It only answers in motion

Episode 18 ended clean: circuit rebuilt, both hall channels proven, one suspect left — the disc. Tonight was meant to be a fifteen-minute victory lap: plug the second motor’s encoder into the proven rig, hand-turn its disc, watch it count, pocket a working spare.

It did not go like that.

First, a fair question I should have asked weeks ago: do I even need these encoders? I have a LIDAR. But LIDAR-only odometry drifts badly during pure rotation — and a vacuum turns constantly — and it leans harder on the Pi. Wheel odometry is the backbone; the LIDAR corrects its drift. So yes, I need them. Onward.

Meter on the sensor, power confirmed — 5.00 V at the pads, both signal lines idling at 4.40 V. Textbook. Then I pressed a magnet to the chips and… nothing. Swapped to the second motor — my clean control, the disc I’d been careful never to wave a magnet near. Hand-turned it slowly. Nothing. 4.40 V, flat.

That was the moment the “I wiped disc #1 myself” theory died. One wiped disc, fine. Two independent discs, both silent? That’s not a coincidence, that’s a pattern — and a pattern usually means I’m the common factor.

So I went back to first principles and re-verified the wiring from scratch, sure I’d crossed something. I hadn’t. brown is Vcc, orange is ground, blue and yellow are the two outputs — and the proof was in the readings: two signal lines sitting together at 4.40 V, just below the rail. Crossed wires can’t produce that. The wiring was innocent.

Then I got out the big gun — a proper neodymium magnet — and waved it at the sensors. Still nothing. And that’s when I noticed something: the roller-brush motor has a hall sensor too, a little tacho on three wires. A third sensor, on a third motor, one I hadn’t touched all day. I wired it up. Same 4.40 V. Same silence.

Three sensors. Three motors. Identical dead-flat readings. At that point you have to stop blaming the hardware. Three things do not break the same way by chance — so the fault was mine, in how I was looking.

I finally did the thing I should have done first: I searched. And the answer reframed the whole week. The Neato wheel encoder, the community threads told me, is built on Allegro A3423-family hall sensors — dynamic, differential devices. Mine are two little three-pin chips rather than one dual-channel part, so I can’t swear to the exact number etched on them. But the behaviour is unmistakable, and dynamic-differential explains every symptom: this kind of sensor does not respond to a static magnet. By design. It only fires on a changing field, as poles sweep past it. Hold a magnet on it dead still and you get exactly what I got all day — nothing.

Everything snapped into place. The reason a magnet gave me edges last episode was that I was moving it in and out — a changing field. The reason a held magnet does nothing is that there’s nothing to change. And the reason my meter reads a flat 4.40 V when the disc spins is that the pulses are real but far too fast for a three-readings-a-second multimeter to show — it just averages them into a blur. I’d built a whole day of tests that a sensor of this type physically cannot answer.

Which stings, because the robot kept trying to tell me. Somewhere back there I’d asked myself “does it need to be spinning?” and talked myself out of it — hall sensors read static fields, everyone knows that. Everyone’s wrong about this kind.

So: you cannot meter these with a DC voltmeter. Not the sensors’ fault, not the discs’ fault — the tool and the method were wrong. The verdict from two episodes ago (“weak disc”) and tonight’s brief panic (“three dead sensors”) are probably both wrong. The sensors are likely fine. I’ve just never once given one a test it could actually pass.

The right test is the boring one I keep deferring to: the logic analyzer, clipped to A and B, motor spinning, watching the raw square waves appear — or not. It’s already in the post. And if the chips really are gone, the fix is an A3423, not the linear hall sensors I nearly bought — a proper dynamic sensor that reads the disc I already own.

I closed a lot of doors today. Just not the one I set out to close.