The question "which bus do we use" for a large building sometimes collapses to "what we used on the last three projects." That's a bad heuristic. The three real-world standing buses (KNX, LON, BACnet) emerged in different decades, optimise for different operational scenarios, and will live in the building 20+ years regardless of which integrator picks up the contract today. The choice is made once, operation runs forever.
Three buses, three philosophies
KNX — the European residential standard
KNX (EN 50090, IEC 14543-3) emerged in 1990 from the merger of three older systems (EIB, BatiBUS, EHS). Goal: distributed intelligence, no central controller, long lifespan. Topology: line-area-backbone, max 64 devices per line, max 15 lines per area, max 15 areas — theoretically 14,400 devices per installation. Physical layers: KNX TP1 (twisted pair, 9,600 bit/s, simple installation over a separate wire), KNX IP (Ethernet, 100 Mbit/s, for the backbone between areas), KNX RF (868 MHz, retrofit).
**Strengths:** 500+ manufacturers, fully multi-vendor, no vendor lock-in. Components from 2000 still talk to components from 2026. ETS6 (engineering tool) is central, vendor-independent. Device prices 30–150 EUR.
**Weaknesses:** 9,600 bit/s isn't enough for today's large HVAC networks with 5,000+ datapoints. KNX IP is only a network bridge, not native high speed. Topology with line couplers becomes demanding to design for large buildings.
LON — the American-European legacy industrial
LON / LonWorks (ISO/IEC 14908) emerged in 1990 from Echelon Corporation. Goal: peer-to-peer industrial automation, distributed control, many physical layers (TP/FT-10, IP-852, power-line). Topology: free, max 64 devices per channel (FT-10 free topology). Speeds: 78 kbit/s (FT-10), 1.25 Mbit/s (TP/XF-1250), 100 Mbit/s (IP-852).
**Strengths:** runs in thousands of legacy industrial / HVAC installations from 1995–2010. Robust, peer-to-peer (no single point of failure). Echelon LonMaker tooling is mature.
**Weaknesses:** Echelon Corporation folded in 2018 (acquired by Adesto, then Dialog Semiconductor, then Renesas). New device development has stopped. Spare parts availability is declining. **A new LON project in 2026 = a mistake**, even though existing LON systems must run another 10–15 years.
BACnet — the American commercial HVAC standard
BACnet (ASHRAE 135, ISO 16484-5) emerged in 1995 as an open protocol for large HVAC + lighting + access control in commercial buildings. Goal: interoperability across many vendors, high speed, scalability to thousands of points. Physical layers: BACnet MS/TP (RS-485, 76 kbit/s, cheap), BACnet/IP (Ethernet, 10–100 Mbit/s, dominant today), BACnet/SC (secure, 2020, MQTT-style).
**Strengths:** today's de facto standard for BMS in commercial buildings (offices > 5,000 m², hospitals, high-rises, retail). BACnet/IP handles thousands of datapoints at very low latency. 800+ manufacturers, a robust multi-vendor ecosystem. The Niagara N4 platform (Tridium / Honeywell) has become the de facto universal BMS frontend that consumes BACnet + Modbus + KNX + others through gateways.
**Weaknesses:** device prices 80–250 EUR (higher than KNX, lower than LON). Configuration more demanding than KNX, requires a BACnet-certified integrator. Without TLS encryption (BACnet/SC), BACnet/IP is a cybersecurity nightmare — in 2020 we saw broad incidents with BMS exposed over the internet.
Real decisions from 2026
Hospital 24,000 m², 4 floors → BACnet/IP + Niagara N4
Project: regional hospital, 480 beds, 3,800 datapoints (HVAC zones, medical gas, sterilisers, cooling, emergency power). The investor originally proposed KNX because "we have it in the central building." We pushed BACnet/IP. Reason:
1. **3,800 datapoints** is above the practical KNX TP1 ceiling (9,600 bit/s) without heavy segmentation via line couplers. 2. **HVAC vendor (Trane, Daikin)** have native BACnet/IP cards on chillers; KNX integration via gateway would double the count of failure points. 3. **Building management software** (Niagara N4) is BACnet-native. KNX integration via the Tridium KNX driver works, but latency + reliability is worse than native BACnet. 4. **Lifespan**: the hospital plans 30+ years of operation. BACnet/IP is in 2026 the only protocol with certainty that it will be maintained into 2055.
CAPEX difference: KNX would have cost 285,000 EUR, BACnet/IP 340,000 EUR (+19%). Operational costs over 20 years: BACnet saves ~180,000 EUR (faster service interventions, cheaper spares, more available integrators).
Residential complex 48 flats → KNX
Project: new apartment block, 4 floors, 48 flats, common areas (lobby, garages, lifts, technical room). Per flat: KNX lights, KNX blinds, AC, zoned heating. Common: BMS for entry, lifts, emergency lighting, ventilation.
The KNX choice was clear:
1. **Per flat**, KNX is very cost-competitive. A KNX switch 80 EUR, a KNX actuator for 8 lighting circuits 280 EUR, a KNX thermostat 110 EUR. Total per-flat CAPEX 1,800–2,400 EUR. 2. **Multi-vendor** ecosystem means an apartment owner in 10 years can buy a new dimmer from a competing manufacturer (Gira, Jung, ABB, Siemens, MDT) and it works with the rest of the system. 3. **Common areas** are small (3–4 zones, ~250 datapoints), KNX handles them comfortably. 4. **Resale value**: a flat with KNX in 2026 has 8–15% higher price; a flat with a proprietary system (e.g. Loxone) has the same or lower price due to vendor lock-in.
Total CAPEX: 165,000 EUR for the complete KNX (per-flat + common). BACnet would have been 220,000 EUR (more expensive components, BACnet/IP infrastructure overkill for this scope).
Legacy 1995 office tower → LON-to-BACnet gateway, not rip-and-replace
Client: 18-storey office tower in Bratislava, built 1996, original LON Honeywell Excel 5000 system with ~1,200 datapoints. Original integrator gone, Echelon dead, spare parts only via aftermarket (eBay, specialist broker firms).
The investor considered full replacement with BACnet (650,000 EUR). We proposed a LON-to-BACnet gateway solution (94,000 EUR):
1. **Working LON devices stay** — thermostats, sensors, actuators in the field still work. Replacement means throwing out functional hardware. 2. **LON-BACnet gateway** (Loytec L-INX, Contemporary Controls BASRT) translates LON datapoints into BACnet objects. The new BMS frontend (Niagara N4) sees everything as BACnet. 3. **Gradual migration**: at every device replacement (sensor fails, thermostat runs its course) it's replaced with a BACnet equivalent. Over 10–12 years the LON layer organically thins out. 4. **Risk**: if the gateway fails, part of the system loses communication. Solution: 2× redundant gateways.
Difference over 15 years: rip-and-replace 650,000 EUR, gateway path 94,000 EUR + ~280,000 EUR of organic replacement = **saving ~270,000 EUR**.
Engineering tools — who works with what
ETS6 (KNX)
The central engineering tool for KNX. Licence: ETS6 Lite (5 devices, free), ETS6 Home (64 devices, 200 EUR), ETS6 Professional (unlimited, 1,100 EUR). Workflow: you import manufacturer catalogues (the ETS6 catalogue has 200+ manufacturers), build the topology, parametrise each device, download the program into devices over USB / IP interface.
**Learning curve**: 40–60 hours for a junior integrator for a basic project. 6–12 months for expertise on complex projects (advanced logic, IP backbone, secure KNX).
Echelon LonMaker / OpenLNS (LON)
Legacy tool, last major update 2015. Today used only to maintain existing LON installations. Licensing: typically bundled with LON hardware purchase, not separately available.
Niagara N4 (BACnet + everything)
The central BMS frontend from Tridium (owned by Honeywell). Runs on Java VM (Jace controllers in the field, Supervisor in the server room). Consumes BACnet, Modbus, KNX, LON via drivers. Workflow: you create a "station" (logical container), import drivers, lay out "points" (datapoints), script the logic in BAjava (Niagara-specific) or in higher versions in Python via nVibe.
**Licensing model**: per-point licensing, often 2–5 EUR per point. On a 5,000-point install that's 15,000–25,000 EUR in licences alone. Plus annual maintenance ~15% of licence per year.
**Competition**: Distech Controls Eclypse, Schneider Electric EcoStruxure Building, Siemens Desigo. All BACnet-native, similar features, similar pricing.
Cybersecurity — the most dangerous part of a BMS
In 2022–2024 we saw broad incidents with BMS exposed to the internet (a Shodan search for "BACnet" surfaces tens of thousands of devices with open ports). Attackers were able to manipulate temperatures, ventilation, access. A handful of rules:
1. **BACnet/IP NEVER directly to the internet.** Always behind a VPN or via OT/IT segmentation under IEC 62443. 2. **KNX IP** the same, plus enable KNX Secure (encryption) for data + auth. 3. **BACnet/SC** (Secure Connect, ASHRAE 135-2020) — use in new builds, TLS 1.2+, certificate-based auth. 4. **Network segmentation**: BMS VLAN separated from the office VLAN. Firewall rules explicit allow-list, no any-any. 5. **Logging**: every write command from the BMS frontend logged to the SIEM with timestamp + user identity.
Decision framework — 5 questions
1. **What type of building?** Residential < 200 flats → KNX. Commercial HVAC > 5,000 m² → BACnet/IP. Industrial / legacy → keep existing LON with a gateway to BACnet/IP. 2. **How many datapoints?** < 800 → KNX handles. 800–3,000 → KNX possible with careful segmentation, BACnet more comfortable. > 3,000 → clearly BACnet/IP. 3. **What BMS frontend do you have / want?** Niagara N4 → BACnet/IP native, KNX via gateway. ETS6 + KNX visualisation → KNX. Vendor-specific (Siemens Desigo, Schneider EcoStruxure) → BACnet (vendor-neutral) or the vendor's proprietary protocol. 4. **What lifespan?** > 25 years → KNX has the richest ecosystem. 15–25 years → BACnet is safer due to vendor diversity. < 15 years → any works. 5. **What integrator is available in your region?** Slovakia has a stronger ecosystem for KNX (residential) than for BACnet (commercial). For a BACnet project it's often Czechs or Austrians, which raises price and service availability.
Anti-pattern: "let's do it in Modbus, it's cheaper"
We regularly run into the proposal "let's deploy Modbus RTU, it's cheaper than BACnet." Modbus is a field-level protocol (PLC → sensor), not a BMS protocol. In small buildings (< 500 m², one HVAC unit) Modbus works. In any building with multiple zones, multi-vendor equipment, schedule programs — Modbus can't carry it. The client will pay in 5 years for both an upgrade to BACnet and throwing out the original Modbus infrastructure.
Our default recommendations
- **Residential < 200 flats, new build**: KNX, ETS6, multi-vendor (Gira / Jung / ABB / MDT).
- **Commercial HVAC > 5,000 m²**: BACnet/IP, Niagara N4, vendor-neutral.
- **Legacy LON installation**: LON-to-BACnet gateway (Loytec, Contemporary Controls), gradual migration over 10–15 years.
- **Industrial OT (production halls, maintenance workshops)**: Modbus / Profinet at field level, BACnet/IP only on the office / HVAC zones of the building.
---
*We design and implement BMS for residential complexes, commercial buildings, hospitals and industrial campuses. If you're considering a BMS for a project, the first call (60 minutes) walks through scope, lifespan requirements and regional integrator availability.*