RF Measurement & Spectrum Monitoring

An RF spectrum display is the output of a receiver and an estimator, not a photograph of the air. Antenna pattern and factor, site, cable loss, preselection, attenuation, preamplification, mixer/ADC range, local-oscillator phase noise, resolution filter or FFT window, detector, averaging, and calibration all shape what appears. Spectrum monitoring adds time and geography: it must detect intermittent emissions without turning missed samples into “unused spectrum.”

The purpose of a measurement decides its settings. Finding a weak spur beside a carrier, measuring channel power, estimating occupied bandwidth, checking frequency error, measuring field strength, direction-finding an interferer, and estimating occupancy are different experiments.

The calibrated reference-plane chain #

flowchart TB
  subgraph FRONT["Calibrated RF front end"]
    direction LR
    F["Incident E-field"] --> AN["Antenna"] --> C["Path + preselector"] --> RX["Receiver"]
  end
  subgraph EST["Estimation and reporting"]
    direction LR
    IF["RBW filter / I-Q"] --> DET["Detector / FFT"] --> COR["Corrections"] --> U["Result + uncertainty"]
  end
  RX --> IF

At a 50 Ω conducted reference plane, power can be reported in dBm. Field strength at a point uses V/m or dBμV/m. For an antenna factor $AF$ in dB/m, receiver voltage level $V_r$ in dBμV, cable loss $L_c$, and gain $G$ entered with explicit signs, a simplified relation is

$$E\,[\mathrm{dB\mu V/m}]=V_r\,[\mathrm{dB\mu V}]+AF+L_c-G.$$

The antenna factor depends on frequency, polarization, impedance environment, mounting, and sometimes direction. ITU-R SM.2138 treats the monitoring system—antenna, feeders, receiver, software corrections—as the tested object. Calibrating the receiver connector alone does not establish field-strength accuracy.

Centre/span and reference level protect meaning #

Centre frequency and span define what frequency interval is observed. Span also affects sweep time or FFT acquisition coverage. Start wide enough to find images, harmonics, and neighboring signals; narrow only after the emission and analyser tuning are understood.

Reference level and input attenuation set mixer/ADC headroom. Too little attenuation lets a strong signal generate compression, intermodulation, and false spurs inside the analyser. Too much attenuation raises the input-referred noise floor. A preamplifier lowers displayed noise for weak-signal work but reduces tolerable input and can itself intermodulate.

A simple overload test is to add 5–10 dB input attenuation:

• a real external signal falls by the expected amount at the input reference plane;
• an analyser-generated intermodulation product often falls faster or disappears;
• the displayed noise may rise after reference-plane correction because the front-end noise figure worsens.

RBW controls selectivity and integrated noise #

In a swept analyser, resolution bandwidth (RBW) is the IF filter bandwidth. In an FFT analyser, the analogous selectivity follows record length and window equivalent noise bandwidth. Narrower RBW:

• separates closer components;
• integrates less white noise, lowering displayed noise power per bin;
• usually lengthens sweep or acquisition time;
• can fail to follow fast-changing or wideband signals.

For flat noise density $N_0$,

$$P_N\approx N_0+10\log_{10}B_{noise},$$

so reducing noise bandwidth by ten lowers integrated noise by about 10 dB without changing physical noise density.

Video bandwidth (VBW) smooths the detected trace after the resolution filter; it does not improve frequency selectivity. A narrow VBW can hide modulation dynamics or peaks while making the trace look cleaner. Rohde & Schwarz's current fundamentals emphasize centre/span, reference level, RBW, and VBW precisely because each changes a different part of the result.

For FFT analysis, window main-lobe width, sidelobes, coherent gain, and ENBW replace the fiction of a rectangular bin. Zero-padding interpolates the trace but does not create resolving power.

Detector and averaging answer different questions #

Trace averaging lowers random variation but can suppress intermittent peaks. Max-hold accumulates the largest observed value and is useful for discovery, but its result depends on dwell time and number of trials. Neither is a probability statement without the observation schedule.

Frequency, power, and bandwidth measurements #

Frequency. A counter estimates cycles over a gate or period against a timebase; an FFT estimates a peak on a finite grid; a phase/frequency demodulator estimates instantaneous deviation. Resolution digits are not accuracy. Timebase calibration, SNR, modulation, gate time, trigger bias, and estimator interpolation set uncertainty. For modulated signals, state whether the result is assigned carrier, centre of occupied spectrum, average instantaneous frequency, or peak deviation.

Channel power. Integrate calibrated power spectral density over the defined channel bandwidth using an RMS/power detector and correct window/RBW normalization. Adding dBm samples arithmetically is wrong; convert to linear power before summing.

Occupied bandwidth. A common method finds the interval containing a specified percentage—often 99%—of total integrated power after defining the total integration span and noise treatment. Another method uses an $x$ dB-down emission mask. These are not interchangeable, and ITU recommendations or the radio standard define which is required.

Spurious emissions. Dynamic range near a strong carrier is limited by analyser phase noise, filter skirts, residual responses, and internally generated distortion. Validate with attenuation, narrower span/RBW, alternate preselection, and—when decisive—a second instrument/path.

Field strength adds site and antenna uncertainty #

ITU-R SM.378-7 addresses field-strength measurement at monitoring stations. The receiving site must control reflections, obstruction, polarization, antenna height, and local emissions. At frequencies where the environment is multipath-rich, moving the antenna can change the result by many dB. One point is evidence about one point.

An uncertainty model can be written

$$E=f(V_r,AF,L_c,G,\text{mismatch},\text{linearity},\text{site},\text{repeatability}).$$

Combine standard uncertainties with their sensitivity coefficients, preserve correlations, and report expanded uncertainty with coverage factor. Cable loss drift with temperature, antenna-factor interpolation, receiver linearity, mismatch, and site repeatability often dominate nominal display accuracy.

Direction finding estimates angle of arrival through amplitude comparison, phase difference, Doppler, time difference, or array beamforming. Reflections can point toward an image rather than the transmitter. Calibrate array geometry and channel phase, then use multiple sites or movement to intersect bearings; one bearing is not a location.

Occupancy is a sampling problem #

ITU-R SM.1880-2 frames spectrum occupancy as repeated observations classified occupied or unoccupied under defined thresholds and schedules. If an emission lasts $T_e$ every repetition period $T_r$, a monitor with dwell/gap pattern can miss it deterministically even over a long campaign.

An occupancy result must record:

• frequency bin/channel definition and RBW;
• threshold or detection algorithm, including noise-floor adaptation;
• dwell time, revisit interval, total observation period, and gaps;
• detector and trace processing;
• antenna/site coverage and sensitivity;
• handling of overlapping/co-channel signals and false detections;
• confidence or detection probability for the emission classes of interest.

“0% observed occupancy” means no samples crossed the rule. It does not prove no transmitter operated between samples or below sensitivity. Randomized or multiple-timescale schedules reduce synchronization blind spots; continuous I/Q capture reduces gaps but creates storage, processing, and calibration burdens.

A reproducible RF measurement workflow #

1. Define the measurand and reference plane: conducted power, E-field, frequency error, bandwidth, modulation quality, bearing, or occupancy.
2. Draw the entire path and load current correction/calibration data for antenna, cables, switches, gain, receiver, and timebase.
3. Survey wide with conservative headroom; identify strong signals that can overload later narrow measurements.
4. Set span, RBW/window, VBW, detector, sweep/acquisition time, attenuation, preamp, trigger, and averaging from the measurand.
5. Challenge the result by changing attenuation, preselection, RBW, time, polarization, and position one mechanism at a time.
6. Capture raw I/Q or trace data plus every instrument setting and timestamp when the decision must be replayable.
7. Apply corrections in linear units where required and build an uncertainty budget at the stated reference plane.
8. Validate with a traceable injected/reference signal and, for field systems, an end-to-end radiated system test.
9. Report what was not observable: gaps, sensitivity limits, overload exclusions, geography, and confidence.

RF monitoring is trustworthy when every displayed feature can be traced backward through estimator, receiver, path, antenna, site, and sampling schedule—and when the absence of a feature is bounded by a measured detection capability.