FFT and Pixracer

Hi Dave,
could you please tell me what IMU_RAW logging is good for?
I just know not to use it if you want FFT logging. :flushed:

OK, I think that is showing Post Filter data, the Static notch just isn’t doing much which has been my experience. You only have 15db of attenuation configured anyway. And back to Vibes you have clipping on 2 axis. Rather large which would be expected with a hard mount.

If this were my craft, and I have one very similar to yours with the same FC, I would soft mount the Pixracer, update to AC 4.0.3 Stable, disable the fixed notch filter, test at hover and review pre-filter data, enable the Dynamic Notch and configure.

Ok, I learned some things about Notch filtering on 3.6.9

  • the bandwidth parameter is questionable because it always gives a very narrow notch in the FFT diagram, maybe 10 Hz wide
  • the INS_LOG_BAT_OPT doesn’t seem to do anything because as soon as notch is enabled there is a notch visible in the FFT data no matter if INS_LOG_BAT_OPT is 0 or 2

Next I will do a soft mount of the FC…

So long :slightly_smiling_face:

soft mounting is done, looks slightly better
clipping appears only in some manouvers and stays below 200 over all

I went a bit overboard with soft mounting the PixRacer on my 210 but performance has been excellent. I’m using tri-blade props now.

your’re not experiencing any clipping at all?
mine seem to appear only when going straight on higher throttle

Yes, occasionally while flying aggressively. I think this is to be expected. The full scale for most of these IMU accelerometers is 16g.

16g is quite a lot!
The flight log is far away from that. But as the IMU is running 4kHz there might be some averaging before the fc sees a downscaled 1kHz signal. Or what is your explanation?

16g can quite easily be exceeded with dynamic events. Frequency is a squared term when considering acceleration. I saw a post recently, I think on the Gitter channel, about a new IMU (don’t recall which one) with potential for use in Flight Controllers with an Accel full scale range of +/-32g.

Edit: It might have been the Bosch BMI088 which is +/-24g not 32. In the Control Zero, Durandal and maybe some other new FC’s.

1 Like

in the FFT display
what is a good value for “bins”?

The default value of 10 bins is plenty good frequency resolution for what we are doing here. I used to do a lot of of NVH and Modal Analysis work with Spectrum analyzers where this was important. That is not the capability available in Mission Planner and it doesn’t have to be for our purposes.

what about the PIDs when updating from 3.6?

Those should be re-factored if required.

yes but for the first try they fit as well as before :sweat_smile:

and I took a look at dynamic notch of 4.0.3
WOW

that’s how it looked before:

I’d like to give a big compliment to the developers :metal: you rock :metal:
so now I think I could sharpen the PIDs a little more …