Dyson Thinks It’s a Jeep Part 2

Recall that my Dyson V10 thinks it’s a jeep. In the state I left off on, it turns out after a little use that the biggest usability concerns were power-related: the tiny little LiPo I was using only lasts 52 minutes in theory, much less in practice. To be fair, I don’t think any of the cells I’ve tried so far have actually been fully, or even substantially, charged. But it’s also annoying to have to plug and unplug the light bar in order to turn it on and off.

Killing two birds with one stone, the obvious next step is to try running the light bar from the same power supply as the motor head. As it turns out, tapping off that power is pretty easy.

Dyson V10 soft roller disassembled

I got this far into the soft roller head before realizing I didn’t need to undo any of the screws for the intake, or the left side.

end of head covering roller motor wires, with mods attached.

If you pull off both roller bars, there’s one screw under the big roller and one screw in the axis of the small roller that hold on the left end plate, which covers the wires going to the motor. All I did was cut a notch with a razor blade in the clear plastic where those wires exit above, and everything goes back together pretty much cleanly.

While I was in here, I fired up the vacuum and measured the voltage on those wires at ~16V (black positive, like household wiring). Boom, switch to a buck converter instead of a boost, and we’re good to go, right?!

Nope. Not so fast.

TL;DW: the spinny brush is a DC motor, and the lines powering it are designed to power all the various heads intelligently with the same power profile. Problem is, the motor heads detect jams and cut power. How do they detect jams? By too much current. What happens when I add a ~50% increase in load? Head no spin.

Normally, the head draws about 0.5A at 16V. If it encounters an obstacle, it turns out, the vacuum will try increasing the voltage to compensate for the load and prevent the motor burning out from too much current. Counterintuitively, this intelligence means that if I start the vacuum with the light bar plugged in, it just won’t power the motor head. But if I start the vacuum and THEN plug the light bar in, it ups the motor voltage to the full 26V pack voltage. Incidentally, the light bar with a buck converter doesn’t care – it’ll just halve its current demand. I could just build a little RC-delay into my power circuit using the enable line of the power converter, but I don’t know about running the motorhead at hyperspeed all the time, and it certainly wouldn’t help battery life on a vacuum that’s already quicker to die than you’d like.

Next Steps

Unfortunately, it looks like this dashes my hopes of running on internal power. I might be able to tweak the circuit and the power demand to trick the vacuum into giving us JUST ENOUGH EXTRA juice, but that seems error prone and unstable at best – I don’t want to alter how the vacuum actually handles real obstacles and surface changes.

Instead, Ill probably try to go back to the battery solution, but use the 16V signal I just built in as an “enable” for the switching converter, to cut power if the vacuum is off and re-enable when it’s on. In the mean time, I’ve got a Labor Day party to get to, so it’ll have to wait.


I tried using the motor head power as an enable signal for the boost converter. Unfortunately, since the converter I’ve got is a simple boost topology and doesn’t implement any niceties like load disconnection, “disabled” simply means that the output is connected through the inductor to the input, and there’s always battery voltage on the LEDs. This doesn’t turn them “on” and doesn’t draw much power, but it does result in a measurable leakage current that will eventually drain the battery.

For now, the solution to that is just to unplug the battery when I’m done vacuuming. If it gets annoying enough, eventually I’ll probably switch to a power converter that has load disconnection, and make a nice enclosure for the power bits.

One Comment

Leave a Reply

Your email address will not be published.