Version 2.5 of the ArduCopter code is now available in the AP Mission Planner and in the downloads area!
The default PIDs are optimized for a 3DR/Jdrones quad with 850 motors and 10" props. If you're using more powerful motors/props, start by turning down Rate Roll P in 25% steps.
While some testers have reported very good flights with the default PIDs, some have reported that this release is a little "sharper" due to the DCM improvements and have found they needed to:
reduce stabilize P by 10% (i.e. 4.5 -> 4.2)
reduce stabilize D by 30% (i.e. 0.15 -> 0.10)
increase rate D from 0 to 0.001
Tuning loiter can be tricky. Refer to the discussions which will appear below for more community feedback on what parameters work best.
Please post your feedback in this discussion. For enhancement requests and bug report, please add them to the arducopter issues list. When possible please include logs (tlog and/or dataflash) and tell us whether you're using APM1 or APM2 and what version of the software you're using (presumably 2.5 but tell us anyway!).
Thanks for this release go to the developers (both in the core team but also those who have provided bug fixes through the issues list) and also the community members who participated in the previous release thread and provided some great detailed information in the form of issue reports and logs which allowed us to nail some bugs!
Is it only me...?
While most of you guys here seems slightly euphoric about the new release, I seem to have issues with it. At this time I address mainly yaw and attitude control, but I have doubts about loiter and auto as well (sorry to sound negative).
3DR stock quad, APM2 (soft suspension), code git_2.5.1_ed52caa6bc0a, 3SLiPo4500mAh
First one can always start blaming the parameters for any anomalies showing up. But after having played with for some flights I would say there is something else going on. AHRS stuff spooking with me...
Yaw: Steep attitudes like 20-30 degrees makes yaw drift several degrees, feels like 10 deg or more. Flying back and forth at aggressive attitude provokes this, every time. Yes, I could make a video, I did not, please believe me anyway. There is no doubt. Yaw rate and stab PIDs seems to do their job otherwise, but I think the reference data (from DCM) is the crook here.
Attitude: Having been observing my quad very thoroughly I am convinced that DCM acts up here too. Often, after returning to level after some attitude, it does not really return to level. There is an artifact making it show some degrees nose up when it is actually nose down (simultaneously observing quad and the artificial horizon). Then soon it returns to true level, airframe and attitude indicator. This seems very much like the artifact that has been there in many previous code versions.
Alt hold works fine, like before.
Loiter, hm. Tried Marcos setup. Could work fine in weak stable wind, but from that, still in loiter mode, climbing to 40 meters and coming up above tree tops, wind increasing, it started circling / oscillating really bad, violently. Tried auto yesterday and it accelerated towards the first waypoint an then it was all over the place, I had to abort the mission.
I have logs, but I the source of error is the DCM then I doubt that logs would provide anything of interest(?)
To wrap it up. I had much better flying with previous 2.2-2.4 versions and there are weird things going on with yaw and attitude, DCM is a suspect. It is not difficult to fly it, but it does not behaving exactly as it should.
I would be VERY interested if any more of you guys observed something like this??
Finally, to cheer you up after my complaints, a fresh pic showing the result of too little focus on the job while tuning and hand testing. It´s on me ;-)
Yes - it hurt. Carbon reinforced props.
Tomas: I have experienced issues with Yaw and a few others have as well so you are not alone. I am still trying to nail down the cause but in my case my logs show a lot of compass interference. That doesn't necessarily mean that it is the cause of the yaw problem, but it is where I have chosen to start debugging. I am working on a compass heading filter to get rid of noise, but it is still very experimental and may prove to be a bad solution to this problem. I expect to flight test my new filter this coming weekend. I will post back then.
Do you have logs that you could share with us? That would be massively helpful in diagnosing the issue. Without logs we are just making guesses which rarely leads to a solution.
You're not alone, welcome to the club :-)
Yes same for me, looks like it knows where to point and locks onto that, but then that heading shifts and the quad rotates to match, then back again. Last time I saw that I had a bent frame, this is a new one and 2.4 worked fine.
Same for attitude, when i tilt to go in one direction, it keeps going until i give it some opposite stick. 2.4 used to stop where it was when i release the sticks.
I have put this down to the change in what would be the right parameters, now DCM is used a lot of the settings need to be different...having some trouble dialling in the right ones though!
Hi Adam, I have mostly got rid of my yaw problem on my Y6 by tweaking ahrs_dcm.h as suggested by Marco.
Before I made changes I was getting up to 30 yaw (mag heading) degree error when I ramp up my motors. It is now down to less than 5 degrees without changing my physical setup. I am going to flight test right now.
Yeah, but I'm trying to use Ch7. to access auto trim, and it just doesn't seem to be working.
I wonder if my problem is actually related to what some others are saying are problems with 2.5.
I just remember, when I first had my APM, I could pretty much let go of the cyclic, in Stab mode, and it would just stay where it was. Now it doesn't do that anymore. It always wanders, and I can't seem to get auto-trim to work.
Also, I can't see those leds, and it takes me maybe 15 seconds to spool up, leaving me with 15 seconds to get into a good hover? Not gonna happen.
Actually guys, I see the yaw issue in 2.4.1. Especially in waypoint execution, moving between 2 waypoints separated by large distances (150+ feet), since the waypoints are separated by large distances and with default waypoint settings, the vehicle speeds off to the waypoint (e.g. 20^ pitch)--if yaw drift occurs, the vehicle veers off fairly fast to the point I need to take it out of auto mode cause it feels like the vehicle is running away. I'm going to repeat those test in 2.5 today...
I'm glad you were able to get it nailed down. I think that it is a functional fix, but I'm not sure it is the right fix for my particular problem. What you are doing by tuning that kP value is just reducing the impact the compass heading has on the overall heading calculation (if I understand it correctly). I am not interested in doing that because i feel that my compass should be reliable in the first place and I will not have to lower the weight it has on the heading if I can filter the noise effectively. I am just experimenting with this heading filter right now, but it might pan out to be something more people can take advantage of.
Would your copter stay on the line from the one waypoint to the other better if you were using simple mode while flying in auto?
Ok, I find it comforting not to be alone with these issues.
Adam, I agree that the compass seems to be involved in the yaw issue. And I agree at least 100% that while it is possible to decrease the symptom by tuning compass authority down, the problem indicates that the compass calibration / decoding is not reliable in the three axis it should be able to handle.
Attaching three logs, from flight with differing parameters.
Cheers / Tomas
Yes of course Graham, although I'll be finishing the tuning guide this weekend, and passing it around the Devs for approval. Maybe include that?
I too am experiencing the same issues when using advanced functions like way-points, RTL and Alt-Hold. It seems that the majority of positive comments on 2.5 are coming from users who are only flying their aircraft in a basic flight mode - up, down, circles etc. I too feel that 2.5 "flies" well using basic joystick flying. However, when using way-points and RTL and ALT-Hold including light winds, the results are not good. In fact I have been spending most of my flying using 2.4.1. I would be interested in seeing a video from Marco that is more than just 5 minutes of loiter and a coke bottle. The same for the other advanced users and testers of 2.5. Do a three way-point flight with a return to landing with momentary position hold at the way-points. This is what is needed before a release is made. Put the video up to demonstrate that these advanced functions are reliable. Then do the release. Just an opinion from someone who has bought a lot of arms, motors and props following every major release. :-)