Yaw tuning with high inertia frames

ArduCopter Copter 4.5
1

I’ve got an unusual 7" quad that I’ve slowly been modifying with kind of unusual additions to the frame. The additions add a fair amount of mechanical and aerodynamic inertia to it - much like a quadplane. Lots of pitch and roll aero damping to go along with the added angular inertia in those axes, but in yaw, there is a ton of added angular inertia, but essentially no aero damping. It is tuned pretty well without any of the additions. With a small addition, it still flies pretty well, but yaw control feels pretty soft. With larger frame additions, it pretty much becomes unflyable. It hunts in yaw continuously. I’ve only attempted to fly it in stab mode, and only for brief moments.

It’s using rather low kv motors and flat-pitch props for 7" (1500 kv on 3S with 7x3.4 props), so very much not a racing copter. I’m not sure if this is a PID problem or a filter problem. I haven’t worked with quadplanes at all, but I suspect, this is similar to yaw issues they experience.

2

It sounds like motors might be getting pushed to maximum, or close, so yaw is suffering.
A .bin log would be handy.

3

Sorry for the slow response. I’ve been swamped, but finally dug the logs out. The mean PWM to the motors is not bad, but there is absolutely divergent yaw behavior that cyclically pushes the motors between min and max. I think it’s a low-pass filter problem in yaw due to the super high angular inertia. There’s a couple of very brief attempts to hover, with commanded and desired yaw shown. I’m currently using the default filter setting for 7" props, and those still seem alright for pitch and roll, but not for yaw.

image
image1204×568 32.5 KB

4

Can you upload that log to a file sharing service and provide a link to it?
Filters are unlikely to be the cause

5

Humm, I just made a super short flight to demonstrate the situation. The bin file is about 1.5MB. The upload tool claims it is bigger than the 4.4MB size limit. Not sure I understand why. I need to find a site where I can share it I guess.

6

Try this:

https://www.dropbox.com/scl/fi/l0k7to07pivjai4ppv8xi/00000035.BIN?rlkey=ssmx2m2a9bydh9fgw8iztpc4y&st=pk5oabwf&dl=0

7

Motors are oscillating and attitude control is quite bad. Yaw might be the least of the issues.
I’ll keep looking and come up with a plan.

These should suit your copter with 7inch props, at least as a reasonable starting point WITHOUT the additions you speak of.

MOT_THST_EXPO,0.54
ATC_ANG_PIT_P,15
ATC_ANG_RLL_P,15
ATC_ANG_YAW_P,8
ATC_RAT_PIT_D,0.0018
ATC_RAT_PIT_I,0.10
ATC_RAT_PIT_P,0.10
ATC_RAT_RLL_D,0.0015
ATC_RAT_RLL_I,0.08
ATC_RAT_RLL_P,0.08
ATC_RAT_YAW_D,0.02
ATC_RAT_YAW_I,0.02
ATC_RAT_YAW_P,0.20
INS_HNTCH_ENABLE,1
INS_LOG_BAT_MASK,1
INS_LOG_BAT_OPT,4
LOG_BITMASK,180222

Reboot the flight controller then set these:

INS_HNTCH_BW,50
INS_HNTCH_FM_RAT,1
INS_HNTCH_FREQ,100
INS_HNTCH_HMNCS,1
INS_HNTCH_MODE,3
INS_HNTCH_OPTS,0
INS_HNTCH_REF,1

Let’s go from there and see the results of another flight test.

Do you have 2 DSHOT ESCs on servo outputs 5 and 6 ? Is this related to the additions operated off RC8 ? Care to divulge any secrets (or what the additions are intended to do) ?
Maybe a photo…
I would tend to get everything working correctly and well tuned before working on the “additions” and that way you are starting from a known-good state.

8

Perhaps it’s worth posting a log for the same copter, but with a lighter weight frame addition, as it does not exhibit this run-away yaw behavior.

https://www.dropbox.com/scl/fi/2cmyn5vqvtpaec3wssf0m/00000033.BIN?rlkey=bl4fejzf6vovj6gtap1y9hcju&st=o0ih9w4y&dl=0

Both of them have quite a bit more angular inertia in yaw than a generic quad, but the one in log 33 has considerably less than the one in lag 35.

Yes - there are two more Dshot ESCs not doing anything in this flight. This is a student research project, where the two additional motors have small props (63mm) to provide lateral thrust. We can either drive them manually through channel 8 and a slider on the Tx, or, the student research bit, through a lua script running a PID controller to drive pitch to zero in cruise flight. We’ve had great success on a slightly more race-like airframe.

1 Like
9

I’ll check the extra log later.
If the motors are operating near full output then yaw will usually be the first to suffer. There could have been some of this in the other log, plus general instability.

10

I think the motors hover at around 50%. I think the only reason we are seeing max throttle is due to the runaway yaw PID loop. In log 33, for the lighter version, it hovers at about 40% throttle, and the heaver one adds about 60 grams, for a gross weight of 711 grams.

The quad itself is pretty lanky, with pretty significant gaps between the props - perfect X. The frame addition is a large plus-shaped 3D printed part with plates extending between the props. The heaver one extends further, and supports CF strips that act as prop guards, with a circular CF strip surrounding the entire thing. Here’s a crude CAD rendering.

image
image1847×1052 312 KB

I’m pretty sure it’s all that mass out at a large radius that is causing problems for the yaw controller. In yaw, I think I need to treat it more like a 10" or 12" quad for filters and acceleration.

Alternatively, I may try to laser cut a foam-board substitute for the huge printed part. Even with the CF strips, it will likely weigh half what the printed shell weighs. Of that 711 gram flying weight, 170g is that printed shell.