RF engineering FAQ

At 290 K, estimate receiver sensitivity from -174 dBm/Hz + 10 log10(B in hertz) + cascaded noise figure + the required SNR, with any implementation or measurement margin stated separately. For coded or spread waveforms, use the correct Eb/N0 or Es/N0 relationship and processing gain rather than substituting a generic SNR. Define the receiver input plane, bandwidth, waveform, error-rate target, gain state and temperature.

For a 20 MHz receiver, thermal noise is approximately -101.0 dBm. Adding 5 dB noise figure, 12 dB required SNR and 2 dB implementation margin gives about -82.0 dBm. Doubling the noise bandwidth to 40 MHz raises integrated thermal noise by 3 dB, so sensitivity worsens by about 3 dB if noise figure and required SNR stay unchanged.

Do not use this arithmetic alone as an installed sensitivity guarantee. Loss ahead of the first LNA, mismatch, filter ripple, temperature, blocker-induced compression, LO phase noise, ADC range and gain-control behavior can move the result. Reject a sensitivity claim that omits bandwidth or error criterion, or that was measured without the blockers and duplex leakage present in operation. Supply the wanted waveform, receiver bandwidth, input losses, NF by gain state, required BER/BLER/PER or EVM, blockers, maximum input, recovery and test uncertainty.

FDD isolation is a simultaneous leakage budget; TDD isolation is a state-and-time budget. In FDD, duplexer rejection, transmitter noise, antenna isolation and receiver selectivity must protect reception continuously while the transmitter is active. In TDD, the switch or circulator, PA turn-off, leakage during the opposite state, timing guard, transient blanking and receiver recovery must all be acceptable before the receive slot is used.

Derive the minimum net isolation from transmitter power and the maximum leakage the receiver can tolerate at each relevant frequency. If +40 dBm is present in the transmit path and the receiver can tolerate no more than -30 dBm at that leakage frequency, the arithmetic minimum is 70 dB before design, temperature, production and measurement margin. Allocate it across duplexer or switch, antenna separation, filtering, shielding and routing; do not assign the entire value to one catalog component.

Reject an isolation claim measured only at center frequency, low power or a static switch state. Verify band edges, transmitter noise outside the wanted band, harmonics, multiple carriers, temperature, mismatch, switching trajectories, control skew, PA decay, LNA protection, AGC recovery and the installed antenna configuration. The RFQ should state duplex mode, bands, guard interval, simultaneous carriers, transmit power, receiver blocker and damage limits, permitted desensitization, switching time and the conducted or radiated reference planes.

Use phase-noise spot or mask limits when particular carrier offsets control reciprocal mixing, Doppler, EVM or close-channel behavior; use integrated jitter when a defined offset band must close a time-domain sampling or timing budget. Neither is complete without carrier frequency, operating mode, offset or integration range, and treatment of deterministic spurs.

Reject a comparison that quotes -dBc/Hz at one favorable offset or an RMS jitter number without lower and upper integration limits. For multiplication by N, reference phase fluctuations rise by roughly 20 log10(N) in the region where the reference dominates. For jitter, integrate the linear phase-noise density over the project band, include or separately budget discrete spurs, then convert RMS phase deviation to time using carrier frequency.

Required inputs are carrier and output state, the system bandwidth or coherent observation time, critical offset regions, modulation or sampling frequency, allowed error, and source architecture. Verify the full mask, the instrument residual floor and the exact integration settings; one number cannot be reused across RF carriers or different integration bands without recalculation.

Initial accuracy, temperature stability, aging and holdover must be separate rows in the frequency-error budget because they describe different mechanisms and time intervals. Initial accuracy applies after a stated warm-up and calibration condition. Temperature stability applies over a defined profile. Aging needs an elapsed interval. Holdover begins when a reference is lost and depends on prior lock history and environment.

Reject a single ppm figure that does not say whether terms are maximum, typical, additive or statistically combined. Also separate supply and load sensitivity, vibration, retrace after power interruption and warm-up. For precision timing, state the observation interval and statistic used; Allan deviation at one averaging time cannot replace a maximum frequency-error limit over an unrelated interval.

The project must provide carrier or clock frequency, allowable accumulated phase or time error, temperature and vibration profile, calibration interval, reference-loss duration and recovery behavior. Acceptance evidence should retain the environmental profile, measurement interval, reference history, frequency record and uncertainty.

PLL settling time is the interval from a named event until the RF output enters and stays within the project’s frequency, phase, amplitude and spectral error limits. Lock detect usually reflects an internal phase/frequency-detector or counter condition and can assert while residual phase, level or spurs remain outside the usable window.

Define the start event, old and new frequencies, power state, reference mode, error bands, observation bandwidth, dwell time and deadline. Test the largest and smallest hops, band crossings, output-level changes, reference reacquisition and temperature extremes. If the system mutes the source, include mute release and transient leakage.

Reject a data sheet that gives only a nominal lock time without step, tolerance or measurement method. The acceptance capture should trigger from the real control command and show RF frequency or phase error and output level through the deadline; for burst, radar or hopping systems, also confirm spurs and modulation quality in the first usable interval.

No. A shared frequency reference can align long-term frequency, but it does not by itself guarantee deterministic or repeatable relative RF phase. Divider state, PLL acquisition, trigger latency, cable electrical length, distribution delay and channel calibration can all change phase while every source remains frequency locked.

Define the required behavior: bounded drift during one dwell, deterministic phase after a common trigger, repeatability after retune, coherence after relock, or recovery after power cycle. State carrier, channel spacing, reference and trigger distribution, cable states, temperature, observation time and allowable phase error.

Reject a coherence claim demonstrated only once after manual phase alignment. Acceptance must repeat the sequence across channels, tuning states and restarts, record relative phase versus time, and show how calibration and delay compensation are applied. If only common frequency is needed, say so and avoid paying for a stronger phase-repeatability requirement.

Insertion loss is the reduction in transmitted power through the intended path, while return loss describes how much incident power is reflected at a port. They are related through the network but are not interchangeable and must be specified at the same reference planes, impedance and termination state.

A receiver filter with low return loss may create standing waves and gain ripple even if its forward loss appears acceptable. A well-matched filter can still have excessive dissipative loss, reducing delivered transmitter power or adding directly to receiver noise figure when placed ahead of the LNA. Reject averages that hide a passband edge peak.

Provide passband edges, maximum S21 loss and ripple, minimum S11/S22 return loss, connector planes, temperature, power and unused-port terminations. Verify with calibrated S-parameters and preserve both measured and de-embedded data when fixtures are removed.

An RF filter needs a group-delay or phase-linearity limit when frequency-dependent phase inside the occupied band can spread pulses, degrade EVM, bias ranging or misalign channels. Constant delay can often be calibrated; delay variation across the signal band is the usual distortion term.

Define the exact occupied bandwidth, maximum peak-to-peak delay ripple, calculation aperture, temperature and power state. Do not apply the requirement blindly to the entire amplitude passband: steep skirts naturally increase phase curvature, and the usable phase-linear region may need extra guard band.

The project must supply waveform bandwidth, pulse width or symbol rate, allowed distortion and equalization capability. Verify unwrapped phase and group delay on a calibrated VNA with adequate frequency resolution; compare channels if differential timing matters.

Out-of-band rejection is attenuation through a stated path outside its passband. Duplexer isolation is leakage from a named source port to a named destination port with the remaining ports terminated and operating as declared. A rejection trace cannot stand in for every transmit, receive and antenna port combination.

Build a leakage budget: maximum transmitter level minus minimum transmit-to-receive isolation gives the residual at the receiver reference plane before other coupling paths. Compare it with damage, compression, blocker and desensitization limits, then include switches, circulators, cables, board and enclosure coupling.

Provide all port names, bands, simultaneous carrier states, antenna mismatch, temperature, power and switch state. Reject an isolation value without a frequency range, direction, terminations or dynamic-range margin. Verify every required S-parameter and the assembled front-end leakage path.

RF filter power handling must be derived from dissipated heat and local voltage or current stress under the real waveform, mismatch, mounting and temperature. A headline connector power is not transferable when insertion loss, airflow, baseplate or duty cycle changes.

As a screen, a 100 W CW signal with 0.8 dB insertion loss delivers about 83.2 W; roughly 16.8 W is not delivered, with the non-reflected portion becoming heat inside the network. That heat can shift resonance and raise loss. Pulse peaks, arcing, dielectric limits, reverse power and PIM may set a lower boundary than average thermal load.

Provide CW, peak and average levels, pulse width, duty cycle, crest factor, VSWR magnitude and phase, simultaneous carriers, ambient/baseplate temperature, cooling and mounting. Acceptance should combine low-level S-parameters, powered thermal soak, post-soak mask verification and nonlinear or PIM tests where relevant.

Insertion loss reduces the wanted signal along the selected route, while isolation and crosstalk describe unwanted coupling from named unselected or simultaneous routes. They solve different budgets and must be stated over the same frequency range, reference planes, temperature and switch state.

Add maximum loss from every switch stage, cable and transition; pre-LNA loss raises receiver noise figure and transmit-path loss reduces delivered power. For leakage, subtract minimum isolation from the maximum aggressor level and compare the residual with damage, compression, blocker or measurement-floor limits. Parallel enclosure and connector coupling can bypass cascaded switch isolation.

Provide the route matrix, aggressor-victim pairs, simultaneous states, source powers, receiver limits and terminations. Verify the assembled network over frequency with enough dynamic range, including adjacent paths and the longest route.

An electromechanical or coaxial switch can be preferable when the project prioritizes low insertion loss, high isolation, broadband behavior or a physically open contact state and can accept slower switching, finite contact life, size and coil or actuator control. A solid-state or PIN-diode switch can be preferable when fast switching, high cycle count, compact integration or silent operation dominates.

These are conditional tendencies, not guarantees. Compare the actual continuous band, loss and isolation at temperature, CW and pulse power, hot-switch limit, linearity, off-port behavior, transition time, RF settling, control bias, leakage current, vibration and expected cycles. MEMS and waveguide technologies add other trade spaces.

Do not choose by technology name. Supply the route states, waveform, switching frequency, power during transition, lifetime target and mechanical boundary, then require evidence under those conditions.