Mixing Matter and SIG Mesh: How Should You Calculate Gateway Needs for a Five-Story Villa?

Calculate gateway capacity across devices, elements and bridged endpoints; separate coverage, throughput and failover, with a runnable model and HVAC selection checklist.

“One gateway per floor” works well for a first-round budget, but it is not enough to justify a purchase. For a five-story villa of about 1,000 square meters, first determine what each device controls and where that control runs. Then determine how many wireless entry points are needed.

This article continues the preliminary design for a five-story unfinished villa. There are no specific product models or on-site tests, so it does not claim that any gateway can cover a particular area or substitute a protocol’s theoretical scale for a product’s capacity.

The first article in this series worked through the wiring, port, and power budgets. This article addresses a harder purchasing question: when a manufacturer provides a “supported device count,” how do you judge whether it is enough for an entire floor? A runnable calculation script is provided below. It checks the budget and logic, not wireless performance in the field.

First, Separate Three Roles Often Called Gateways

A Matter controller handles application-layer management and control. A Thread border router connects a Thread network to the IP network on Ethernet or Wi-Fi; it does not replace a Matter controller.[1][2]

A SIG Mesh gateway is a different role. If it supports Matter Bridge, it can expose supported capabilities of downstream devices to a Matter controller. If it does not, integration requires another supported local interface. Bridging does not turn the original Bluetooth lights into Thread devices.

Device Question to ask
Matter controller Can it manage the target device types, and how is a backup restored?
Thread border router How are network credentials managed, how does it connect to the IP network, and what takes over if it fails?
SIG Mesh gateway / Bridge How many actual devices can it manage, which attributes does it bridge, and how does it coordinate across zones?

One physical device may perform several roles, but the product must explicitly support them. The name alone is not enough to establish that.

Two other quantities are easy to confuse: a Matter fabric and a Thread network. The former concerns a control platform’s trust and management relationships; the latter is a wireless network. Adding another Matter control platform for the same device does not automatically create another wireless path. Conversely, adding a Thread border router does not automatically replicate Home Assistant’s automation logic. Redundancy for control, transport, and automation should be specified separately.

Assign Functions First, So Devices Do Not All Take an Unnecessary Detour

In this proposal, SIG Mesh primarily handles lighting control; low-power sensors can use Matter over Thread; air conditioning and underfloor heating should preferably use documented, manufacturer-supported local interfaces; and cameras use the wired network and an NVR for recording.

This allocation is a starting point for product selection, not a ranking of protocol performance. Lighting control calls for checking consistency across panels, drivers, groups, and dimming, while HVAC needs complete operating status, zone control, and protective logic. Home Assistant’s official documentation also notes that a device’s native local integration may expose more proprietary capabilities than its Matter interface.[1]

Even when a product supports Matter, check “can turn on and off” separately from “has complete functionality,” item by item. An air conditioner might allow temperature setting but omit required fault codes; a light might support brightness but not the manufacturer’s special scenes. The quotation should include a feature matrix, not just protocol labels.

Are Five SIG Mesh Gateways Five Entry Points or Five Systems?

This must be answered during the quotation stage.

If each floor has an independent network, local configuration and the scope of failures are clearer, but cross-floor operations depend on higher-level coordination. If multiple gateways join the same Mesh network, verify that the manufacturer supports shared configuration, coordinated control, and duplicate-message handling. Do not casually add an unfamiliar gateway to the same network.

Bluetooth SIG’s official introduction explains that Mesh supports publish/subscribe messaging and multihop forwarding, enabling distributed control without putting all logic in a central controller.[3] But “the protocol allows it” does not mean the purchased panels have been configured to do it. Verify that panels can still control lights directly when the central controller and gateway are off.

For initial quotations, this project assumes one gateway entry point per floor while reserving a second candidate installation location. Ask manufacturers to provide:

  1. Limits on device count, elements, actual management capacity, and bridged endpoints.
  2. The configuration method for direct panel-to-light control.
  3. A clearly specified architecture for multiple gateways on one network or for zoned management.
  4. Where scenes execute, and what changes when the internet, central controller, or gateway goes offline.
  5. Procedures for configuration backup, gateway replacement, and reconnecting devices.

Separate logical networks by floor do not mean freedom from radio interference. With dense nodes, relaying, retransmissions, and status reports all need tuning. Enabling relay on every mains-powered device should not be treated as a guarantee of greater reliability.

Use a Hypothetical List to Define What “Enough Capacity” Means

Suppose one floor has 48 lighting drivers, 12 panels, and 20 sensors: 80 physical devices in total. This is only an illustration, not a lighting inventory for this villa.

Suppose the manufacturer also specifies these hypothetical limits: 100 managed devices, 200 elements, and 100 bridged endpoints per gateway. For this example, further assume that each lighting driver uses one element, each panel uses four, and each sensor uses one. The total element count is:

Physical devices: 48 + 12 + 20 = 80
Elements: 48 × 1 + 12 × 4 + 20 × 1 = 116
Bridged endpoints: assuming only 48 lights and 20 sensors are exposed, 68 in total

Element and bridged-endpoint counts must come from the actual product configuration and mapping. Do not assume “one physical device equals one element and one endpoint.” The figure of 68 depends on keeping panels in the local control layer; it could change if panel button events also need to be exposed to the higher layer.

Now suppose only 80% of each stated limit is used to leave room for expansion. This is not the same formula as “add 20% to current demand”:

Dimension Demand Stated limit Available at 80% utilization Preliminary result
Physical devices 80 100 80 Exactly at the planning limit
Elements 116 200 160 Capacity remains
Bridged endpoints 68 100 80 Capacity remains

If these loads can reasonably be distributed among multiple independent gateways, a lower bound on the number required is:

Capacity lower bound = max(
  ceil(physical device demand / available device capacity),
  ceil(element demand / available element capacity),
  ceil(endpoint demand / available endpoint capacity)
)

The result in this example is 1. Add one more sensor of the same kind, and the device count becomes 81, making the result 2. Expansion therefore occurs before the stated limit of 100 is exceeded.

But this applies only if the load can be split and each gateway has independent capacity. If two gateways both manage the same complete set of devices, each may need to accommodate all 81 devices; their capacities cannot simply be added together. Nor can rounding up conceal an indivisible room group that exceeds a limit on its own.

The 1 or 2 produced by the calculation is therefore only a capacity lower bound, not a purchase quantity. Coverage, scene load, product architecture, and failure scope must also be considered.

Verify Coverage, Throughput, and Failover Separately

First, test coverage: put the selected devices in their actual rooms, close the doors, and test at normal installation heights and orientations. Being able to commission a node or control it occasionally is not proof of stable coverage. Wall finishes, metal in ceilings, and cabinet doors may change the conditions, so repeat testing after finishing work even if tests in the unfinished building pass.

Second, test load: switch a scene involving 48 lights on the same floor while sensors report data and devices return status. Per-light control produces a different message pattern from a single group-address command. Confirm from the gateway implementation and logs whether grouping is actually used and whether many replies are awaited at once. Do not estimate protocol load by simply equating “48 lights” with “48 wireless packets,” and do not infer response time from a device-count limit alone.

Third, test failover: turn off one entry point and check whether affected devices recover through another. If the second entry point depends on the same floor-level PoE switch, it may reduce the effect of a single gateway failure, but it cannot withstand that switch losing power.

The remedies differ: inadequate coverage may call for relocation or suitable forwarding nodes; message congestion may call for fewer unnecessary retransmissions and polls; failed failover calls for checking backup paths and configuration. Treating all three as “add another gateway” can cost more without fixing the problem.

Multiple Thread Entry Points Do Not Automatically Form Redundancy Either

Home Assistant’s documentation says a Thread network can have multiple border routers, but configuration methods differ across ecosystems.[2] OpenThread documentation describes capabilities related to service discovery and connecting Thread partitions through the infrastructure network.[4] Whether a specific commercial device provides those capabilities, and how to enable them, must be checked against its model and firmware.

If each floor has one border router and thick floor slabs prevent radio signals from reaching between floors, “there are four others” does not answer whether another floor can take over when one floor’s entry point loses power. Perform a power-loss test using the most distant sensor on that floor, and record reachability and recovery time.

At the pilot stage, start with two adjacent floors. Verify credential synchronization, commissioning, service discovery, normal access, gateway failure, and restoration of power before expanding to five floors. Also confirm whether mains-powered devices can themselves route Thread traffic; battery sensors should not be assumed to form the backbone of cross-floor forwarding.

For this villa, I would use three on-site questions to decide the number of Thread entry points rather than fixing it at five in advance: Can the most distant device on basement level two reliably reach its floor’s entry point? After one entry point is lost, what other path is actually reachable? Do those paths share a power supply or switch? Passing only the first supports calling the setup “usable.” Passing the other two provides a basis for describing redundancy under the failure conditions tested.

Why I Would Initially Put Matter on the Same Automation VLAN

For this project, whose equipment choices are not yet finalized, I would initially put the Matter controller, relevant IP devices, and border routers’ infrastructure interfaces on the same automation network, retaining the necessary IPv6 and multicast communications. The commissioning phone would join that network temporarily; afterward, I would verify the normal day-to-day access path.

This does not mean removing isolation throughout the house. Cameras, guests, and management interfaces can still occupy separate zones. It simply reduces variables along the Matter path first, before rules are tightened one at a time. Official troubleshooting material explicitly emphasizes local IPv6 communication and explains that IPv6 internet service from the ISP is not required.[1]

Air Conditioning and Underfloor Heating: Put Interface Capabilities in the Quotation

If the requirement is “put every room into energy-saving mode after leaving home,” confirm at least the control targets, settable values, returned status, and failure behavior. A thermostat’s set temperature, room temperature, actuator position, and the equipment’s actual operating status are different pieces of information. Reading one of them does not justify claiming complete system feedback.

Must ask Acceptable, verifiable answer Insufficient answer
Control path Specifies the interface model, documentation, and local access method Supports smart home
Status source Distinguishes actual device reports from gateway-cached values and specifies the update interval Looks normal in the app
Central controller offline Original manufacturer panels and local temperature control work independently, demonstrated on site It should not go offline in theory
Command failure Timeouts are visible; an unconfirmed value is not shown as executed Successful transmission counts as success
Power restoration Specifies the recovery policy so an old scene does not overwrite a newer manual setting Restores automatically by default

A local interface may use a manufacturer’s proprietary protocol or an approved standard interface. Do not promise that “all air conditioners support Modbus” without model documentation. If infrared is a fallback, state its feedback limitations. Freeze protection, overtemperature protection, and pump-and-valve interlocks remain the responsibility of the relevant specialist systems; the central controller adjusts operating strategy only within the agreed scope.

What Evidence Turns Budget Figures Into a Purchasing Decision?

I would first give the contractor a table of candidate gateway locations and conduit provisions, give suppliers a capability checklist, and then purchase only the equipment needed for a pilot. Pilot results should record each model and firmware version, capacity definitions, bridged-attribute lists, cross-floor load behavior, failure effects, and recovery steps.

If one gateway meets the capacity requirement but fails coverage testing, purchasing just one is still unjustified. If a second gateway merely duplicates the same failure path, it should not be sold as redundancy at a premium. If two options both pass the requirements, compare maintenance complexity, manufacturer support, replacement costs, and licensing fees.

The final purchase quantity should follow from three results: whether device capacity is sufficient, whether room coverage passes, and whether the effect of an entry-point failure is acceptable. Floor count is only a planning starting point. Without records for those three factors, “five” and “ten” are both unverified numbers.

Supporting materials: capacity calculation and path-evaluation script, run results, supplier capability checklist template, and preliminary design. The script uses the assumptions in this article and checks only the arithmetic and dependencies; it includes neither radio propagation nor a Matter/Mesh protocol stack. Sources checked as of 2026-11-15:

Series reading: Part 1: Wiring and Capacity Budget · Part 2: Mixed Protocols and Gateway Selection · Part 3: Failure and Recovery Acceptance Testing

The templates and design document in the attachments are in Chinese, and the Python script depends only on the standard library. Software calculation results do not represent on-site measurements.

AI-generated scene illustration, not a photograph of this project or a wiring instruction.

Discussion

Share a question or your experience. Comments appear after moderation. Your email is never displayed publicly. Comment privacy

Loading comments…