PWM output glitches

[tkloss]

After changing the pwm output rate in fmu.cpp manually from 50hz to 200 or 400hz, the engines of my quadcopter made a periodically abnormal sound, like a loud click. I analyzed the pwm output signal with a logic analyzer and found variations in the pulse width. For example, the width suddenly changed from 909 microseconds to 1320, but just for one pulse.
I think I found the reason: Every second, an arming state update is sent to the pwm driver
up_pwm_servo_arm(aa.armed && !aa.lockdown);
After modifying the code to only send the update on a changed state of either aa.armed or aa.lockdown, the pwm output signal seems to be stable and the abnormal sound of the engines disappeared.
Maybe we should discuss the exact reason and how to solve the problem correctly.

[LorenzMeier]

Which branch are you using? master or px4io-i2c? The behavior on px4io-i2c should be different, and its the rewrite of the IO interface. Its good to know though and we'll check.

[tkloss]

I am working on the master. The pwm output comes from the PX4FMU, no PX4IO installed.

[px4dev]

Thanks for the detailed investigation; I'm pretty certain I know what's going on and why you're getting these bogus pulses. It should be easy to fix the PWM drivers to also ensure that they can't generate them by accident; currently the enable code in the PWM drivers does not check whether it's active, and so you can get a situation where a pulse that is in progress can be restarted.

[ghost]

Hi! Any updates? I have a strange issue with PWM outputs: when switching between SAS modes PWM output goes to zero or full throttle for ~0.1s...1s sometimes. I tried to catch this on "actuator_controls_0" topic but it seems to be clean. Maybe, it's the same issue?

[DrTon]

This issue can be closed, I think?

[px4dev]

Thanks Anton, I'll do that.