Excited!!

[mwendling95]

Hey Gbro,

saw you post where you announced this and wanted to stop by and say hello.

Im setting up my bluetooth network soon, I moved into a house and have not reconfigured it yet. I was designing something myself similar to this and with our move and life never finished it. Regardless of current quality will be my testing and primary system going forward. I'll work on debugging and sending you any thing I see. I would love to see this be THE solution over esp presence and Bermuda. Could be a huge game changer.

another thing, im watching UWB chips and tech very closely. in my opinion they will probably end up being the long term solution but is far off from ready for deployment. What do you think?

[gbroeckling]

Hey — sorry for the slow reply, and thanks for stopping by. Testing help from someone who's already tried to build this is worth a lot, so yes please. The most useful reports are the boring ones: what you expected, what you got, and the Advanced → Health output at the time.

On UWB: I think you're right about the precision, and I still don't think it replaces BLE for this use case. The reason is less about maturity than about how the radios differ.

BLE is passive — a phone, watch, tag or headphone advertises constantly and any scanner in range just overhears it. That's what makes whole-home presence tractable: you can locate things that have no idea you exist. UWB isn't passive. Ranging is a two-way, session-based exchange, which means the thing you're tracking has to actively participate — a running app, a paired accessory, or a purpose-built tag. There's no equivalent of "overhear the phone in the next room." Add that phone UWB radios are gated behind fairly closed APIs, and that a tag ranging continuously burns battery far faster than one advertising BLE, and the "just put UWB anchors in every room and track everything" future doesn't really arrive on its own.

Where UWB is genuinely superb is the cooperating-tag case: 10–30 cm on something you chose to attach a tag to. So my read is that the endgame is hybrid rather than a replacement — BLE as the wide passive net that sees everything, UWB for the handful of things you want to the centimetre, and probably mmWave radar for tagless "is a person in this room" where you don't care who. Different questions, different radios.

That shape is why the spatial model underneath PadSpan is deliberately radio-agnostic: rooms, barriers and scanner positions are all stored in real-world metres rather than as RSSI-space artefacts, so a source publishing actual positions is a smaller lift than a rewrite. To be clear about status though — the current ingestion path is RSSI-shaped and no UWB work is scheduled. It's an architecture I'd like to keep open, not a roadmap promise.

One thing worth setting expectations on since you're deploying fresh: RSSI gets you roughly 1–2 m median with well-placed scanners and per-device calibration, 2–4 m without. Occasional 5–10 m single-reading jumps are multipath physics, not bugs — so don't spend your debugging energy there. Room-level correctness is the metric that matters and the one I want your reports on.

As for being the go-to option — that's for users to decide, and ESPresense and Bermuda both did real work I benefit from. I'd rather earn it on accuracy and setup experience than claim it.

[gbroeckling]

You offered to help test back in May — if that still stands, now's the moment.

I'm in the middle of a major core upgrade: the photo is no longer a coordinate authority anywhere in PadSpan. Rooms, scanners, beacons, calibration and history are all metres in an independent fabric store, and the floor plan is only a display surface.

Testing is underway now, and the next few days is the useful window. Beta channel — enable pre-releases in HACS.

[gbroeckling]

You offered to help test back in May, and I asked again in August while I was mid-rework. That rework is done: v0.34.3, beta channel.

What's most useful to report, in order:

1. Whether anything you'd already set up survived the upgrade — calibration points, room shapes and scanner positions should all be where you left them.
2. Multi-floor, if you have it. v0.34.0 fixed a bug where the floor list never reached the positioning code, so every multi-floor install was being scored as a single storey.
3. Anything that renders blank. Four views were parsing and then throwing silently — they're fixed and now covered by a test that loads every view under a DOM shim, but that class of failure is easy to miss by eye.

Pre-release, so you'll need pre-releases enabled in HACS. Stable is still v0.21.13.

The boring reports are still the useful ones.