Dual-motor tailsitters

@tridge Glad to see the behavior on Linux is reproducible. Is the Windows version of GE older or newer? I believe @Arne-W reported that the plane icons were much larger than I’m seeing, and I think he’s also using Ubuntu. My PR for apmplanner2 is:

@tilt Just had a look at your 21.BIN log of the Illusion flight! It sure was a wild wide!
The most obvious thing is the tilt gains are far too high. I see you had Q_TAILSIT_VHGAIN set to 1.4, which I think is much too high. I’d suggest trying 0.5, and you may need to make it even lower.
I suspect you had it that high so that it would tilt the motors up a lot when on the ground. To do that I’d suggest raising the pitch IMAX a lot instead. So set Q_A_RAT_PIT_IMAX to 1 (up from 0.44).
Note that I have fixed the issue we had with you forwarding telemetry from your machine using MissionPlanner. If you’d like to try again and have me connected while you try a flight I’d be happy to do that.

@tilt note that your log shows servo outputs beyond the 1000 to 2000 range, despite the MIN/MAX being configured. I’ve just pushed a patch from @khancyr that fixes that.

Hey, Tridge ,
thought something was not right,how do I get that bug fix
went back to 4000 on both and hovered ok
will try the new ajustments you suggest
the weather has been windy and found that I am a pinch flyer and that limited my forward stick, corrected that and it handled gusts of 15+ mph but could use more forward tilt
how do we set up to fly,it was very dark last time we worked mission planner
Mr.Hall asked for bin21, maybe wait for his input ?

also,my download page keeps opening when i’m on this site (get player)?any idea what that is?

@tilt ahh, just seen your flight video here:

looks great!
I’m trying to work out which flight log that video corresponds to.
ahh, I think it may be 19.BIN. Yes, that looks like a nicely stable flight in QSTABILIZE.
The attitude tracking looks really good. It is definately suffering from a couple of bugs that have been fixed in master.

just wait a couple of hours and then load the “latest” firmware with control-Q in MissionPlanner. The fix is building now.
Now that I’ve found that log where its flying well, I have to adjust my recommended tuning. It’s actually tracking really well. Just adding the output servo saturation bug fix will help a lot, and raise Q_A_RAT_PIT_IMAX to 1 will probably help a lot.
I suspect that if you lower Q_TAILSIT_VHGAIN as I suggested earlier then you’ll have to also raise Q_A_RAT_PIT_* gains, and maybe Q_A_RAT_YAW_* gains too. But just fly again with the saturation output fix first, as really its flying well, so make small adjustments and see how it goes.
I’m really happy its flying. Yipee!

no? what browser are you using?

Great, Thanks,
I put bigger rudders on but that made for single point contact on the t/e and yaws around a bit on t/o I think I will go back to the smaller rudders and add a challange to code,stright wing yaw control in ff. would love to remove the rudders

chrome, well, I won’t download it

@tilt I’ve had a bit more of a look, and I see what the issue is with takeoff. You needed really high Q_TAILSIT_VHGAIN to get the motors pointing up. If you instead set Q_A_RAT_PIT_IMAX higher that would get them pointing up, but I think it may flip over as it doesn’t de-saturate the integrator fast enough.
So for now what you’re doing with setting Q_TAILSIT_VHGAIN really high and the other gains low is the best approach until we fix it to handle this better.
I’ll discuss with Leonard soon

wow,ok,I think,wish I spoke apm better

@tilt Leonard and I discussed it, and we think we have a solution, but its a bit late for me here. We’ll code it up tomorrow. We’ve reproduced the problem in the simulator.
Night!

@tilt I’ve built a version of the code for pixracer with a fix for angling the motors straight up on takeoff. It should help a lot.
I’ve put a binary for pixracer for you to test here:
http://uav.tridgell.net/RickYonika/plane-vectored-takeoff-fix.px4
You should be able to use it with the same gains you successfully used before. It will just add extra vectoring when at large pitch errors (like when on the ground).

good morning ,will try ASAP… i’ll let you know how it works as soon as I know .
been puttin pixhawk in gray as well, and building a tilt styrker

1 Like

great, good luck with the flight!
The PR with the change is here:


I think the same idea could help with non-vectored tailsitters too, to give them more pitch authority without integrator windup. If it works for you I’ll experiment with trying it for normal tailsitters - maybe @kd0aij can try on his stryker

@tilt one more thing - I notice you are hovering at about 70% throttle. If you want to try QHOVER or QLOITER then you should set Q_M_THST_HOVER=0.7 so that it gets the right hover throttle without it having to learn it. It will make for better takeoffs in QHOVER.
I’ve also been testing QLAND in tailsitters, and I think there is a bug for tailsitters. So you’ll probably find that automatic landing doesn’t work yet.

great, thanks also would a usb camera be helpful when we remote fly, if so any suggestions?

Hi,loaded and ran in hand ,sudden full yaw uncommanded?

logs on the way, and mr hall

I had a quick look, and the output doesn’t look strange to me. The yaw kicks in the first time when the board is rotated in yaw, and its trying to rotate back. Then it sees yaw input from RC channel 4, and again yaws a lot.
If you still find it strange then ping me on hangouts and I’ll do a shared screen and I’ll have a look over your shoulder.
Cheers, Tridge