IndieRF Match

Smith chart calculator and impedance matching

Plot an impedance on the Smith chart, read its Γ, return loss, and VSWR, then match it. Type the load or drop a Touchstone file, and get an L-section, Pi, T, single stub, or quarter-wave transformer with real part values and the matched band.

Loading the Smith chart…

How to use

  1. 1Set your load. Demo antenna, a typed Z, or a .s1p file.
  2. 2Set frequency and source. f0, Zs, and the chart reference.
  3. 3Pick a network. L-section, Pi, T, a single stub, or a quarter-wave transformer.
  4. 4Compare solutions. Signs, stub positions, and the minimum-Q L, on the chart and the cards.
  5. 5Pick real parts and Q. Snap to E24 or E96, then set inductor and capacitor Q. For a line, set Z0 and εeff.
  6. 6Check bandwidth and export. Read the shaded band, then copy a link or download.

Question or a correction?

Smith chart matching examples by network type

Seven worked cases, covering the L-section, Pi, T, single stub, and quarter-wave. Each button opens the tool with the case loaded, so you can change the load or the frequency and watch the answer move.

Which matching network to pick
NetworkPartsPick it when
L-sectionTwo chip partsYou want the fewest parts and the widest band the resistance ratio allows.
PiThree chip parts, shunt at each endYou want a narrower band than the L-section, usually to reject harmonics.
TThree chip parts, series at each endBoth ends want a series element, or you need a DC block on both sides.
Single stubOne open or short line, no chip partsBoard length is cheaper than parts. It does not roll off harmonics.
Quarter-waveOne line section, no chip partsThe load is real and one narrow section is enough.
  • L-section

    Match a 2.4 GHz chip antenna to 50 Ω with an L-section

    The synthetic 2.45 GHz demo antenna reads 10.64 + j5.558 Ω into a 50 Ω source. It is a model, not a measured part.

    Low-pass L, from the source: shunt 2.499 pF, series 968.4 pH. The ≥10 dB band is 2.311 GHz to 2.584 GHz.

    Nodal Q is 1.92, so the band is 273.2 MHz wide. An L-section is the widest two-part lossless match for a given resistance ratio, which makes this the number to beat.

    Open in MatchRead the note

  • L-section

    Match 50 Ω to 75 Ω with an L-network

    A real 75 Ω load into a 50 Ω source at 100 MHz.

    Low-pass L, from the source: series 56.27 nH, shunt 15.01 pF. The high-pass pair is 45.02 pF and 168.8 nH. The ≥20 dB band of the low-pass is 71.24 MHz to 122.2 MHz.

    Connected directly, the pair reflects |Γ| = 0.20: 14.0 dB of return loss and 0.18 dB of mismatch loss. Many links can skip the network. Build the L (nodal Q 0.71) when you need better than 14 dB.

    Open in Match

  • Pi

    Match a 915 MHz antenna with a Pi network

    A typed example load of 20 − j40 Ω at 915 MHz, into 50 Ω. It stands in for a short whip and is not a measured antenna. The load is held at that value at 1.83 GHz.

    Low-pass Pi at loaded Q 4, from the source: shunt 9.527 pF, series 6.895 nH, shunt 3.479 pF. The two-part low-pass L is shunt 4.261 pF, series 11.22 nH.

    At 1.83 GHz the Pi reads −22.3 dB and the L reads −10.1 dB. The price is band: ≥10 dB over 115.9 MHz for the Pi against 341.8 MHz for the L. Take the Pi when the second harmonic is in the specification.

    Open the PiOpen the L

  • Pi and T

    Pi or T network for a 10 Ω PA load

    A real 10 Ω optimum load into 50 Ω at 2.4 GHz. It is a typed value, not a vendor transistor file.

    Pi at loaded Q 5: 6.631 pF, 899 pH, 13.59 pF, ≥10 dB over 2.266 GHz to 2.516 GHz, −25.5 dB at 4.8 GHz. T at loaded Q 5: 6.795 nH, 1.798 pF, 3.316 nH, ≥10 dB over 2.266 GHz to 2.516 GHz, −25.5 dB at 4.8 GHz.

    A low-pass Pi and a low-pass T at loaded Q 5 read the same, −25.5 dB at 4.8 GHz, so harmonic rejection does not decide between them. A Pi puts a shunt part at each end and a T puts a series part at each end, so pick by which element your board wants at the load.

    Open the PiOpen the TRead the note

  • Single stub

    Single-stub match for 60 − j80 Ω

    A textbook-style 60 − j80 Ω load on a 50 Ω line, drawn at 2.4 GHz on 1.6 mm FR-4 (εeff 3.351).

    Shunt stub 0.110 λ from the load, which is 7.53 mm. The shorted stub is 6.48 mm long. The open stub at the same spot is 23.54 mm.

    The short is 17.1 mm shorter than the open, so it is the smaller build. A stub spends board length instead of parts, and it does not roll off the second harmonic the way a low-pass Pi does.

    Open the shorted stubOpen the open stubRead the note

  • Quarter-wave

    Quarter-wave transformer for a 100 Ω load

    A real 100 Ω load into a 50 Ω line at 1 GHz.

    Z1 = √(50 × 100) = 70.71 Ω, 90° (a quarter wavelength) at the design frequency.

    One section sets the match at one frequency, so the band is narrow. Pick it for a real load when line length costs less than parts.

    Open in Match

  • L-section with port extension

    Match an antenna from a NanoVNA reading

    A NanoVNA calibrated at the SMA reads 24.6 + j54.8 Ω at 2.45 GHz. 50 ps of lossless 50 Ω line (15.0 mm of air) sits between the connector and the antenna pad.

    With a +50 ps shift the pad reads 10.65 + j5.54 Ω. Low-pass L, from the source: shunt 2.498 pF, series 969.9 pH.

    Enter the one-way delay: the reflection crosses the line twice, so the phase moves twice as far. Skip the shift and the solver sizes parts for 24.6 + j54.8 Ω, an impedance the antenna never presents.

    Open the shift in MatchRead the note

Last updated 11 November 2026.

How to read the chart

Smith chart, 2.4 GHz chip-antenna example
Synthetic chip-antenna load and the low-pass L-section, drawn from the load toward 50 Ω.

Every point on the chart is one load impedance at one frequency. The center is a perfect match to the chart reference, 50 Ω here, and the rim is total reflection. The left end of the horizontal line is a short and the right end is an open. The upper half is inductive and the lower half is capacitive. Open a 50 Ω load and it sits on the center with VSWR 1.00.

To read a load, divide by the reference. 25 + j50 Ω is z = 0.5 + j1, where the r = 0.5 circle crosses the x = +1 arc in the upper half. Its distance from the center is |Γ| = 0.620: return loss 4.15 dB, VSWR 4.27. Turn on VSWR circle and every point on that ring has the same VSWR. Read 25 + j50 Ω.

Then move it. Lossless 50 Ω line turns the point clockwise on its VSWR circle, toward the generator: 0.1 λ takes it to 184.8 + j70.3 Ω at the same VSWR. A series part slides it along its resistance circle: a series 3.2 pF pulls it down to 25 + j0.26 Ω, VSWR 2.00. A shunt part slides it along its conductance circle: a shunt 2.5 pF swings it clockwise to 124.8 + j4.6 Ω, VSWR 2.50. Two moves that end on the center are an L-network. The Smith chart note walks each step.

The chart plots the reflection coefficient Γ = (Z − Zref) / (Z + Zref). Return loss in decibels is −20 log10|Γ|, and VSWR is (1+|Γ|)/(1−|Γ|). The chart reference Zref is independent of the source impedance: the match target is the conjugate of the source, Zs*, which is the center only when the source is real and equal to Zref. Admittance circles are the same geometry read as y = 1/z. A constant-Q circle through z = 1 + jQ is centered at (0, −1/Q) and passes through the short and the open.

From a file to an impedance

Drop a 1- to 4-port Touchstone file. The chosen port’s reflection is interpolated in Γ, not in Z, and the frequency is not extrapolated past the file. On a multiport file the other ports are terminated. “file” keeps each port’s own reference so the one-port result is Sii. A numeric termination reduces the network through its Z matrix. If the phase steps more than 20° between samples, the sweep is too coarse to trust the band edges. The reference plane starts as the plane in the file. A signed length then slides it along one uniform line: positive removes a feed, negative adds line. Z0 defaults to 50 Ω. εeff or a velocity factor sets the delay, and an optional loss is in dB/in. Degrees and picoseconds are the one-way time at the design frequency. “Use the open Trace Studio line” copies Z0, εeff, and loss from the calculator in this browser, or a fab preset locks a 50 Ω microstrip (Zdiff when the saved line is a pair).

Which network this tool builds

The network control solves an L-section, a Pi, a T, a single stub, or a quarter-wave transformer. An L-section is two lumped parts and the widest lossless match for the resistance ratio. A Pi or a T adds a third part so the loaded Q can sit above that minimum. A stub or a quarter-wave line spends trace length instead of a chip L and C. A microstrip width for the line Z0 is on the microstrip calculator, and εeff for the physical length comes from Trace Studio. Measured files that you only want to plot belong in the S-parameter viewer.

Pi and T networks

A Pi or a T is two L-sections through a virtual resistance, following Bowick. The loaded Q you type is the larger of the two section Qs, so it has to exceed the L-network minimum Q = sqrt(Rhigh/Rlow − 1). For a Pi the virtual resistance sits below both parallel resistances, Rv = Rhigh / (Q² + 1). For a T it sits above both series resistances, Rv = Rlow · (Q² + 1). Each end has two signs, so a real-to-real step has four networks. The low-pass Pi is shunt C, series L, shunt C. The L-section cards stay on the chart as the minimum-Q reference.

For 50 Ω to 250 Ω at 100 MHz and loaded Q = 10, the minimum Q is 2.00 and the low-pass series reactance is 35.6 Ω. These figures are computed by this engine. Typing a bandwidth instead of Q sets Q = f0 / bandwidth. That is the design estimate. The shaded span on the plot is the measured return loss.

Lumped solutions also list the nearest E24 and E96 values. Those snaps use the same component-Q model as the curve: inductor series resistance ωL/Q and capacitor ESR 1/(ωCQ). The plotted curve still follows the series selected in the form, including the neighbor search on the RF grid.

Single-stub matching

A single stub, following Pozar, moves along a line until the real part of the normalized admittance is 1 (shunt stub) or the real part of the normalized impedance is 1 (series stub). The stub then cancels the imaginary part. Open and short stubs are both solved, and each placement has two distances inside a half wavelength. The chart draws the through line as a constant-|Γ| rotation when the line Z0 is the chart reference. The stub length itself is the rim rotation from the open or the short.

For 60 − j80 Ω into 50 Ω the nearer shunt position is 0.110 λ and the farther is 0.259 λ. At the nearer position the shorted stub is 0.095 λ and the open stub is 0.345 λ, a quarter turn apart. Distance d and length l are also shown in degrees and in millimetres from Z0 and εeff. Velocity factor is 1/sqrt(εeff). The match sets Zin equal to the real line Z0, so a complex source is warned rather than treated as a conjugate target.

Quarter-wave transformer

A real load uses one section, Z1 = sqrt(Z0 RL), λ/4 long. A 100 Ω load into 50 Ω needs 70.71 Ω and 90°. A complex load is moved on a Z0 line to a real-axis impedance first. There are two such points, and the shorter section is listed first. Z1 is usually not the chart reference, so that arc is not a constant-|Γ| circle. Physical length uses the same εeff as the stub.

Worked example: 2.45 GHz chip antenna

The demo load is synthetic, not a vendor file. A 10 Ω resistor, 3.2 nH, and 1.5 pF in series, with 0.35 pF in parallel, gives 10.64+j5.558 Ω at 2.45 GHz. Unmatched into 50 Ω that is about 3.7 dB return loss. Only topology A is available, because the resistance is below 50 Ω. These figures are computed by this engine.

Scroll sideways for more columns

Computed chip-antenna L-sections. The snapped row searches neighboring RF values for the lowest worst-case |Γ| from 2.2 to 2.7 GHz, so it can leave the nearest 0.1 pF step.
NetworkParts, source to loadRL at 2.45 GHz≥ 10 dB
Low-pass, ideal2.499pF, 968.4pHmatched2.311GHz – 2.584GHz
High-pass, ideal1.689nH, 2.496pFmatched2.331GHz – 2.59GHz
Low-pass, RF snap2.3pF, 1nH23.5 dB2.32GHz – 2.606GHz
Low-pass, QL = 30, QC = 3002.3pF, 1nH27.0 dBGT -0.24 dB

The RF grid is 0.1 nH and 0.1 pF through 9.9, then E24. The combined search may pick a neighbor of the nearest value when that neighbor improves the worst |Γ| on the sweep. On this antenna the low-pass capacitor moves off the nearest 2.5 pF step. Constant Q is the default and is labeled that way. Inductors can instead use Q(f) = Q0 · √(f/f0). Capacitors stay at the Q you type. Leave the Q field blank, or type ∞, for an ideal part.

Component Q

Q is the ratio of reactance to series loss: ωL/R for an inductor and 1/(ωC·ESR) for a capacitor. In this tool that resistance is applied when the network is evaluated, not inside the topology solver. It sets the insertion loss, which shows up as transducer gain GT below 0 dB, and it moves the match a little, so the return-loss peak is no longer infinite. A higher nodal Q also narrows the band. The presets in the form are typical starting points near 1–6 GHz. They are not datasheet values. Read the Q-versus-frequency curve for the part and the value you will buy.

  • Multilayer chip inductors are the low-Q end, often about 10–25 near 1–6 GHz. Murata’s LQG family is this construction. Reference specification JELF243B-0010 lists a minimum Q of 8 for the 10 nH part LQG15HS10NJ02. The product page for LQG15HS47NJ02, the same series and the same minimum of 8, lists the Q test frequency as 100 MHz. That minimum is not the Q at 2.4 GHz.
  • Thin-film inductors, such as Murata LQP, sit above multilayer parts. Murata’s LQP03HQ reference specification lists a minimum Q of 20 for LQP03HQ10NH02 (10 nH), measured at 500 MHz for the 0.5 nH to 30 nH range.
  • Wirewound chip inductors are higher, often about 30–70 and sometimes more. Murata notes that the LQW wire-wound structure can reach a higher Q than LQP or LQG. The LQW15AN specification measures Q at 250 MHz for 1.5 nH to 43 nH; LQW15AN9N0 (9.0 nH) has a minimum Q of 25 at that frequency. Coilcraft’s 0402HP series, document 526 revised 20 December 2023, lists a typical Q of 62 at 900 MHz and 90 at 1.7 GHz for 0402HP-10N, measured on an Agilent/HP 4287A with a 16197 fixture.
  • C0G/NP0 MLCCs in small picofarad values are often Q of 100 or more in the low GHz. High-Q RF capacitors are higher still. Murata GJM is a high-Q C0G/C0H series aimed at about 500 MHz to 10 GHz. Johanson’s high-Q catalog quotes Q greater than 1,000 at 1 MHz, typically 10,000 below 1,000 pF. ATC’s 600S series quotes Q greater than 2,000 at 1 MHz and plots Q against capacitance at 450 MHz and 2 GHz. Use those curves. A 1 MHz catalog Q is not the Q at 2.4 GHz.

Sources: Murata inductor FAQ, “Why is the Q characteristics of the LQW better than the LQP and LQG”; Murata reference specifications JELF243B-0010 (LQG15HS) and JELF243C-0021 (LQP03HQ); Murata LQW15AN reference specification; Murata GJM series lineup; Coilcraft document 526, 0402HP Series; Johanson Technology high-Q MLCC catalog; ATC 600S Series datasheet. Preset buttons such as “Multilayer ~15” are inside the ranges above. They are not a quoted Q from any of those documents.

Limits

  • Ideal lumped elements, plus a series loss from Q and an optional inductor self-resonance. No pad or via parasitics and no electromagnetic simulation.
  • No stability circles. A conjugate match of a transistor is not a stability analysis. For a transistor, start in Amplifier Design, which places Γ_S and Γ_L on the stable side of the stability circles and links back here to build each network.
  • The reference shift is one uniform line. It does not model a pad, a via, or a fixture with more than one section.
  • Exports are free, no account needed. Share links, on-screen results, Touchstone .s2p/.s1p, and the BOM CSV download without an account. Measured points stay in the URL hash and are not sent to the server.

References

  • P. H. Smith, “Transmission Line Calculator,” Electronics, vol. 12, pp. 29–31, January 1939.
  • D. M. Pozar, Microwave Engineering, chapter 5, single-stub tuning and the quarter-wave transformer.
  • C. Bowick, RF Circuit Design, chapter 4, L, Pi, and T matching networks.
  • R. Ludwig and P. Bretchko, RF Circuit Design: Theory and Applications, matching networks.
  • R. M. Fano, “Theoretical limitations on the broadband matching of arbitrary impedances,” Journal of the Franklin Institute, 1950.

The L-network page lists a computed check of Pozar’s Example 5.1. Printed book figures are not copied here until the edition and page are transcribed. Design frequency on the demo is 2.45GHz.

Notes

Checked against scikit-rf: the cases, the tolerance and what the check does not cover.

FAQ

How do I use an online Smith chart?

The chart is the reflection-coefficient plane. The center is a perfect match to the chart reference, usually 50 Ω. Circles of constant resistance sit to the right of the short, and arcs of constant reactance leave the right-hand open. Click the chart or type a load, set the design frequency, and each element is drawn from the load back toward the source. A through line rotates on a constant-|Γ| circle when its Z0 is the chart reference. A stub’s length is the rim rotation from the open or the short.

How do I get antenna impedance from a NanoVNA?

Save a 1-port Touchstone file and drop it here. Pick the frequency you want to match. The tool reads Γ from the file and converts it with the file’s reference impedance. A NanoVNA calibration is at the SMA, not at the antenna pad. A positive length removes one uniform feed (port extension). Degrees and picoseconds are the one-way delay at the design frequency; the reflection phase is twice that. It is not a pad or a multi-section fixture.

Why does an L-network have two or four solutions?

Each allowed topology has two signs for the reactance that lands on the match circle. A load whose resistance is below the source uses the series-at-the-load topology. A load whose conductance is below the source uses the shunt-at-the-load topology. A complex source can allow both, so up to four networks. A load already on the match circle collapses to one element, and a second valid network can remain.

Can I match a purely reactive load?

No. A lossless network cannot create the real part a conjugate match needs. A negative resistance is rejected for the same reason: this tool matches passive loads.

Which solution has the most bandwidth?

The list is ordered by the measured return-loss band that contains the design frequency, then by nodal Q. The f0/Q figure is only an estimate. The band edges come from the sweep, with linear interpolation of return loss. An L-section is the widest lossless match for a given resistance ratio. A Pi or T is narrower because you asked for a higher loaded Q. A single stub or one quarter-wave section is narrow as well.

When should I pick a Pi, a T, a stub, or a quarter-wave transformer?

Pick an L-section when you want the fewest parts and the widest band. Pick a Pi or a T when you want a loaded Q above that L-network minimum, usually to reject harmonics: a low-pass Pi is shunt C, series L, shunt C. A Pi or T uses three parts. A stub or a quarter-wave transformer uses no chip L or C and spends PCB length instead. It does not give that low-pass shape. The shorter stub fits a smaller board.

What does loaded Q mean on a Pi or T?

It is the larger of the two L-section Qs in Bowick’s design, and it has to sit above sqrt(Rhigh/Rlow − 1). You can type that Q, or a bandwidth, in which case Q becomes f0 divided by the bandwidth. That bandwidth is the design estimate, not the shaded return-loss span on the plot. Below the minimum, the tool keeps the L-section and does not invent a Pi or T.

How do I read a single-stub length?

Distance d is how far the stub sits from the load. Length l is the stub itself. Both are shown in electrical degrees, wavelengths, and millimetres. The millimetres use the line Z0 and εeff you type. Velocity factor is 1/sqrt(εeff). Estimate εeff for a microstrip in Trace Studio. Open and short stubs at the same d differ by 90°. The two distances are the nearest solutions inside a half wavelength.

How does a quarter-wave transformer match a complex load?

A real load needs one section, Z1 = sqrt(Z0 RL), a quarter wavelength long. A complex load is not real, so a Z0 line moves it to a real impedance first. There are two such points, one high resistance and one low, and each then gets its own Z1. The transformer Z0 is usually not the chart reference, so its arc is not a constant-|Γ| circle.

How accurate is this compared with a circuit simulator?

The ideal parts are the same closed forms used in textbooks, and every solution is checked by cascading ABCD matrices. Finite Q and an optional inductor self-resonance are lumped models. Pad, via, and layout parasitics are not included, so a simulator or a VNA measurement is still the check before a board spin.

How do I match 50 ohms to 75 ohms?

You may not need to. A direct 50 Ω to 75 Ω junction reflects |Γ| = 0.20, which is 14.0 dB of return loss and 0.18 dB of mismatch loss. When you want better, use an L-network. At 100 MHz the low-pass version is a 56.27 nH series part at the 50 Ω side and a 15.01 pF shunt part at the 75 Ω side, with a nodal Q of 0.71, and return loss stays above 20 dB from 71.24 MHz to 122.2 MHz. The high-pass version is 45.02 pF and 168.8 nH.

When should I use a T network instead of a Pi?

A Pi has a shunt part at each end and a T has a series part at each end. At 2.4 GHz into a real 10 Ω load, A low-pass Pi and a low-pass T at loaded Q 5 read the same, −25.5 dB at 4.8 GHz, so harmonic rejection does not decide between them. Choose the one whose end elements suit the board. A T is the one to reach for when both ends want a series part or you need a DC block on both sides.

How long is a quarter-wave transformer, and what impedance does it need?

The section impedance is Z1 = √(Z0 × RL) and the length is a quarter wavelength, 90°, at the design frequency. For a 100 Ω load into 50 Ω, Z1 is 70.71 Ω. The physical length is c / (4 f √εeff), and εeff belongs to the 70.7 Ω line you draw, not to the 50 Ω line next to it, because a wider or narrower trace has a different εeff.