RF engineering FAQ

An RF power divider creates two or more controlled output paths from a common input; a directional coupler extracts a defined fraction of a travelling wave while preserving a main through path. Equal dividers are selected by port count, theoretical split, excess loss, amplitude/phase balance, output isolation, return loss and power. Couplers are selected by coupling factor, coupling flatness, directivity, main-line loss, return loss, orientation and power.

Use a divider when multiple branches need known relative amplitude and phase, such as parallel receiver paths, local-oscillator distribution or coherent test channels. Use a directional or bidirectional coupler when the system must monitor forward or reflected power, level a source, detect mismatch or feed a measurement channel without taking half of the main-line power.

Do not substitute one only because the connector count looks similar. A coupler normally provides unequal outputs and a defined isolated port; a divider normally provides comparable branch outputs. The RFQ should state signal direction, every port function, system impedance, full band, required ratio, loss definition, isolation or directivity, match, average/peak power, mismatch condition and the termination used on unused ports.

For an equal N-way RF divider, split loss is the unavoidable 10 log10(N) dB reduction caused by sharing the input power among N outputs; excess insertion loss is the additional loss of the real device beyond that ideal value. A 2-way ideal divider has 3.01 dB split loss, a 4-way has 6.02 dB and an 8-way has 9.03 dB.

If a 2-way divider measures 3.55 dB from common input to one output, the excess loss on that path is approximately 3.55 - 3.01 = 0.54 dB. The 3.55 dB value is the total path loss. A data sheet that lists 0.54 dB may be reporting excess loss only, so the column definition must be confirmed before products are compared or a link budget is calculated.

Excess loss does not replace the other acceptance checks. Measure both output paths under the same terminations and reference planes, then verify amplitude imbalance, phase imbalance, return loss and output isolation. At high power, repeat the check under representative average, peak, duty-cycle and mismatch conditions because heating or termination stress can change the practical result.

Often yes, but only when the divider is reciprocal, the manufacturer permits reverse use, and the input signals meet the required amplitude, phase and power conditions. A 0-degree divider used in reverse produces the vector sum at the common port; it does not simply add two headline power ratings.

Equal coherent inputs that are in phase combine efficiently. Phase or amplitude error reduces wanted output and can send power into the isolation resistor or another source path. Uncorrelated signals, a missing source, a reflected load or an unlocked phase relationship therefore create different internal dissipation and fault stress.

For an RFQ, state frequency, signal coherence, relative phase and amplitude tolerance, average and peak power per input, output VSWR, fault duration and allowable source-to-source coupling. Confirm both the matched power rating and the isolation-resistor or internal-load dissipation rating before acceptance testing.

An image response is an undesired input frequency that reaches the same desired IF through another valid sum or difference relationship with the LO. For low-side downconversion with IF = RFwanted - LO, the corresponding image commonly lies at RFimage = LO - IF. For high-side injection, the image moves to the opposite side of the LO and the wanted spectrum may reverse frequency order.

Calculate the complete image band at every RF, IF and LO edge. Then derive required rejection from the maximum credible image or blocker level, allowed converter output interference and downstream gain. Allocate that rejection between the RF preselector, mixer or image-reject/IQ architecture, calibration and any post-conversion filtering. A converter data-sheet image number is not sufficient unless its frequency, gain state, temperature, input level and output reference plane match the project.

Reject the plan when the image overlaps the wanted preselector passband, when quadrature mismatch cannot maintain rejection across production and temperature, or when the required filter transition is physically unrealistic. The RFQ should include wanted and image bands, LO states, blocker levels, preselector response, gain/phase calibration method, residual-image limit and the production verification sweep.

Choose by the required phase relationship, bandwidth, loss and isolation rather than by port count alone. A Wilkinson network is the usual starting point for equal 0-degree split or combine paths with useful output isolation and low excess loss over its designed band.

A resistive divider can cover very wide bandwidth, including low frequency or DC in some implementations, but adds unavoidable resistive loss and usually offers weaker branch isolation. A 90-degree hybrid creates quadrature paths for balanced amplifiers, image-reject architectures or phased networks; a 180-degree hybrid creates sum/difference or differential paths.

Compare every candidate at the same frequency, impedance, reference plane and termination state. The acceptance set should include total and excess loss, amplitude and phase balance, isolation, return loss, average and peak power, temperature, load mismatch and the rating of any isolation termination.

LO drive level establishes the mixer switching condition, so too little or too much drive can change conversion gain or loss, P1dB, IP3, noise, offsets and leakage. Specify minimum, nominal and maximum LO power at the converter LO connector after the complete distribution path. Include source tolerance, splitter and switch loss, cable loss, temperature drift and mismatch.

LO leakage is LO energy appearing at the RF input, IF/output or other protected interface. At an antenna it can become an emission; at an ADC or amplifier it can consume headroom, mix again or create DC and baseband errors. Measure RF-to-LO, LO-to-IF and RF-to-IF isolation separately because they protect different paths. Add filtering, shielding, balanced topology or cancellation only after the required leakage limit and reference plane are known.

Verify conversion response, compression and two-tone intermodulation at the LO drive extremes, not only at nominal drive. Also check phase noise and reciprocal mixing with the real blocker plan. Reject the converter when acceptable linearity depends on an LO tolerance the distribution network cannot maintain, or when leakage remains above the antenna, ADC, amplifier or channel-allocation limit.

Coupling factor sets the sampled level, directivity limits separation of forward and reverse waves, and return loss determines how strongly the coupler and its interfaces disturb or re-reflect the signal. Detector accuracy alone cannot define the accuracy of a forward/reflected-power measurement.

Finite directivity leaks part of the much larger forward wave into the reverse channel. That leakage adds vectorially to the true reflected wave, so the error changes with frequency, cable length and mismatch phase. Coupling must also place both samples inside the detector's calibrated dynamic range without compression or noise-floor loss.

Calibrate to the named DUT reference plane, include adapters and cables consistently, terminate unused ports, and verify coupling flatness, directivity, main-line loss and port match over the full band. State the required return-loss or VSWR accuracy and reject any setup whose directivity floor is too close to that target.

Amplitude and phase errors prevent array channels from adding with their intended complex weights, which changes boresight or scan direction, coherent gain, sidelobe level, beam shape and null depth. The impact depends on where errors occur across the aperture, whether they are correlated, the commanded beam, element pattern, frequency and calibration strategy. A single typical channel value does not establish array performance.

Build an error budget for manufacturing spread, temperature gradient, frequency slope, control quantization, LO/clock skew, mutual coupling, power-dependent drift and calibration residual. Use the same RMS, worst-case or statistical assumptions in hardware requirements and the pattern simulation. Specify gain/phase range, resolution, monotonicity, settling and repeatability at the calibrated reference plane.

Reject the requirement if it gives only room-temperature center-frequency accuracy, if the control range is consumed by nominal correction, or if the calibration system cannot observe a failed or drifting channel. Production evidence should retain per-channel raw data before and after calibration plus selected array patterns at band edges, scan extremes, temperature corners and representative fault states.

Receiver survival is only the first boundary; the protected receive path must also recover to specified gain, noise, phase and linearity before the earliest required echo is processed. State transmit peak and average leakage at the protected reference plane, pulse width, PRF, rise/fall behavior, frequency, source/load mismatch, switch or circulator isolation, limiter threshold, limiter leakage and the maximum non-damaging input of every exposed stage.

Define recovery from the end of the actual transmit event to a measurable receive criterion: gain within tolerance, noise floor restored, DC or baseband offset settled, phase coherent if required and no residual compression. A component may survive while limiter memory, LNA overload, AGC or bias recovery masks close-range targets.

Verify with the real waveform or a justified equivalent over frequency, temperature, supply and component tolerance. Include repeated bursts and worst-case VSWR. Reject a design when residual leakage or recovery time is specified only typically, when the test pulse has less energy than the real waveform, or when protection behavior cannot be correlated to the nearest usable range gate.

An RF measurement reference plane is the electrical boundary at which the corrected quantity is defined. A VNA calibration may establish correction at cable ends, while the project needs S-parameters at connectors, probe tips, package pads or an internal fixture boundary. State both the calibration plane and the reporting plane, then list every adapter, cable, bias tee, probe and fixture section between them.

De-embedding is required when the result must exclude a stable, linear and characterized network between those planes. Use measured or defensible fixture parameters with the correct orientation and port mapping. Retain the fixture file, method, frequency range, connector torque, temperature and verification result. If the project accepts the complete evaluation assembly, reporting the assembly result may be more honest than removing an uncertain fixture.

Do not de-embed a nonlinear, time-varying, poorly repeatable or inadequately characterized path. Reject a corrected result when it creates nonphysical gain, passivity or causality behavior without explanation, becomes extremely sensitive near fixture nulls, or cannot reproduce a check standard. Fixture removal also carries uncertainty; it is not a free accuracy improvement.

P1dB and two-tone IMD are separate nonlinear tests. P1dB uses one tone and a power sweep. It compares measured gain with a stated small-signal reference and reports the input- or output-referred level where gain has fallen by 1 dB. The plan must define frequency, sweep direction, dwell, bias, thermal state, load, reference planes and how the linear gain fit is established.

A two-tone test applies two frequencies with stated spacing and per-tone power, then measures fundamentals and intermodulation products, commonly the third-order products near the tones. State whether power is per tone or combined, provide source isolation, and show that source IMD and receiver distortion are below the DUT products. A residual-system measurement with a through path or suitable attenuation is part of the evidence.

Do not use an extrapolated IP3 as a substitute for compression, or either test as a substitute for modulated EVM, adjacent-channel power or memory-effect testing. Compare devices only when reference planes, tone spacing, power convention, bandwidth, averaging, bias and temperature match. If the real application uses pulsed or modulated signals, add that waveform rather than assuming the two-tone result predicts it.

A wireless link budget is the directional sum of every gain and loss from a named transmitter reference plane to a named receiver input; link margin is expected received power minus the receiver level required for the stated performance. Include conducted output, transmit feeder and connector loss, transmit antenna gain, path loss, polarization and pointing loss, receive antenna gain, receive feeder loss and the required receiver input. Keep free-space loss, rain, clear-air multipath, diffraction, obstruction, interference and equipment tolerance as separate rows.

For example, +30 dBm conducted power, 1.5 dB transmit loss, 34 dBi transmit gain, 129.6 dB free-space loss, 34 dBi receive gain and 1.5 dB receive loss produce about -34.6 dBm at the receiver input. Against a -70.5 dBm requirement, the arithmetic margin is 35.9 dB. Do not call all 35.9 dB fade margin until the project has applied the approved rain, multipath, clearance, pointing, polarization, interference and tolerance model.

Reject a budget that mixes typical and worst-case values, combines directions, hides reference planes or treats a regulatory EIRP ceiling as guaranteed radiated power. Required inputs are frequency, distance, path profile, antenna heights and patterns, feeder construction, polarization, climate, availability target, waveform, bandwidth, receiver criterion and coexistence environment. The final evidence should include the calculation revision, source of each value, installation measurements and the method used to predict outage.