ZigBee BMS Retrofit: Cut Cabling in Existing Buildings

October 28, 2026
Latest company blog about ZigBee BMS Retrofit: Cut Cabling in Existing Buildings

A wireless BMS retrofit keeps the existing fan coil units, valves and pipework and replaces only the control layer: wireless room nodes, wireless I/O and a ZigBee-to-BACnet gateway, so no new cable trays enter an occupied building. The ZigBee automation market is USD 2.73 billion in 2026, forecast to reach USD 5.68 billion by 2032 (360iResearch, September 2026).

Short answer

ZigBee wins a retrofit when cable routes, not device price, set the budget: it deletes tray, containment and ceiling-access work from an occupied building. It loses for plant safety interlocks, hardwired trips and anything needing deterministic sub-second timing between controllers. Keep those on BACnet MS/TP or Modbus RTU.

The retrofit maths: cable vs radio

A hard-wired retrofit is rarely expensive because the room device is expensive. It is expensive because of everything between that device and the riser: cable, containment, fire-stopping at every compartment crossing, local isolators, access to a finished ceiling, and the hours the room cannot be sold.

What actually costs money in a refit

A 20-room hotel floor means 20 cable runs, 20 ceiling penetrations and 20 chances to damage a corridor decorated last year. A wireless BMS retrofit keeps the fan coil unit (FCU), valves and pipework and takes only the signal path off that list: the thermostat still takes mains at the old back-box, but nothing travels back to the riser.

So the comparison is not wireless versus wired thermostat. It is one trade on a ladder for an hour per room versus a cabling package, an access plan and a phased handover.

Criterion ZigBee retrofit Wired BACnet MS/TP
Cabling None: 2.4–2.485 GHz radio, 10–20 dBm, up to 100 m free space (Matinal datasheet) Twisted pair to every device plus tray, containment and fire-stopping; 76.8 kbps MS/TP (Matinal datasheet)
Downtime Minutes per room: swap the thermostat on the existing back-box and mains feed Segment by segment, with ceiling and riser access, usually out of hours
Latency No datasheet figure, no deterministic bound; depends on hops, retries and poll interval Token-passing: scan time follows node count and baud rate, and repeats
Determinism Best effort with retries — fine for setpoints, fan speeds and schedules Deterministic token rotation — usable for sequences and protective actions
Best for Occupied hotels, heritage interiors and phased campus refits where containment dominates New-build risers, plant rooms, safety interlocks and healthy existing MS/TP segments

Where ZigBee loses (and you should say so)

ZigBee is a low-power mesh on IEEE 802.15.4, and its own figures say where it belongs: 2.4–2.485 GHz at 10–20 dBm, up to 100 m in free space, and 250 kbps on the ZigBee variant of the IOM-633 platform (Matinal datasheet). Short range, narrow bandwidth, shared medium.

  • Not for plant safety interlocks. Trips, freeze stats, high-limit cut-outs and smoke-mode overrides belong on copper; radio must never be the only path for a protective action.
  • Not for fast closed loops. A valve loop with a few-second dynamic needs local I/O and a fixed scan, not a mesh.
  • Not through metal. Risers, lift shafts and foil-backed finishes are RF-hostile; free space does not survive a concrete frame.
  • Not as a protocol island. Wireless only earns its place if the head-end sees ordinary BACnet objects.

Anatomy of a ZigBee retrofit

Three layers, three jobs: rooms collect and control, the riser translates, and the legacy layer keeps whatever copper you are not replacing.

Room layer: wireless thermostats and sensors

The room device carries most of the specification risk because it is what the client sees. The TFHV-SK-FVVMD-S3-CK-Z fan coil thermostat is built for this work: 90–265 VAC so it reuses the existing back-box feed, an 86 * 86 mm panel that fits the standard back box, dual-sensor cascade PI control (SHT4x room sensor plus NTC coil probe) with 0.5 °C setpoint steps, IP20, EC fan PWM and PICV 0–10 V or relay valve output, and a ZigBee 2.4 GHz link instead of the control cable (Matinal documentation).

Everything else in the room should be justified by a control sequence. The ZB-TnRH measures temperature and humidity only — no CO₂ anywhere in the Matinal range, so do not let an air-quality clause promise it. In the ZigBee end devices range, the ZBSC-16A is a 16 A ZigBee switch with Modbus RTU and 2000 V isolation (Matinal datasheet).

Riser layer: the ZigBee-to-BACnet gateway

The gateway is where wireless stops being a gadget. Matywirt is the ZigBee protocol converter in the range: BACnet/IP, BACnet MS/TP at 76.8 kbps, Modbus RTU at 38.4 kbps, RS-485, a ZigBee radio on an external antenna, a 1 ms real-time cycle and 3 V FLASH (Matinal datasheet). It turns room points into the AI, AV, BI and BV objects the head-end already understands, so the BMS never learns a room is wireless.

Two notes save arguments. A gateway is a translator, not a supervisor: schedules and sequences belong in the head-end. And MQTT is not a gateway feature here — the cloud uplink lives in MatyContrE software, so an "MQTT gateway" specification is describing a head-end.

Where a site wants local I/O in the room, the ZB-IOM-633 offers 6 AI, 3 AO, 6 DI and 3 DO over BACnet MS/TP, Modbus RTU or ZigBee (Matinal datasheet). Where a plant room needs more analogue points and local logic, the PIOM-843 programmable I/O module adds built-in PID control and up to 16 Function Blocks with 8 AI, 4 AO and 3 DO.

Legacy layer: ZigBee-to-RS-485 for the meters you keep

Most retrofits keep something: a heat meter, an energy meter, a VAV box. Those stay on Modbus RTU, bridged by the ZigBee-to-RS-485 converter (MW-ZB-485) at 9600, 19200 or 38400 bps, 8-N-1, device ID by 8421 coding, 24 V AC/DC at 2 VA (Matinal datasheet). It is a transparent bridge, not a scheduler: set the master timeout and retries for the radio.

Two application scenarios show the shape of the work rather than a claim of savings: a hotel guest-room fan coil control and energy management scenario, and a smart campus central air conditioning control scenario with wireless rooms under a BACnet head-end. Both are scenario write-ups, not client references.

Latency, node density and mesh design

ZigBee promises no latency figure, and no honest vendor should hand you one. The delay between a room changing state and the BMS seeing it depends on four things you do control:

  • Hops. Every mains-powered ZigBee device in range can route; sleeping battery devices cannot. Give each floor several routers so no room depends on a single relay.
  • Node counts. The number of room devices per ZigBee coordinator is a per-zone allowance: size per zone, record it, leave headroom.
  • Traffic. At 250 kbps the mesh shares one narrow medium: poll in tens of seconds and send changes, not streams.
  • Channel. The 2.4 GHz band is shared with Wi-Fi. Use channels clear of the site's Wi-Fi plan — 15, 20 and 25 are the usual low-overlap choices — and record it.

The 100 m free-space figure is an upper bound, not a design distance. Behind a concrete frame and a bathroom pod, expect a fraction of it: design one hop between adjacent rooms plus a router at each riser, and survey with live nodes.

Failure modes and fallback strategy

Wireless fails differently from wired, and the difference is visibility: a cut bus is obvious, a mesh quietly retrying is not. Design for five failures.

  1. Gateway or coordinator offline. Rooms must keep running on local logic — schedule, setpoint and fan speed held in the thermostat — not freeze the last value sent.
  2. Router offline. The mesh should re-route, and rooms that lose their path must fall back locally. Test by powering down one router at a time.
  3. Interference or congestion. Expect retries and stale values; give every wireless point a communication-fault state in the BMS and never let a last-known-good value drive a loop.
  4. Power loss. Configuration and clock should survive it: the TFHV-SK-FVVMD-S3-CK-Z retains time and setpoint across power failure through its real-time clock, and the Matywirt ZigBee protocol converter keeps configuration in 3 V FLASH (Matinal datasheet).
  5. An uncommissioned device joining. The failure nobody notices until an audit.

One sentence belongs in the specification: on loss of communication a room device must keep controlling its fan coil at a defined local setpoint or schedule, must not hold a valve open or a fan indefinitely, and must alarm at the head-end.

Commissioning and network security

Wireless commissioning is a discipline, not a line item you drop when handover slips:

  1. Survey the finished building. RF changes once ceilings close and doors are hung; a survey in a stripped shell describes a different building.
  2. Join with credentials, not open joining. Commission each device deliberately, then disable permit-join so a stray device cannot attach.
  3. Replace default link keys. ZigBee 3.0 networks ship with default Trust Center keys; leaving them is a blank admin password. Record the network name and key handling in the handover file.
  4. Validate the point map. Every wireless point should appear as the expected BACnet object with the expected units and scaling, and writable points (mode, setpoint, fan speed) must accept a write from the head-end through the coordinator.
  5. Test the failure cases. Gateway off, router off, device off, channel jammed: each needs an expected behaviour and a measured actual one.
  6. Soak the network. A full week of normal occupancy with trends on communication faults, before handover rather than after the first complaint.
  7. Hand over the radio as a system. The O&M manual should carry the channel plan, the Wi-Fi coexistence note, node counts per zone, firmware versions and the interlocks that stayed wired.

The last item matters most. Wireless gets a bad reputation on some sites not because of the radio but because nobody documented the network and IT later reused the channel.

The regulatory tailwind: EPBD 70 kW and market growth

Regulation pulls the retrofit case forward as much as cabling cost. Under Article 13(9) of the recast Energy Performance of Buildings Directive, (EU) 2024/1275, non-residential buildings above 290 kW of HVAC output have needed a building automation and control system since 31 December 2024, and the threshold drops to 70 kW by 31 December 2029, with transposition due by 29 May 2026 (eu.bac, 29 May 2026).

Two consequences follow. A long tail of smaller buildings — the 70 to 290 kW band of small hotels, clinics, schools and offices — now needs a BACS it never budgeted for, and many have no spare containment. And Article 13(10)(c) requires interoperability: a system that only talks to its own app fails that test, while a ZigBee mesh that lands in BACnet passes it, because radio becomes transport under a standard supervisory protocol.

The market data points the same way. Building automation is put at USD 124.73 billion for 2026, up from USD 110.73 billion in 2025 and tracking toward USD 204.14 billion by 2030, with tariffs on controllers, sensors and networking devices pushing buyers toward local sourcing (Research and Markets, January 2026). ZigBee automation is estimated at USD 2.73 billion in 2026 and forecast to reach USD 5.68 billion by 2032 at a 13.00% CAGR (360iResearch, September 2026).

None of that is a savings claim. The Directive requires capability and interoperability, not a protocol or a percentage. If a business case needs a number, measure the building: fan coil runtime and cooling energy before and after the change, with tariff and weather alongside.

Frequently asked questions

Is ZigBee reliable enough for building control?

Yes for zone-level temperature, fan speed, occupancy-driven setpoints and lighting — provided the network is surveyed, documented and alarmed. For hard interlocks and plant safety, no: keep those wired, where a fault is visible and timing repeats.

How does ZigBee data reach a BACnet BMS?

Through a ZigBee-to-BACnet gateway such as Matywirt, which exposes wireless points as BACnet/IP or BACnet MS/TP objects (MS/TP at 76.8 kbps, Modbus RTU at 38.4 kbps, 1 ms real-time cycle — Matinal datasheet), so the head-end reads ordinary BACnet objects.

What range can I expect?

Matinal's ZigBee modules specify a 2.4–2.485 GHz link at 10–20 dBm with range up to 100 m in free space (Matinal datasheet). Treat it as an upper bound through walls, floors and shafts, plan for far less, and rely on mesh routing.

How many nodes per room controller?

Node counts are set by the ZigBee coordinator and by how many routers each zone can host, not by the thermostat alone. Size per zone rather than per floor, record the count, and keep headroom for rooms subdivided later.

What happens when the wireless network drops?

Whatever you specified — if the answer is "we are not sure", the design is unfinished. Specify local fallback so fan coils fail to a safe, energy-sane state rather than freezing the last setpoint, and alarm at the head-end.

ZigBee or BACnet MS/TP for a retrofit?

MS/TP if the bus already exists and is reachable; ZigBee where pulling new RS-485 through finished ceilings is the dominant cost. Many sites end up with both.

Sources

Previous Post
 
Next Post