Write the source requirement at the system boundary
An RF source is acceptable only when its carrier, amplitude, spectrum and timing meet the system limit at a named output plane under stated operating conditions. Record the connector, impedance, load match, cable state, output power, frequency, temperature, supply, reference mode and control state. Then specify the phase-noise mask, deterministic spurs, frequency error and settling behavior that matter to the receiver, transmitter, radar, converter or test path.
Freeze the source contract before comparing architectures
Start with use, not topology. A fixed reference clock, agile local oscillator, coherent radar source and substitution generator can share a carrier frequency while requiring completely different close-in noise, tuning speed, restart phase, output control and environmental behavior. State every required carrier range, forbidden range, step size, sweep or hopping sequence, simultaneous output state and modulation condition before deciding between XO, TCXO, OCXO, VCO, PLL or direct-synthesis paths.
Define output power as a range with accuracy, flatness, harmonics, compression margin and settling criteria. Include the real load: a poor match or reverse signal can change level, pulling and spur behavior. If an external reference is used, give its frequency, level, waveform, impedance, phase-noise profile, loss-of-reference behavior and whether the source must hold, mute or free-run.
| Requirement block | What must be stated | Reject the proposal when |
|---|---|---|
| Carrier and output | Frequency ranges, resolution, power, flatness, waveform, load and reference plane | Typical range is quoted without forbidden bands, accuracy or output conditions |
| Spectral purity | Phase-noise mask by offset, spurs, harmonics, subharmonics and broadband noise | One favorable phase-noise point is used as the complete purity specification |
| Frequency stability | Initial tolerance, temperature, supply, load, aging, warm-up, vibration and holdover | All effects are collapsed into one ppm value without interval or conditions |
| Switching | Start trigger, frequency/power step, error band, observation bandwidth and deadline | Lock detect is accepted without measuring residual frequency, phase, level and spurs |
| Coherence | Shared reference, trigger, deterministic phase, drift, relock and restart behavior | Common frequency is assumed to guarantee repeatable channel phase |
Specify phase noise as a system mask, not a brochure spot value
Single-sideband phase noise is meaningful only with carrier frequency, offset frequency, output level and operating mode. Close-in noise can affect coherent integration, Doppler discrimination and reciprocal mixing; farther-out noise can enter the demodulation or converter bandwidth and degrade EVM, SNR or adjacent-channel measurements. Build a mask over the offset regions that map to those mechanisms. Record whether the limit is typical, maximum or measured, and whether options such as low-noise or fast-switching modes change the profile.
Do not compare phase-noise plots unless the carrier, reference, output mode and instrument residual floor are compatible. Multiplication raises phase fluctuations by approximately 20 log10 of the multiplication ratio within the region where the multiplied reference dominates. A PLL then combines reference, detector/divider, loop-filter and VCO contributions differently across offset frequency; no single spot proves the complete loop is suitable.
Worked boundary: converting integrated phase noise to time jitter
Assume the integrated single-sideband phase noise over the project-defined band is -60 dBc. Under the small-angle approximation, RMS phase deviation is approximately sqrt(2 × 10-6) = 0.001414 rad. At a 100 MHz carrier, RMS time jitter is 0.001414 / (2π × 100 MHz), or about 2.25 ps. The number is valid only for the stated integration limits and included spurs; changing the lower or upper offset can change the result materially. A buyer should reject an integrated-jitter headline that omits those limits.
Separate deterministic spurs from random phase noise
Harmonics sit at integer multiples of the carrier. Nonharmonic spurs can come from references, PFD activity, fractional synthesis, DDS images, power rails, digital clocks, mixing products and leakage. Broadband noise and residual AM or FM are different again. Search each over a declared frequency span, resolution bandwidth, detector, averaging method and output state. A clean close-in plot does not prove the wideband spectrum is clean.
Define exclusion zones around the carrier separately from far-out spur limits. For a tunable source, test a matrix of carrier, output power, reference mode and temperature rather than one convenient frequency. If the source drives a mixer or converter, inspect spur combinations in the actual RF/LO/IF plan. If it clocks a data converter, include the aperture-jitter band and deterministic clock products that fold into the sampled band.
Keep accuracy, stability, aging and holdover as separate budgets
Initial frequency accuracy describes the error after the specified warm-up and calibration condition. Temperature stability describes variation over a stated temperature profile. Aging requires an interval and usually behaves differently over the first year and later years. Supply, load, vibration and retrace are separate sensitivities. Holdover describes behavior after a reference is lost and must name the elapsed time, prior lock history and environmental state.
Long-term stability cannot be reduced to one unexplained ppm number. For precision timing, state the observation interval and statistic, such as Allan deviation, and preserve the measurement setup and preprocessing assumptions. For ordinary RF equipment, a bounded maximum error over temperature and time may be the useful acceptance metric. Choose the metric that closes the system budget rather than the one that produces the smallest headline.
Measure settling after the event that actually disturbs the source
Frequency switching can disturb carrier frequency, phase, amplitude, spurs and modulation simultaneously. Define the start event: command receipt, trigger edge, reference change, power step, mute release or relay action. Define completion as entry into and continued residence within a frequency, phase and amplitude error band, with any spur or modulation requirement restored. The observation bandwidth and dwell time belong in the requirement.
A digital lock indicator usually reports an internal condition and may assert before the RF output has reached the required residual error. Test the largest and smallest frequency steps, band crossings, output-level changes, reference reacquisition, cold start and relevant temperatures. If a system opens a receive window or samples immediately after a hop, measure the source at that exact deadline rather than after a generous delay.
Define what phase coherence must survive
Sharing a 10 MHz reference aligns long-term frequency but does not by itself equalize cable delay, synthesizer divider state, trigger latency or phase after retune. Decide whether the project needs simultaneous frequency lock, bounded relative phase during one dwell, deterministic phase after each trigger, repeatable phase after frequency changes, or phase recovery after power cycle. These are different acceptance tests.
For multiple channels, record reference and trigger distribution, cable electrical length, channel delay adjustment, residual drift and calibration interval. Exercise retune, mute, relock and restart. A system that is coherent once after manual alignment may still fail an automated beamforming, MIMO, radar or coherent measurement sequence.
Verify source behavior without hiding the measurement floor
- Draw reference planes. Mark external reference, source output, distribution paths, load and measurement plane.
- Freeze operating states. Include carrier, power, mode, waveform, reference, supply, temperature and load.
- Measure the phase-noise mask. Show instrument residual floor, cross-correlation or reference method and raw traces.
- Search deterministic products. Cover harmonics, nonharmonic spurs, subharmonics and wideband noise across the tuning matrix.
- Run stability profiles. Record warm-up, temperature, aging interval, reference loss and recovery.
- Capture switching transients. Trigger from the real command and measure frequency, phase, amplitude and spectrum at the required deadline.
- Test coherent channels. Include cable states, trigger alignment, retune, relock and restart.
- Retain acceptance evidence. Store settings, calibration state, uncertainty, raw data, limits and software revision.
Frequent causes of a source specification failure
- Quoting phase noise without carrier frequency, offset, output state or maximum/typical status.
- Converting phase noise to jitter without stating integration limits or spur treatment.
- Using one ppm value for initial accuracy, temperature, aging and holdover.
- Calling the output settled when an internal lock bit changes state.
- Assuming a common reference guarantees deterministic phase after retune.
- Testing one carrier while the actual source crosses bands or changes divider architecture.
- Ignoring load mismatch, reverse power, cable phase and reference-distribution noise.
- Reporting a measurement at or below the analyzer residual floor without margin or correlation.
Minimum information for an RF signal-source RFQ
- Carrier ranges, forbidden bands, resolution, tuning sequence and simultaneous channels
- Output power range, accuracy, flatness, waveform, impedance, VSWR and reverse-power condition
- Phase-noise limits by carrier and offset region, plus integration limits where jitter matters
- Harmonic, nonharmonic spur, subharmonic, broadband-noise and residual AM/FM limits
- Initial accuracy, warm-up, temperature, supply, load, aging, vibration and holdover conditions
- Switch trigger, step size, error bands, observation bandwidth, deadline and dwell
- External reference and trigger levels, loss behavior, distribution and phase-coherence requirement
- Supply, control interface, mute, fault, alarm, shielding, grounding and environmental limits
- Measurement plane, fixtures, calibration, uncertainty, raw-data format and acceptance matrix
