Erratic motor throttle - PWM update rate?

[AndrewRadford]

I have an issue where if I give my quad any throttle at all, the motor speeds are not really constant, they jump all over the place, it's like they are jittery or have glitches. Generally the roll pitch and yaw work, and the right motors speed up etc, however it's not possible to have a stable hover.
The min throttle when armed, does not seem to suffer this issue.

I suspect that it may be that the ESCs cannot handle the 490Hz PWM - They seemed to be OK just off the receiver, although they need a fair bit of gas to start (MINTHROTTLE=1300).

Here is the setup:

Snelflight frame (small carbon frame about 8" motor to motor)
Hobbywing Pentium 6a ESC
HobbyKing 10g motor
WMP, Pro mini, all wired direct.


I've tried setting P and I to 0, and D to 50 to eliminate the WMP and just have raw throttle, and I get the same issue.

Props are all balanced etc, and I'm running 2S for now just to be safe.

Is there any other test I can do? The only thing I can think of is to swap out one of the escs with a 10A one, just to test.

Unfortunately I don't seem to be able to run the GUI when the flight battery is connected, must be something to do with my FTDI, so I can't inspect anything with the motors going.

Please anyone help - would like to get this puppy flying! If necessary I can do a youtube vid or something.

Andrew

[captaingeek]

which version are you running.

lets take the ESC's out of the equation. In the GUI, are the motor output's dancing?

[AndrewRadford]

Unfortunately I can't see the motor outputs while the ESCs are live and connected. With just the Arduino, Rx and WMP powered up though, yeah the throttle for each axis looks stable to the eye. It could be possible that the glitches are very high and very quick though and the GUI or my eye misses it.

[AndrewRadford]

Update:

I hacked my FTDI so I can now connect the GUI while the thing is live and spinning props in my hand.
It's not the ESCs at all - the Arduino is actually sending jumpy commands to the ESCs - jumping from 1300 to 1750 - this is what I have set as throttle max.
This is happening while I am holding the thing steady, it never tilts more than 5 degrees or so.

The Gyro graph plot seems to respond correctly to the movement however I notice it is VERY sensitive - I can easily get the chart line to go way up out of the chart area and into the buttons and controls etc! Is this is the cause? Can I reduce the gyro sensitivity somehow?

[Hamburger]

The Gyro graph plot seems to respond correctly to the movement however I notice it is VERY sensitive - I can easily get the chart line to go way up out of the chart area and into the buttons and controls etc! Is this is the cause? Can I reduce the gyro sensitivity somehow?

1. you did try and calibrate gyro manually after powerup? yaw-left + elev-back
2. in the main thread at RCG you may find vairous posts that suggest dividing sensor input by a factor of 2 or 3 like here http://www.rcgroups.com/forums/showpost ... count=9723
3. what did you do to the ftdi to make it attachable during runtime (I suppose you mean no reset)?
Hamburger

[AndrewRadford]

Thanks Hamburger - problem identified and solved.

As the RCG thread says, there are some clone WM+ boards that use the 3 axis ITG gyro, however the firmware is bogus on them and they are WAY too sensitive. Scaling down the gyro readings does work, however you lose resolution, and also there are overflow artefacts at large rotation rates which manifest as unwanted flips.

Eventually the solution can be seen here:

viewtopic.php?f=6&t=27&start=20

This basically bypasses thw WM+ micro and talks to the gyro direct. Some minor mods are required to code and hardware. There is an advantage where cycle time is reduced a lot however.

Now I have this puppy almost flyable, seems only a question of tuning PIDs, and perhaps pilot learning