Multi-Reader Deployment

Production RFID deployments typically involve multiple readers working in concert. A typical warehouse might have 4–8 readers at dock doors and 2–4 per conveyor line — all feeding data into a central middleware that deduplicates, filters, and routes tag events to business systems (WMS, ERP, TMS).

The architecture has three layers: Edge (readers + antennas at physical read points), Middleware (event processing, deduplication, business logic), and Integration (API connections to WMS/ERP/TMS). The middleware layer is critical — it transforms raw tag reads (EPC + antenna + RSSI + timestamp) into meaningful business events like 'pallet received at dock 3' or 'case loaded onto truck B'.

Network design: Each fixed reader connects via Ethernet (preferred for reliability) or Wi-Fi. Use a dedicated VLAN for RFID traffic to isolate it from general network traffic. Typical bandwidth: 1–5 Mbps per reader during active inventory. Ensure ≤50ms network latency for real-time applications. Use heartbeat monitoring to detect reader failures — a reader going offline at a dock door means missed shipments.

When multiple readers operate in close proximity, their RF signals can interfere. Three primary coordination strategies exist, each with trade-offs:

Frequency Hopping Spread Spectrum is the primary interference management mechanism in regions like Vietnam (920–925 MHz). The reader rapidly switches between channels during inventory rounds, ensuring that even if two readers collide on one channel, they separate on the next hop.

Practical FHSS configuration: Configure each reader with a channel mask defining which channels to use. For 2 adjacent readers, assign complementary masks — Reader A uses channels [0, 2, 4, 6, 8] and Reader B uses channels [1, 3, 5, 7, 9]. This guarantees zero overlap. For 3 readers, split into groups of 3–4 channels each.

Channel hopping speed matters: faster hopping reduces the probability of sustained collisions but adds overhead. Most readers hop after each inventory round (every 100–400ms). The NRN protocol SET_WORKING_FREQUENCY command configures the channel list — e.g., bytes [0, 2, 4, 6, 8, 10] set channels 0 through 10 with 1 MHz spacing.

SET_WORKING_FREQUENCY payload:

2 readers (zero overlap):
  Reader A: [0, 2, 4, 6, 8]   → 920.0, 921.0, 922.0, 923.0, 924.0
  Reader B: [1, 3, 5, 7, 9]   → 920.5, 921.5, 922.5, 923.5, 924.5

3 readers:
  Reader A: [0, 3, 6, 9]      → 920.0, 921.5, 923.0, 924.5
  Reader B: [1, 4, 7, 10]     → 920.5, 922.0, 923.5, 925.0
  Reader C: [2, 5, 8]         → 921.0, 922.5, 924.0

Dense Reader Mode is an EPC Gen2 feature specifically designed for environments with many closely-spaced readers (>2 readers within 3m). DRM uses narrower channel bandwidth and Miller-encoded tag responses to reduce inter-reader interference.

DRM trade-offs: Enabling DRM improves multi-reader coexistence significantly but reduces single-reader performance — the narrower bandwidth means lower data throughput per reader. In practice, a reader in DRM mode inventories tags about 20–30% slower than in standard mode, but the system-level performance improves because readers no longer block each other.

When to enable DRM: More than 2 readers within 3 meters of each other. Readers at adjacent dock doors that can 'see' each other's tags. Dense ceiling-mount retail installations. When to keep DRM off: Isolated readers with >5m separation. Single-reader handheld applications. Conveyor tunnels with good RF shielding.

Tag starvation occurs when certain tags in a population are consistently skipped during inventory rounds. This typically happens because stronger tags (closer to the antenna, better-oriented) dominate the reader's attention, and weaker tags never get a chance to respond.

Detection: Monitor your unique-tag-count vs total-read-count ratio. If you're reading 50 unique tags but getting 5000 total reads, the strong tags are being re-read 100× while weak tags are starving. A healthy ratio is unique-tags × 3–10 = total reads.

Mitigation strategies: Use proper Q value (too low = collisions cause weak tags to lose, too high = slow rounds). Enable session persistence (S2/S3) so already-read tags go silent. Rotate antenna focus by sequencing through antenna ports. Adjust power levels to create more uniform coverage — reduce power on antennas pointing at nearby tags, increase power on antennas covering distant areas. Use the 'target' flag to alternate between A→B and B→A inventory directions.

Advanced technique: Implement 'select' commands to partition the tag population into groups and inventory each group separately. This is particularly effective for mixed populations where small item-level tags coexist with large pallet-level tags.

These configurations have been validated in production deployments and represent best practices for common scenarios.