3.5" freaks out when performing roll flip

Hi, I’m trying to add support to an AIO board (JHEF405PRO) and everything seems to work except when attempting a roll/pitch flip. I built from master (2a5bc6223ebbec935789b9cabf608dd3469a3afc)

The quad (3.5"/ 4S / 3800kv) freaks out. I tried the flip by quickly throttling to zero then making a quick roll movement. This same quad flies well when using INAV without even tuning.

Things I’ve verified:

  1. I’m in Acro
  2. Acro trainer is disabled
  3. Air mode is on
  4. The quad flies find and alt hold / RTL both work
  5. ANGLE_MAX is 8000 but I read that doesn’t affect Acro

I heard my radio say Stablized Mode when this happened so I don’t know if Arducopter wanted to avoid being upside down or it’s just a glitch in telemetry.

Blackbox isn’t optimal and has some missing chunks in ESCs 3 and 4

It was a windy day but maybe there are too many vibrations?
I have bidirectional dshot and notch filters enabled. No FFT


Params
29-03-2025 ACRO.param (25.2 KB)

Many thanks!

First, your logs are really trashy, possibly you are trying to log too much to a flash memory which cannot handle such speed. One possibility is INS_LOG_BAT_OPT=5, where the first bit causes a really high logging volume, which may be useful to see the full freq picture, but is quite heavy on the logging system. It may be that the flight controller is quite overloaded, which may cause random bugs.

Second, you may want to set the first bit of ACRO_OPTIONS to 1 (currently 0) to force all the components of air mode in Acro. The matter of air mode used to be quite complicated in ArduCopter, because there are multiple components that require attention when air mode is on. To the best of my knowledge, updated by chaotically jumping over the sources right now, simply arming with function 154, as you do, does not yet enable the full air mode, just a part of it. Without this part of the air mode, when the throttle stick is near zero, all sticks no longer control the craft but rather can be used in arm-disarm gestures. Another part, priority mixing (see ATC_THR_MIX_* parameters), can only be enabled via ACRO_OPTIONS or by using a switch with the function 84 “AirMode”. If you dropped throttle shortly before that maneuver, it could be enough to switch into the ATC_THR_MIX_MIN mode, 0.1 in your case, which would leave only a little room for the actual roll, because the machine thinks it should descend quickly and should not care much about angles.

You also have notch filtering misconfigured, INS_HNTCH_REF = 0 is incorrect, it should be 1. Essentially, this is a multiple that transforms freqs reported by DShot into freqs that are filtered out. For RPM-based (e.g. not throttle-based) setups this should generally be 1 unless something unusual happens. Maybe you have more problems, but to diagnose these, I would want logs that are not trash. At least yaw tuning seems quite poor for what I would call proper acro.

So, before Amilcar comes, I would also gently suggest to set up the machine properly, maybe using the methodic configurator indeed :slight_smile:

1 Like

Thanks a lot for the detailed reply.
ACRO_OPTIONS seems like a good candidate. I’ll try it when the wind calms down! If it doesn’t work I’ll play with ATC_THR_MIX_MIN

Logging worked fine for quick indoor hover tests (for tuning) but yeah for real flights it’s missing data and is choppy. I’ll reduce it after checking the yaw pids and fixing the INS_HNTCH_REF. Chriss Rosser never mentioned this parameter

For RPM and ESC telemetry based tracking, this parameter is set to 1 to enable the Harmonic Notch Filter using the RPM sensor or ESC telemetry set to measure rotor speed.

I’ll fix it and check tuning again. yaw does behave weirdly sometimes.
Thanks again!