RF Calculators

Γ, VSWR, return loss, mismatch loss — and the Smith chart — are expressions of the same reflection — use whichever the instrument or datasheet hands you; the calculator converts between all of them. The anchors worth keeping in your head: return loss 10 dB (VSWR ≈ 1.9) is exactly 10% of power reflected, 90% delivered — the customary "acceptable antenna" line — while 6 dB (VSWR 3:1) already reflects 25%, enough to stress PAs and corrupt data. VSWR survives from slotted-line days; a VNA's S11 is return loss directly. On the Smith chart at right, distance from center is |Γ| — every point on the dashed circle has the same VSWR — and the position of the point encodes the complex load.

In antenna test, mismatch counts twice: once at the DUT, once in the measurement path. A poorly matched adapter or cable ripples the frequency response and lands in the pattern data as uncertainty. Drag the slider to see how quickly reflected power grows.

At 28 GHz, λ = 10.7 mm; at 77 GHz, 3.9 mm — and nearly every physical dimension in an antenna design is a fraction of it: patch elements ≈ λ/2 across, array pitch ≈ λ/2 for grating-lobe-free scanning, transformers and stubs λ/4. Hence the λ/2 and λ/4 outputs above.

Velocity factor is what shortens those dimensions inside a medium: solid-PTFE coax propagates at ≈ 66% of c, low-density PTFE at ≈ 85%, so a coax quarter-wave stub is physically shorter than its free-space length suggests. k = 2π/λ and ω = 2πf are provided because array factors, near-field transforms, and reactance calculations consume them directly.

The working outputs here are the resonant frequency and Q: L and C are chosen to put reactance cancellation exactly at the operating frequency — the bread and butter of matching networks and feed tuning — and Q sets the bandwidth you pay for it: high Q selects narrowly, low Q covers a band.

The practical mmWave warning: Q is loss-limited, and at these frequencies fractions of an ohm of trace resistance, via inductance, and bond-wire parasitics routinely dominate — datasheet component Q rarely survives the layout. Susceptance is included for parallel-topology analysis.

The measurement question NF answers: can you see the sidelobe? The floor is −174 dBm/Hz + NF + 10·log₁₀(B); resolving a −40 dBc sidelobe at 110 GHz stands or falls on that sum, and narrowing the IF bandwidth is often cheaper than buying a lower noise figure.

Noise temperature is the same quantity on a linear scale — Te = 290·(F−1), so 1 dB NF ≈ 75 K — preferred in satcom and radio astronomy where differences of tenths of a dB matter. The Friis cascade's practical content: the first stage dominates, and every dB of loss ahead of the LNA adds a full dB to system NF — which is why the LNA belongs at the antenna, and why mmWave frequency extenders sit directly at the range probe rather than behind a cable run.

FSPL is the term people mis-model: it is not absorption, just spherical spreading, and it grows 20 dB per decade in both distance and frequency — 67 dB at 28 GHz over 2 m, 76 dB at 77 GHz over the same span. That 9 dB step between bands is why mmWave systems lean so hard on antenna gain.

In a test chamber the link budget is the measurement dynamic range: generator power + horn gain − path − cables must leave the DUT signal comfortably above the receiver floor, or sidelobes and back-radiation disappear into noise. 40–60 dB above the floor is a sound design target. The calculator runs the full Friis form — every term in dB, so the budget is pure addition.

Atmospheric absorption is separate from FSPL and is computed above with the full ITU-R P.676-13 line-by-line model. Over the short paths of a test chamber it is usually negligible — a couple of hundredths of a dB — but it dominates long outdoor links, and near the 60 GHz oxygen complex it reaches roughly 15 dB/km at sea level. Use the button to fold it into Other Losses rather than guessing a figure.

EIRP = P_tx − L_cable + G_ant is the regulatory quantity — FCC and ETSI limits are written against it, so a measured antenna gain is directly a compliance number. The recurring trap is ERP vs EIRP: ERP references a half-wave dipole, ERP = EIRP − 2.15 dB, and standards differ on which they specify — check before you certify.

The patch estimator gives Balanis first-order dimensions — W = (λ₀/2)·√(2/(εr+1)), L just under a dielectric half-wave less 2ΔL of fringing — good enough to start layout, not to sign off; final dimensions come from full-wave simulation. Complete feed and inset-matching design lives on the companion Phased Array Calculators page.

Antenna gain enters the radar equation squared (transmit and receive), so R ∝ √G: 3 dB of verified gain is +41% range — and 2 dB lost to a fabrication issue is range you can catch in the chamber before it becomes range lost on the road.

This calculator gives the single-pulse range at SNR_min = 0 dB — about 25 m with the 77 GHz defaults. Production automotive radars reach 100–200 m on a 10 m² car because FMCW coherent integration adds 25–35 dB of processing gain on top; the antenna-gain scaling holds unchanged underneath. RCS anchors: truck ≈ 200 m², car 10–100 m², pedestrian 0.5–1 m², bicycle 1–2 m² — the binding requirement is the small targets, which is what drives both antenna gain and integration time.

At mmWave, waveguide is how signal reaches the DUT with acceptable loss: frequency extenders connect through WR-28 (26.5–40 GHz), WR-12 (60–90 GHz), or WR-10 (75–110 GHz). Band selection is a hard constraint — below the TE₁₀ cutoff nothing propagates; above the TE₂₀ cutoff higher-order modes corrupt the measurement — so each band's usable window sits between the two.

λg always exceeds the free-space wavelength and diverges toward cutoff, which is why calibration reference planes, quarter-wave chokes, and transitions must be computed in guide wavelength, not λ₀. Wave impedance sits above 377 Ω and varies with frequency — it is the match a VNA sees at horn and transition interfaces. Defaults are WR-28, the 5G NR FR2 workhorse; drag the slider to walk the bands.