SONOFF MINI-ZBRBS-E (Zigbee roller shutter wall switch) — field notes: rocker quirks, calibration, and traps

Just installed the new SONOFF Fusion MINI-ZBRBS-E (the wall-switch version of the MINI-ZBRBS shutter relay) and hit a few things that aren’t in the manual and weren’t searchable anywhere yet. Posting so the next owner finds them. Setup: ZHA, firmware 0x00001005, works natively — no custom quirk needed (you get the cover entity plus calibration sensor, “Configure cover limits” button, inverted switch, and external trigger mode select out of the box).

1. Percentage control does nothing until you calibrate

Fresh out of the box the device only knows fully-open/fully-closed: partial positions are ignored and the UI arrow only re-enables after a complete run. Check sensor.<device>_calibrated — if it says not_calibrated, press the “Configure cover limits” button entity (or hold the physical back button 10 s). The shutter runs a self-calibration cycle (~1–2 min) and after that the slider and set-position work perfectly.

2. Short press on the rocker does nothing at the limits — by design

The rocker behaves like this: mid-travel, short press = start, short press again = stop, long press = hold-to-run. But when the blind is fully open or fully closed, a short press does nothing — only a long press moves it away from the limit.

It’s not a dead button: recorder history shows the device starts closing/opening and cancels within 20–200 ms of each short press. The firmware appears to treat short press as “resume last target”, and at a completed limit the last target equals the current position, so it starts and instantly stops. The manual only documents “Long press: open or close the curtains” — so press-and-hold is the intended gesture at the extremes. Counterintuitive, but consistent. (I’ve sent SONOFF a feature request to make short-press-at-limit move in the only possible direction.)

Don’t bother with the external trigger mode select as a workaround: switching it to “edge trigger” suppresses short-press control entirely. Leave it on “pulse trigger”.

3. Trap: 6 quick presses = pairing mode = device drops off your mesh

Pressing a rocker direction 6 times within 6 seconds puts the device into pairing mode — it leaves the Zigbee network. I did this by accident while testing short presses and the device went unavailable for ~50 minutes. Recovery is painless: enable permit-join (zha.permit), do the 6-press gesture again (or power-cycle it at the breaker — it wakes in pairing mode), and it rejoins with the same identity — entities, name, area and calibration all intact. But if your cover suddenly goes unavailable after someone “played” with the switch, this is why.

4. Two hardware notes for installers

  • The six round holes on the back are the terminal screws; the wires enter cage slots on the edge. Four of the six screws are used (N, L, L Out1, L Out2); two are unlabeled and unused.
  • The L (live) wire entry is the round port inside the raised white column that looks exactly like a mounting-screw boss. The manual’s wiring diagram does show the L wire going there, but it’s very easy to mistake for the claw-screw hole and conclude “there is no L terminal”.

Overall: recommended device — 1 A relay, acts as a Zigbee router, native ZHA support, solid percentage control once calibrated. You just need to know the three quirks above.

Quick clarification regarding your setup. As it’s a dual button device is a short press on any of the two buttons closing the same relais (i.e. when it was stopped moving upwards, both buttons would continue upward movement with a short press) or is it a short press on the down button just does nothing when blind is completely open?

I’ve observed that the relais internally does not cut power after reaching the 100% or 0% position. It seems to rely on the motor to turn off. So two taps on the opposite rocker would also initiate a movement - first to turn off, second to turn on opposite direction.

This was only observations with bulbs as I am waiting for the actual motors to arrive, though. Maybe the bulbs don’t draw enough energy for the relais to detect end-position. But then again I would assume the calibration should also allow for it to turn off (I did the manual calibration, which is time-based).

If the switch really is going to act like this when wired to a real motor I was thinking about a workaround automation that would send a stop-command to the device after 100% or 0% are announced.

Same problem
You have to press the right button a couple of times to turn it back on when it’s all the way off
Maybe we should submit a request to Sonoff