Repository navigation
Replies: 11 comments 6 replies
|
On advanced->health there does appear to be a Z but which is tight to the height of the floor/stack it's on, not the actual Z position: It might make sense to just 'slap' 'm on the floor, but wouldn't it be more useful for the calculations if a user can opt to enter Z coordinates? |
|
Good questions, and your Health-tab read is exactly right. Ceiling height: put in the interior floor-to-ceiling height (2.6), not 2.8. Slab thickness isn't modelled anywhere — that field's only job today is to supply the default Z for scanners synced from that map, and floors aren't stacked in metres (the 3D stack view's separation is a cosmetic pixel slider). Floor separation in the RF math is a flat per-level dB penalty (20 dB in the heatmaps, 10 dB in the live positioner), independent of ceiling height, so adding the slab just biases the scanner Z without buying anything. Sensor Z: there's no UI field, and you inferred the reason correctly — nothing consumes Z yet. It is stored per scanner ( You're right that it should matter, and I'll take this as the feature request. Two pieces are needed: 3D distance in the positioning path, and per-floor base elevation — which is where your slab thickness would finally have a home (floor-to-ceiling + slab = floor-to-floor). Worth being honest about the size of the win within a floor: a 1.4 m Z error across a 5 m horizontal span is ~4% distance error, well under RSSI multipath noise. The real gains are your ceiling-mounted bathroom scanners and cross-floor, where a true Δz would replace that flat penalty. For now: ceiling height = 2.6, and leave the sensors where the sync puts them. |
|
Relevant to the floor and sensor-height questions you raised here: I'm in the middle of a major core upgrade where all geometry — rooms, scanners, beacons, calibration — became metres in an independent fabric store rather than anything derived from the floor plan image. Testing is underway now and I'd value your eyes over the next few days. Beta channel — enable pre-releases in HACS. |
|
Sure, happy to help! |
|
Thanks for the offer, here's the concrete ask. Install: v0.33.2, beta channel. What changed: all geometry is metres in an independent fabric now. The floor plan image is a backdrop and no longer a coordinate authority anywhere. Rooms, scanners, beacons and calibration points carry real coordinates, and scanners have a real Z, so multi-floor fits slant range instead of pretending one plane. What I need eyes on, in order:
Anything that looks off, a screenshot of Overview next to Mapping tells me more than a log does. |
|
Correction to what I posted above — install v0.34.3, not v0.33.2. v0.34.0 went out a few hours later and it's the one that matters for your case. The floor list was never actually reaching the positioning code, so every multi-floor install was being scored as a single storey and cross-floor paths all fell back to a flat one-slab penalty. That's the exact thing #54 was about, so testing on v0.33.2 would have shown you the old behaviour and wasted your time. Point 3 in my list above stands, just do it on v0.34.3. |
|
If you haven't wiped your calibration yet — please don't. And if you already have, no harm done, but read on before you re-capture, because there's a setup question that changes what the new points will be worth. Why I'd rather you didn'tA wipe-and-recalibrate has a decent chance of fixing your middle-floor stick. That's the problem. There are two live candidates for it and they'd both go quiet, so I'd learn nothing and you'd have spent an afternoon:
Recapturing fixes (2) silently and leaves (1) in place. So if the wipe helps, I still can't tell you which it was — and if it doesn't help, you've lost the points that would have told me. What I'd like instead, in orderYou're on v0.34.5. Go to v0.36.1 (beta channel) first and hard-refresh — it carries the trim/stack fix and its repair button, plus a fix for v0.35.0 shipping a broken Mapping tab (clicking a sub-tab threw before re-rendering, so the previous tab stayed on screen and it read as a hang — that one wasn't you). Then, before touching any calibration:
If it's still stuck after that, then wipe — and at that point the recapture is worth something, because we'll know it wasn't the geometry. Separately, thanks for the earlier note that the stack renders correctly now and that devices migrate in from "outside" → middle floor → correct position over a few minutes. That settling behaviour is useful on its own; I want to know whether it still does that on 0.36.1. |
|
Separate ask, and entirely up to you — it has nothing to do with the above and nothing changes if you say no. v0.35.0 added an opt-in usage report, so it'll be there once you're on v0.36.1: Settings → Update Check & Privacy → Help improve PadSpan. It's off by default and does nothing until you turn it on. I'm asking you specifically because of what this thread is. You raised floor and sensor height, that became #54, and the whole Z and floor-stacking design came out of it — and I have no idea how anyone's floors are actually configured. PadSpan is developed against one house. I can't see whether people set What it sends, once a day: counts, versions and flags. How many scanners, floors, rooms, placed lights, calibration points, IRKs and objects (by kind), which integrations are installed, which feature switches are on, which tabs and tools got used, a few health flags (is the BLE callback alive, are rotating addresses resolving), and how many warnings each module logged. What it never sends: MAC addresses in any form, IRKs or licence keys, device / room / floor / entity names, coordinates, or timestamps finer than the day. There's a Preview what would be sent button that shows you the exact JSON before you decide, and the code refuses to transmit a report at all if it finds anything identifier-shaped in it. The report carries a random ID so installs can be counted rather than pings, and you can replace it any time. No pressure at all — the button is there either way. |
|
Was about to start wiping but spotted your message in time this morning (my TZ) - I already was on 35 - but I'll push for the new one as we speak. I'll add the telemetric, no problem. FYI starting with the previous version it already showed the floors tapered (which is correct for my house as each floor is slightly less long than the one below it. |
|
That rules out the one I was most worried about. Nothing trimmed, no Rebuild alignment offered, 👁 Show agreeing — candidate 1 is gone, and the taper rendering correctly on the older version says the same thing independently. Your floors are set up right too: 2.8 / 2.6 / 3 giving 0 / 2.8 / 5.4 is exactly what it should be. So go ahead and re-measure. The hold I put on that was for a different install and doesn't apply to you. Which leaves the measurement as the interesting part. "I measured but I don't see any reflection of it" may mean there's no metre anchor stored at all — in which case scale is coming from a fallback and every calibration point is worth less than you think. When you re-measure, tell me whether the figure survives a reload. That's the bit I need. Forget the Health screenshot, that was a vague ask on my part. One number decides this instead: how many of your calibration points have no floor id. I described what happens to those wrongly on the 20th — since 0.32 a blank-floor point isn't dropped to the ground any more. It's measured flat instead, with no vertical leg, so its distance to every scanner on another floor comes out short. Quieter than a phantom storey, and still enough to bend one floor's fit while the others stay fine. I'd rather the software told me than have you dig for it, so 0.36.7 (beta, out today) counts those points in the usage report, along with whether any map is actually measured. You said you'd switch the report on — once you're on 0.36.7 with it enabled, I get both numbers the next morning without you doing anything. It's Settings → Presence → Help improve PadSpan, as you pointed out. I've fixed the path everywhere I'd written it wrong. Ceiling 3 vs 2.4: leave it. Ceiling height only draws the stack; the Z model reads floor_to_floor_m and nothing else. You were right not to tinker. |


Uh oh!
There was an error while loading. Please reload this page.
Hi @gbroeckling thanks very much for your effort, coming of a few other attempts this look really promising. Question about heights though, given a multi-floor house, what is the expectation for height settings? Including the height of the floor itself or not? I.e. concrete slabs between floors? So say bottom floor is 2.6m physical height and a 20 cm slab, should we put in 2.8?
The other one would be about sensor height, due to some physical restrictions and previous project placement I have most sensors on the bottom floor at the bottom, at approx 1m/3ft on middle layer and about 2m on the top floor, but some bathroom sensors are at the ceiling as well. I guess there is a reason for not adding height (z) to the sensors looking at the configs?
All reactions