Autotuning Difficulty on 10" Props

ArduCopter Copter 4.7
1

I’m having difficulty autotuning my two 10" quadcopters that have the same low-vibration frame. One is 6s2p and the other 8s2p. Both have the same Matek H743 slim, one RM3100 compass, and a second compass less immune to EMI and iron. Both, after Magfit, then selecting only the rock-stable RM3100 compass, reducing PIDs to 59% of defaults, experience yaw and opposing-axis jerks during autotune of roll and pitch which never finish. In one case, autotune kept increasing roll D endlessly. In all cases, I could only finish autotuning yaw E and D.

Before I commit to learning how to manual tune, I’m looking for opinions and/or guidance on what I might attempt to do differently to fix my issues with autotuning my two 10" AOS UL10 rigs running latest FM. They range from about 2.2 to 2.5kg. I have a third, very light UL10 that did not experience this issue, but it was on older FM.

If a dev, or someone knowledgeable, believes that producing a .bin file would be helpful, please explain what log settings are critical. I might already have a suitable log to share (preferrable, but need to figure out how to make it anonymous), or I will obtain a new log (risky). In all cases, I begin with gyro filter at 80 and a multi-notch filter to keep the gyro well-filtered below -60db.

I have some experience manual tuning back when I was using INAV 7, but ideally I would like to figure out why these two 10" rigs will not complete an Ardupilot autotune on roll and pitch. I am dedicated to producing the exact type of log file that would be most useful to helpful others to help diagnose my issues. I have a lot of experience with 5" and 7" quads on Ardupilot and cannot bring myself to learn/suffer through the AMC. Again, the PIDS I start with are 59% of defaults (on roll and pitch this is P=0.08, D=.0021). Again, I have no issue with yaw autotuning. I’m using fast rates with 2k gyro / 2k rate, and sometimes 4k gyro / 2k rate. I lock fast rates at arm given the low pre-arm CPU utilization in the high 30’s to low 40’s. Logs confirm low CPU utilization, nLlon below say 40 (usually zero), ROUT less than 0.05, vibes generally below 10 with no spikes above 30. Could I be missing narrow spikes that occur as the copter twitches during AT without raw logging? Without raw logging, I generally log at 2048hz.

Edit: both copters fly quite well and are responsive with no strange noises, twitches, or motor overheating with PIDs at 59% of defaults. Also, there is no ‘failed to level’ evets during autotuning attempts.

2

gyro filter at 80 seems to by on a high side for 10" quad. Assuming You have all filters set, You should have some log suitable for filter tool. If not log, You can paste how pre and post filter chart looks.

I did autotune on 8" 2kg without any issues with 4.7.0 and 4.7.1 beta

3

2026-08-25 13_23_25-Filter Review_ 12 8-22-2026 10-51-54 AM.bin — Mozilla Firefox
2026-08-25 13_23_25-Filter Review_ 12 8-22-2026 10-51-54 AM.bin — Mozilla Firefox2116×1147 226 KB

This is a test flight where I reduced to gyro filter 70 and did not attempt AT. When I click to show aliasing, the post filtered data remains below -60db above 60hz. From zero to 60hz, it stays below -52db. Sometimes, the post-filter data is “mirrored” about the midpoint like this log shows. Why?

4

Filtering seems to be OK.

Did autotune increase PIDs to values high enough to cause uncontrolled oscillations? Not sure what do You mean by jerks?

Yaw can change during autotune.

If that is first autotune sometimes it takes some time as PIDs are way off from the optimal. I did experience that sometimes it can drain the whole battery before autotune is done for single axis. Perhaps You set them too low on startup.

5

The ‘jerks’ are typically about 5-10 degrees yaw movement combined with a simulaneous 10-30 degree movement on the opposite roll/pitch axis (depending on whether it is a roll or pitch AT). It starts almost immediately (within about 10 seconds) after switching to AT. I let one AT attempt on roll go for about 20 minutes, and I’m a little hesitant on increasing PIDS at this point. I only attemp a roll or pitch AT after sucessfully completing a yaw AT. DSHOT 300, one rig on BLheli32 and one on AM32. Both with two Rubycon 1000uf capacitors. When I fist had this problem on the 6s2p, I used default PIDS and it flipped into the pool. The ‘jerks’ get progressively worse, then a little better, then a lttle worse, and it just goes on repeatedly like this.

6

OK then, what I would look next is ATT in log to see if DesRoll/Pitch and ActRoll/Pitch behaves correctly, and after than RCOUT to check if motors output changes accordingly. From what I understand During pitch AT you also have roll movements, is that correct?

Dest would be post .bin log from AT flight.

7

Thanks for this, Adam. I will plan another AT attempt with GPS logging off since I cannot figure out how to make my .bin anonymous. I’m using MP in Windows and I’m not a scripter. Also, I will confirm roll/yaw jerks during pitch AT attempt. I’ll log attitude, RCin, RCout, motors, notch, compass, system performance, fast IMU…

8

One more question - in hoover, if You move sticks to command pitch movement, does it also make roll?

9

Negative. It is quite smooth and stable in hover, even when rolling / pitching / yawing.

10

I mean, no uncommanded movements when not autotuning.

11

Share a .bin file. I bet you can turn off that second notch filter.