Link budgets, interference, reliability and system testing

Name the reference point and unit of every term

Power in dBm is 10 log₁₀(P/1 mW); a power ratio in decibels is 10 log₁₀(P₂/P₁). Antenna gain in dBi uses an isotropic reference. A decibel gain or loss can be added to a compatible logarithmic power account. Adding two dBm powers directly is incorrect: convert to linear powers, add them and convert back.

A budget should identify transmitter output, feeder losses, directional antenna gains, propagation loss and receive-path loss without counting the same effect twice. State the receiver reference plane. The required input level depends on modulation, coding, bit rate, bandwidth and the accepted error criterion. A generic sensitivity number without those conditions can be misleading.

Compare signal, noise and interference carefully

Margin is the difference between predicted available signal and a stated required level in matching units. Extra loss reduces that difference. Required fade allowance depends on the actual environment and availability goal; a nominal number does not identify an outage probability. The reverse direction may have a different power, antenna or noise account.

Signal-to-interference-plus-noise ratio compares signal power with the combined unwanted power over a consistent bandwidth. Independent noise-like components can often be combined by linear power addition. Coherent or structured interference can require a more specific model. A strong out-of-band signal can also affect a nonideal receiver; simply assigning it zero in-band power does not describe every impairment.

Measure the service at the right boundary

Count bit errors over a known tested bit count and distinguish pre-decoder from post-decoder errors. Packet failures, loss, duplication and delay are separate outcomes. Goodput counts useful delivered payload over elapsed time; coded-bit rate includes other work. A short test with no observed errors is not proof of zero error probability.

Vary conditions that challenge the model: attenuation, pointing, interference, timing and traffic, within an appropriate controlled test arrangement. Record settings and failures. A successful laboratory condition does not establish field coverage, every receiver configuration or long-term availability. Preserve the tested denominator rather than quietly excluding intervals where the receiver lost synchronization.

A six-decibel nominal margin can be exhausted

For an original budget, transmitter 10 dBm plus 6 dBi transmit gain, minus 1 dB transmit feeder loss and 80 dB path loss, plus 8 dBi receive gain and minus 2 dB receive feeder loss gives−59 dBm at the receiver. Suppose its supplied requirement is−65 dBm for the stated signalling and error target. Nominal margin is 6 dB.

An additional 7 dB loss makes received power−66 dBm and margin−1 dB. Separately, a test with 17 wrong decoded bits among 1, 000, 000 compared bits has observed fraction 1.7×10⁻⁵. Report the tested conditions and loss-of-sync intervals; neither calculation alone gives a deployment reliability guarantee.

Try the reasoning

1. Two independent noise-like components are−90 and−85 dBm. What is their combined power?

   Show the answer and reasoning

   Add 10^(−90/10)+10^(−85/10) in milliwatts, then convert: approximately−83.807 dBm, not−175 dBm.

2. At−59 dBm desired signal, what is the ratio above that combined unwanted power?

   Show the answer and reasoning

   Approximately 24.807 dB under the stated same-bandwidth power-addition assumptions. Mapping it to errors needs the receiver and signalling model.

3. Do 17 errors in a million compared bits establish the fraction of bad packets?

   Show the answer and reasoning

   No. Packet sizes, error clustering, coding and lost packets matter. Bit and packet denominators answer different questions.

What to carry forward

A budget and a test become useful when their reference planes, conditions, uncertainty and denominators agree.