=0]Screen Shot 2015-07-29 at 8.04.47 PM.png[/attachment]

5

I a feel you, I am having very similar problems.

Wes

6

[quote=“Just for fun”]I a feel you, I am having very similar problems.

Wes[/quote]

Hey Wes,

Randy replied on the diydrones 3.3 thread :

[quote]It looks to be a control problem instead of a sensor or attitude estimation problem. This means that it can likely be fixed with tuning. Below is a graph of the Desired Yaw Rate (red) vs actual (green) and we can see that they’re not tracking well at all.

I’m far from an expert on tuning but instead of using the defaults, I’d recommend using the MP’s Extended Tuning screen and greatly increase the Yaw Rate’s P, I and IMAX values.
Rate Yaw P: 0.2 -> 0.3
Rate Yaw I: 0.02 -> 0.04
Rate Yaw IMAX: 100 -> 200 (it’s strange how MP seems to be scaling this value, the raw parameter RATE_YAW_IMAX should increased to 2000).
Looks like you’re using the defaults in this flight at least, so try increasing[/quote]

So I will give that a try and report back.

7

I tried to up the yaw I to .04, but mission plan would not let me go higher than .02.

Are you using the stable mission planner or the beta?

8

[quote=“Just for fun”]I tried to up the yaw I to .04, but mission plan would not let me go higher than .02.

Are you using the stable mission planner or the beta?[/quote]

yep, I am using the beta (and currently Copter 3.3rc10). I can make some screenshots of the setup I’m currently using if it would be helpful.