Tuning the Harmonic Notch

Sorry for the OT question!
I remember that somebody wrote in a post the parameter values that can be reset after setting the dynamic notch in order not to get excessive logging, but I can’t find it anymore. No mention in the docs. Andy, could you help?

LMGTFY

https://ardupilot.org/copter/docs/common-imu-batchsampling.html#common-imu-batchsampling

Thank you so much. asking to understand. for example, 33 Hz ACC0 highest peak amplitude (3. harmonic) level is lower than 32 Hz. But i need to look at first peak amplitude level for ACC0 and GYR0, right ?

Accels are not directly affected by the harmonic notch, it is only a second order effect. You should only look at gyros

1 Like

Okay last question . what do i need to look for to decide better. peak amplitude level at INS_HNTCH_FREQ value for on GYR0 or higest peak ?

That’s really down to experience and looking at a lot of graphs. You should always be targetting the fundamental frequency - the question is which one is it? Usually governed by prop sizes.

fundamental frequency is INS_HNTCH_FREQ, right ? It would be very helpful if there was a document to let us decide which results are good or which one is better :slight_smile:

@andyp1per it is tipicaly either:

  1. the lowest-frequency peak (sometimes hard to see, you need to zoom in and see if the other peaks are at multiples of this frequency) or,
  2. the highest-amplitude peak

right?

Usually the lowest frequency peak, but sometimes its difficult to see

1 Like

Edited the question to add more detail :slight_smile:

You could take this class after fulfilling several class pre-requisites :grinning:
https://ece.engin.umich.edu/academics/course-information/course-descriptions/eecs-351/

1 Like

I can recommend the book “Feedback control of dynamic systems” by Franklin and Powell

1 Like

I see that’s available as a used book at a very good price!

Hi Andy @andypiper . I was reviewing a log today after some notch filter tweaking and noticed Nlon is reported as very high. Load is nominal. Not sure if it’s a scaling issue with MP or it’s something I should worry about. Pixracer running FFT referenced notch ,Window size 64, Double Notch and Dynamic Harmonic. What do you think?

I think you are asking too much of an F4, you should cut out some notches somehow

OK thanks. I’ll disable Double Notch and Dynamic Harmonic.

One more for you Andy. This is a Pixhawk1 with just Throttle referenced Notch so as simple as the notch filter get’s. The Long Running Loops are indicated very high. I’m just trying to understand where this should be relative to the Wiki comment about “low 10’s”. Should there be a concern here or again might this be a scaling issue with how MP is presenting it?

Personally I would say that’s ok. I run my Pixracer at 75% load with a 128 length FFT

After looking at a few more logs from various craft perhaps this Nlon criteria of “low 10’s” is misleading. I just looked at a Cube Orange with no Notch filter at all and it was ~500. At least that’s what the GCS’s are reporting in the graphs.

1 Like

If you are on master then yes - it needs some more tuning