Impedance matching is the practice of making a load, source or line present the impedance that its neighbour needs, so that the desired power crosses the interface instead of reflecting from it. It matters in RF design because a mismatch costs delivered power, adds gain ripple across a band, stresses power amplifiers and degrades receiver noise. A lossless matching network of inductors, capacitors or transmission lines removes the mismatch at the design frequency and over a limited bandwidth around it.
The idea is simple, but it hides several different goals. A designer may want the most power into a load, no reflection on a line, the lowest noise figure from an LNA or the best load line for a power amplifier, and these do not all call for the same impedance. This article sets out what is actually being matched, how much a mismatch costs, and how a practical network is designed, using a complete 868 MHz example.
Why Impedance Matters at Radio Frequency
Impedance Z = R + jX combines resistance with reactance, and the reactance changes with frequency. A capacitor has X = -1/(2Ï€fC) and an inductor has X = 2Ï€fL, so a component that is unimportant at audio frequencies can dominate at 2 GHz. Real parts also carry parasitic inductance and capacitance that shift their behaviour again.
At low frequency a connecting wire is short compared with the wavelength, and voltage is nearly the same along its length. At RF the same wire may be a sizeable fraction of a wavelength, so signals behave as travelling waves. Once that is true the interconnect has a characteristic impedance Z0, and any load that differs from Z0 sends part of the incident wave back. See the relationship between frequency and wavelength for the point at which this starts to matter.
Most RF systems use a 50 Ω reference. The value is a historical compromise for air-spaced coaxial cable, sitting between the impedance for lowest loss (near 77 Ω) and the one for highest power handling (near 30 Ω). Connectors, cables, instruments and catalogue components all share it, which is why so much matching work means getting to and from 50 Ω.
Three Meanings of a Matched Load
Conjugate Matching for Maximum Power Transfer
A source with impedance ZS delivers its maximum available power when the load is the complex conjugate of the source, ZL = ZS*. The reactances cancel and the resistances are equal. For resistive parts, the delivered fraction of available power and the efficiency are:
PL / Pavail = 4r / (1 + r)2 and η = r / (1 + r), where r = RL / RS
Figure 1 shows the point that is often misunderstood. Conjugate matching maximises power into the load for a fixed source, but half of the power drawn is dissipated in the source resistance. It does not maximise efficiency, which keeps rising as the load resistance increases.
Reflectionless Matching to a Line
On a transmission line the aim is usually different: make ZL = Z0 so that the reflection coefficient is zero and no standing wave exists. When the source and the line are both 50 Ω and real, this condition and the conjugate condition give the same answer, which is why the two are often confused. They separate when the source impedance is complex or differs from Z0.
Matching for Noise or Linearity
Active devices add two further targets. An LNA has an optimum source reflection coefficient Γopt that gives its minimum noise figure, and this is generally not the conjugate of its input. A power amplifier has an optimum load found by load-pull, which sets output power and efficiency and is again not the conjugate of the output impedance. In both cases a deliberate mismatch is the correct design. Impedance matching in RF filters and amplifiers covers these trade-offs in detail.
Reflection, VSWR and Mismatch Loss
The reflection coefficient describes a load relative to Z0, and the other familiar figures all follow from it:
- Reflection coefficient: Γ = (ZL – Z0) / (ZL + Z0)
- Voltage standing wave ratio: VSWR = (1 + |Γ|) / (1 – |Γ|)
- Return loss: RL = -20 log10|Γ| dB
- Mismatch loss: ML = -10 log10(1 – |Γ|2) dB
Return loss says how far down the reflected wave is. Mismatch loss says how much delivered power falls. They are easy to mix up because a poor-looking return loss can correspond to a modest mismatch loss: 10 dB return loss, equal to a VSWR of about 1.9, costs only about 0.46 dB. The VSWR, return loss and mismatch loss page gives the full conversion and a calculator.
The lost power is not always the main cost. Reflected power returns to the source, where it can change a power amplifier’s load line and raise device voltage or dissipation. In a cascade, two mismatches separated by a line form a weak resonator and put ripple on the gain and group delay. A filter designed for 50 Ω terminations also changes shape when terminated in anything else.
Worked Example: Matching 15 – j20 Ω to 50 Ω at 868 MHz
Consider a small antenna or device port that measures ZL = 15 – j20 Ω at 868 MHz, driven from a 50 Ω source. The first step is to quantify the problem, and the second is to design an L-network.
Size of the Mismatch
Γ = (15 – j20 – 50) / (15 – j20 + 50) = (-35 – j20) / (65 – j20), which gives |Γ| = 0.593 at an angle of -133.2°. From that value:
- VSWR = 3.91
- Return loss = 4.54 dB
- Power delivered = 1 – 0.5932 = 64.9% of the available power
- Mismatch loss = 1.88 dB
L-Network Design
Because the load resistance of 15 Ω is below 50 Ω, the series element goes next to the load and the shunt element goes on the 50 Ω side. The Q of the transformation is set only by the resistance ratio:
Q = √(50/15 – 1) = 1.528
The series branch must have a total reactance of +Q × 15 = +22.9 Ω. The load already contributes -20 Ω, so the added series element must supply Xs = 22.9 + 20 = +42.9 Ω. That is an inductor of L = 42.9 / (2π × 868 MHz) = 7.87 nH.
The shunt element has reactance magnitude Xp = 50 / Q = 32.7 Ω. Using a capacitor, C = 1 / (2π × 868 MHz × 32.7 Ω) = 5.60 pF.
As a check, the series branch is 15 + j22.9 Ω, whose admittance is 0.0200 – j0.0306 S. The capacitor adds +j0.0306 S, leaving 0.0200 S, which is exactly 50 Ω. Figure 2 shows the same two moves on a Smith chart: the inductor travels along a constant-resistance circle and the capacitor along a constant-conductance circle, finishing at the centre.
Bandwidth and Real Components
The match is exact at one frequency only. Modelling the load as 15 Ω in series with 9.17 pF, the 10 dB return loss bandwidth of the network runs from about 746 MHz to 978 MHz, as Figure 3 shows. Low-Q transformations like this one are comparatively wideband; a larger impedance ratio raises Q and narrows the band.
With standard values of 8.2 nH and 5.6 pF, the return loss at 868 MHz is 24.4 dB and the deepest point moves down to about 849 MHz. That is still a good match, and it shows why tolerance matters less than it seems for a low-Q network. Losses are a separate matter. An inductor Q of 40 at 868 MHz adds about 1.07 Ω of series resistance, which in series with the 15 Ω load dissipates about 0.30 dB. Choose high-Q components whenever the load resistance is small.
Choosing a Matching Network
The right topology depends on the frequency, the impedance ratio and the bandwidth required. These are the main options:
- L-networks: two reactive elements. Simple and low loss, but Q is fixed by the impedance ratio.
- Pi and T networks: three elements, which allow Q to be chosen independently. Useful for setting bandwidth or harmonic rejection.
- Quarter-wave transformers: a line of impedance √(Z0ZL) a quarter wavelength long matches real impedances without discrete parts.
- Single and double stubs: open or shorted line sections provide the reactance at microwave frequencies, where lumped parts stop behaving well.
- Transformers and baluns: wideband ratio changes, often with a change from balanced to unbalanced as well.
- Broadband, multi-section networks: several sections trade size for bandwidth, up to the Bode-Fano limit for a given load.
As a rough guide, lumped elements suit frequencies up to a few gigahertz, and distributed lines take over above that because an 0402 part is no longer small compared with the wavelength. The impedance matching calculator gives L, Pi and T values directly.
Common Matching Problems
A paper design rarely survives first contact with a board. Pads and traces add reactance, so the impedance should be measured at the reference plane where the network will sit. A vector network analyzer, calibrated to that plane, is the usual tool.
Loads also move. Antennas detune when placed near a hand or inside a case, and antenna matching in products often leaves a Pi footprint so that values can be tuned on the final hardware. Over a wide band, the load itself changes with frequency, and the best achievable match is limited by how much energy the load stores. Finally, high power can break a match: inductors saturate, capacitors heat, and an amplifier’s output impedance changes with drive level.
Frequently Asked Questions
What happens if impedance is not matched?
Part of the incident power reflects from the mismatch and less reaches the load. In a transmitter the reflected power can stress the amplifier. In a cascade, reflections cause gain and group delay ripple, and filters lose their designed shape.
Is a perfect match always the goal?
No. An LNA is normally matched closer to its minimum noise figure, and a power amplifier is matched to the load that gives the required output power and efficiency. Both are deliberately mismatched in the conjugate sense, and isolators or balanced stages can be used to keep their ports well behaved.
What is a good VSWR or return loss?
It depends on the application. A VSWR of 2:1, about 9.5 dB return loss, is a common working target for antennas, where it costs roughly 0.5 dB. Instrument ports and cascaded stages usually aim for much better, often 15 dB to 20 dB return loss or more.
Why is 50 Ω used in RF systems?
It became the standard compromise between the lowest-loss and highest-power impedances of coaxial cable. Now that instruments, connectors and components are all specified at 50 Ω, using it makes parts interchangeable. Other systems exist, such as 75 Ω for video and cable television.
Related Topics
For the basics behind this page, start with the RF fundamentals hub and S-parameters explained, then work through the RF design guide. More design tools are on the RF calculators page.