Hi Alexey. I have a dart250 too and try with autotune level 6 and 7; result is not so stable. Perhaps we could share mutualy our logs to learn and try optimsiation ? My dart is 285g, Li-ion 2s.
hi, here my latest PIDs 4.0.9 fw. Not move to 4.1 yet
After 2 session with autotune (first was at level 6), second at AUTOTUNE_LEVEL_7:
This is the fligth with autotuning level 7.
I’m using ArduPlane V4.1.0dev (7eb4e9b3) from Yaapu
Thanks for those PIDs. I’m close to finishing up a Dart250G, and while I know that PID sharing has a bad reputation, using other peoples’ values as a starting point has worked well for me in the past.
PS: What coast is that? Very nice.
Hi, idea is not to use other PID directly. It’s a bad idea. But understanding in which direction improvements could be done.
What is really important is to see how different the throw (servo course travel) is between 2 identical plane.
Well, on copter there’s the Initial Settings Calculator, integrated into MP, and so far I regard using other peoples’ PIDs as starting point as similar. I still think it can be helpful, especially on a tiny plane that’s way off the AP defaults which are meant for much larger ones. We’ll see when I try it.
Anyone got a decent PIDs for 4.1?
Alexeys PIDs are very shaky in 4.1 although they look brilliant in 4.09. When I use Autotune it results in a similar unsatisfying values and behavior as macfly. Idea anyone?
Contrary to my expectations, I had no problems autotuning the Dart250G even in a very tight flying spot, starting off from scratch with only INS_GYRO_FILTER raised to 40 and SCHED_LOOP_RATE raised to 200. Even though I couldn’t fly in a straight line for more than 5 seconds, the result was convincing even when I flew in wider areas later. Very impressive tuning routine in 4.1. So I’d suggest just trying it.
The PID controller changed significantly from 4.0.x to 4.1. I know on the plane I moved to 4.1 everything changed.
I finally uploaded a video containing an 4.1 Autotune (level 7) result:
With no wind, very pleased with the result. But as soon there is the lightest breeze, the shaking horizon comes back.
What optionsapart from carefully raising D gain on roll is useful?
What is the point of raising gyro filter or schedule loop rate?
ONE_Shot for servos for increased speed?
Thanks for your suggestions.
Share a log from the flight if you want. If the shaking is actually an oscillation the PID loop (and it looks like it might be) you can turn down RLL_RATE_SMAX. That will allow the performance possible with the higher gain values, but will turn down the gains if there is an oscillation.
Sure! here It comes (big one):
Overall I’d say that plane is flying well. Just looking at the demand vs actual pitch&roll, the tune looks pretty good. There is certainly some noise in the outputs. I don’t see big oscillations or anything too scary. Now’s the hard part, the tweaking to make it better and I always find this so subjective. I’m going to stick with my earlier suggestion, try RLL_RATE_SMAX = 140. There seems to be more noise in the roll and this might just clean it up a bit. I don’t think increasing the auto tune level will do much to help because it looks like it’s tracking the desired angles pretty well already.
thanks for your analysis, I will try RLL_RATE_SMAX=140 out.
I agree that it flies well with no or very low wind.
What else options can I playwith:
Increase D term manually?
Noise on Output?
→ OneShot or other options
→ Gyrofilter to also smooth out outputs? → no noise in loops visible so far, so no influence here…
→ Increase schedule_loop_rate? Don’t want to overload my Omnibus F4pro
Please don’t get annoyed, all these questions are not only valid for the Dart250g, but also for my other and future builds.
Once you have the tune sorted you might want to fly a few minutes straight and level in manual mode. Have a look at the logs and see what the actual pitch and roll are doing. This will give you an idea of what is possible. A 250g plane will be kicked around by the slightest breeze, and will not feel as smooth in the FPV as you would have in a similar weight quad.
The new filtering may help, but I haven’t had need to tinker with it yet so I can’t offer any advice there. (RLL_RATE_FLTE and FLTT)
OneShot or Dshot are ESC protocols, not for servos and while they make all the difference in a quad, they don’t do much for planes. (Not counting VTOL)
Have a look at this comment from Tridge.
Thanks,I will keep you updated;)