Hello guys.
im trying to help a couple of friends with TBS discovery quads on V3.1.5.
both of them fly pretty well in stabilize, and went through auto tune.
despite two very different setups (one 600KV 13" and one 900KV 10") both of them exhibit similar behaviour in LOITER - and that is a slow wobble on the arm axis (45 degree axis) - so loiter seems unstable and almost useless for AP for now.
Can you guys share if you found a similar problem? and how did you setup yours?
thanks.
Replies
Hi All,
First, I am sorry for not uploading a movie from this weekend.
I would like to thank Lior for all his help with configuring the PID parameters following Edgar's screen shots (thank you too).
The TBS is flying stable, I think we need to adjust the high gain a little bit but it is flying great now, much better lock in Loiter comparing to the NAZA lite.
Thanks all for their kind help.
Best Regards,
Amir.
Amir,
Just wondering...are you using the clone or the disco?
If you auto tuned the Disco without the GoPro last time that sure could have caused this as the CG would be off
The other issue I had was the yaw, these yaw PIDs made it fly real straight (with no wind), before it was rocking yaw (right/left) a lot.
This is a raw video, no stabilization added. The video title has the Yaw PIDs in it, Disco Pro has no yaw gimbal so this is actually how straight it flew. :)
Hi Edgar,
According to your questions, It is a TBS Disco (no clone).
I made the first auto tune with the Go Pro but the CG was not center caused by the battery location.
It was fixed.
I promise to upload a video on Friday.
Thanks again.
Best Regards,
Amir.
Hi Lior,
1 quick question, are your arms real DJI Flamewheel arms?
If they are not, especially running heavy battery it is probably contributing to your problem.
3rd party (not DJI) Flamewheel arms are pretty much always much more flexible, and especially running your 5200mah battery that is going to allow the motors to move around excessively.
The only satisfactory fix is to get real DJI Flamewheel arms.
Other than that it really is likely to be just a matter of tuning, especially Rate P and it will probably require some manual trial an error with the Rate roll Pid P value and the Rate Pitch Pid P value in particular.
You do need to have a solid compass mot, but if you aren't having yaw problems that probably isn't it.
One of the great things about the Hoverthings Flip series is that their CNC machined G10 Fibergalss arms are seriously stiff.
One other thing that is important, long skinny mass notwithstanding, the CG should be under the center of the copter front to back & side to side, if the CG is offset (easy to happen on a dead cat front to back) it is unlikely to be stable in loiter and very hard if not impossible to tune for stability).
If necessary it would be better to offset the flight controller forwards or backwards a bit in order to allow you to achieve CG balance in the middle - that is very important and a little offset of the flight controller will have negligible effect. (The flight controller does need to be very closely aligned with the copter axes however.)
Best,
Gary
Wow guys, thanks so much for all the info - much appreciated!
these 2 machines aren't mine, they belong to Amir and Eitan which will step into this thread soon and share info.
one machine is a genuine Disco and the other is a RCT Spyder (similar to disco) - and both exhibit the same behavior.
we will look into unlocking the pitch and roll P values - that makes a lot of sense and i think this is the first place to look.
why did you mention pitch P should be higher? on the pitch axis youre moving more mass away from center, but the radius of the moving force (pitch part of each motor's thrust) has a longer distance compared to roll motion (hope my explanation is clear).
Hi Lior,
Since the mass is more spread out in the pitch direction it requires more force to get it moving (or to stop it moving for that matter) and that is what the rate PID is for.
So with Pitch rate PID P set higher than Roll Rate PID P it will apply throttle at a higher rate to change pitch than to affect roll.
Maybe easier to visualize as 2 same weight flywheels.
If you visualize pitch rate P as a flywheel with the weight all at the edge and roll rate P as the same weight flywheel with the mass all at the center, you can see you need to apply considerably more force to the pitch flywheel to get it moving or to stop it than you do for the roll flywheel.
I reasoned this out and applied it to Oliver's FLIP FPV Pro and it immediately eliminated the twitchy jerking around problem. In his case roll rate P is about 1/2 the value of Pitch rate P.
Best Regards,
Gary
Hi Alll,
I have just joined this forum recently.
With my original TBS Discovery with Original DJI arms just installed the APM 2.5 Version 3.1.5.
The Apm set as quad V, all Compass calibrations have been made.
Last Friday we made the auto tune procedure (with GoPro installed) and it went pretty well.
Because I had only one buttery to test the quad I cant tell a lot of information now, but I didn't faced wobble
The Loiter was not hoovering at one spot it was travel a little bit (right and left sides) We also noticed that we had differences high difference (On Loiter and Alt hold) when I was hoovering on sunny area and shadow area.
Few facts:
1. I have mounted the APM on a peaty hard damping - I have replaced it with softer ones.
2. I have covered the Barometer with Thicker and darker sponge.
3. The Drone was not actually CG centered so I fixed it too.
On this Friday I will do a new auto tune, and another test that will be filmed as well so I could upload it here for a demo.
Dear Edgar, my PID's are really close to yours after the auto tune.
Thanks to Lior for opening this discussion.
and thank to all of you that trying to help us here.
Thanks so much Gary,
I have mine flying exceptionally well with V as you can see in the video, It looks like the Y is the same as the dead cat (TBS and others) and for me it works like a charm, very smooth...Lior seems to have some kinda "slow wobble on the arm axis (45 degree axis)", his is the same as mine but his does not have the weight of the gimbal.
Whats odd is he has 2 motor configs and they seem to do the same thing...so any advice you can give him to tune it out would be wonderful.
Thanks again, Ed,
@Edgar,
At one time there was a separate tune for H which had the props rotating in a different configuration, but that has been eliminated in favor of just standard X, (it added needless complexity and provided nothing of value).
The main difference between normal X and H or Dead cat functionally is the fact that instead of a relatively symmetrical mass front to back and side to side centered in the frame you now have a mass that is narrow side to side and long front to back necessitating a different response for pitch as opposed to roll.
And I do not think autotune is currently set up to fully optimize this difference although you can start with it and then finish by adjusting the Roll rate P and the Pitch rate P afterwards.
Best Regards,
Gary
Hi Lior,
I don't know if this will help, but it is really important for Dead Cat Quads and is not at all discussed in the wiki.
My friend Oliver has a Hoverthings Flip FPV Pro which is very similar to the TBS quad and I have one now to which I am assembling.
In any case, it became very clear it is necessary to unlock the Rate roll and Rate pitch PIDs and to supply different values for each.
The logic is as follows: The mass for the Dead Cat (battery - Camera) is long and skinny this means a lot more force is required to change pitch in the copter than is required to make it roll - in fact a lot more.
Basically the rate Pitch Pid P value needs to be quite a bit higher than the rate roll PID P value.
I believe that Oliver's Rate Pitch P is now twice that of his Rate Pitch roll.
He had a very unstable loiter prior to doing this and now has a rock steady loiter.
And it makes sense.
It will probably take some experimentation to get this right and I would use your current best Loiter rate P value as a center point and adjust the Pitch rate P higher and the Roll rate P lower around that center point.
I think this will probably fix the problem and I don't think that Auto-tune is currently at all set up to handle the wide disparity in a dead cat's front to back versus side to side mass distribution.
Best Regards,
Gary
-
1
-
2
of 2 Next