I was curious if anybody else was having yaw problems on their pixhawk multirotor.
I just finished another scratch build quadcopter and it seems that when using the pixhawk regardless of PID settings or auto tune attempts the quad oscillates left and right continuously in flight (Roughly 5-15degrees) in the yaw axis . Sometimes it is a slow oscillation other times it is possible to hear opposing motors winding up while holding heading only to eventually give up and rapidly yaw off in one direction or another. While the yaw oscillations are present sometimes it induces problems in the roll and pitch axis.
Some info regarding my build.
-KDE 515kv motors with 12" carbon props
-Afro series ESC’s 20A Opto
-Openlrs UHF Control sometimes, other times Frsky 2.4Ghz
-Pixhawk using latest firmware (i’ve used beta as well and the problem persists)
-Vibration levels measured from the pixhawk via data logging is at most ~.6 in -flight.
-Magnetic interference was reported at 9% from the mission planner auto analysis.
-Ferrite rings were placed on the data link wiring 900mhz
-GPS Sat lock varies from 6-9 sat in flight
-When the mag was turned off via Use_compass the problem persisted.
-With Yaw Stab P maxed the problem is slightly improved. (I believe 6 is the max even though auto tune defaulted to 11 on yaw. I’ve since turned it down to 10.
-in any winds the multirotor is barely controllable. in no wind it sometimes flies well?
-The log file from my most recent flight is too large to upload but i’ll happily take screen shots of whatever may help.
I have plotted RCOU 1-4 and there seems to be a slight gap ~100+units of spread between RCOU 12-34 but I am not sure if that is the problem.
The DesYAW plotted against Yaw shows the problem as percieved by me the best.
Just completed another short flight. Log file attached. The yaw issue was at its greatest point about half way into the flight.
It almost feels insane for me to be trying to help you out when you probably know at least twice as much as I do about multirotors, but I had a thought I figured I’d share that might prove useful.
Given the erratic, unpredictable nature of the problem you’re experiencing, added to this being a new build, I suspect this may actually be a hardware issue. I don’t know if you have a watt meter or not, but I’d consider hooking that up to your power system and only connect one motor. See if the motor pulls a consistent amount of energy at various speeds. Try this test with each motor isolated and see what you get. I suspect that one or more of the motors may be acting up and RCOU isn’t telling the whole story.
A complete shot in the dark, but hey, sometimes even a blind squirrel finds a nut!
All help is always appreciated when it comes to troubleshooting.
- I’ll swap the motors out first. I happen to have plenty of this brand and its an easy way to investigate the problem before troubleshooting the motors themselves.
- Second step will be to to change the speed controllers for a different brand. I have heard that some SimonK firmware based speed controllers have issues with this controller. Currently i’m using Afro20A Opto ESC’s and i’ll probably switch back to some Plush.
All of the troubleshooting above will give me a chance to verify the wiring and guts of the quad as well. Probably wouldn’t hurt to twist the ESC - Motor wires a bit more and clean up my power distribution wiring while the weather is bad.
Everything considered, I also like the features of the 3.3Beta firmware and it seems to run just fine vs the current release 3.2 version. Not to mention many of the auto tune issues appear to have been resolved in the beta release 5. When in doubt, tear it all apart and build it again never seems to fail.
An observation that hopefully leads to a solution:
Since my problem started
- I have since replaced everything with regards to my motors and ESC’s.
- I have tried different brands of ESC’s to include plush, afro, and multistar brands.
- I have rebuilt the frame 4 different times to ensure all bolts and nuts are in fact tight.
- I have Auto tuned in 3.3 and 3.2. Autotune aggressiveness was .05 using 3.3RC-7
In all of my log files vibrations are minimal +/- .3 to .6. Magnetic interference ranges from 4-13% depending on where the multirotor is flown.
None of the above seemed to help.
Finally i decided to video tape the craft while flying and play it back in slow motion. It appears that each arm is possibly moving about its own axis while flying. The carbon booms seem to twist ever so slightly with every input from the FC.
Now for some questions:
Carbon tubes are supposed to be resistant to twisting, that being said obviously everything has its limits. Keeping in mind that cheaply made carbon tubes will twist more than ones from reputable sources Is there a recommended boom size (12mm, 14mm, 16mm, 18mm and above) for a given motor size / boom length?
My craft was using 14mm 11" carbon weave tubes for arms with an inner thickness of ~1.2mm. Each motor attached to the arm is capable of producing 3kg of thrust but in normal operation produces <1kg each. The craft is designed for AP so it is not aggressively maneuvering. The frame center is rock solid and does not flex. I am currently in the process of upgrading the booms to a larger diameter and I was curious if anybody has any specific recommendations as of to the size of the tubes. Along those same lines has anybody used Square carbon tubes instead of circular? Are these more prone to twisting than tubes in the .75in range?
Having been building these things for so long a often overlooked problem can be related to version control. In this case the problem was exactly that regarding my ESC’s.
I have a stock of about **** (Number removed for my own sanity) or so ESC’s spreading across many brands purchased from recently to many years ago. The problem that was inducing strange yaw / flight performance in this case was related to the AFRO series ESC’s that I had been using. Despite all of them having similar labels on the outside upon closer inspection it became apparent that what was under the hood was vastly different. Some of the ESC’s had different firmware and some of the ESC’s even had different MOSFETS (Not that it should have mattered. Electrically equivalent components are substituted all the time. ) Once all of the firmware was updated to the most recent stable releases and internal components matched up, the craft is now flying exceptionally well. Quieter even.
Some lessons learned:
Learn to flash your ESC’s. Even coming from the manufacturer / supplier, many ESC’s can have different firmware loaded onto it, resulting in sub optimal performance. This becomes even more important once you start to build up a collection through various means.
STAY AWAY FROM Multistar ESC’s! Following my ESC learning experience I loaded a batch of Multistar, plush, and AFRO ESC’s all with stock firmware onto a mini fpv racer with known good components and the results between all 3 of them were astounding.
The Multistar ESC’s ran hotter and consumed significantly more amperage than the other 2. Using multistar ESC’s there is a lengthy delay between throttle input and response even when directly connected to the Rx. This lag caused the craft to not fly under almost every PID setting testable. Using extremely weak PID settings it was flying but unstable. The end result was one of the ESC’s overheating and shutting off while coming in for landing. (No damage occurred thankfully.) This result was then repeated with the 20A, 30A and 45A (Now discontinued) versions. I was about to flash these ESC’s with simonK but why put good firmware onto a bad base product when other users are reporting that these ESC’s only run hotter with SimonK. These ESC’s once performed well but it appears that subtle manufacturing changes has resulted in a poor performing product. At best these will only be used in RC boat / plane projects where greater cooling can be achieved.
The plush series performed well but just was not as crisp as the AFRO series.
The Afro Series ESC’s (With matching firmware) Performed the best out of all of them. Throttle response was instantaneous (perceived to be), PID values increased, and motors ran smoother even across multiwii, naze, KK2, and pixhawk flight controllers. For the foreseeable future all of my 30A and below ESC’s (Where is the 45A Afro!!) will be using AFRO series speed controllers. I’ll proudly support SimonK and his efforts to drive the multirotor community further.