I have a Heewing T1 ranger FPV plane running Arduplane 4.4-beta3 on a Matek H743 Mini which I mainly fly in FBWA and ACRO mode like it was a 5" quad.
The T1 Ranger is a light and small plane, but it has a lot of control surface authority (full roll rate in manual is >720 degrees/sec for example), a tiny bit of surface deflection goes a long way and it’s servo’s are fairly quick (70msec/60deg).
It flies very well after initial tuning, but it’s stability in windy conditions is not as good as I would like it to be. I think there is more to be had with a bit of manual tuning.
I am not looking for log review/help yet, I just want to know what the general procedure is to accomplish this goal.
But although the plane flies like it’s on rails I still think external disturbance rejection can be better.
What I did so far.
I increased scheduler loop rate to 200Hz, switched the motors to bidirectional DShot600 and the servos to Oneshot.
I increased INS_GYRO_FILTER to 60Hz.
I enabled the FFT-based harmonic notch (INS_HNTCH_ENABLE=1, INS_HNTCH_REF=1, INS_HNTCH_HMNCS=3, INS_HNTCH_MODE=4, INS_HNTCH_OPTS=2)
I verified the effectiveness of the filters using IMU batch sampling. Motor peaks are around 350Hz (small motor, 4" props) and these are squashed nicely. There are no peaks larger than -80dB above 100Hz.
(side note: I initially tried using bidirectional Dshot. And although I do see RPM telemetry coming in for both motors, the notch filters seem not to track these and stay parked at their lowest frequency. After this I decided that FFT-based notches should work just as well on a plane)
Next, I did a couple of autotuning flights, progressively increasing autotune level to 9.
After this, I manually increased PTCH/RLL_RATE_P/D up to 150% of the autotune results keeping the ratio the same. 150% of tune results seems a safe level; 200-250% results in an incidental jitter in the FPV view. This increase of P/D is worthwhile; it does make the controls a bit sharper, which is nice when flying low and fast. But it does not help wind resistance.
Keeping P and increasing D, which is what I would do on 5" race/freestyle quads seems to make things worse although in the logs the contribution of the D term is still quite low when compared to P and especially FF. I did not yet try to decrease D.
How to continue from here? Or should I just accept that it is a plane and not a quad?
Full param file here for the curious, but it is still a work in progress:
Something as light as a T1 Ranger is going to get kicked around in the lightest wind. I had similar revelations when I started flying planes FPV. The plane looks solid LOS, but when you put the goggles on you see how much it bounces.
Here is one, but it is from general putting around a bit, not anything special: Nextcloud
I would prefer to take a log, make an improvement, take a log again before I ask you guys to spend time on it.
Regarding the RPM telemetry: yes, when enabled I do see the ESCx_RPM numbers in Mission Planner’s Status pane changing. These are in the log above too.
I have not found yet what log item provides harmonic notch center frequency, but setting INS_HNTCH_MODE=3 (and the other parameters as described in the ’ ESC Telemetry Based Harmonic Notch Setup’ documentation) did not provide a good motor frequency rejection and a dip at the lowest harmonic notch frequency which led me to believe that it was not tracking.
The FFT-based setup does provide a clean plot.
I probably have to. But speedy motors/ESC’s, excessive power and the excellent filter chain and control logic in Betaflight works very well for smaller racing quadcopters. You can hear the motors reacting violently to disturbed air, but I can still fly a racetrack and hit the gates in 8-15m/s wind without jitter in the FPV view.
I am just wondering if the same cannot be done for a plane this agile, at least on the pitch and roll axes. I thought that reducing phase delay and increasing the D-terms would help, but so far I am not seeing much improvement.
As said before, it is probably how things work; >10:1 thrust/weight and little rotational inertia on an quad is different from a plane. I just have an itch that needs scratching.
Here’s my theory and it’s worth everything you paid for it.
I also fly betaflight 5" quads. I feel they have a couple of advantages compared to equivalently sized planes. They have a very small profile relative to the external forces. All their lift, and thrust is produced by the props, which we have very high rates of control as you pointed out.
If you flew “older” quads, back when all the ESCs were just left overs from RC planes, you remember they bobbled around the sky. I keep an old DJI F550 with a Naza flight controller around, and every couple of years I take it out. It’s a good reality check for how far we’ve come.
Planes and wings can be tuned in to be incredibly sharp, but it will take a lot of work and there are more variables. The mechanical and aerodynamic elements are just as important as getting the D-term set right. C of G is critical. Some planes have very different characteristics as they move through their speed range. Arduplane will fly most planes on defaults out of the box. But to get that perfect tune takes a long time, compared to what it takes for a quad. And “perfect” will depend on what you want to do with it.
To prove a point, put the plane in manual for a while and see how it feels. If the plane doesn’t fly well in manual, then you’ll never be able to get the tune perfect. Get the mechanicals sorted first before you pull your hair out over PID tuning.
As a side note, I also fly some DJI Matrice drones at work. M210, M300, M600. The 210 and the 300 have fixed angle FPV cameras. I’m always shocked at how much those “professional” drones bounce around on a windy day when compared to my Ardupilot and Betaflight drones.
[quote=“Allister, post:6, topic:104207, full:true”]
Here’s my theory and it’s worth everything you paid for it. [/quote]
Fair!
Exactly. We have increased motor- and battery power and stiff(er) carbon fiber frames. But the biggest change that transformed quadcopters from wobbly barely not crashing into the stable and agile vehicles they are today are the softwares. On the ESC, on the radio receiver, on the flightcontroller, etcetera.
Probably fixed wing cannot benefit to the same level due to, well, physics, but still…
One day I want to build myself a <250g FPV plane suitable for low-and-fast flight. Airspace in my part of the world becoming more and more restricted for everything above that. But my part of the world (The Netherlands) is also quite windy; 3-4Bft is fairly normal, 5-6 is no exception. Not really a good match for a small and light plane, unless the flightcontroller is capable of handling the disturbances better than a human can. Power is not an issue with modern motors, maybe a bit of thrust vectoring on the mechanical side and a tightly tuned flightcontroller on the electronics side will make such a thing behave like a much larger plane regarding stability.
For me that is the fixed-wing equivalent of a freestyle quad. Low and fast flight, crisp response, and when I get better flying fixed wings: shooting gaps. Officially I am not allowed to fly BVLOS anyway.
Here is a video of an ad hoc post-dinner flight with my flying brick (an overweight AtomRC Dolphin). Not the most interesting scenery, but it was a nice evening: https://www.youtube.com/watch?v=SnbnnOPETm0&t=186s&ab_channel=DaBit
Feels fine after I moved the CoG a bit back from the manufacturer recommendations. I also tried to keep the plane light.
I would say that the plane in manual is less ‘bumpy’ than in FBWA, and about as bumpy as in ACRO (but acro corrects). It is a handful for me in manual since I lack the flight hours and it is pretty sensitive to the controls (sure, expo or reduced throws help). As said before; full roll rate at half throttle is above 700 deg/sec, pitch is only a bit slower. That T1 Ranger is pretty close to a 5" quad with wings. I like it!
