Where SFDR Shines—and Where It Does Not

I like Spurious-Free Dynamic Range when the question is precise: with this signal, at this level, under these settings and across this declared spectrum, how far below the reference is the largest unwanted discrete component? That is useful engineering. Trouble starts when the answer is promoted into a universal radio ranking.

A receiver can have very clean converter spurs and still be limited by reciprocal mixing, blocking or intermodulation. A transmitter can have a large single-tone carrier-to-spur ratio and still generate poor two-tone IMD or wideband spectral regrowth. SFDR has not failed in either case; it was asked a different question.

One Acronym Covers Several Test Families

For an analogue-to-digital converter, the standard single-tone idea is the ratio between the fundamental and the largest harmonic or other spur in a specified frequency range. Results may be stated relative to the applied carrier in dBc or relative to converter full scale in dBFS. Those references are not interchangeable when the carrier is below full scale.

Receiver designers also use instantaneous SFDR or SFDR3 for the range in which a weak signal is above the receiver noise while third-order products from stronger signals remain below that noise. That is a two-tone, noise-and-linearity construct derived from noise floor and intercept behaviour; it is not the same measurement as the largest line in a single-tone converter FFT.

On a transmitter spectrum, people sometimes use “SFDR” informally for the carrier-to-largest-spur separation. That can summarize a declared scan for harmonics, synthesizer products, clock leakage or other discrete emissions. It does not by itself prove compliance, because the applicable spurious and out-of-band domains, modulation, detector, reference bandwidth, limits and reference plane come from the governing rule or standard.

A Converter SFDR Number Needs Its Spectrum

The active IEEE 1241-2023 ADC test standard exists to make converter terminology and test methods reproducible. A useful converter SFDR record states at least:

• input frequency and amplitude, converter sample rate and clock source;
• analogue input path, gain, common-mode condition and converter mode;
• the observed frequency range or Nyquist zone;
• whether DC, harmonics, images or interleaving products were included or excluded;
• record length, FFT window, coherent-sampling treatment and averaging; and
• whether the result is referenced to the carrier in dBc or to full scale in dBFS.

The largest spur can change with input frequency, level, sampling rate and observation band. A narrowband result that excludes an out-of-band harmonic is valid for that stated band; it is not automatically a full-Nyquist result. Analog Devices’ wideband ADC SFDR guide gives practical examples of harmonic and interleaving products becoming the limiting line under different definitions.

Reference Bandwidth Is Part of the Evidence

A mathematically discrete spectral line does not gain power merely because the analysis bandwidth is widened, but its measured amplitude and detectability can still be affected by FFT leakage, window correction, analyser filter shape, detector choice, phase noise and the noise under the line. Noise-like results do change with measurement bandwidth. A reported spectrum therefore needs resolution bandwidth—or FFT bin width and window—plus video bandwidth, averaging and detector where those apply.

Regulatory reference bandwidth is a separate concept. ITU-R SM.329-13, for example, defines recommended reference bandwidths for spurious-domain emissions and measurement methods for narrowband and broadband emissions. An analyser screenshot set to a convenient RBW is not automatically a result in the legally applicable reference bandwidth.

The same discipline applies to receiver noise. A noise floor quoted in dBm needs the receiver bandwidth and settings. A density in dBm/Hz needs a documented normalization. Without that information, a spur-to-noise or dynamic-range comparison can move merely because the measurement window changed.

Single-Tone Spurs Are Not Intermodulation

A single spectrally pure tone is ideal for exposing harmonics, aliases, clock products, reference leakage and other discrete artifacts. It does not exercise the envelope-dependent nonlinearity that creates intermodulation between two or more signals.

A two-tone test applies two known signals and observes products such as 2f₁ − f₂ and 2f₂ − f₁. Those third-order products are troublesome because they can fall close to the wanted frequencies. Tone spacing, per-tone level, source purity, combining isolation, measurement bandwidth, device gain state and the chosen endpoint all belong with the number.

This is why “SFDR” cannot stand alone as a transceiver adjective. The mechanism that creates one fixed spur is not necessarily the mechanism that creates two-tone IMD, reciprocal-mixing noise or broadband overload.

DR3 Measures a Two-Signal Receiver Failure

Third-order intermodulation dynamic range, often shortened to DR3 or IMD DR, uses two equal off-channel signals. Their third-order products are positioned in the wanted channel and increased until they reach a defined response or noise-related endpoint. The reported dynamic range depends on tone spacing, receiver bandwidth, preamplifier or attenuator state, AGC and other signal-processing settings.

DR3 answers: how much separation exists between the receiver’s weak-signal floor and this particular two-tone third-order failure? It does not answer how the receiver behaves with one blocker, a field full of many signals, or a discrete internal birdie. An extrapolated intercept result also assumes a linear slope region; direct overload, AGC action or ADC clipping can invalidate that extrapolation.

RMDR Separates Phase-Noise Desense

Reciprocal Mixing Dynamic Range applies one strong signal at a declared offset from the tuned frequency. The blocker mixes with local-oscillator or sampling-clock phase-noise sidebands and raises noise inside the wanted channel. The public ARRL receiver procedure uses a 3 dB rise above the receiver noise floor as its endpoint and reports offsets such as 2, 5 and 20 kHz.

RMDR is therefore an offset-specific, bandwidth-specific strong-signal test. It is not simply another name for blocking, and it should not be compared across radios unless roofing filters, bandwidth, preamp state and the test endpoint are aligned.

Blocking Looks for Gain Compression

A blocking-gain-compression test keeps a weak wanted signal present while one off-frequency blocker is increased. The endpoint is a declared reduction in wanted-signal response, commonly 1 dB in the ARRL method. This probes compression or desensitization in the receiver chain.

Reciprocal-mixing noise can mask that reduction by making the output rise instead. That is why ARRL separates blocking gain compression from reciprocal mixing. A “noise-limited blocking” figure and a true gain-compression endpoint are not equivalent evidence.

NPR Loads Many Frequencies at Once

Noise Power Ratio is a different stress test again. Band-limited noise with a deep notch is applied to the receiver or converter. Added device noise, quantization, clipping, nonlinearity and many-signal interactions produce residual energy in the notch. NPR compares the noise reference outside the notch with the residual measured inside it.

The ARRL QEX NPR measurement article describes that reference-and-notch method. A defensible result states total applied noise power, loading or backoff, statistical properties and crest factor, occupied bandwidth, notch width and location, receiver configuration, analysis bandwidth and averaging. NPR is especially useful for broadband and many-signal loading, but it does not identify which individual discrete spur would dominate a single-tone SFDR scan.

A Transmitter Needs Both Spur and Linearity Tests

For a transmitter, a single unmodulated carrier is good at revealing harmonics and discrete clock, local-oscillator or reference products. It is a poor stimulus for judging SSB envelope linearity. A two-tone SSB test deliberately drives the chain with a changing envelope so odd-order intermodulation products appear beside the two fundamentals.

A useful two-tone report states the tone frequencies and spacing, equal-tone drive, transmitter output or PEP, operating mode, ALC and processing settings, duty and thermal state, analyser reference plane, RBW and the IMD order being reported. It must also say whether dBc is referenced to either test tone or to PEP; for two equal tones those reference levels are 6 dB apart.

For a digitally modulated waveform, occupied bandwidth, spectral-emission mask, adjacent-channel power and in-band error metrics may be more informative than a two-tone number. ETSI EN 301 783 and the applicable national rules illustrate why modulation, necessary bandwidth, RBW and equipment configuration belong to an emissions result.

Predistortion Needs a Controlled Before-and-After Test

Digital predistortion tries to pre-correct a power amplifier’s nonlinear response. If it works, intermodulation or spectral regrowth can improve even while a fixed synthesizer spur remains unchanged and still limits a single-tone SFDR number. The reverse is also possible: a clean discrete-spur scan says nothing about whether predistortion remains stable across waveform, temperature or load.

Judge predistortion with the stimulus and feedback conditions the implementation supports. Compare enabled and disabled cases at the same final output power, waveform, drive, load, sample point, bandwidth and thermal state. Record both out-of-band measures such as IMD or adjacent-channel leakage and an in-band fidelity measure where the waveform permits it. The Analog Devices DPD engineering guide makes the same point: one static spectrum does not capture the operating-condition matrix or stability of the correction.

The Final Reference Plane Belongs to the Station

Once an external amplifier and output filter are added, the radiated-signal chain includes the exciter, interface and drive level, amplifier, bias and supply, ALC or other control loops, predistortion feedback path, filter, connectors and load. A transceiver-only result cannot certify that complete chain.

Measure conducted emissions at a declared final connector through suitable high-power sampling, attenuation and filtering that protect the instrument and do not create their own products. Radiated compliance and station interference also depend on the antenna system, coupling paths and local law. Do not substitute a bench carrier-to-spur ratio for the method and limit that your jurisdiction requires.

A Result Worth Comparing Comes With a Test Record

My conclusion is deliberately narrow: SFDR shines wherever the largest unwanted discrete component is the thing you need to know. Keep it there. For receiver crowding, many-signal loading or transmitter envelope cleanliness, choose the test that recreates that failure and report enough conditions for somebody else to repeat it.

Primary Technical References

• IEEE 1241-2023: ADC Terminology and Test Methods
• Analog Devices: Understanding SFDR in Wideband GSPS ADCs
• Analog Devices MT-012: Intermodulation Distortion Considerations for ADCs
• ARRL Laboratory Test Procedures Manual
• ARRL QEX: Noise Power Ratio Receiver Testing
• ITU-R SM.329-13: Unwanted Emissions in the Spurious Domain
• ETSI EN 301 783: Commercial Amateur-Radio Equipment

Questions, antenna-factor records or height trials to share? Contact RF.Guru.