Category: Gear

The Home Assistant Green launched at $99 in 2023. In January this year, Nabu Casa raised the suggested price to $159 / €139. In April they raised it again, to $199 / €179. Two increases in four months, on the one piece of hardware whose entire pitch is that it is the cheap, unintimidating way in.

What I appreciate is that nobody dressed it up. The pricing post names the cause in the second paragraph — RAM — and says the component cost of building a Green “has nearly doubled” since production started. No supply-chain word salad, no vague gesturing at tariffs. And then, in the same post, Nabu Casa point readers at the installation docs and suggest you may already have something lying around that would run Home Assistant perfectly well.

That is more or less the advice I have been handing out for two years, from a house where Home Assistant OS runs bare metal on a used HP EliteDesk 800 G4 mini that cost me a fraction of a Green. So you would expect this post to be a victory lap. It is not — because the same shortage that repriced the Green has been quietly repricing my side of the argument too, and I only noticed when I went looking for a spare box to test restores on.

The used market stopped being a fixed bargain

Here is the thing about a used enterprise mini PC: the most valuable component in it is no longer the CPU. It is the memory. DDR4 kits that sold for well under a hundred dollars in late 2025 are now going for double that or worse, and the analysts covering it are talking about the squeeze lasting into 2027 rather than easing this year. Refurbishers price their stock accordingly. The listing that was CHF 90 with 16 GB eighteen months ago is not CHF 90 with 16 GB today.

Which kills a piece of advice I used to give freely: buy the cheap 8 GB machine, drop a 32 GB kit in later. At current prices that upgrade can cost more than the computer did. If you go used in 2026, buy the RAM already installed, in the box, from the seller. Treat the memory as the thing you are actually shopping for and the chassis as the free accessory.

What the box under my stairs actually buys me

Not speed. Home Assistant itself is not a demanding piece of software, and a Green runs a normal house without complaint. What an x86 mini PC with real RAM and an NVMe drive buys is headroom, and headroom only matters if you are going to spend it.

I have spent mine three times this year. Once when my recorder database climbed to 5.8 GB and the repack needed free space equal to the file itself — trivial on a 512 GB SSD, genuinely awkward on 32 GB of eMMC. Once when 2026.8 shipped a native llama.cpp integration and I pointed a CPU-only conversation agent at it. And once when I wiped the machine deliberately to find out whether my backups were a safety net or a hypothesis, which is much less frightening when you can pull the drive and keep the old one on a shelf.

None of those are things a beginner needs in month one. All of them are things I would have hit a wall on within a year.

What the Green buys that my EliteDesk does not

Start with the number I keep having to eat. The Green is specified at roughly 1.7 W idle. My EliteDesk sits around 10 W at the wall on a plug meter. Call it 8 W of difference, running every hour of every year: about 70 kWh, which at Swiss household rates is somewhere near CHF 20 a year. Over five years that is a meaningful slice of the price gap I was feeling so clever about. It does not erase it, but anyone quoting a purchase price without the electricity is only telling you half the story — and I have been that person.

Then the unglamorous parts. The Green is a supported platform with a warranty, a known-good power supply, and no BIOS to think about. My machine needed a firmware update I put off for months, which is exactly the drawer of neglected devices I complained about a few days ago. It is silent in a way a repurposed office PC with a small fan is not, and it does not look like a repurposed office PC, which matters in a shared flat more than enthusiasts like to admit.

And the part I refuse to hand-wave: Nabu Casa contributes the majority of its profit to the Open Home Foundation. Buying a Green funds the software I am running for free on hardware I bought from a stranger. That is not a spec-sheet line and I am not going to pretend it is worth exactly CHF 80, but it is not nothing either.

The rule I would give someone asking today

Two questions, in this order.

Do you want a hub, or do you want a server? If the honest answer is that you want lights, sensors and thermostats to work locally and you have no appetite for maintaining a small computer, buy the Green. The premium is real, the outcome is a supported box that gets you to a working smart home in an evening, and you can stop reading here.

Are you going to run something else on it within a year? Frigate with a camera or two, a local voice pipeline, an LLM conversation agent, Zigbee2MQTT plus an MQTT broker plus a Grafana you swear you will look at. If yes, a used x86 mini with 16 GB is still the better buy, because the alternative is buying a Green now and a second machine in eight months.

If you land on used, the 2026 checklist is shorter than it used to be and mostly about memory:

• 16 GB installed, in the listing, from the seller. Do not plan to add it later.
• An SSD included and sized for the job — 256 GB is fine, 512 GB gives your recorder database somewhere to breathe.
• 8th-generation Intel or newer. Older silicon idles high enough to eat the savings you came for.
• Original power brick included. Replacements for these little machines are annoying and overpriced.
• A seller with a returns policy and a rating you would trust with a bank transfer.

The uncomfortable part

I am biased. I already own the machine, I have already spent the headroom, and it would be very convenient for me if the used route were obviously correct. It was obviously correct when I bought in. It is now a closer call than it was, and it gets closer every month that memory prices go the wrong way.

Nabu Casa ended their pricing post with advice that felt strange to read from a vendor: if you want something, buy it sooner rather than later, because prices will stay volatile. I keep turning that over. It applies just as much to the second-hand listing you have had open in a tab for three weeks. Whichever side of this you land on, the hardware is not getting cheaper while you deliberate.

I will keep running the EliteDesk. If it died tomorrow and I had no plans beyond lights and sensors, I would probably buy a Green and be quietly relieved about the firmware updates.

For most of the time I’ve been doing this, “battery-powered presence sensor” was a contradiction in terms. mmWave radar — the technology that can tell a person sitting perfectly still on a sofa from an empty room — drinks power. So the rule was simple: if you wanted real presence detection, you ran a wire. If you wanted a battery, you got a PIR motion sensor and you accepted that the lights would go off on you mid-paragraph because you dared to read a book without waving your arms.

That rule quietly died over the last year, and I’ve spent the past week reading teardowns and reviews to figure out what my own shortlist looks like now. Here’s where I landed — and what I’m actually going to do about it, which is not “buy everything”.

The trick: PIR wakes the radar

The breakthrough isn’t a magically efficient radar. It’s a division of labour. The new generation of battery sensors — Aqara’s FP300 and SwitchBot’s Presence Sensor are the two prominent ones — pair a nearly-free-to-run PIR sensor with a 60 GHz mmWave module that sleeps most of its life. PIR spots you entering the room and wakes the radar; the radar holds the room “occupied” while you sit still; when it decides the room is empty, it goes back to sleep and hands the watch back to the PIR.

Both devices use the same radar chip, a Possumic RS6130, which SmartHomeScene’s teardown of the FP300 found is built around an ultra-low-power mode that lets it stay useful for years on coin cells. Aqara claims up to three years on two CR2450s — for a device that also reports light, temperature and humidity. The SwitchBot runs on ordinary AAAs, which I appreciate on principle; coin cells are the one battery format I always forget to stock.

I want to be honest about the trade-off baked into this design: the radar only wakes when the PIR triggers. Walk-in detection is PIR-grade, not radar-grade, and there are edge cases — a room you enter very slowly, or re-enter within the radar’s sleep window — where the handoff can stumble. Reviewers report it works well in practice. But it’s a duty-cycle trick, not free energy, and it behaves like one.

My 2026 shortlist, by job

Battery, no wires, mainstream: the Aqara FP300 is the obvious pick, and not just because everyone says so. It speaks Zigbee or Thread — you choose — which means it drops straight onto my existing Zigbee2MQTT network with no app, no cloud, no account. Five sensors in one housing (mmWave, PIR, lux, temperature, humidity) also means one battery to manage instead of three separate devices. Derek Seaman’s Home Assistant-focused review covers the setup details well.

Wired, local-first, tinkerer-friendly: the Everything Presence Lite remains the best value in the ESPHome camp at around $35. It’s an ESP32 with a 24 GHz radar that tracks up to three targets with live X/Y coordinates, and because it’s ESPHome, the firmware is yours — no vendor app, no phoning home, updates on your schedule. Apollo Automation’s MSR-2 and R PRO-1 play in the same league; the R PRO-1 even takes Power-over-Ethernet, which is the kind of thing that sounds excessive until you’ve fished a USB cable through a finished wall.

Budget: the Sonoff SNZB-06P at roughly $15 is still the honest cheap option. It’s a 5.8 GHz radar, so it’s less precise than the 24/60 GHz crowd and it has no PIR, but for a hallway or bathroom where you just need “is someone in here, including standing still at the mirror”, it does the job — and as a mains-powered Zigbee device it doubles as a router that strengthens the mesh.

DIY: an ESP32 plus a Hi-Link LD2410 module costs a few francs and twenty minutes of your evening. This is how I first learned what mmWave could do, and the community support in ESPHome is still excellent. It won’t win on looks. It will win on understanding exactly what your sensor is doing and why.

The local-first fine print

A few things I weigh that spec sheets don’t mention. Battery Zigbee devices are end devices, not routers — an FP300 joins your mesh but doesn’t strengthen it, while a wired Sonoff or a plug-in Apollo does. ESPHome devices are the gold standard for ownership: the configuration lives in my Git repo next to everything else, and no firmware update ever arrives without my say-so. And Aqara’s older FP2 — still a fine sensor with its 30 configurable zones — needs the Aqara app for initial zone setup. You can uninstall the app afterwards and run it locally via HomeKit, but “install our app first” is exactly the kind of small concession I’ve been trying to stop making.

Placement matters more than brand, by the way. mmWave sees through more than you expect: a fan, a curtain breathing near a tilted window, even a large houseplant in an air current will hold a room “occupied” all night. Every one of these sensors has sensitivity and zone controls for a reason. Budget an evening of tuning per room, whatever you buy.

What I’m actually doing

Not replacing everything, that’s certain. The PIR sensors in the hallway and the storage room are staying — for rooms you only walk through, presence detection solves a problem that doesn’t exist, and a PIR on a lithium cell will outlive some of my appliances. The office and the living room are different: those are the two rooms where somebody regularly sits still for an hour and gets plunged into darkness for it, and those are the two rooms that justify real presence hardware.

The office already has a wired ESPHome radar and it isn’t going anywhere. For the living room — where my better half has opinions about visible cables that I have learned to respect — the FP300 is the first battery sensor I’d actually consider, precisely because it asks for no wire, no app and no cloud. That combination didn’t exist two years ago at any price. That’s the quiet story of 2026 so far: the compromise between “works properly” and “goes where I want it” is finally dissolving, one duty-cycled radar at a time.

Good roundups if you want to go deeper: SmartHomeScene keeps a continuously updated best-and-worst presence sensor list — including the ones to avoid, which is half the value.

For as long as I have been putting radios in this flat, the advice has been the same and it has been boring: pick one protocol per device and stop fiddling. A bulb is either a Zigbee bulb on my Zigbee2MQTT network or it is a Thread device on somebody’s Matter fabric. Not both. That constraint is the reason my mesh is stable, and it is the reason I can answer the question “what happens if the internet dies” with a shrug.

Silicon Labs and Signify announced on 23 June that they are removing the constraint. Select Philips Hue bulbs built on Silicon Labs’ MG26 and SiMG301 wireless SoCs will run Zigbee and Matter over Thread concurrently — the same bulb, the same radio, two networks at once. Today those bulbs still make you choose at commissioning time. The firmware update that flips on concurrent operation is promised “later this year.”

My first reaction was that this is a marketing slide. My second reaction, after reading how the silicon actually does it, was more interesting: it is real, it is clever, and it quietly changes what “my device” means.

Why one radio can do this at all

Zigbee and Thread are not rival radio technologies. They are rival network layers sitting on the same 802.15.4 PHY, both using OQPSK modulation in the same 2.4 GHz band. That shared foundation is the whole trick. Silicon Labs’ own multiprotocol documentation spells out the easy case: if both networks happen to sit on the same 802.15.4 channel, a single radio can receive Zigbee and Thread frames concurrently with no time slicing at all. It is just listening. The upper layers sort out which stack gets the packet.

The hard case is the realistic one, because your Zigbee coordinator and your Thread border router almost certainly did not pick the same channel. For that, the chips use what Silicon Labs calls concurrent listening: the radio flips between two channels on the order of tens of microseconds, just long enough to sniff for a preamble, and parks on whichever channel is actually talking. It is not the coarse time-slicing that Bluetooth-plus-Thread devices have to do, because Bluetooth uses GFSK and genuinely requires the radio to change modes. Two 802.15.4 networks are close enough cousins to share.

There is a cost, and it is documented rather than hidden. Silicon Labs notes that enabling concurrent listening degrades PHY performance slightly, with 802.15.4 sensitivity dropping to around −98 dBm. That is a general figure for the feature, not a measurement of any specific Hue bulb, and I would not pretend otherwise. But directionally it is what you would expect: a radio that spends part of its life looking somewhere else hears a little less well. In a dense mesh with a mains-powered bulb every four metres, a decibel or two is noise. At the ragged edge of coverage — the garage, the far end of a concrete corridor — a decibel or two is the difference between a device that responds and a device you have to go and reset.

The channel-planning problem nobody wants to talk about

Here is the part that actually worries me, and it has nothing to do with Hue. Everything above lives in 2.4 GHz. My Zigbee network is on channel 20 specifically because it sits in a gap between the Wi-Fi networks my neighbours run, and finding that gap took an evening with a spectrum scan and a bad mood. A Thread border router commissioning itself onto a channel of its own choosing is one more radio competing for the same air.

Until now, a dual-stack device was a device that had picked a side, so its traffic all went to one place. A bulb that is genuinely live on two 802.15.4 networks is emitting for both, answering both, and being polled by both. Multiply by thirty bulbs in a well-lit flat and the aggregate is not nothing. I have already written here about eight months I wasted blaming my walls for interference that turned out to be a USB 3 port. I am not eager to hand myself a new class of ambiguous RF problem to misdiagnose.

What it actually solves, and for whom

Be fair about the problem being fixed. Hue’s differentiated features — dynamic scenes, entertainment sync, the gradient stuff that genuinely does look good — live in the Hue Bridge and speak Zigbee. Matter gives you cross-ecosystem control and easy setup, but a Matter-over-Thread Hue bulb was a dumber Hue bulb. Signify’s CTO for Hue, George Yianni, framed the collaboration as supporting “the interoperability consumers want through Matter while preserving the advanced capabilities” of the Hue ecosystem. Stripped of the press-release grammar, that is an honest description of a real trade-off customers have been complaining about for three years.

And for someone in a mixed household — an Apple Home user married to a SmartThings user, say — being able to bind the same bulb into Hue’s own world and into a Matter fabric without buying two bulbs is a genuine improvement. This is a good feature. It is aimed at people who are not me.

What I will actually do about it

Nothing, for now, and I want to be precise about why rather than waving the local-first flag.

My bulbs are joined directly to Zigbee2MQTT. There is no Hue Bridge in this flat, which means the entire premium half of the pitch — the scenes, the entertainment sync, the Hue app experience — is something I have already declined. What CMP offers me is the ability to also put a bulb into a Matter fabric I do not run, on a border router I do not own, for control I already have through Home Assistant. The value is genuinely zero here. That is not a criticism of the feature; it is just a reminder that “more connectivity” is only a benefit if you wanted the second connection.

The part I will be watching is the security surface, because it is new. A device that maintains simultaneous membership in two networks has two sets of credentials, two commissioning paths, and two ways to be talked to. Whether a Matter fabric can be joined to one of these bulbs without the Zigbee network noticing — and what a Zigbee2MQTT user would see if it happened — is not something the announcement addresses, and it is the question I would put first if I were reviewing one. My instinct is that Signify has thought about it. My instinct is not evidence, and the industry’s recent track record on default-open commissioning states is not spotless.

So the buying advice stays where it was. If you are building a local-first house, buy the plain Zigbee device, join it to one coordinator, and enjoy having exactly one thing to debug. Concurrent multiprotocol is not a reason to buy and not a reason to avoid — it is a capability that will quietly arrive in silicon you were going to buy anyway, and the interesting question is not whether it works. It is who else gets to talk to your lights once it does.

I will report back when the firmware lands and I can put a scanner on it. Until then: one radio, one network, one thing to blame.

Switzerland’s electricity regulator publishes a number every year that quietly undermines half the reasons people give for buying a UPS. In 2024 the average end consumer here experienced 21 minutes of interruption in total — 11 minutes of it planned, 10 minutes unplanned — across an average of 0.34 outages. Ten unplanned minutes. For the whole year. Spread across everyone.

So when I put a battery behind the HP EliteDesk 800 G4 that runs my Home Assistant OS install, I had to be honest about what I was buying. Not insurance against the grid — the grid is fine. Insurance against a much dumber category of failure. Framing it that way made the sizing questions easier and the shopping questions harder.

The grid is not what takes my house down

Every power loss my automation stack has actually suffered originated inside the building envelope. An RCD tripping because a cheap appliance decided to leak to earth. An electrician flipping the wrong breaker, because the labelling in this building was done by an optimist. Me, moving a freezer and discovering that the socket I unplugged shared a spur with the cupboard the server lives in.

None of those show up in a SAIDI figure. They are house events, not grid events, and they are far more frequent than ten minutes a year. Where supply is genuinely shaky, the case for a UPS writes itself. Here, the case is subtler, and it is entirely about what an ungraceful shutdown does to software that was mid-write.

What you are protecting is a SQLite file

Home Assistant’s recorder writes constantly. Every state change lands in home-assistant_v2.db, and with a few dozen Zigbee devices and some power meters that is a steady trickle all day. A mini PC with an NVMe drive shrugs at the wear — one of the reasons I moved off a Pi and an SD card. What it does not shrug at is losing power mid-write.

Modern SQLite in WAL mode is resilient and most hard cuts cost you nothing. “Most” is the operative word. The failure mode is not a dramatic explosion; it is a database that comes back slightly wrong and a history graph with a hole in it you notice three weeks later. Add a Zigbee coordinator yanked mid-firmware-operation, or an add-on’s config rewritten at exactly the wrong instant, and the argument becomes simple: I would like the machine to be told about the outage and shut itself down, rather than being executed.

That is the whole feature. Not “keep running through a blackout.” Just: find out, and land the plane.

Sizing is the easy part, and it is where everyone overspends

The EliteDesk 800 G4 Mini idles around 10 to 13 watts depending on configuration — HP’s Energy Star figures land just above 13 W, reviewers measure a little lower. Mine sits at roughly 12 W with a Zigbee stick and an NVMe drive, and does not meaningfully move under Home Assistant’s workload.

Twelve watts is a rounding error to any UPS on the market. The smallest sensible line-interactive units start around 650 VA — roughly 360–400 W of real power — so the box loads one to about three percent. The temptation is then to buy either the cheapest thing on the shelf or something enormous because the runtime charts look impressive. Both are wrong. Cheap units frequently have no data connection worth using, and a UPS your automation system cannot query is just a slightly better power strip. Enormous units are wrong because runtime at three percent load is not something manufacturers characterise honestly — published curves usually start at 50 percent — and because sealed lead-acid packs in this class want replacing every three to five years regardless of how gently you treat them. Twice the capacity mostly buys twice the eventual replacement cost.

The box is not the only thing that needs to stay up

This is the part I got wrong on the first pass. I plugged in the server, felt clever, then realised that during an outage I would have a perfectly healthy Home Assistant instance that could not reach anything and that I could not reach either.

The Zigbee mesh illustrates the problem. My coordinator is a USB stick on the server, so it survives — but every mains-powered Zigbee device in the house is a router, and all of them die with the power. What is left is a coordinator talking to whichever battery sensors are close enough to reach it directly. Door contacts keep reporting; anything routed through a smart plug does not. Fine, as long as you know it — and a good reason not to build outage-critical automations on top of mains-powered relays.

More importantly: the router, the switch, and the fibre ONT. Without those, the mobile app cannot reach the box, the notification you carefully configured to warn you about the outage cannot leave the house, and remote access is gone. Together they are another 15–25 W in my case. That is what pushed me from “any small UPS” to “one with enough outlets and enough runtime to cover the whole networking shelf for long enough to matter.”

Sine waves, transfer times, and when the spec sheet earns its keep

Line-interactive units — the ordinary consumer kind — pass mains through directly and switch to the inverter in a few milliseconds when the supply drifts out of range. That transfer time is a non-issue for computer power supplies, which hold up easily across it. Online double-conversion units remove the transfer entirely at the cost of price, noise, and standby draw. For a home server cupboard, line-interactive is the right answer. The waveform question is the one worth reading carefully. Many line-interactive units output an approximated or stepped sine wave on battery rather than a true one — Eaton’s Ellipse PRO range, for instance, is line-interactive with an approximated sine output, and you have to move up to something like the 5SC to get pure sine. Power supplies with active PFC can behave badly on a stepped waveform: coil whine, extra heat, and in bad cases a reset at the moment of transfer — precisely the moment you needed it not to reset.

Here I have to be honest rather than dogmatic. The EliteDesk Mini runs on an external 65 W laptop-style brick, not a desktop ATX supply with active PFC, and small bricks are generally far less fussy about waveform — same for the router and the ONT. So for my specific load, the stepped-sine objection is weak. It stops being weak the moment a NAS or a desktop joins the same UPS, and most cupboards grow in exactly that direction. If there is any chance yours will, pay for pure sine now.

Making Home Assistant actually see it

A UPS your system cannot query is worth very little, and this is where the local-first setup pays off. Home Assistant’s Network UPS Tools integration is a Platinum-quality, local-polling integration used by about 4.1 percent of installs. It exposes battery charge, load, estimated runtime, input and output voltage, and a human-readable status sensor, polling every 60 seconds by default.

The detail the docs state plainly and people still miss: the integration cannot talk to a UPS directly. It is a NUT client. Something has to run the NUT server that owns the USB connection — a NAS, a separate small machine, or the Home Assistant host itself via the community NUT add-on. Running HA OS bare metal with the UPS plugged into that same box, the add-on keeps everything in one place. Point the integration at localhost:3493 and the sensors appear.

The automation in the official docs is the right starting point — trigger on the status sensor moving to On Battery, Battery Discharging, send a notification. Then extend it: wait for battery charge or estimated runtime to cross a threshold before shutting anything down. React instantly to every brownout and you will shut down cleanly for a two-second flicker, which is worse than not reacting at all.

The chicken-and-egg nobody warns you about

If Home Assistant shuts itself down gracefully and the UPS keeps happily supplying an empty socket, then mains returns two minutes later, you have a house that is dark in a different way: the automation server is off and will stay off until someone walks over and presses a button. That is a worse outcome than the corruption you were avoiding.

Solving it properly means the UPS cuts its own output once the shutdown completes and restores it when mains returns, with the machine’s BIOS set to power on when AC is applied — on the EliteDesk, an “After Power Loss: Power On” firmware setting worth checking before you need it. The UPS half varies by manufacturer, and NUT’s shutdown handling is well documented but genuinely fiddly. Budget an evening for it, not twenty minutes.

What I would tell someone starting this

Buy for the network shelf, not the server. The server is a 12-watt afterthought; the router and ONT decide whether you ever find out there was a problem. Check the NUT hardware compatibility list before you buy, because a supported data connection is the entire difference between a UPS and a heavy power strip. Prefer pure sine if a NAS or desktop might join the cupboard later. And size for a graceful shutdown plus margin for flickers, not for riding out a blackout — riding it out is a different and much more expensive project.

And accept that in a country with ten unplanned minutes of outage a year, this will look like a waste of money for a long time. It is the same bet as a backup routine: the value is entirely in the one evening it saves you, and you will not know which evening that is until it arrives.

For years, my answer to “which presence sensor should I buy?” came with an asterisk the size of a power brick. mmWave radar — the technology that can tell a sleeping human from an empty couch — was simply too hungry to run on batteries. If you wanted real presence detection instead of PIR motion blips, you accepted a USB cable snaking down your wall. In my house that meant exactly two rooms got proper presence detection, because those were the two rooms where I could hide the wire.

That constraint quietly died over the past year, and I don’t think enough people noticed. So today: a look at where presence sensing stands in mid-2026, what I’d actually put on my own walls, and the category of devices I’d still avoid even though they’re a third of the price.

The trick that made batteries viable

Quick recap for anyone who hasn’t fallen down this rabbit hole. PIR sensors detect changes in heat, which means they need actual movement. Sit still reading a book and your lights go out — the classic smart home embarrassment. mmWave radar detects micro-movements like breathing, so it knows you’re still there. The catch: radar transmits constantly, and constant transmission eats coin cells for breakfast.

The solution that finally shipped in current-generation hardware is almost annoyingly obvious: use both. A PIR sensor sips microamps and acts as the doorbell — the moment it sees motion, it wakes the mmWave radar, which takes over and holds presence until the room is genuinely empty. Then the radar goes back to sleep. The 60 GHz radar chip both Aqara and SwitchBot use in their latest battery sensors is built around exactly this duty-cycling idea, and it’s the reason we’re suddenly seeing multi-year battery claims on devices that would have lasted weeks before.

What I’d buy for battery-only spots

The Aqara FP300 is the one that’s been going in and out of stock since launch, and having read the teardowns and long-term reviews, I understand why. It packs the 60 GHz radar, a PIR trigger, plus light, temperature and humidity sensors into one puck running on two CR2450 cells, with a claimed battery life of up to three years. It speaks either Zigbee or Thread — you pick the firmware — and pairs with Zigbee2MQTT without any Aqara hub in sight, which is exactly how I like my devices: local, hub-free, and talking to the coordinator I already own.

One detail that matters if you’re deciding between the two radio modes: reviewers consistently report that the full set of configuration parameters is only exposed in Zigbee mode. The Thread/Matter firmware is far more limited. That tracks with my broader experience — Matter keeps promising the world and keeps shipping the lowest common denominator. For a Home Assistant household, Zigbee mode is the obvious choice, and it’s not close.

The SwitchBot Presence Sensor is the budget sibling worth knowing about: same radar chip, same PIR-wakes-radar principle, but powered by standard AAA batteries rated for up to two years. No environmental sensors, so it does one job — but AAAs are cheaper and easier to source than coin cells, and with fewer extras on board it may well outlast the FP300 in practice. If I were covering a hallway or a guest room where temperature data is irrelevant, this is where I’d save the money.

Where wires still win

Battery sensors answer “is someone in this room?” Wired sensors can answer “where exactly, and how many?” — and if you have power available, that’s still a meaningful upgrade.

The Aqara FP2 remains the zone-mapping king: up to 30 detection zones in a single room, each exposed as its own occupancy entity in Home Assistant. Desk occupied, couch occupied, dining table empty — three different automations from one device. The honest caveat: zone setup happens in the Aqara app. It’s a one-time thing and you can uninstall the app afterwards, but for a no-cloud purist it’s a compromise, and I’d rather tell you that up front than have you discover it during onboarding.

For the ESPHome crowd, Apollo Automation’s R PRO-1 is the interesting one — dual mmWave radars (an LD2412 and an LD2450) with Power-over-Ethernet, natively ESPHome so it’s fully local and endlessly tweakable. PoE on a presence sensor sounds like overkill until you remember that anyone running cameras already has a switch with spare ports. And if your budget says otherwise, the Sonoff SNZB-06P has quietly become the default cheap wired Zigbee option — nothing fancy, but it holds presence reliably.

The DIY corner

An ESP32 plus a Hi-Link LD2410B/C module is still the cheapest way into mmWave, and the ESPHome integration is mature to the point of boring — which is the highest compliment I can pay infrastructure. The newer LD2412 and LD2450 modules now have official ESPHome support too, with the LD2450 adding multi-target tracking for those who want FP2-style capability without the app. My own bare-metal Home Assistant box already runs the ESPHome add-on for other projects, so the marginal cost of one more DIY sensor is a few francs of hardware and an evening of fiddling. That evening is the fun part, if we’re honest.

What I’d still avoid

The flood of cheap Tuya-based Zigbee presence sensors on AliExpress and Amazon is still going strong, and my advice hasn’t changed: skip them. It’s not that they can’t detect presence — many do it fine. The problem is how they behave as network citizens. Several of the popular models spam the Zigbee network with constant status updates, and testers have found this essentially unfixable. A single chatty sensor can degrade a mesh that dozens of well-behaved devices depend on. Saving twenty francs on a sensor isn’t worth destabilizing the network your whole house runs on.

The bottom line

For the first time, my recommendation doesn’t depend on where your power outlets are. Battery mmWave is no longer a compromise — the FP300 in Zigbee mode is a genuinely complete presence sensor you can stick anywhere. Wires still buy you zones and multi-person tracking, and DIY still buys you the most capability per franc. The only losing move in 2026 is the bargain-bin Tuya puck that turns your Zigbee mesh into a shouting match.

Next on my list: replacing the PIR sensor in the office — the room where I sit stillest, longest, and get plunged into darkness most often. If the FP300 survives a Swiss winter of radiator-adjacent mounting, you’ll read about it here.

The Shelly 1 Gen4 and 1PM Mini Gen4 cover most of my in-wall switching, but readers keep asking the obvious follow-up: is there anything else out there? Specifically, is there a real alternative to the full-size Shelly 1PM Gen4 — the 16A relay with built-in power metering — from another manufacturer? I went looking. The short answer: yes, but almost nobody matches all of what the Shelly does at once.

Disclosure: some links on this page may become affiliate links. If you buy through them, I earn a small commission at no extra cost to you.

What you’re actually trying to replace

The reason the 1PM Gen4 is hard to beat isn’t any single spec — it’s the combination. In one wall-box-sized module you get 16A switching, real power and energy metering, no hub required (it exposes a local HTTP API and MQTT over Wi-Fi), and multiprotocol flexibility — Wi-Fi and Bluetooth out of the box, switchable to Zigbee or Matter. Most competitors nail two or three of those and quietly drop the rest. So the right alternative depends entirely on which of those four things you’re willing to give up.

The five alternatives worth knowing

1. Aqara Single Switch Module T1 — the closest true match

This is the most direct like-for-like: a Zigbee in-wall relay with genuine power and energy metering, in a similar compact size. If you already run Aqara or a solid Zigbee mesh, it does almost exactly what the Shelly does. The catch is that it’s Zigbee-only and happiest through an Aqara hub, so you lose the Shelly’s hub-free Wi-Fi independence and its protocol-switching trick.

2. Sonoff MINIR4 / MINIR4M — cheapest, but no metering

Sonoff is the usual price-fighter, and this is exactly where it stumbles. Its in-wall minis are excellent little relays — the MINIR4M even adds Matter — but they have no power metering at all. Sonoff’s energy-monitoring lives in its plugs and DIN-rail modules (POWR3, POW Elite), not in a wall-box relay. So Sonoff is only an “alternative” if you’re happy to drop the metering requirement to save money.

3. EVVR Energy Monitoring Relay — 16A metering with HomeKit

EVVR built this one specifically around accurate energy monitoring at a full 16A, with Matter and native HomeKit support. It’s a genuine 16A-plus-metering competitor and a strong pick if you live in the Apple Home world. Downsides: it’s pricier than the Shelly and has a larger footprint, so back-box space matters.

4. Tuya / MOES / Zemismart Zigbee relays — cheap and metered, if you’re brave

There’s a whole class of white-label Zigbee in-wall modules with power metering, sold under Tuya, MOES, Zemismart and others, often for very little money. In Zigbee2MQTT or ZHA they can work well. But build quality and firmware are inconsistent, and there’s no clean local Wi-Fi/HTTP API like Shelly’s — so treat them as a budget/tinkerer route, not a set-and-forget one.

5. Staying inside Shelly’s own range

Sometimes the best “alternative” is a different Shelly. The 2PM Gen4 gives you two metered channels at 16A each in one box — great where two loads sit together. And the DIN-rail Wave Pro 1PM covers the same 16A-with-metering job in a consumer-unit form factor. Same ecosystem, same app, no compromises on local control.

Quick comparison

Ratings vary by exact SKU and region — always check the specific model’s datasheet before wiring it to a real load.

So which should you buy?

If your requirement is the full set — in-wall, 16A, metered, hub-free, and multiprotocol — the Shelly 1PM Gen4 is genuinely still close to unique, and switching brands mostly means giving something up.

Pick Aqara’s T1 module if you’re already invested in Zigbee/Aqara and want the closest metered equivalent. Pick EVVR if you want 16A metering with proper HomeKit. Reach for Sonoff only when you can live without metering and just want cheap reliable switching. And go Tuya/MOES/Zemismart only if you enjoy the Zigbee2MQTT rabbit hole and want metering on a shoestring.

For my own house, the Shelly stays the default — but it’s healthy that the alternatives finally exist. Competition is how these things get cheaper.

If you’ve shopped Shelly’s Gen4 line recently you’ve probably hit the same wall I did: there are now four “Shelly 1” relays that all sound nearly identical. The two people ask me about most are the full-size Shelly 1 Gen4 and the tiny Shelly 1PM Mini Gen4. They look like the same product in two sizes — but they actually differ on three things at once, and picking the wrong one means either a relay that won’t fit your wall box or one that can’t carry your load. Here’s how I decide.

Disclosure: some links on this page may become affiliate links. If you buy through them, I earn a small commission at no extra cost to you.

Same family, three differences at once

The confusing part is that these two aren’t a clean “big vs small” pair. Going from the 1 Gen4 to the 1PM Mini Gen4 changes the size, the maximum current, and whether you get power metering — all in one step. So it helps to name what each letter and word in the product name actually does.

The “PM” means Power Metering: the device measures real-time voltage, current, wattage, and accumulated energy, and reports it to Home Assistant or the Shelly app. The plain 1 Gen4 has no metering at all — it only switches.

The “Mini” means the module physically shrinks to fit inside a crowded back box behind an existing switch or socket, at the cost of a lower current rating.

Shelly 1 Gen4: the full-size dry-contact workhorse

The Shelly 1 Gen4 is the one I reach for when something needs real switching capacity or galvanic isolation. Its relay is a potential-free (dry) contact rated to 16A, which means it doesn’t pass your mains through to the load — it just opens and closes a contact. That’s exactly what you want for switching a separate circuit, a boiler call-for-heat input, a garage door motor, an electric gate, or anything running on a different voltage than the module itself.

What it does not do is tell you how much power that load is drawing. There’s no energy reporting, because there’s no metering hardware inside. For a boiler relay or a gate, that’s fine — you rarely care about the wattage of a dry contact.

Shelly 1PM Mini Gen4: metering in a matchbox

The Shelly 1PM Mini Gen4 takes the opposite priorities. It’s small enough to disappear inside a junction box behind a light switch, and it adds the power metering the plain 1 lacks — so every light, fan, or appliance it controls becomes a live energy sensor in Home Assistant. The trade-off is capacity: it’s a switched output rated to 8A / ~2000W, not a 16A dry contact.

For the bulk of a smart home — ceiling lights, extractor fans, a desk-lamp circuit, under-cabinet LEDs — that’s plenty, and the per-device consumption data is genuinely useful for spotting a fan left running or building an energy dashboard.

Head to head

Don’t forget the other two in the grid

Because the difference is really two independent choices — size and metering — Shelly fills out the full 2×2 grid, and the other two corners are often the better answer:

If you want metering and 16A capacity, that’s the full-size Shelly 1PM Gen4. If you just need a small dry-contact switch with no metering, that’s the Shelly 1 Mini Gen4 (8A). Laying it out as a grid makes the choice obvious:

How they fit a Home Assistant setup

Both join Home Assistant the same way and share the Gen4 multiprotocol trick: out of the box they speak Wi-Fi and Bluetooth, and you can flip them to Zigbee or Matter from the button or app. If you’ve already set up a Thread/Zigbee border router like the ZBT-2, the Zigbee and Matter modes drop straight into that mesh. In Wi-Fi mode you also get MQTT, webhooks, and the local web UI, so automations keep running even if your HA server is down — the same independence that makes the Shelly Flood Gen4 such a strong fail-safe leak sensor.

The practical difference is what each one contributes to your dashboard. The 1PM Mini gives you a per-device power sensor for free — useful for energy automations like “tell me if the bathroom fan has been on for more than an hour,” and exactly the kind of tile worth surfacing when you’re building a clean Home Assistant dashboard. The plain 1 Gen4 gives you a clean, isolated switch entity and nothing else, which is exactly right when it’s wired to something like a boiler input where mixing voltages would be dangerous.

My rule of thumb

Reach for the Shelly 1 Gen4 when the job needs muscle or isolation: a high-current load, a separate circuit, or a dry-contact input on a boiler, gate, or garage motor. Metering would be wasted there anyway.

Reach for the Shelly 1PM Mini Gen4 when you’re retrofitting behind an existing switch and the load is light — lights and fans — and you want that circuit to double as an energy sensor. The small body is the whole point: it goes where the full module physically won’t.

And if you find yourself wanting both metering and 16A, or both small size and a dry contact, that’s your sign to look at the other two corners of the grid rather than forcing one of these two to do a job it wasn’t built for.

A small box from Nabu Casa landed on my desk this week: the Home Assistant Connect ZBT-2. It’s the official Home Assistant radio stick — successor to the SkyConnect/ZBT-1 — and it’s about to become the anchor of my Thread network. Here’s what it actually is, why I’m using it as a Thread border router instead of an Apple or Google hub, and how the setup works.

Disclosure: some links on this page may become affiliate links. If you buy through them, I earn a small commission at no extra cost to you. I bought this unit myself.

What the ZBT-2 actually is

On paper it’s simple: a USB-C dongle that adds a 2.4 GHz 802.15.4 radio to your Home Assistant server. Inside there’s a Silicon Labs EFR32MG24 (Cortex-M33) doing the radio work and an ESP32-S3 acting as the USB-serial bridge. It speaks either Zigbee 3.0 or Thread — you choose during setup, and it can’t do both at once. That one sentence saves you a lot of forum reading.

Compared to the old ZBT-1/SkyConnect, you get more transmit power (up to 10 dBm), a proper detachable 4.16 dBi antenna instead of a PCB trace, and noticeably better resilience against interference. The box includes a desk stand and a USB extension cable — and that cable is not decoration. USB 3.0 ports and SSDs are loud neighbours at 2.4 GHz; putting half a metre between the radio and your server is the single cheapest range upgrade you’ll ever get.

Everything is open source — hardware and firmware — and the enclosure opens without breaking clips, with test pads exposed. Very on-brand for Home Assistant.

Quick specs

• Radio: Silicon Labs EFR32MG24, 2.4 GHz 802.15.4, up to 10 dBm TX
• Bridge: ESP32-S3, USB-C
• Protocols: Zigbee 3.0 or Thread (one at a time)
• Antenna: 4.16 dBi, detachable, omnidirectional
• Works with: ZHA, Zigbee2MQTT, OpenThread Border Router
• In the box: stick, antenna, desk stand, USB-A→C extension cable
• Price: ~$49 / €49

Why Thread, and why not just use a HomePod?

If you own an Apple TV, HomePod, or a Google Nest hub, you technically already have a Thread border router in the house. So why buy one?

Control. With a big-tech hub, the Thread network credentials live in their ecosystem, and your border router stops working the day you change phone platforms or the vendor changes its mind. With the ZBT-2 running Thread under Home Assistant, the entire network — credentials, routing, commissioning — lives in HAOS, on my hardware, with no account attached. Matter-over-Thread devices get commissioned straight through the Home Assistant Matter integration and never touch a cloud.

That’s the same reason this blog runs Home Assistant in the first place: local control isn’t a feature, it’s the point.

How the Thread border router setup works

The concept trips people up, so here’s the short version: the ZBT-2 is not a border router by itself. It’s the radio. Your Home Assistant server plus the ZBT-2 together become the border router — the device that routes packets between the low-power Thread mesh and your normal home network.

1. Plug it in via the extension cable, away from USB 3.0 ports and metal cases. Home Assistant discovers it automatically.
2. Choose Thread during setup. HA flashes the OpenThread RCP firmware onto the stick. (Pick Zigbee here instead and it becomes a ZHA/Zigbee2MQTT coordinator — same hardware, different life.)
3. Install the OpenThread Border Router add-on. It claims the ZBT-2 as its radio and forms the Thread network. The Thread integration in HA shows your network and its credentials.
4. Commission Matter-over-Thread devices with the Home Assistant companion app. Scan the QR code, the device gets the Thread credentials, joins the mesh, and shows up as a Matter entity. Fully local.

From there, every mains-powered Thread device you add (plugs, bulbs) also acts as a router and strengthens the mesh — Thread’s quiet superpower over Wi-Fi gadgets.

Things worth knowing before you buy

• One protocol per stick. If you already run Zigbee, keep your existing coordinator for Zigbee and dedicate the ZBT-2 to Thread (my setup), or vice versa. Don’t plan on one stick doing both.
• Use the extension cable. Yes, I’m repeating it. It matters that much.
• Channel planning: Thread, Zigbee and Wi-Fi all share 2.4 GHz. Keep your Wi-Fi on a fixed channel and let Thread/Zigbee live in the gaps.
• Matter-over-Thread is the future-proofing play. The interesting new sensors and locks are increasingly shipping as Matter-over-Thread first.

Verdict so far

At ~$49 it’s not the cheapest 802.15.4 radio you can buy, but it’s the one built and maintained by the people who build Home Assistant itself — firmware updates arrive through HA, setup is genuinely plug-and-pick-a-protocol, and the open hardware means it won’t become a paperweight when priorities shift. As the anchor for a local-first Thread network, it’s exactly the boring, reliable piece of infrastructure I want.

The full setup walkthrough — OTBR add-on, first Matter commissioning, and the inevitable mistakes — will be its own build log soon. Subscribe to the newsletter below if you want it when it lands.