RF measurement errors and corrections
Recognise the common errors in RF and EMF surveys, from overload, geometry control and detector mistakes to bad unit conversions and poor documentation.
Offline pages
Saved on this device only. Connect at least once every 30 days to confirm Premium. Sign-out or a change of account removes saved copies.
On this page
You stopped at:
ContinueAn unexpected RF reading is a reason to test the measurement, not proof of one particular fault. Start with its quantity, coverage and operating conditions, then choose a controlled check whose result would distinguish competing explanations. Change one relevant variable at a time and keep the original and corrected records.
Use instrument and chain records, broadband–selective comparisons and uncertainty. A valid result requires more than plausible size or repeatability. A missing band, wrong detector or common correction error can give a stable number.
Protect the measurand before interpreting a symptom
Record field quantity and units, band coverage, RMS or peak definition, time window, source state, geometry and enabled corrections. Compare like with like: a peak-hold scan and a time-averaged field measurement need not agree. If a required part of the method is unknown, keep the result provisional rather than assigning it an unsupported assessment conclusion.
Check instrument ratings and overload indications before using data. Follow the equipment’s supported range or attenuation procedure; a setting cannot protect a sensor located before it. Keep probe position, operator distance and source state fixed during a range check where the method permits. Moving away changes the field you are measuring, so it is not by itself a controlled test of instrument linearity.
Test overload with a reference-plane comparison
An overloaded stage can compress the signal or create spurious responses; its bias need not be upwards. If raw input power is measured through a known external attenuation A, recover the upstream power level by adding A in dB. If the display already compensates it, do not add it again.
For a stable input and a linear supported chain, increasing external attenuation by 10 dB should lower the uncorrected receiver power by 10 dB while leaving the corrected upstream estimate unchanged within uncertainty. Agreement supports consistency over those tested settings; it does not prove calibration, complete spectral capture or exposure compliance. Keysight’s analyzer documentation illustrates why attenuation, preamplification, correction tables and acquisition settings belong to the measurement method.
Guided practice 1
Use a controlled range check instead of a symptom guess
Problem
A supported receiver chain is tested at one fixed point with a stable source and unchanged detector and band integration. Its raw, uncorrected readings are −15 dBm with 0 dB external attenuation (overload warning on), −22 dBm with 10 dB attenuation, and −32 dBm with 20 dB attenuation (both warnings off). The two added attenuators have the stated characterised losses and all input ratings are respected. Find the corrected upstream levels from the 10 dB and 20 dB settings. Then find the ratio of the original displayed power to their common corrected estimate, using P₁/P₂ = 10^((L₁-L₂)/10). State which data support a linearity cross-check and what remains unproven.
Work through the model, then compare your reasoning
Hints
Hint 1
Check the model
The meter readings have not been corrected automatically. Compare the two warning-free estimates before interpreting the overloaded one.Show solution step by step
Reconstruct comparable upstream levels
Method
Add each characterised external attenuation to its raw level.Reason
Both corrected results refer to the same upstream plane and input state.Working
Answer
At 10 dB attenuation, the estimate is -22 + 10 = -12 dBm. At 20 dB it is -32 + 20 = -12 dBm. The raw reading falls by the expected 10 dB, while the corrected estimates agree.
Interpret the invalid initial reading
Method
Compare the initial display with the candidate linear-region estimate, then limit the conclusion.Reason
An overload warning invalidates the original measurement; agreement between corrected settings supports only the tested range consistency.Working
Answer
The ratio is 10^([-15-(-12)]/10) = 10^(-0.3) = 0.501187…. The original display is about half the candidate upstream power, illustrating a downward compression bias. Do not treat the overloaded value as a valid lower exposure result. Repeat final measurements with a supported valid chain and its uncertainty; the range cross-check does not establish all-frequency coverage, detector adequacy or a conformity decision.
Choose checks that distinguish other causes
| Observation | Competing explanations | Controlled evidence to seek |
|---|---|---|
| Reading changes when a person moves | Body coupling, changed position or real source variation | Fixed probe and source conditions; documented operator distance and cable routing |
| Reading changes with small displacement | Real gradient, multipath, poor position repeatability | Marked positions and repeated geometry, using the prescribed spatial plan |
| Rotation changes the result | A directional sensor, residual anisotropy, fixture or handling effects | Sensor response specification and stable multi-axis comparison at fixed geometry |
| Sweeps vary or omit bursts | Capture timing, bandwidth, detector or source-state mismatch | Acquisition suited to the signal’s timing and full bandwidth, with documented controlled comparisons |
| Broadband and selective disagree | Missing bands, unequal averaging, response weighting, overload or corrections | Matched quantity and state, verified coverage and one chain/configuration check at a time |
A longer sweep is not a universal cure for intermittent signals: it can visit a frequency at the wrong time. Capture probability, triggered or time-domain methods and integration need to match the signal and instrument capability. A max-hold trace can help locate peaks without representing the required time average. Establish those distinctions before comparing numerical levels.
Fix the mathematical model, not only the arithmetic
For a required linear power average, convert logarithmic readings to power, apply the specified time or spatial weights and only then convert back if needed. An arithmetic dB average produces a geometric mean of the linear powers. A voltage-level difference uses 20-logarithm scaling, while a power-level difference uses 10-logarithm scaling.
Plane-wave conversion such as S = Eᵣₘₛ²/η₀ needs justified locally plane-wave conditions. It cannot reconstruct a near field’s energy flow from one E reading when the E–H relationship differs. Measure or model the quantity required by the method with appropriate characterisation. Likewise, a transmitter’s guided output power is not a local exposure reading.
A weak distance trend alone does not prove overload; near fields, extended sources and reflections can violate an inverse-distance model. Check the supported instrument response and physical field model separately. Once a fault is identified, correct its actual owner—coverage, geometry, settings or chain—not the result by arbitrary scaling or averaging.
Form a testable explanation for a cross-check gap
Check your understanding 1
At the same fixed point and stable source state, a characterised flat-response broadband meter reads 13 V/m RMS. A valid selective measurement covers only band A and gives 12 V/m RMS. The source inventory also lists a distinct-frequency band B outside that selective span. Range checks are consistent, and compatible averaging makes the root-sum-square model applicable. If the sole missing contribution were B, what RMS field would explain the gap? Propose a controlled follow-up and state why averaging 13 and 12 into one result would be inappropriate.
Show answer
Under the stated hypothesis, EB = square root of (13²-12²) = 5 V/m. Extend the characterised selective coverage to B using suitable settings and corrections, with the point, source state and averaging controlled, then test whether the combined contributions reconcile within uncertainty. The inferred 5 V/m is a prediction conditional on the model, not a measurement or proof that B is the sole cause. The 13 V/m total and the 12 V/m band-A result are different measurands; averaging them would hide missing coverage rather than estimate either quantity honestly. If the prediction fails, investigate the remaining response, state and chain hypotheses instead of forcing agreement.
Review the measurement decisions
Check your understanding 2
An analyzer gives an overload warning and corrected readings change when a supported input attenuator is increased. What should you do before using the data?
Show answer
Treat the original data as unreliable. Set adequate attenuation or range, verify linear operation with the correction applied and repeat the measurement. Do not report the overloaded readings; a weak distance trend alone would not prove overload.
Check your understanding 3
Why is operator-body coupling a real error source?
Show answer
Your body perturbs fields and can couple energy into probes/cables, changing the measurand. This causes different techs to get different numbers at “the same spot” unless geometry is controlled.
Check your understanding 4
What should you do when broadband and selective disagree beyond expected uncertainty?
Show answer
Check supported range and overload indications, then ensure comparable time behavior, coverage and chain settings. Choose a controlled check that distinguishes the plausible causes; do not average different measurands to force agreement.
Check your understanding 5
Linear vs dB combining. Why is it wrong to average two power density values by averaging their dB values?
Show answer
dB is logarithmic, so the arithmetic mean of dB values represents a geometric mean of the linear powers. The required arithmetic power average is different. Convert to W/m², calculate the required average with its time or spatial weights, then convert back only if needed.
What to keep
Treat a symptom as evidence for a hypothesis, not a diagnosis. Compare the same quantity and reference plane with controlled geometry, state and settings. Reject overloaded data, test coverage and capture, and correct the actual model or chain before drawing an assessment conclusion.
Spotted an error? Report a correction
Syllabus and review details
No official syllabus alignment is listed for this lesson.
Last reviewed: