Coexistence & Desense: Reproducible Multi-Radio Interference Testing for Private 5G & IoT

Why coexistence and desense decide user experience

Private 5G devices rarely operate alone. They sit beside Wi-Fi/BLE/UWB/Zigbee/LoRa/GNSS and often host multiple cellular carriers/CA. The result: your receiver’s effective sensitivity is dictated less by textbook NF and more by in-band/adjacent interferers, self-jammers, and platform noise. If your lab cannot reproduce these conditions, your field performance will be unpredictable.

Working definition:

• Coexistence: Performance with external or co-located radios active (time/frequency/space sharing).
• Desense: Degradation of receiver sensitivity (ΔEIS or ΔTIS) caused by internal emissions, leakage, or external interferers.

KPIs that matter (and map to real UX)

• EIS / TIS vs interferer power (ΔdB at fixed PER/BLER target)
• Throughput & BLER vs SINR with interferers present (CDFs, not single points)
• Latency/jitter distributions under traffic load + coexistence
• Blocking, ACS, IM3 rejection: OTA analogues of conducted metrics
• EVM degradation of the wanted signal vs the interferer scenario

Rule of thumb: publish the desense curve—EIS degradation vs interferer level—per scenario. It’s the quickest way to compare designs and sites.

Architectures for reproducible tests

Anechoic (MPAC/CATR) with Controlled Injectors

• Pros: Deterministic angles/polarisation, precise power at DUT.
• Use: Adjacent/co-channel blocking, angle-of-arrival studies, beam vulnerability.

Reverberation Chamber (RC)

• Pros: Fast, statistically rich, excellent for throughput robustness under multi-radio stress.
• Use: High-volume coexistence sweeps, production screening, “will it break” regressions.

Hybrid is common: characterise mechanisms in anechoic → harden designs and vendors with RC stress tests.

Interferer menu (cover the real world)

• Co-channel NR: same band, same SCS/BW; wanted vs interferer PRBs overlap.
• Adjacent-channel NR: set Δf and ACLR/ACLR′ masks.
• Wi-Fi 6/6E/7: 80/160/320 MHz OFDM OBW near FR1/6 GHz devices.
• Bluetooth Classic/LE: advertising vs data; duty cycles that hit LNA AGC.
• UWB: short bursts with high PAPR; test ranging-on + 5G Rx.
• Zigbee/LoRa: narrowband blockers; watch intermod with 2.4 GHz/BLE.
• CW tones: worst-case blockers and 2-tone IMD diagnosis.
• Self-jammers: on-device TX (Wi-Fi hotspot, NR uplink bursts) that desense own Rx path; include CA 2CC/3CC/4CC and hotspot tethering.

Vary modulation, duty cycle, bandwidth, polarisation, AoA, and time alignment (co-existence managers often depend on MAC-level arbitration).

Power at the DUT: calibration you can defend

1. Conducted reference: validate generator level & waveform EVM.
2. OTA transfer function: measure path loss from interferer antenna to DUT phase centre using a calibrated probe/reference.
3. Set-and-lock: drive interferer to achieve Pin(DUT) within ±0.5 dB at each frequency/angle.
4. Log environment: temperature, chamber state, stirrer speed (RC), positioner pose (anechoic).

Without traceable power at the DUT, desense curves are meaningless.

The core experiment (repeatable, portable)

Step 0 – Baseline
Measure EIS/TIS and throughput/BLER for the wanted link alone.

Step 1 – Add an interferer
Enable one interferer at a defined Δf / bandwidth/duty cycle / AoA. Sweep Pin(DUT), e.g., from −90 to −30 dBm (or RC field strength equivalents).

Step 2 – Record Δ
At a fixed PER/BLER target, compute ΔEIS (dB) vs Pin. Repeat for 2–3 SINR anchors.

Step 3 – Build the matrix
Expand across interferer types, angles, polarisations, and device orientations (hand/head, wall/ceiling for CPE).

Output
A desense surface: ΔEIS(Pin, scenario) + throughput/latency CDFs per scenario. This becomes your regression dashboard.

Self-desense diagnostics (the silent killer)

Symptoms: good conducted specs, mysterious OTA fails when another on-board radio or display/PMIC is active.

Checklist:

• Time-correlate sensitivity dips with internal TX bursts or digital activity.
• Near-field scan around antennas, display ribbons, and PMIC inductors.
• Mode sweep: Wi-Fi tethering, BT audio, UWB ranging, GNSS+NR DL, CA uplink.
• Mitigation trials: ground stitching, filter swaps, shielding foils, DC-DC frequency plan changes, antenna isolation tweaks.

Capture before/after desense curves to quantify fixes.

Throughput correlation (don’t stop at sensitivity)

• Compute effective SINR including desense (noise rise + interference coupling).
• Plot Throughput vs effective-SINR with and without interferers; compare error bands to field logs.
• In RC, pair with coexistence traffic (e.g., saturated Wi-Fi AP near the DUT) to emulate real contention.

Uncertainty & repeatability (what to publish)

• Contributors: OTA power uncertainty, chamber field uniformity, positioning, instrument linearity, reference antenna factor, and temperature drift.
• Report: Expanded uncertainty (k=2) for EIS and interferer power; attach calibration certificates and daily quick-check logs.
• Golden setup: lock DUT build, seeds, profiles, calibration files; exchange across sites for cross-lab correlation.

Automation that scales

• Script interferer sweeps (type, Δf, duty, power), wanted-link traffic, and logger hooks (KPIs, spectrum snapshots).
• Export single-file bundles (configs + results + environment) so vendors/partners can replay.
• Produce one-click reports: desense curves, KPI CDFs, uncertainty tables.

Common failure signatures (and quick fixes)

• Cliff at specific Δf: front-end filter skirt or duplexer leak → verify mask, try steeper filters/isolation.
• Only at high duty cycles: LNA AGC or baseband overload → adjust AGC timing, baseband clipping margin.
• Uplink collapses when Wi-Fi is on: PA supply droop/PMIC noise → re-plan DC-DC frequency, add decoupling/shielding.
• Great EIS, poor throughput: MAC/stack coexistence arbitration → tune coexistence manager and retry with realistic traffic models.

Procurement & lab build checklist

• Interferer sources: NR/Wi-Fi/BLE/UWB/Zigbee/LoRa generators with precise power control and APIs.
• OTA plumbing: calibrated antennas, combiners/couplers (for conducted baselines), low-leak feedthroughs.
• Chamber capability: anechoic (angle/pol control) and/or RC (fast statistical stress), with field calibration kit.
• Power-at-DUT metrology: reference probes, path-loss calibration, temperature logging.
• Lifecycle: drift monitoring, re-cal kits, SLAs, spares; portable test scripts and golden DUTs for cross-site correlation.

Coexistence/desense is not a side test—it’s the test that aligns lab results with the real world of crowded spectra and multi-radio platforms. Make power at the DUT traceable, design scenario-rich sweeps, publish uncertainty, and automate the workflow. Your reward: devices and networks that keep their promises—even when the air is full.