Pi-Network Matching: Choosing Q and Splitting the Ratio

A pi network matches through a virtual resistance R_v = R_high/(1 + Q²) set below both terminations. For 50 Ω to 200 Ω at Q = 5, R_v = 7.69 Ω and the second section runs at Q = 2.345. Q must exceed the L-network value of 1.732.

Engineer’s note: Three parts give a Pi network a free choice of Q, but only upward: its virtual resistance sits below both ends, so its Q is always higher than a single L-network's. My advice is to pick the lowest Q that still gives the harmonic rejection you need. A lower Q is less sensitive to part tolerance, and that pays off directly in yield.

For practitioners

Why add a third element

An L-network has two elements and two unknowns. Once you fix the two resistances, the Q is fixed too:

QL=RhighRlow−1Q_L = \sqrt{\frac{R_{\text{high}}}{R_{\text{low}}} - 1}

For 50 Ω to 200 Ω that gives √3 = 1.732. You can't change it. The only choice left is whether the L is low-pass or high-pass. Often you need a narrower response than that, for harmonic rejection or for selectivity ahead of a sensitive stage. A third element gives you the extra degree of freedom, and in a pi network that freedom is the Q.

The cleanest way to design a pi is to stop treating it as a three-element network. Split it down the middle of the series arm. Each half is then an ordinary L-section: a shunt element at an outer port and a series element facing the centre. The two halves meet at an internal node where both see a purely resistive impedance. That is the virtual resistance, R_v.

The virtual resistance sits below both ends

In an L-section, the shunt element always sits on the high-resistance side. A shunt-first L transforms downward. In a pi, both outer ports have shunt elements, so both halves transform down toward the middle. That forces the rule:

Rv<min⁡(R1,R2)R_v < \min(R_1, R_2)

Each section then has its own Q, set by its own ratio:

Qi=RiRv−1Q_i = \sqrt{\frac{R_i}{R_v} - 1}

The section on the higher-resistance side has the larger ratio, so it has the larger Q. That higher Q governs the bandwidth, so it's the one quoted as the network Q. The T-network is the mirror case: its virtual resistance sits above both terminations.

Choosing Q and solving for R_v

Pick the design Q and apply it to the high side:

Rv=Rhigh1+Q2R_v = \frac{R_{\text{high}}}{1 + Q^2}

Then check that R_v is below R_low. This is the same as requiring

Q>RhighRlow−1Q > \sqrt{\frac{R_{\text{high}}}{R_{\text{low}}} - 1}

The boundary case shows why. If Q equals the L-network value, R_v comes out equal to R_low. The low-side section then has Q = 0 and its shunt reactance R_low/0 goes to infinity, so that element disappears. The pi turns back into an L. Any Q below that limit puts R_v above R_low, and no pi can do that.

Worked example: 50 Ω to 200 Ω at Q = 5

The minimum Q for this ratio is √3 = 1.732, so Q = 5 is valid.

  1. Virtual resistance: R_v = 200/(1 + 25) = 200/26 = 7.69 Ω. That's below 50 Ω, so the design is valid.
  2. Low-side section Q: 50/7.69 = 6.5, so Q₂ = √(6.5 − 1) = √5.5 = 2.345.
  3. Shunt reactances, X_shunt = R_i/Q_i: 200/5 = 40 Ω on the load side and 50/2.345 = 21.32 Ω on the source side.
  4. Series reactances, X_series = Q_i × R_v: 5 × 7.69 = 38.46 Ω and 2.345 × 7.69 = 18.04 Ω.
Q = 2.345(50 to 7.69 Ω)Q = 5(200 to 7.69 Ω)50 ΩsourceX = 18.04 ΩX = 38.46 Ω200 ΩloadX = 21.32 ΩX = 40 Ωvirtual R_v = 7.69 Ωsame type: 18.04 + 38.46= 56.50 Ω
The 50 Ω to 200 Ω pi network at Q = 5 drawn as two L-sections that meet at the virtual resistance R_v = 7.69 Ω. Concept drawing.

The two series reactances sit in series with each other, so they merge into one physical component. If both are the same type, they add: 18.04 + 38.46 = 56.50 Ω. In the classic low-pass pi, with shunt capacitors and a series inductor, the circuit is a 21.32 Ω capacitor to ground at the 50 Ω port, a 56.50 Ω inductor in series, and a 40 Ω capacitor to ground at the 200 Ω port. If one section is low-pass and the other high-pass, the series arms have opposite signs and you take the difference, 38.46 − 18.04 = 20.42 Ω. The net is still inductive, because the larger arm is inductive.

To get component values, convert at the design frequency f₀: C = 1/(2π f₀ X) and L = X/(2π f₀).

What the extra Q buys and what it costs

The loaded Q of a matching network sets its fractional bandwidth, roughly BW ≈ f₀/Q. Raising Q from 1.732 to 5 gives a much narrower match and, in the low-pass form, steeper attenuation above the passband. The second section adds some selectivity of its own, so the response is a little narrower than the single-number estimate suggests.

There are costs:

  • Circulating reactive current rises with Q. A rough estimate of the extra dissipative loss scales with the ratio of loaded Q to component unloaded Q, so high-Q pi networks punish lossy inductors.
  • Sensitivity to tolerance rises with Q. Small errors in a 21.32 Ω or 40 Ω shunt reactance shift the match more when the network is sharply tuned.
  • A pi can only make the bandwidth narrower than an L. If you need a wider match than Q = 1.732 allows, use cascaded L-sections with an intermediate resistance between the two ends. A pi can't do that.
Quick check for any pi design: R_v must be below both terminations, and the section on the higher-resistance side must carry the larger Q. If either check fails, the arithmetic is wrong, usually because the design Q was applied to the low side.

In plain words · for students and beginners

Pi-network matching in plain words

An everyday picture

You need to get from the 2nd floor of a building to the 8th. The direct staircase has one fixed steepness, and that's the L-network. Its Q (how sharply tuned the match is) is set entirely by the two floors. A pi network takes a detour instead. It goes down to the basement first, then climbs to the 8th floor. Because you choose how deep the basement is, you choose how steep both climbs are.

5102050100200300Resistance (Ω), log scale200 Ω50 ΩRv = 7.69 Ωsingle L-network:Q = 1.732 (fixed)Q = 2.345Q = 5
Going lower than both ends lets you choose Q (must be above 1.732). Concept drawing.

Where the picture breaks down: in a building, a steeper climb only costs effort. In a circuit, a steeper "climb" (higher Q) makes the match work over a narrower range of frequencies (narrower bandwidth). It also makes the parts lose more energy as heat.

The same example, step by step

We want to match 50 Ω to 200 Ω. Ohms (Ω) measure resistance.

  1. Pick Q = 5. It has to be more than 1.732, the fixed Q of the direct L-network.
  2. Find the basement level: 200 ÷ (1 + 5 × 5) = 200 ÷ 26 = 7.69 Ω. This is the virtual resistance. It isn't a real part. It's the impedance (the AC version of resistance) you would see at the middle of the circuit.
  3. Find the Q of the other half: 50 ÷ 7.69 = 6.5, and √(6.5 − 1) = 2.345.
  4. Parts to ground (shunt parts): 200 ÷ 5 = 40 Ω and 50 ÷ 2.345 = 21.32 Ω.
  5. Parts in the signal path (series parts): 5 × 7.69 = 38.46 Ω and 2.345 × 7.69 = 18.04 Ω. If both are coils, they combine into one coil of 56.50 Ω.

These Ω figures are reactances. Reactance is how strongly a coil or capacitor opposes AC at a given frequency. You turn them into actual coil and capacitor values once you know the frequency.

Mistakes beginners make

  • Putting the middle level too high. In a pi, the virtual resistance has to be lower than both ends. If yours comes out between 50 Ω and 200 Ω, something is wrong.
  • Choosing Q too low. Any Q at or below 1.732 doesn't work for 50 Ω to 200 Ω. At exactly 1.732, one part disappears and you're back to an L-network.
  • Applying Q to the wrong side. The design Q goes with the larger resistance (200 Ω here). The other half always ends up with a lower Q.
  • Adding series parts that don't add. Two coils add up. A coil and a capacitor partly cancel, so you subtract.

Remember

  • A pi is two L-networks back to back that meet at a resistance lower than both ends.
  • R_v = R_high ÷ (1 + Q²). For 200 Ω and Q = 5 that's 7.69 Ω.
  • Higher Q means a narrower bandwidth and more loss. A pi can only make the match narrower than an L, never wider.

Frequently asked questions

How do I calculate the virtual resistance in a pi matching network?

Choose the design Q and apply it to the higher-resistance side: R_v = R_high/(1 + Q²). For 200 Ω and Q = 5 this gives 200/26 = 7.69 Ω, which must also be below the other termination.

What is the minimum Q for a pi network?

The Q must be higher than the L-network Q for the same ratio, √(R_high/R_low − 1). For 50 Ω to 200 Ω that is √3 = 1.732. At exactly that value the pi turns back into a plain L-network.

Why does a pi network need R_v lower than both terminations?

Each half of the pi is an L-section with its shunt element at the outer port. A shunt-first L-section always transforms down, so both halves end at a resistance lower than their own termination.

How do I combine the two series reactances in a pi network?

If both series arms are the same type, add them. In the 50 Ω to 200 Ω, Q = 5 example, 18.04 Ω + 38.46 Ω = 56.50 Ω. If one is inductive and the other capacitive, take the difference, which here is 20.42 Ω.

References and further reading