Drop SWR as a Verdict—Keep the Measurement

Here is the provocative version: drop SWR as a universal pass/fail verdict. Keep it when it is the clearest field indication. In a laboratory specification, return loss often fits the dB language of filters, cables and amplifiers better. But replacing SWR with return loss does not create new one-port information. The engineering upgrade is to declare the mode, reference impedance, reference plane, frequency, power and uncertainty—and to retain complex phase and two-port data when the decision needs them.

The Physics Did Not Begin with a Vacuum Tube

A standing wave is produced when incident and reflected waves coexist on a transmission line. That mechanism is not tied to a particular transmitter technology, a narrow band or a purely resistive load. Vacuum-tube transmitters helped make practical SWR indicators familiar to generations of operators, but they did not define the boundary of the quantity.

Modern vector instruments can display the same calibrated reflection as complex Γ, impedance, a Smith chart, return loss or SWR. Keysight’s VNA reflection documentation states explicitly that these display forms are calculated from the same reflection measurement. The useful question is therefore not which era a label belongs to. It is which representation exposes enough information for the decision.

One Reflection Magnitude, Two Scales

Start with one propagating mode at one declared reference plane and a real, positive reference impedance Z_ref. For a load or one-port with complex reflection coefficient Γ and |Γ| < 1:

Return loss and SWR both discard ∠Γ. A VNA may plot 20 log₁₀|S₁₁| as a negative number; positive return loss uses the opposite sign. State the convention so that “−20 dB S₁₁” and “20 dB return loss” are not mistaken for different results.

Those last two columns use P_ref/P_inc = |Γ|², mismatch efficiency η_m = 1 − |Γ|² and ML = −10 log₁₀(1 − |Γ|²). They describe incident and accepted power at the same reference plane, for the declared mode and wave normalization. They do not prove where accepted power is dissipated or radiated, and they do not include later re-reflections from a mismatched source.

The Reference Impedance and Plane Are Part of the Result

A number without Z_ref is incomplete. Conventional 50 Ω S-parameters answer a different normalization question from 75 Ω data, a balanced 100 Ω differential port or a waveguide mode. For a real reference impedance, the complex impedance corresponding to Γ is:

That equation requires complex Γ, not return loss or SWR alone. Every point on a constant-|Γ| circle has the same SWR and return loss but a different resistance and reactance.

The reference plane matters just as much. On an ideal uniform lossless line, moving the plane rotates Γ while preserving |Γ|. The displayed impedance changes, while SWR and return loss remain constant. On a lossy line, a reflection measured farther from the load is attenuated on its return trip; return loss can look better and SWR can look lower even though the load did not improve.

That is why cable length is not a mysterious SWR cure. A plane shift, loss, connector or fixture may change what the instrument sees. Calibrate at the decision plane or de-embed only a characterized fixture. IEEE 370-2020 formalizes measurement-quality and fixture-removal practices for high-frequency interconnect data; the same discipline—defined planes, validated fixtures and consistency—also improves lower-frequency RF work.

Return Loss Does Not Contain the Missing Phase

Changing the vertical scale from SWR to dB does not recover ∠Γ. Phase is what locates the impedance around a constant-magnitude circle and what determines coherent addition after another reflection.

For a cascade, two modest discontinuities can reinforce or cancel depending on their separation, propagation constant and reflection phases. Scalar return loss and insertion-loss limits may bound a component, but they cannot reproduce the installed ripple. Preserve complex S-parameters and cascade them with a suitable network representation. The Touchstone 2.1 specification requires the reference resistance and supports complex, mixed-mode network data precisely because normalization, port order, magnitude and phase belong together.

Insertion Loss Answers a Different Question

Return loss and SWR are one-port reflection quantities. Insertion loss is a two-port comparison: how much the received power or wave magnitude changes when a device is inserted between stated source and load conditions. For a passive two-port measured in the usual matched real-reference system, the common VNA display is:

That convention is extremely useful, but it is not a universal synonym for heat loss. The measured S₂₁ includes the transmission behaviour of the device under the analyzer’s port normalization and termination conditions. Input/output mismatch, internal dissipation, leakage, radiation, mode conversion and coherent re-reflections can contribute differently depending on the network and setup.

Mismatch loss is narrower: it converts the unaccepted fraction at one declared plane into decibels. Transducer gain is broader: it compares power delivered to the actual load with power available from the actual source, including source and load reflection coefficients and their interaction with all four two-port S-parameters. In an installed mismatched chain, transducer gain—or a full signal-flow calculation—is usually closer to the system question than simply adding return loss and insertion loss as unrelated scalars.

Keysight’s RF power-measurement fundamentals uses signal-flow graphs to keep generator mismatch, load mismatch and re-reflection distinct. That boundary is essential whenever “power getting through” is the claim.

A Power Amplifier Sees Magnitude and Phase

An SWR or return-loss limit defines the radius of a load-reflection circle, not one load impedance. Rotating Γ around that circle changes resistance and reactance at the PA reference plane. The output network and device can then see very different voltage, current, dissipated power, efficiency, gain compression and stability conditions even though SWR and return loss are unchanged.

Keep the transmitter manufacturer’s stated SWR or foldback limit—it is a valid operational boundary under its declared power, supply, temperature, modulation and duty-cycle conditions. Do not promote that single scalar into a complete stress model. For design or qualification, sweep reflection magnitude and phase at rated operating conditions, or use calibrated load-pull and ruggedness testing. Keysight’s active load-pull guidance explicitly varies load conditions to assess performance degradation and breakdown stress.

Balanced Networks Need a Modal Definition

A balanced structure does not make SWR impossible. It makes an unqualified single SWR inadequate. A multi-conductor network supports defined differential and common modes, each with its own reference resistance and reflection coefficient.

• S_dd11 describes differential-to-differential reflection at the differential input.
• S_cc11 describes common-to-common reflection at the common-mode input.
• S_cd and S_dc terms describe conversion between differential and common modes.

If a modal reference resistance is real and positive and |Γ_mode| < 1, the same SWR and return-loss transforms can be applied to that modal reflection. But “differential SWR” derived from |S_dd11| does not include power converted into common mode. The conversion terms must be retained when balance and unwanted current are part of the requirement.

Rohde & Schwarz’s balanced-device measurement guide treats the differential, common and conversion parameters together. Touchstone 2.1 likewise defines how mixed-mode ordering and modal reference resistances are recorded. The engineering move is to name the mode—not to declare SWR undefined everywhere a balanced pair appears.

Measurement Uncertainty Sets the Useful Digits

A return-loss trace is not automatically more accurate because it is shown in decibels. Residual directivity, source match, reflection tracking, receiver noise, calibration-standard models, connector repeatability, cable movement, drift and reference-plane error all contribute. At high return loss, the desired reflected signal is small, so an error vector that was negligible at 10 dB can dominate the result at 40 dB.

Report the calibration method and plane, reference impedance, frequency grid, source power, IF bandwidth or averaging, connector/fixture state, verification result and uncertainty appropriate to the claim. Do not advertise a universal VNA return-loss floor. The NIST covariance-based VNA uncertainty work shows why correlations among complex S-parameter errors matter when uncertainty is propagated into a derived quantity.

De-embedding also has uncertainty. Moving a plane through a well-characterized line is not the same as mathematically erasing an unknown adapter, flexible cable or fixture. Verify the final plane with a check standard whenever the result will become a limit or a cascade model.

Why a Field SWR Display Still Earns Its Place

A directional wattmeter or transmitter display can turn forward and reverse indications into SWR quickly. That is useful for tuning, spotting a wet connector or broken antenna, watching band-to-band changes and staying inside a transmitter’s protection envelope. The Bird Model 43 manual, for example, provides SWR conversion graphs from directional forward and reflected readings.

Use that indication within its instrument boundary:

• correct frequency and power element or sensor range;
• adequate directional-coupler directivity and calibration;
• enough reflected-channel resolution for the claimed ratio;
• known measurement plane and stable cable configuration;
• awareness of harmonics, modulation and peak/average detector behaviour; and
• no inference that low SWR proves antenna efficiency, pattern or safe internal component voltage.

Forward minus reverse power can represent net power accepted by the downstream network when the directional quantities are calibrated consistently at the same plane. It is not automatically radiated power. Feedline, matching network, transformer, ground and antenna losses remain downstream.

Specify the Decision, Not a Favourite Display

For a component or broadband subsystem, a useful specification usually includes:

• frequency range, port mode, reference resistance and physical reference plane;
• return-loss or |S₁₁| limit, with the sign convention stated;
• complex S-parameter data when impedance, cascade ripple or de-embedding matters;
• insertion loss or gain with source/load conditions and direction declared;
• conversion terms for mixed-mode devices;
• power, bias, temperature and linear/small-signal or large-signal state;
• calibration, verification and measurement uncertainty; and
• separate PA ruggedness, radiation, thermal or safety evidence when those are the real requirements.

For installation and service, SWR may remain the fastest useful display. For a dB link budget, return loss may read more naturally. For cascades and impedance, retain complex Γ. For delivered power through mismatched stages, calculate transducer gain or use the complete signal flow. The metric follows the decision.

The Bottom Line

Drop SWR when it is being used as a universal quality score. Keep it when it communicates an operational limit efficiently. Return loss is not richer one-port magnitude data; it is the same |Γ| on a logarithmic scale. Insertion loss is a different, two-port quantity, and neither scalar retains the phase needed for impedance, cascades or PA stress.

A modern RF specification is not modern because it uses dB. It is modern because it declares the mode, impedance, plane, frequency, operating state and uncertainty—and preserves complex data wherever the physics requires it.

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