Support per-motor throttle-based harmonic notch

[andyp1per]

To enable set the mult-source bit in INS_HNTCx_OPTS and use the throttle notch

[lthall]

This all looks right to me.

[andyp1per]

Flight tested on a quad with ESC telemetry. First notch uses ESC telemetry, second notch uses throttle - both multi-source, outcome seems really pretty good:

[andyp1per]

I've added this to DevCall so that @IamPete1 and @lthall can duke it out.

[IamPete1]

I really think this should be useing the _actuator value and removing the thrust linearisation with the actuator_to_thrust method.

See discussion: #26674 (comment)

[IamPete1]

I have done the change over to use _actuator here: https://github.com/IamPete1/ardupilot/tree/actuator_throttle_notch

This is the key commit: IamPete1@c70db86

[tridge]

@andyp1per nice work! I'd love to see some test logs

[tridge]

i think we need a pre-arm check for the ratio of bw to freq for multi-source
we could have a notch filter option to disable the check

[tridge]

@andyp1per @lthall maybe we should pre-arm warn if you have multi-source and bw > 0.75*freq/nsources ? we could have a INS_HNTCH_OPTS bit to disable this check

[andyp1per]

@andyp1per @lthall maybe we should pre-arm warn if you have multi-source and bw > 0.75*freq/nsources ? we could have a INS_HNTCH_OPTS bit to disable this check

We don't currently do this for ESC notches and I am currently against it. Please can we move the discussion of that to a separate PR so that we can have a good discussion and not hold this PR up?

[tridge]

We don't currently do this for ESC notches and I am currently against it. Please can we move the discussion of that to a separate PR so that we can have a good discussion and not hold this PR up?

this PR (which I love btw!) will make this situation a lot more common. We need to fix it before we merge this PR. If we defer this test we'll never put it in and we'll get crashes

[andyp1per]

i think we need a pre-arm check for the ratio of bw to freq for multi-source we could have a notch filter option to disable the check

This has been split out into #27068

[IamPete1]

LGTM

[tridge]

@andyp1per I put some logs here for a hexa that has ESC telemetry:
https://drive.google.com/drive/folders/1fveumuzs7zw7zCqw7jKLHqiaXxx3k5nC?usp=drive_link
I used a small lua script to allow switching between the ESC notch and the per-motor throttle notch:
#27163
the match is OK for some motors, less good for others. Here is motor3 for example:

and motor6 is good too:

motor5 is less good:

this is all from 1.BIN, which was the last flight we did (microSD swap before this flight)
we couldn't try the octo as it has an orangeplus on it, so no bdshot for 8 motors

[andyp1per]

I had a look at the BobaHex logs using this. Unfortunately the ref was slightly wrong leading to about a 15Hz separation between the actual motor frequency and the frequency being used by the throttle notch. This has the effect of effectively doubling the bandwidth being used leading to better noise attenuation than would have otherwise be attained using the ESC notch only with a relatively narrow bandwidth (Q=6.2). I'm not sure this proves much other than flying with wider notches is ok. My goto notch width is Q=5 which is not far away so really need an octo test with narrower notch to see the differences.

[lthall]

This really should go in.

given up on pre-arm :-)

[andyp1per]

Thanks @tridge !