FFT PkAvg: Should I enable in-flight fft harmonic notch filter?

Hi, following instructions here: In-Flight FFT Based Harmonic Notch Setup — Copter documentation

I got data for the first step and it looks like this:

Which shows that the PkAvg hits the minimum of 80 Hz. Does this signify harmonic notch filter wont help?

I attach the bin file, but here are more plots for your convenience:

motor out:

vibe check:


roll rate:

bin file (a fortuitous link name): https://paste.c-net.org/DonaldBegins

What did the gyro FFT look like pre and post filter with INS_LOG_BAT_MASK set to 1? In the log you posted the Notch filter isn’t enabled so the FFT data is just being processed but not used.

i thought i was supposed to log without enabling fft-based filtering to inspect the logs first and see if enabling fft-based filtering makes sense. So I set FFT_ENABLE to 1 as per In-Flight FFT Based Harmonic Notch Setup — Copter documentation

Apparently I forgot to set INS_LOG_BAT_MASK to 1 as per Initial Tuning Flight — Copter documentation

But nevertheless i thought i am not supposed to enable it yet, until i know it will help?

You need to review the pre-filter data 1st . Perhaps the 80hz floor is too high.

that seems so. i may reduce to 50Hz and see what the chart looks like

See what the pre-filter FFT graph looks like and then decide what to set the low end for. I’m talking about the Gyro FFT not the FTN data.

oh ok. so i dont understand. I set INS_LOG_BAT_MASK = 1, do a test flight, and then what chart do i look at and what do i look for?

oh also im seeing pm.nlon 200 when it is suggested it should be low 10’s here: Advanced In-Flight FFT Setup — Copter documentation

This one. It’s what you use as reference to configure the Notch Filters. Only look at the Gyro data, notch filters are not active on the accels.

1 Like

This is very helpful by the way. I think it should be in the docs.

  1. Determine how you would benefit from FFT filtering with INS_LOG_BAT_MASK = 1.
  2. Verify that compute is reliable with FFT_ENABLE = 1.
  3. Activate filtering and inspect results to verify it is working as expected.

Here is what i got… I can clearly see harmonics of about 100 hz but not sure if amplitude makes it a concern.

Right. Try MINHZ at ~50 and MAXHZ at ~250 and see what the Post filter data looks like with INS_LOG_BAT_OPT to 2.

BTW-Some of the pre and post analysis info is in the Throttle Based section of the Wiki. It may answer some of the points you mentioned.
Filter Analysis

1 Like

heres what i got enabling dynamic fft:

attenuation successful i guess. the accel amplitudes went down too, i guess because the Copter’s changed reaction to less noise?

btw how did you get dB plot?

that throttle section is helpful and should be in more general section, applying to other methods of attenuation as well.

Looks good. The accel went down because there is a low pass filter present which is represented in the post filter data. And some response to less noise as you suggest.

The Y axis scale is really unitless. It’s good for A/B comparative measurements. There may be a more sophisticated answer to this but that’s what I know.

Yea, the documentation has grown over time. It wasn’t that long ago when Throttle based Dynamic Notch was the only option. It’s typical that new features lead the Doc process.

1 Like