LINK BUDGET
Start with power gains and losses.
A basic link budget adds transmit power and antenna gains, then subtracts propagation, cable, connector, mismatch, polarization, obstruction, implementation, and other losses to estimate received power.
Useful margin is the difference between expected received signal and the receiver performance needed for the chosen modulation/data rate under the required error rate. The arithmetic is simple. The hard part is choosing realistic losses and fade margin instead of filling the spreadsheet with free-space optimism.
RSSI
Received power is not the same thing as decodable signal.
RSSI is device-specific received-signal-strength information. It can be useful for trend and threshold behavior, but it does not by itself tell you signal-to-noise ratio, interference type, demodulation margin, packet error, or whether the reported energy belongs to your transmitter.
A strong interferer can produce impressive RSSI while destroying the desired link. A weak desired signal in a quiet channel may perform better than a stronger signal buried in noise.
PACKET ERROR
The application experiences retries, not dBm.
Track packet error rate, retransmission count, acknowledgement loss, forward-error-correction state where available, data-rate adaptation, and the latency distribution those behaviors create.
If a protocol retries automatically, an application may see near-perfect delivery while airtime and energy cost quietly triple. Reliability statistics without retry statistics can hide a deteriorating radio environment.
FADING
Moving a few centimeters can change the channel.
Multipath propagation lets reflected copies of a signal combine constructively or destructively at the receiver. Indoor environments, vehicles, industrial spaces, and urban geometry can produce deep local fades that are not captured by a smooth distance-only path-loss model.
Diversity in antenna location, polarization, frequency, or time can reduce dependence on one bad channel state. The system should be tested in the geometry where it actually operates, including moving people, doors, machinery, and vehicle orientation.
ANTENNAS
The antenna installed in the product is not the antenna from the datasheet.
Nearby batteries, ground planes, displays, enclosures, cables, hands, bodies, mounting hardware, moisture, and manufacturing variation can detune or shadow an antenna. Orientation changes polarization alignment and radiation pattern.
Prototype RF performance with the production-like enclosure and mounting condition. An evaluation board on a foam block is a reference point, not a field configuration.
INTERFERENCE
Noise occupancy can matter more than distance.
Other radios, switching electronics, motors, USB clocks, DC/DC converters, harmonics, and external emitters can raise the noise floor or create burst interference. Shared unlicensed bands can be crowded even when your own RF design is excellent.
Measure channel occupancy and error behavior over time. A failure that appears every few seconds may be periodic interference rather than fading or firmware scheduling.
CONGESTION
A healthy RF path can still have no airtime left.
Contention protocols, gateway capacity, duty-cycle limits, hidden nodes, retransmissions, and too many clients can saturate a network without any receiver sensitivity problem.
Track channel utilization, queue depth, retry backoff, and time-to-transmit. A network that succeeds eventually can still be unusable for control because worst-case latency exploded.
POWER
Weak links consume battery.
Retries require transmitter on-time, receiver listening, processor wake time, and sometimes higher transmit power or slower robust data rates. A marginal radio link can therefore show up first as unexpected battery drain.
Energy-per-delivered-message is often a more useful system metric than transmitter current alone.
DEGRADATION
Design lower-bandwidth modes before the link forces them on you.
When link quality falls, a system can reduce telemetry rate, compress messages, prioritize control/status data, postpone bulk transfers, lower video quality, increase local autonomy, increase buffering, or change modulation/rate where supported.
The important distinction is intentional degradation. The operator should know that the link is impaired and which capability was reduced. A machine that silently drops half its telemetry while still displaying “CONNECTED” has chosen optimism as a protocol.
FIELD METHOD
Measure link behavior as a system.
1. Build a realistic link budget. Include cable, antenna, body/enclosure, and fade margin.
2. Log RSSI/SNR plus packet outcomes. Power without error behavior is incomplete.
3. Record retries and latency distributions. Averages hide collapse.
4. Test orientation and enclosure state.
5. Survey interference over time.
6. Measure energy per delivered payload.
7. Force degraded modes. Confirm the system preserves the most important traffic before total loss.
BOTTOM LINE
Connectivity is an operating envelope.
Useful radio health combines received signal, noise/interference, packet success, retries, latency, airtime, energy, geometry, and application requirements.
The link is failing when communication becomes too expensive or too late, not only when the bars disappear.
SOURCE TRAIL
Technical starting points.
NASA — Space Communications and Navigation background
Analog Devices — RF path and link concepts