What this calculator does
A matching network is a lossless arrangement of inductors and capacitors placed between a source and a load so that the source sees the complex conjugate of its own impedance, . That is the condition for maximum power transfer. For a lossless network it also means the load sees looking back toward the source. The calculator designs three classic lumped networks:
- the L-network (two elements). For a given and there are two solutions, and the calculator gives both;
- the Pi network (shunt, series, shunt) and the T network (series, shunt, series), where you choose the loaded and so set the bandwidth.
Enter the source and the load as series resistance and reactance (, with negative for a capacitive part). For each element, the results give its reactance and its inductance or capacitance at the design frequency. Each label tells you whether the element is in series or in shunt, and where it sits: on the source side, in the middle, or on the load side.
The L-network
Pozar (§5.1) uses two L-section shapes. When the load resistance is at least the source resistance (), the shunt element goes across the load and the series element sits on the source side. Otherwise the series element sits next to the load and the shunt element goes across the source. The calculator picks the shape that is always realisable. Within that shape, the quadratic for the shunt susceptance (or series reactance) has two roots. These are solution 1 and solution 2. One is usually low-pass (series L, shunt C) and the other high-pass. With a reactive load, the absorbed reactance can make an element switch type.
Once and are fixed, so is the L-network's Q. For resistive terminations it is (Bowick Eq. 4-1). For complex terminations, the calculator reports the node Q at the shunt element, or .
Pi and T networks
A Pi network is two L-sections back to back. Each one transforms its termination down to a "virtual" resistance in the middle, and the two series arms combine into one element. Bowick defines the loaded Q through the larger termination:
The T network is the dual. Both terminations are transformed up to , and the two shunt arms combine. A Pi or T network exists only when is larger than the L-network's Q, . Below that the calculator warns you and reports the minimum. At exactly , one shunt (Pi) or series (T) element disappears, and the network becomes an L-network. The Pi/T form selector chooses the low-pass form (series inductors, shunt capacitors, which also filters harmonics) or the high-pass form (series capacitors, which also block DC).
The bandwidth estimate is . It is a good guide for and only a rough one for lower Q.
Worked example
Match a transistor input of to a 50 Ω source at 900 MHz (the defaults).
L-network, solution 1: the load resistance is lower than the source resistance, so the series element is on the load side. The calculator gives a series inductor of 7.3654 nH (+41.651 Ω, which includes the 20 Ω needed to cancel the load's capacitive reactance) and a shunt capacitor of 6.1259 pF (−28.868 Ω) across the source. The network Q is 1.7321, so the estimated bandwidth is 519.62 MHz.
Solution 2: a series capacitor of 107.13 pF (−1.6506 Ω) at the load and a shunt inductor of 5.1049 nH (+28.868 Ω) across the source. The Q is the same. This version blocks DC, but the 107 pF part is almost a short at 900 MHz, so its self-resonance matters.
Pi, low-pass, Q = 5: Ω. The network is a shunt 17.684 pF across the source, a series 3.3005 nH, and a shunt 12.340 pF across the load. The bandwidth estimate drops to 180 MHz.
Solving and sweeping
There are no solve-for targets here, because the outputs are the element values themselves. To design for a bandwidth, set directly. Use Sweep & plot on the frequency to see how the component values scale. The reactances do not depend on frequency, so and .
Accuracy and validity
| Item | Accuracy |
|---|---|
| Element reactances (ideal L and C) | Exact: the network presents to better than (checked by an independent cascade over random complex terminations) |
| Pozar Ex. 5.1 and Bowick Ex. 4-4/4-5 | Reproduced within the books' rounding (≤ 1 %) |
| Bandwidth | Approximation, best for |
| Real components | Not modelled: parasitics, self-resonance, finite |
The calculator warns about inductors below about 0.5 nH and capacitors below about 0.3 pF, which are comparable to pad parasitics. It also warns about Q above 20, where component loss becomes significant (loss grows roughly as , where is the components' unloaded Q).
Practical notes
- Absorb the parasitics. If the load has a 0.5 pF package capacitance in shunt, subtract it from the calculated shunt capacitor rather than adding a separate element (Bowick's approach).
- Pick the solution for the job: low-pass to suppress harmonics after a power amplifier, high-pass or series-C to block DC bias, and the one with the most practical component values otherwise.
- Complex source. Enter the real output impedance of the driving stage. The network matches to its conjugate, not to its value.
- Check self-resonance. Every chip inductor and capacitor must have a self-resonant frequency well above , usually at least twice it.
- Multi-section for bandwidth. If the L-network's Q is already too high for the bandwidth you need, cascade L-sections through intermediate resistances (geometric steps) to lower the Q.