DJI Zenmuse H3-3D – repair (Part 1?)

Big Drama…

I had a pretty horrible crash with my DJI F550. There was a catastrophic resonance which made the copter unstable and uncotrollable. The Drama was, that my beloved DJI Zenmuse H3-3H Gimbal was underneath, when it crashed on the roof. This is how the gimbal looked… (Thank god, the GoPro was OK)

The first problem was a sticking ‘roll’-axis. It was easily fixable. The axis of the roll axis has an easily reachable screw which claps the motor to the axis. The arm squeesed onto the regulator PCB due to the crash. Fixing was as easy as loosening the screw and tightening it again.

But, at a close look, I found a defect halleffect sensor on the ‘nick’-motor of my gimbal. Normally this is a death-sentence. Googling A search online yielded that the DJI motors are sold and made nowhere. Not even Alibaba.

After I tore down the gimbal and removed the motor, halleffect sensor crumbs fell out of it. The other sensor were labelled ‘1432’, which yielded exactly no results, aside from some analog Sensors from Allegro. Important to note was that Pins 1 and 3 of the 3 remaining sensors were connected to each other. Also, Pin 2 of these SOT23 sensors were connected to something what seemed to be a ground pour. Some cross-referencing later it turned out that the pinout 1: VCC, 2: OUT, 3: GND i a pretty usual pinout for halleffect sensors in SOT23 package. Conveniently, there also was a small ceramic cap on the flex-PCB which was connected to Pins 3 and 1 respectively GND and VCC. I traced one of the remaining sensor outputs to the flex-connector, where I removed the kapton and connected a wire (please re-check the pinout on the picture). I set a current limit of 5mA and slowly raised the voltage to 4.5V. But, unfortunately, there was no recation to stimulation with a neodymium magnet. It was not before i connected my scope and found some interesting spikes. The sensors on the motor have open-drain outputs.

But which kind of halleffect sensors are they?

There are several types of hall effect sensor. Bipolar, unipolar, latching, analog, … luckily, there are very nice videos on Micronas’ Youtube Channel – [1] [2], which neatly explain the behavior of bi- and unipolar hall effect sensors. In our case, the output of the sensors was consistent with unipolar halleffect sensors. Quickly checking out Farnell I found the SS360xT (x for North or South) from Honeywell as up to the job. The only thing still unknown was the polarity of the sensor (north-on or south-on). So I ordered both categories…

After testing which polarity was the right one (magnets!) (NORTH in our case – so it’s SS360NT) I mounted the sensor in the motor, which was not quite easy duue to some kind of melting plastic, and attached a wire to it’s output in the usual way.

I mounted ‘freestyle’ pullups for open-drainability and hooked everything up for a test….

Actually looks pretty nice. Yelloow is our new sensor. Interestingly, our new sensor seems to be ‘high’ on power-on, which the others aren’t.

Apart from that the signal looks as we would imagine such a signal, showing magnet position in a imaginary circle, measuring always 120° apart.

Now of course it is on to putting it all back together. It’s important to note that the notch in the little gear which turns the pot on the controller PCB must be aligned with the notch on the motor axle. Otherwise, the closed-loop-control isn’t working right. Apart from that, always align the screws with the notches in the axles…

Tada – The gimbal is working again! 🙂

The (absolute) angle of the ‘roll’ axis is still off because it’s mount is bent. But that maybe is

part of part 2… 🙂