Debugging the tricopter mini racer

[bhuism]

Thanks @planb, worked like a charm, tri_tail_servo_speed went from

# get tri_tail_servo_speed
tri_tail_servo_speed = 300

to

# get tri_tail_servo_speed
tri_tail_servo_speed = 263

Now I now servo feedback is working 🙂 thanks!

I still need to do it in the air I guess 🙂

[RCExplorer – David]

Absolutely love the new firmware! The rewrite PID controller is fantastic! Great tune as well from Lauka. Flies fantastic on 3 blade 6×4.5 props as well 😀

Attachments:

1. 

   12794478_573924866108503_5075260160485782321_n.jpg
   

[swissfreek]

6×4.5x3s?! Wow, is that with 30A ESCs? I was going to swap my 6×4.5s for 5×4.5x3s when I get back to my Tricopter in May, but maybe I should adjust my plans 😉

[dustin]

I don’t ever fly anything without a current sensor. With 6045 on 2208s, I get > 140A on a quad: http://gfycat.com/MaleColorfulEgret — so that’s around 35A per arm. My experience with x3 is… more current.

[RCExplorer – David]

Didn’t get that much more current draw with the 6×4.5×3 props. The 1800kV motors have plenty of torque to drive them. Motors came down cold as usual 🙂

Yes running LittleBee 30’s all around 🙂

[Joost]

If you’re running Cobra 2208 2000kV they might run a bit more hot, but shouldn’t be too big a difference. Those are still HQ props? Glassfibre reinforced nylon?
Just got my bags of DAL 6×4.5BN so I’m curious if they are just as responsive/generate similar thrust. Have been eyeballing triple bladed props for a while now, so will probably get them before the end of the summer.

[hertz]

I want to try HQ 4-bladed props as well, maybe can even go with 5″ ones.
David, when are going to get an integrated F3/F4 flight controller for the tricopter? 🙂

[swissfreek]

Well then. I was going to upgrade to the LittleBees anyway. Now I have another reason. Oh David, always finding ways to take my money. I love it!

[PlanB]

Back to an issue mentioned earlier.

There is definitely some bug with virtual servo on F3 flight controllers. I’m almost 100% shure that it isn’t hardware problem. I’ve changed flight controller to RMRC DODO, changed rear ESC (yes, it’s calibrated properly), but same issue with pitch axis arises – tricopter tilts forward on takeoff and on quick trottle increase.

Below are screenshots of logs of two consequent flights, one with “tri servo feedback = VIRTUAL”, other with feedback wire.

I term on pitch axis has constantly high value when virtual servo is used. That means it compensates some constant force (excessive thrust from tail motor in this case). When feedback wire is connected, I term is close to zero while hovering.

As seen from logs, tail motor initially gets more thrust from flight controller, which is later compensated by PID loop.

Also I got unusual results from TailTune: “tri_tail_motor_thrustfactor” value is around 60, while normally it should be around 95.

Triflight usage has gone wider than just RCExplorer specific firmware, a lot of people, including me, are using it on other tricopters with different flight controllers and it’s very discouraging to get such not evident misbehaviour.
While nobody else reports this error I may presume that it’s some configuration error on my side, but I have no clue what.

Logs:
blackbox_log virtual servo
blackbox_log servo feedback wire

Attachments:

1. 

   virtual-servo-1-expl.png
   

2. 

   virtual-servo-2-expl.png
   

3. 

   virtual-servo-3-expl.png
   

4. 

   feedback-wire-1-expl.png
   

5. 

   feedback-wire-2-expl.png
   

6. 

   feedback-wire-3-expl.png
   

1. blackbox_log-servo-feedback-wire.txt

[lauka]

@PlanB: very good investigations on the issue! I think I will unit test the virtual servo function and see if I can find something. If someone else has time to check thay would be awesome, I’m quite busy atm. Leo got time? 😀

On what fw versions have you experienced this?

[PlanB]

That’s happening on 0.5 beta 2, RMDO and SPRACINGF3 targets are affected. Strangely, on beta 1 there were no such a problem.

settings dump

BTW, thank you for all your efforts, Lauka 🙂

[lauka]

Thanks for the feedback :).

If it’s only on beta 2 it must be the new integer based virtual servo implementation.

Checked the code quickly on my phone and probably spotted the issue.

I’ll try to release a new beta soonish when I have time to check it more thoroughly.

This is why we have betas 🙂

[LeoTheHuman]

Hey guys!
[offtopic]
I’m back from my get-away from Russian winter in Florida, had quite a practice there with its sunny and warm days. Here are a couple videos from there:


Unfortunately, it was not logistically wise to bring the tricopter with me (needs larger batteries and breaks on crashes more than Armattan F1-4B, unfortunately), so I was only using quads at that time. I’m still not at home and plan to be there in a week. Weather here in Russia got better for flying, so…
[/offtopic]
I will be back to testing in a week or two. Will take a look at the issue if it persists, but, unfortunately, I have no F3 boards or tricopters not from @David, so I’ll have to test bad configs on my mini trying to reproduce the problem, or just stare at the code, which I’m quite good at. (:

[PlanB]

Hi, that’s me again 🙂

Another problem that I faced with is heavy tail drift on throttle drop. Any time i punch the throttle and then drop it to a minimum, tail blows away counterclockwise. I can not correct this with PIDs no matter what i try. Worth to mention I’m using braking ESCs (it’s important in this case).

I investigated logs and got some conclusions, let me explain…

I found that on fast throttle drop tail drifts CCW, probably, due to a force caused by ESC braking.
I mean that when throttle is increasing and motor is accelerating, counteracting force rotates tricopter clockwise and PID loop successfully corrects this, because there is enough tail authority on high RPM. When you suddenly drop throttle to a minimum, ESC brakes, motor decelerates, and counteracting force rotates tricopter CCW, but that can’t be easily corrected due to a luck of thrust.

Just in case – tri_tail_motor_thrustfactor has nothing to do in this situation, it only compensates aerodynamic forces caused by propeller moving through the air, not forces caused by acceleration/deceleration of motor and propeller.

@Lauka, you wrote that Triflight 0.5 beta 2 uses integrator disable instead of the original Cleanflight Air mode one?
I really have no clear solution to propose, but I think that active I-term may help, disabling integrator in airmode is probably not best solution.

What do you think about this, guys?

@Lauka, it would be absolutely great if you could make firmware build with default cleanflight airmode behaviour for me to try. I can check it on saturday and prove or deny my findings.

Attachments:

1. 

   yaw-integrator-issue-expl.png
   

[lauka]

Hi again 🙂

The tail drift after throttle drop is a known issue that’s next on the list to try to fix. I think it’s exactly because of what you said, the braking of the rotating mass + loss of authority. I like the way you’re thinking. I have been thinking of using dynamic servo rate based on the tail motor speed. When it’s spinning slowly the servo would move lot more. It might help but it’s hard to fix when there’s no authority. Maybe in that case we could allow servo to travel fully left to help with the counteract.

Would be good to test to get some altitude and let the copter drop for longer time. Does the drifting slow down? In other words how much of it is because of the motor braking? In that case another thing would be to try to model the effect of the motor speeding up/slowing down to the yaw axis, and input this as an ecpected error to the PID.

Faster servo and higher min throttle should also help. The latter reduces the inverted hang time.

I tested the default integrator functionality of cleanflight and it does not support tricopters well. The I term windup issues that caused drifting were back.

The integrator handling does not affect in this case, it’s only in greater y/p/r stick deflections or when the gyro error grows too big.

[Author]

Posts