4+1 Quadplane uncommanded roll on transition start

[harkhka (Divyanshu Harkhka)]

Hi experts,

I’ve been working on a 2.9KG 4+1 Quadplane with 2 elevons and no rudder design, as shown below:

The details of my build are as follows:

• Wingspan: 1130mm.

• Wing Area: 31dm2.

• Wing Propulsion: 13 Inch pusher propeller.

• Battery: 6S Li-Ion.

• For the WING mode, the COG point is what is 60mm from the leading edge (as suggested by the designer).

• For VTOL mode, since the battery is placed inside the fuselage on the front and not at the centre of the VTOL motor booms, the drone is front-heavy in VTOL mode, and the front motors have to work harder than the rear motors to maintain level in Q modes.

• Using Arduplane 4.6.3.

I’ve tuned the VTOL mode, and it flies really well and stable. I’ve tried Qloiter multiple times, and it performs amazingly. No vibrations or oscillations whatsoever.

Next, I went to try the forward flight.

I took off in QLOITER and switched to CRUISE mode at a decent altitude.

(Using CRUISE mode since I have a spring-centered throttle on my RC, and I’m not confident in using it for FBWA, as it might cause a stall in case I leave it during flight).

So, upon entering CRUISE mode, I noticed that the pusher started rotating, and the drone started moving forward and gaining speed.

After gaining some speed, the VTOL motors started spooling down (I could hear), but the drone started rolling left and yawed left, all uncommanded. I immediately switched to QLOITER in order to prevent any crash. It rolled to around -40 degrees, all uncommanded.

The next day, I went again for the test. This time, I increased the PID Roll gains and reduced the throttle skew, thinking that it might help, but got the same result. When the VTOL motors spooled down, it again rolled to around -35 degrees, all uncommanded.

My pusher prop is rotating clockwise (when seen from the back), so can the left roll be a result of counter torque produced by it? I’m thinking that as long as the VTOL motors are active, they counter the torque, but when the VTOL motors spool down, the torque rolls the body?

I’ve checked, and the servo numbering is correct, and the elevons are working in the correct direction for both pitch and roll movements.

What else can cause this?

Please note: Airspeed sensor is installed, but currently I’m not using it as it’s not calibrated yet. ARSPD_USE is set to 0 for both flights.

Here are the flight logs from my test flights:

FL20_Pv36_2026-05-02 18-27-09.bin

FL21_Pv37_2026-05-11 18-02-25.bin

Please help.

Thanks
Divyanshu

[Michail_Belov]

Seems you may have inverted the ailerons… i.e. you probably have to change in the servo menu the ailerons. Also, the first test flight you should do in FBWA mode, not cruise. Try to roll the airplane in FBWA mode holding it in hands to see if ailerons counteract the roll correctly

[harkhka (Divyanshu Harkhka)]

Hello,

Thanks for your response.

I have checked the elevon direction many times, and it looks correct to me. Here are the details:

• When I pull the pitch stick, both elevons go up. When is push the pitch stick, both elevons go down.
• When I move the roll stick towards right, the right elevon moves up, and the left moves down. When I move the roll stick towards left, the left elevon moves up, and the right goes down.
• When I pitch the nose down, both elevons move up. When I pitch the nose up, both elevons move down.
• When I roll the body towards right (left wing up), the left elevon moves up and the right moves down. When I roll the body towards left (right wing up), the right elevon moves up, and the left moves down.

Is this correct?

If you want I can also upload a video for the same.

Thanks,

[Michail_Belov]

I do not have experience with flying wings, but, the behaviour you describe is absolutely correct, both for stick movement and for stabilizing action.

However, you log file indicates that the desired roll and actual roll begin to diverge, and the PWM signal to elevon also diverge, i.e. it seems that the action is inverted.

Your GPS speed grows, so it is not a problem of stall.

Pusher prop you have generates very little torque to be of any meaning, especially at the high speeds you have been transitioning.

I suggest that you do not try to fly the plane and get another opinion from someone more knowlegeable.

[Rolf]

@harkhka

Is this correct?

Based on your description, it doesn’t seem wrong at first glance. Still, a video would be helpful, because judging by the log file (FL20_Pv36_2026-05-02 18-27-09), I find it hard to believe:

During the transition, it is noticeable that the quad motors reduce their power relatively suddenly:

Even at that speed, the airspeed is already sufficient for the control surfaces to begin to be effective:

The cause is likely a sudden pitch-up, which requires reducing the quad’s power during the transition. For clarity, the average of the quadcopter motors is shown. You can see how the pitch deviation is corrected while the roll deviation gets worse and worse:

During the transition, the quad motors or the rudders could cause erratic roll behavior. First, the quad motors. The graph shows the average of the right and left quad motors. The VTOL is rolling further and further to the left, even though the quad motors are working against it; it is clear that the right motors are almost shut down and the left ones are delivering significantly more power.

What do the control surfaces do?

While the quad motors counteract the roll, the rudder inputs seem to have the opposite effect and actually intensify the roll in the wrong direction.

That’s why I’m actually looking forward to seeing a short video showing the control surfaces moving left and right as the fuselage rolls in FBWA mode

Rolf

[Michail_Belov]

There is another detail I see, that is, the desired yaw changed strongly just before the roll, so it could be that CRUISE mode was executing a roll to achieve the yaw it thought was needed. Not sure why. And then, the desired and real roll should be the same.

If you want to take a risk, I would try to make a transition to FBWA mode, there, the yaw would be irrelevant.

[Rolf]

@Michail_Belov

There is another detail I see, that is, the desired yaw changed strongly just before the roll, so it could be that CRUISE mode was executing a roll to achieve the yaw it thought was needed.

That’s a good idea, but in my opinion, the change in flight direction is a result of the strong roll, not the other way around, if you look at the sequence: The desired yaw change does not immediately affect any of the control surfaces.

The question certainly remains as to why Desired Yaw suddenly diverges from Yaw by nearly 180° when the transition is initiated. I’m stumped by that, too.
I, too, recommend that for first-time transitions and aircraft that have never flown in fixed-wing mode, you perform the maiden flight or the first transition in FBWA mode.

Rolf

[Michail_Belov]

By the way, I had an interesting situation with yaw somewhat similar to what is going on here, where there was an uncommanded yaw after a failed transition from Qhover to FBWA when going back to Qhover. The transition failed because the forward motor did not start due ESC having a low voltage protection. But after returning back to Qhover from non-completed FBWA mode, there was a sudden change in yaw, and violent reaction, but in qhover mode. Not sure if there could be a bug somewhere…

[harkhka (Divyanshu Harkhka)]

Hello @Rolf and @Michail_Belov

Firstly, thank you for taking your time to view the logs. I really appreciate it.

I have made a video showing the elevon directions with both stick inputs, and body movements.

Please check the video here:

20260514_121640.mp4

I really hope we can solve this issue.

Thanks,

Divyanshu

[Rolf]

The direction of the control surfaces is undoubtedly correct. Sorry I insisted on that. It just occurred to me that it might be the mixer. The fuselage was tilting up a bit before, causing the pitch to deflect downward. You could manually check if the roll function has enough deflection when the elevator stick is pressed at the same time (pitch stick full down and roll stick left or right)

If that’s okay, I’d try a transition flight to FBWA. If that works, the desired yaw difference will still be the root of the problem.

Rolf

[Nemozoli (Zoltán Németh)]

I have a slightly different take to this erratic behaviour. You can see in both logs that one of the motors is completely shut off when CRUISE is switched, while the others are still working and not spooling down. If this motor is the left rear one, this tilts the plane to the left and yaws it to the left also (the front left motor is still working, rotating to the right, and the counter-rotation twists the plane to the left). The motor on RCOUT 4 still doesn’t switch back on for a while after switching to QLOITER mode… so the roll and the yaw to the left continues until it spins back up. I would look at that ESC or motor… (but mainly the ESC calibration).

[harkhka (Divyanshu Harkhka)]

Hello @Nemozoli,

Thank you for taking a look at the logs.

M4 is the rear-right motor.

My analysis is that since the quadplane is rolling left, the right side of the motors (M1 and M4) has to reduce, and the left side (M2 and M3) has to increase to try to balance roll.

Also, since the drone is also front-heavy in VTOL mode, the front motors (M1 and M3) have to work harder to maintain pitch, so M1 can’t go to minimum to maintain pitch, so only M4 goes to minimum

Regarding ESC calibration, all ESCs are well calibrated.

What are your thoughts on this?

[harkhka (Divyanshu Harkhka)]

Hi @Rolf @Michail_Belov ,

No need to be sorry. I’m glad that the aileron direction doubt is now clear by the video.

I have seen in logs that the desired yaw goes to 0, exactly when the plane mode (transition) starts, and goes back to normal when the Q modes kick back in.

I think it is because (as stated in the docs): “ArduPilot, in fixed wing modes (except ACRO mode, if yaw rate controller is enabled), provides NO direct rate stabilization or heading angle hold for yaw, as it does for pitch and roll”. Source: Yaw and Yaw Control in Plane — Plane documentation

So I think that the desYaw is useless in plane modes.

Also, let’s say for a moment that the plane was trying to follow the desired yaw set to 0, then also, the initial yaw was around 300° before transition start. If it really wanted to go to 0°, it should have turned right, not left, correct?

Secondly, to support this, when the transition starts and till it ends, the “RATE.YDes” is not logged, or is 0. This confirms that the rate controller was not working for Yaw.

What are your views on this?

Also, regarding elevon movements based on PWM, here are the details:

• For the right elevon, it moves up when the pulse decreases from 1500us, and moves down when the pulse increases from 1500us.
• For the left elevon, it moves up when the pulse increases from 1500us, and moves down when the pulse decreases from 1500us.

So, for countering the uncommanded left roll, the plane should try to roll right, that is, right elevon up (pulse decreasing), and left elevon down (pulse decreasing). I can see in the logs that both the “RCOU.C6” and “RCOU.C7” are decreasing when the plane is rolling left.

I’m not sure what to do next to resolve this roll issue

Thanks,
Divyanshu

[Rolf]

I took a look at the speeds. Sorry I didn’t do this sooner. AIRSPEED_MIN is 17 m/s, and you listed the stall speed as 14 m/s. The quad motors start to cut out very early, at around 10 m/s. This is likely caused by the pitch suddenly increasing. In any case, it’s likely a tip stall. Now we need to figure out why your VTOL is experiencing such high pitch fluctuations already during the transition.

Rolf

[harkhka (Divyanshu Harkhka)]

Yeah, I also noticed just now that the quad motors start to cut off around 11m/s. I don’t know why. Could it be possible that the quad motors spooled down early, and the wing started stalling because of less airspeed?
Also, during the uncommanded roll left, the drone also loses altitude, which I forgot to tell you earlier (sorry for this).

Could a premature quad motor slowdown, followed by a partial stall, be the reason for all this?

[Rolf]

How stable is the VTOL’s forward flight in QMODES (without initiating a transition), especially in QLOITER ?

@harkhka

Could a premature quad motor slowdown, followed by a partial stall, be the reason for all this?

Until someone comes up with a better explanation, I’ll accept that as a hypothesis. As I said, the reason why the quad motors throttled back so early still needs to be clarified.

Rolf

[harkhka (Divyanshu Harkhka)]

@Rolf

How stable is the VTOL’s forward flight in QMODES (without initiating a transition), especially in QLOITER ?

Very Stable.

During the transition attempt, the drone moves ahead to around 200m from takeoff. When the drone rolls left, I switch to QLOITER and fly back home in QLOITER only. But that is pure quad flight.

[Allister (Allister Schreiber)]

I think that C of G issue is causing more issues than you think. If this is a kit, and the VTOL motors are in the correct position then the hover values should be closer. It also suggests that the fixed wing C of G isn’t correct. When the plane starts to roll and pitch, the elevons aren’t saturated, but may be limited by the mix to get enough authority to recover especially at low speeds. Delta wings are really touchy when it comes to C of G.

Q_Assist is turned on (good), and it’s set to 15m/s. The plane never really get’s past that so it’s just on the edge of getting out of transition.

I agree that FBWA is the better choice for a transition on an unknown plane, but I’d follow that up by FBWB. Like Cruise, FBWB will try to hold a height, but it won’t fight the heading and try to introduce more roll. If the plane yaws during transition it won’t fight it as long as it can level the wings and maintain height.

As @Rolf said, I think the plane stalled. My suggestions: Fix the C of G. Transition in FBWA or FBWB.

[Michail_Belov]

a 3 kg plane with 31 dm2 wing would have stall speed around 11 m/s…

[harkhka (Divyanshu Harkhka)]

@Michail_Belov

stall speed around 11 m/s…

It is the minimum theoretical value, and the real-world value might be higher. That’s why I set AIRSPEED_STALL=14; AIRSPEED_MIN=17 in my params for both the flights.
Currently the question is why the VTOL motors started spooling down even before he set stall speed?