The Nature of the Problem
GNSS (such as GPS and BeiDou) receivers operate in the L-band around 1.5 GHz, with received power far below the thermal noise floor, requiring high-gain front-ends and correlator accumulation. When cellular modules transmit LTE/NR, power can reach several hundred milliwatts, and out-of-band noise, harmonics, and intermodulation products may fall into the GNSS band or couple directly into the receiver through the PCB, causing sensitivity degradation (desense) or even inability to locate. Automotive, logistics, and shared devices often require both cellular and GNSS simultaneously, making coexistence design critical for mass production.
Spatial and Antenna Isolation
Increasing the distance between cellular and GNSS antennas, using orthogonal polarization, and using the ground plane for isolation can reduce coupling. Shared-antenna solutions require high-isolation duplexers and filters, with costs and insertion loss needing trade-offs. Antenna feedlines should avoid parallel routing, and connectors should be well-shielded models. The metal structure of the device may form cavity resonances, altering isolation, which needs verification in the assembled device state.
Frequency-Domain and Time-Division Strategies
Placing a high-Q bandpass filter at the GNSS front-end can suppress out-of-band interference but introduces insertion loss, which needs to be designed together with the LNA noise figure. In some scenarios, time-division is used: GNSS tracking is paused during cellular transmission slots and concentrated integration is performed in idle slots, relying on assisted positioning and ephemeris prediction compensation. Protocols and baseband need to support coordinated scheduling to avoid simple "all-on-at-once" operation that collapses both performances.
Testing and Certification
Evaluate the GNSS C/N0 degradation under the maximum cellular transmission separately in conduction and radiation environments. In road tests, focus on harsh conditions where urban canyon multipath and high cellular load occur simultaneously. If exports require RED/FCC, confirm that GNSS is not the main transmitter but may still be affected by EMC test setup, and communicate the test plan with the laboratory in advance.
Engineering Practice Supplement
From the perspective of "GNSS and cellular coexistence," common on-site problems often concentrate on "parameters meet laboratory standards but are difficult to reproduce in the field." Beyond RF and protocols, also check for power-ground bounce, connector contact, and firmware race conditions. On the cellular side, it is recommended to collect RSRP/SINR, cell reselection, and PDP reconstruction records; on the application side, there should be clear strategies for retries, idempotency, and queue persistence. When involving data security and compliance, include the scope of data collection, storage location, retention period, and third-party sharing in the privacy and contract annexes to avoid high costs for patching after deployment.
Out-of-band energy falling on navigation frequencies
Power from a navigation satellite at the antenna port is far below the receiver’s own noise. The correlator lifts the signal out of that noise with a long coherent integration. A cellular transmitter works in another band, but its noise floor, harmonics, and intermodulation, once they fall into a narrow band such as GPS L1 or BeiDou B1, raise the noise floor of the navigation front end. Carrier-to-noise ratio falls, and in a severe case the whole search fails. Coupling is not only through space. A slot in the ground plane, an unshielded RF connector, or ripple on a shared supply can deliver transmit energy into the navigation low-noise amplifier. Coexistence is three things at once: transmitter out-of-band energy, a coupling path, and linearity of the navigation front end. Adding a navigation antenna by itself does not solve it.
A coexistence design is required when cellular and navigation run at the same time in one enclosure, especially at high transmit power, in a hot vehicle, or inside a metal cavity. If the service can pause tracking during a transmit slot and integrate again in an idle slot, time division can be added, but the baseband schedule must actually separate them, not leave both chips running at full rate. If an export test places navigation reception and cellular transmission in one setup, the product state must match production routing.
Isolation that falls back inside the product cavity
Isolation that passes with two antennas spread apart on a development board often fails once they are inside a metal shell, because a cavity resonance sends the energy back. That is the most common reversal. Too little filter insertion loss already costs navigation sensitivity when cellular is off; too much still fails to stop intermodulation near the passband. Measuring desense only at a conducted port, and ignoring spatial coupling between antennas, still loses satellites on the road. A time-division scheme that cancels the idle gap during retransmission or heavy uplink eats the navigation integration time, and assistance ephemeris cannot restore the tracking loop.
A comparison should hold transmit power, band, and navigation constellation fixed and record the change in carrier-to-noise ratio, not only whether a position appears. Thresholds in a weak-signal setting are covered in weak-signal GNSS. A-GNSS shortens a cold start, but assistance data does not fix a transmitter that has raised the noise floor. On a revision, lock antenna position, shield can, and battery bay as one variable, or it is impossible to tell whether the match improved or the coupling happened to shrink.
Hold the cellular band, transmit power, and navigation constellation fixed, and record only the change in carrier-to-noise ratio. “Whether a point appeared” is not the only conclusion. Lock the metal shell, battery bay, and shield can as one set of variables. Isolation with antennas spread on a board does not represent the enclosed product. Look at filter insertion loss and out-of-band rejection together, and retest intermodulation near the passband at maximum transmit power. A time-division scheme has to confirm that retransmission and heavy uplink do not consume the idle slots. Road tests should use urban-canyon routes with high load. A conducted-port pass in the lab does not replace spatial coupling.
Record the enclosure state in the coexistence test. When the board passes and the product loses satellites, check the cavity and the routing before the navigation algorithm. Use the carrier-to-noise drop at maximum transmit power as the bar for whether a revision is acceptable, and do not mix it in one table with the low-power result. If the slot schedule cannot actually separate the two, do not claim time-division isolation in the software description.