I got low frequency spike at around 24 Hz which probably is due to fuselage resonance (the image in camera shakes at some throttle settings). The vibes are low, below 10.
These are the notch filter settings, but I think I haver to change the secondary notch filter to fixed and increase attenuation to 30 maybe.
Any suggestions?
First of all, the second notch should not have INS_HNTC2_MODE = 3 if you want it to use that way.
Second, the answer would quite likely depend on the characteristics of the craft, and you provided no context. In some cases, this may be caused by control loop feedback, and putting a notch there would make things much worse.
And in any case, a log would be helpful.
That mode I want to change to fixed.
It is a VTOL 6 kg fiberglass plane, 2.2 meter span, the arms, wings, and fuselage are glued together.
Well, with such a size, it may indeed be a resonance, and may indeed be worth suppressing with a notch filter. (I would also consider some kind of vibration mitigation in the design, but it can just be a bit too late)
Since the VTOL part will be used only for short periods, the vibrations per se do not bother me. The vibes are surprisingly pretty low, around 5. I am having a rather hard time setting it up, it hovers like a brick…
At the end of this hover I had small crash, due to ESC overheating, with T cutoff set at 130 in ESC, and displayed T around 120..122. (using a Lua script to watch out for the overheating). I ordered already another ESC of the same type, and I plan to run two channels on each of the ESC, I hope that will solve the overheating issue.
I started with 4S. Did a qwik tune where the qwik tune suggested rather high P for pitch of 0.21 and 0.13 for roll, although the plane quad motors are almost square (65 cm for roll, 75 cm for pitch). It was hovering marginally.
Then I decided to change to 5S hoping to decrease current and ESC overheating. If it did improve the overheating issue, that was marginal.
I reduced manually the P to 0.16 and 0.07 for pitch and roll respectively, and reduced D proportionally, but the hover flight was terrible, had strong oscillations of about 5..8 degrees in roll and 1..2 degrees in pitch, despite the fact that I reduced P for roll more tha n for pitch.
I am not sure if the oscillation is due to too high D for the roll, a P of 0.07 to me looks like too low.
Yes, some of these 24Hz frequency oscillations can come from two high or too low PIDs.
Use the PID review tool to look at the PID resonances, you may find that one axis is kicking it off in which case you can dial down the PIDs on that axis.
I decided to change props from 15 x 4 to 13 x 6.5. The tuning results were very very much better. I got the feeling that it could be due to extreme inertia of the props (these were APC 15 x 4, 45 grams each). The 13 x 6.5 prop weighs about 31 grams. The motor is V3115 1050kV T-motor
The Qwik Tune did this time an excellent job. Still, I do not understand why there is such a difference between the axis PIDs (0.2 for pitch, 0.07 for roll, and also why P/D ratio for pitch is so extremely low (1:14 for pitch, 1:38 for roll). The flight feeling was very good.
Another issue is the ESC overheating. When I changed to smaller prop, the average current remained about the same (I expected an increase of about 20 % due to prop inefficiency). The current swings were reduced, and RCout which was oscillating plus minus 100..150 uS narrowed down to plus minus 30 uS. However, the air time to overheat (120 C) was reduced strongly, from 2.5 minutes downto to 1.5 minutes.
That one I do not understand… If everything improved, the ESC should be overheating less, but it does not…
What I plan to do is to place two 4 x 1 ESCs at the trailing edge on the boom, each one for 2 channels, so that prop wash would hit the ESC. That involves some reworking, and long battery to ESC cable (about 45 cm.). I hope that with a few capacitors that will be OK.
Trying to sort the ESC overheating… I kept the plane in ground effect where it could stay at 550..600 W and that resulted in a stable 80 C ESC temperature, i.e. the temperature did not rise anymore. For a hover flight, it needs about 800 W, but this brings the temperature very quickly (in under 2 minutes) to over 120 C.
The ESC now is mounted in the belly, flush with the fuselage skin.
I want to divide the control over two ESCs, but the doubt is if it is enough to install the ESCs in the belly, or they must be placed in prop wash.
The power difference between ground effect and normal hovering hardly could explain the temperature difference. I suspect that in ground effect, the prop wash hits the belly and produces a much more efficient cooling. That could imply that exposing the ESC to direct prop wash could be much more efficient for cooling than installing two ESCs in the belly.
I have the following options:
Not sure which way to go…
The plane was designed to 6.3 kg, the test flying I was doing was at 5.2 kg (with 40 18650 cells)…But the motors are luke warm, certainly under 60 C and hover throttle around 0.22. That suggests that I probably have a lot of reserve for going up in All up weight. The fuse has space for about 70 cells, so I could push for 6.8 kg… That would result in 50 % battery weight.
When you use that tool, how can you tell if the spike is a PID resonance vs some other cause?
Can you explain this a little more?
I’m not a good expert in all this, but various parts of the system, including the particulars of ESC firmware and software, motor inertia, soundwave propagation through frame materials and suspension, delays in signal travel times, and all of that interacting with PID control, can cause interesting phase lags not quite captured by the PID web tool, and as a result self-excitation at some frequencies.
In multirotors, this is probably most often seen in yaw, because of actuator dynamics happening in the same plane as the controlled axis. This in particular is addressed by this additional yaw filtering and Yaw(E) autotune axis. In other designs, such as single/coax/bicopters, this can also happen in other axes. Depending on the particular design - which includes all bicopters with gimbaling motors above the CoG - this can come in an opposite direction, which complicates tuning beyond any sensible level and requires advanced frame code.
For instance, I have noticeable problems getting rid of a peak at around 45-50Hz in yaw in two of my quads. One has a very tiny peak of about 3dB, another (a very small one without a compass) of about 6dB, that all despite both Yaw(E) and Yaw(D) autotune runs produced something sensible. While this does not ruin my life completely, I suspect it’s precisely because of these peaks autotune cannot tighten pitch and roll (looks pretty tight by the looks of it, but you know), so that PID contributions on these axes are unusually small, and steady desired/actual angle deviations in windy conditions reach up to 1-1.5 degrees.
And yes, I tried to add a filter on this frequency, which resulted in more peaks, worse in amplitude, in lower frequencies and a much worse control.
This is a reasonable explanation. What I have found is that there is often coupling between the axes - so an overtune on one can cause oscillation to appear on another. Usually if you use the PID review tool you can tell which axis is driving it. Quite often this can be an angle P oscillation so always worth trying to reduce that to see if it helps (for instance values of 15 on roll and pitch rather than the 30 that autotune might give you).
