Silicon Labs has executed a 200-node Matter-over-Thread validation network deployment to prove scalability in commercial IoT hardware.The semiconductor manufacturer announced the physical deployment at the Connectivity Standards Alliance (CSA) inaugural Unify event. The engineering data provides documented evidence that the Matter standard possesses the capacity to support expansive environments associated with commercial buildings, factories, and multi-dwelling units. The industry is pushing the technology past initial interoperability testing toward executing physical validations at true enterprise scale.Daniel Cooley, CTO at Silicon Labs, stated: “Matter is rapidly evolving from a smart home technology into a platform capable of supporting much larger deployments.“This work demonstrates not only that Matter-over-Thread can theoretically scale to thousands of devices, but also how Silicon Labs is helping customers deploy, manage and future-proof those networks through innovations spanning Matter, Thread, and Concurrent Multiprotocol technologies.”Production environment architectureEngineers constructed the validation network directly inside the Silicon Labs Boston Connectivity Lab and the surrounding corporate office spaces to evaluate performance under real-world conditions. The deployment team deliberately avoided a sterile laboratory simulation, choosing instead to distribute hardware throughout an environment containing heavy interference from active office devices, including twelve separate Wi-Fi networks, active Bluetooth devices, and standard corporate Thread network traffic.The physical layout comprised 40 distinct connectivity clusters distributed across the facility floor. Each cluster contained six ‘Wireless Starter Kits’ hosting a specific radio board, creating a dense mesh network spanning a maximum distance of approximately 50 metres between the furthest nodes. All devices were configured as Full Thread Devices.While commercial facility installations often mix edge endpoints with routing devices, the engineering team opted to configure every node as a router. This architecture deliberately stressed the system’s ability to process heavy background traffic generated by standard network routing maintenance operations.The underlying hardware consisted of BRD4187C development boards equipped with EFR32MG24 wireless systems on chips. The software infrastructure ran on top of FreeRTOS, utilising thirteen separate operating system threads to service concurrent operations, message parsing, and security routines.Silicon Labs’ Bluetooth stack remained active on the chips throughout the test window but did not process active data. Central control was handled via a four-gigabyte Raspberry Pi device running Ubuntu 22.01.2 LTS and an OpenThread Border Router implementation.Remediating technical constraintsImplementing a 200-node network in an active corporate space exposed several technical constraints that required engineering adjustments.Initial deployment strategies relied on Bluetooth Low Energy for commissioning individual devices onto the network fabric. Testing revealed distinct limitations regarding the range and device capacity of Bluetooth Low Energy when managing bulk automated setup cycles. The localised signal range frequently caused automated deployment scripts to fail before establishing fabric credentials across all nodes.To bypass these hardware constraints, the engineering team transitioned to an on-network commissioning methodology. Under this remediation framework, devices are first attached to the underlying Thread mesh structure using a network dataset provided via the OpenThread Border Router command-line interface. Once attached to the mesh, each hardware node registered a Service Registration Protocol instance. The deployment system then issued an on-network commissioning command, completing the fabric registration securely over the established Thread infrastructure without requiring proximity-dependent Bluetooth interactions. This altered approach yielded a 100 percent commissioning success rate across multiple test iterations, averaging seven seconds per individual node.A second technical hurdle emerged regarding data collection and latency instrumentation. Initial test phases relied on standard command-line interface printouts and software-level logging to track message propagation speedsThe engineering team discovered that the local processing time required to generate text logs via the device software stack inflated application-layer latency by approximately 50 percent. This overhead distorted the accuracy of the performance metrics, making them unsuitable for assessing true user experience, such as the exact time delay between a switch press and a light array changing state.Engineers resolved this measurement distortion by disabling all software-level Matter logging and routing data through the physical ‘Packet Trace Interface’. This interface is a hardware-level debugging system that captures live radio frequency traffic and internal debug events directly from the wireless system on chip via dedicated general-purpose input-output pins.By extracting timing data through a physical backchannel connected to a host computer, the team isolated application-layer execution events with microsecond-level accuracy, completely eliminating software logging overhead from the final performance data.Empirical performance metricsThe validation parameters targeted multicast messaging functionality, unicast communications, and long-term network stability under actual deployment conditions.Performance logs from the 50, 100, and 150-node test iterations revealed a highly stable latency curve. Across these three network sizes, the mean application-layer latency remained steady at approximately 91 milliseconds, with 95 percent of all commands processing within a 110-millisecond window across multiple payload sizes.The performance curve altered noticeably during the transition to the full 200-node network. Mean application latency climbed to 116.51 milliseconds for 8-byte payloads, 125.45 milliseconds for 16-byte payloads, 124.11 milliseconds for 32-byte payloads, and 129.76 milliseconds for 64-byte payloads. More importantly, the 95th percentile latency increased sharply, reading at 320 milliseconds for 8 bytes, 350 milliseconds for 16 bytes, 330 milliseconds for 33 bytes, and 340 milliseconds for 64 bytes.Investigation into the raw data exposed a distinct physical explanation for this performance variance rather than a core protocol failure. Automated deployment scripts activated devices sequentially based on network address order. The final 50 nodes required to scale the network from 150 to 200 endpoints were physically positioned in a corner hallway zone of the Boston facility. This structural area possessed only a single line of sight to the primary office floor, turning the entry to the hallway into a severe wireless congestion point.The physical infrastructure forced the network to rely on multi-hop forwarding and extensive packet flooding to reach the isolated hallway nodes. Structural concrete columns within the office space further blocked direct radio frequency paths, triggering additional network hops to navigate the perpendicular walls. This localised topological restriction directly caused the latency spike seen at the tail end of the data distribution.Despite these physical constraints, the network sustained continuous operation, demonstrating less than one percent packet loss for standard payload sizes under a three-hour continuous stress test.Controlled unicast testing and long-term stabilityTo isolate the exact incremental latency introduced by multi-hop forwarding away from open-air office interference, engineers conducted separate unicast testing within a controlled laboratory environment. This evaluation utilised Ramsey STE330 isolation boxes conductively linked via shielded coaxial cables to form a fixed, multi-hop topology ranging from one to seven hops.The unicast testing monitored the latency of Certificate Authenticated Session Establishment alongside application-layer command round-trip times. Security session setup represents an infrequent but mandatory computational operation required before application messaging can proceed.Laboratory data indicated that session setup latency scaled linearly with distance, tracking at 65 milliseconds for a single hop, 150 milliseconds at three hops, 251 milliseconds at five hops, and peaking at 352 milliseconds across a seven-hop span.Application-layer round-trip times followed a similarly predictable trajectory based on payload size and hop count. A single-hop unicast message required an average of 62 milliseconds for an 8-byte payload, rising to 105 milliseconds for a 296-byte payload.At a distance of four hops, the average round-trip time grew to 204 milliseconds for 8 bytes and 324 milliseconds for 296 bytes. At the maximum distance of seven hops, the hardware recorded mean round-trip latencies of 341 milliseconds for minimal 8-byte payloads and 473 milliseconds for maximum 296-byte payloads.The system maintained zero packet loss across all tested hop counts during these unicast evaluations. The gradual rise in response times stems directly from standard cryptographic decoding overhead, packet parsing within the multithreaded software layer, and transmission fragmentation mandated by the underlying Thread radio layer.This data establishes clear deployment parameters, confirming that while Matter-over-Thread naturally handles large physical areas without manual routing configurations, localised physical layout decisions and structural materials govern final network responsiveness.See also: Industrial automation drives private 5G past 2,000 deploymentsWant to learn more about the IoT from industry leaders? Check out IoT Tech Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including AI & Big Data Expo and the Cyber Security Expo. Click here for more information.IoT News is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.