Triflight 0.7 Beta 2

[Kevin_Erik]

@lauka

Any chance of there being something to dampen down tail-wag?

This has become a great source of frustration for me as I know that my Tri-Baby is working perfect otherwise. Only time I can get it to stop wagging is if I set the Yaw I-gain to zero, after which it will slowly rotate in place as there is nothing to counteract the random variations in the coopters Yaw orientation.

Keep in mind that my Tri-Mini runs smooth while the Tri-baby’s tail slowly oscillates in flight no matter the settings, flight controller or Servos used. (Both use the same Firmware, triflight_bf_3.2.2_v2_MATEKF405.hex) I’m not 100% sure but I think that the motor spacing of the copter may have a big impact on the Servo’s performance. Furthermore the Tail-Mechanisms Yaw CG (Center of |Gravity) is higher than the Y-axis of the Flight Controllers IMU. Which i believe leads to the mass being sloshed from side to side in such a way as to cause the copter to Roll slightly off its axis. This condition can may cause the Wag to feedback on both the Yaw and Roll axis. The amount of Roll is dependant on both the CG difference and the rotational mass of the tail motor / propeller.

Note: Motors are perfectly balanced, showing less than 0.001 % accelerometer deviation at 50% power. Motors run with almost no filtering and only just get warm after a minute or two of hovering. (LS2207-1900KV @ 4S and sometimes 6S, with 5x4x3 Dal props, 4-In-1 ESC running DShot-1200)

Things tried:
-With and without notch Filtering.
-P, I & D changes from 0 to 100.
-Added caps to main ESC but still this should be relatively clean if it’s just hovering. (Still wags though)
-Added filtering to the servo and or Yaw specific filtering.
-Lowered / Raised the Tri_Yaw_Boost with very minimal effect.
-Changed the Servo update rates.
-Changed the Gyro / PID loop rates.
-Isolation pads for the motor and flight controller.

Only things I’ve not tried:

-Invert the Tail-Mechanism and motors, effectively flipping the entire copter. I’ve read somewhere that this was effective but the amplifying details and follow on video were absent.
-Change the D-Shot to Multishot as I know that works on the Tri-Mini.

Video of flight: https://www.youtube.com/watch?v=WqnOFVTbxAs

Cheers!

Kevin

[samp38]

Hi again,

Thanks Lauka for your answer. Here the test :

0.7beta3 code with one little modification to log the triGetCurrentServoAngle return value. It is stored in debag[2] when TT is ACTIVE. I commented the line 814 and added
DEBUG_SET(DEBUG_TRI, DEBUG_TRI_YAW_ERROR, triGetCurrentServoAngle());
after line 527 of mixer_tricopter.c

Servo bench tune ok
servo reversed with cli command

Small flight with TT activated a few seconds after take off. I waited for success beep code and landed/disarmed. Then, as expected, i had the TT fail beep code so i turned TT off and unplugged the battery.

As you can see in the log the angle during TT is around 85 and can’t meet the condition line 556 (between 90 and 120). That’s why reversing the return value of the feedbackServoStep function who sets tailServo.angle worked with beta2. For some reason it doesn’t with beta3.

I had bad fast servo wobbling on arming, with throttle low (motors spinning at min speed). It disappears as throttle rises but 0.7 beta2 doesn’t produce that effect.

Can this angle shift be the reason for the low throttle wobbling and the bad yaw behavior (only with beta3, beta2 seems ok)?

Thanks :-),

link to log :

https://drive.google.com/open?id=1G1KWhYqs4rpX2cY6LxqNoouNi8a-kIY1

Attachments:

1. diff-1.txt

[Kevin_Erik]

I see from your log that the RSSI is set to channel 11. What are you using for Servo feedback?

NM i see it…

[samp38]

I use CURRENT. I used to have analog RSSI but now I use telemetry.

[samp38]

When I change the line 426 of mixer_tricopter.c of the beta3 version to :

const float currentAngle = ((endAngle – TRI_TAIL_SERVO_ANGLE_MID) * (-(ADCFeedback – midValue))
/ (endValue – midValue) + TRI_TAIL_SERVO_ANGLE_MID);

instead of

const float currentAngle = ((endAngle – TRI_TAIL_SERVO_ANGLE_MID) * (ADCFeedback – midValue)
/ (endValue – midValue) + TRI_TAIL_SERVO_ANGLE_MID);

My angle is around 95 and I can complete the TT. But I still have low throttle servo wobbling, that I don’t see with beta2. I tried lowering Yaw PIDs but no change. At the end of yaw the tail motor gives a punch so the copter pitch forward. This behavior is no shown with beta2.

[samp38]

Here is my diff and a log of a little flight just after successful TT with modified 0.7beta3 version. I tweaked tri_motor_acceleration to 0.3 so it seems to better keep orientation during punch in. But as you can see the servo oscillates at punch out and the tail motor jump at the end of fast yaw rotation.

link to log :

https://drive.google.com/open?id=1ChsH1xkwLtNpQlCgZGaMBQ2s3_21m6Hc

Attachments:

1. diff-2.txt

[Kevin_Erik]

I may opt to roll back to 7 Beta 0.2 just so i can see if the Tail problems persist or not.

[MadPoet]

I had the same tail tune problem with beta 3 on my Baby Tri. I’ve also noted that beta 2 succeeds without an issue.

[lauka]

@Kevin_Erik: Do you have a blackbox log available from a flight with the baby?

[Kevin_Erik]

@lauka

It’s on the 3.2 page along with a video and Diff report.

Testing – Triflight rebased on 3.2.2 Betaflight

Keep in mind that I’ve had this configuration all over the place and have never found anything allowing the tail to remain stable in flight. (groung wag is managable) Originally, I was running the Matek F405-AIO FC and thought that the 32k gyro was making the tail freakout so i swapped to the F405-CTR with the 6000 series gyro. Aparently, it’s not the FC nor the servo’s used, as I’ve tried three different servos thus far. Currently have the Bluebird BMS-385D installed with feedback wire attached to the RSSI pad.

Note: only time I’ve had the tail stable is when the Yaw I-Term is at ZERO. Even a setting of “1” will cause the jitter / wag to return no matter what the other system settings maybe. I could be wrong here but the I-term shouldn’t be causing the copter the freak out.

[Kevin_Erik]

If needed i can DEFAULT the firmware and do a quick flight to produce a baseline LOG to compare with.

[lauka]

@Kevin_Erik: Okay, the I-gain is a good find. I’ll concentrate my investigations on there then :). There are some new functionality that was implemented for the I-term for quads. Not saying that the fault couldn’t be in the tricopter code.

So the tail oscillation is completely removed when I gain is at 0?

[Kevin_Erik]

@Lauka

Yes, the tail is totally smooth but obviously I can’t fly it like this as the Tri would just spin CCW in place as there is nothing to counteract the motor torque. Yes, I tried it once and got about a meter off the ground before quickly realizing that I’d not be able to manually compensate quick enough.

Note: I am about to reflash the Baby and set everything to Defaults. Will only add whats needed to get it working and then attempt to fly. This should provide you with a solid baseline vs what I submitted previously.

[lauka]

Could you also get a log with the I term at zero?

The copter shouldn’t spin too fast if the thrust factor is set correctly, obviously not flyable still.

[Kevin_Erik]

@Lauka

Sorry but I can’t… the USB connector for the FC just came off.

[Author]

Posts