RF Matching Networks: L, Pi and T Compared
An L-network has its loaded Q fixed by the resistance ratio: Q = sqrt(R_high / R_low - 1), which is 1.73 for 50 ohm to 200 ohm. Pi and T networks add a third part so you can choose a higher Q and a narrower, more selective match.
For practitioners
What a matching network does
A matching network transforms one impedance into another so that a source delivers its power to a load without reflecting it back. This page covers four common forms for real resistances: the L-network, the Pi network, the T network and the quarter-wave transformer. They differ in how many parts they use, whether you can choose the bandwidth, and what they pass or block besides the signal.
The common thread is loaded Q. The loaded Q of a match is roughly the center frequency divided by its 3 dB bandwidth:
A higher Q gives a narrower match. The question for each topology is whether you get to pick Q or whether the circuit picks it for you.
The L-network: two parts, Q fixed
An L-network uses two reactive parts, one in series and one in shunt. It matches two real resistances. With R_high above R_low, the loaded Q is fixed by the ratio:
You get no freedom here. Two parts give two equations, and the two equations are exactly the two things you must satisfy: cancel the reactance and set the resistance ratio. That is why the L-network is the simplest match and also the one you cannot tune for bandwidth.
It comes in two versions. The low-pass version passes DC and attenuates high frequencies. The high-pass version blocks DC. Both give the same match at the design frequency, so the choice depends on what else the network must do: block a bias voltage, pass a DC path, or reject harmonics.
Worked example: 50 ohm to 200 ohm at 1 GHz
The ratio is 200 / 50 = 4, so:
The series reactance on the low-resistance side is X_s = Q × R_low = 1.73 × 50 = 86.6 ohm. The shunt reactance on the high-resistance side is X_p = R_high / Q = 200 / 1.73 = 115.5 ohm.
At 1 GHz, ω = 2π × 10⁹ rad/s. The two solutions are:
- Low-pass: a series inductor of 86.6 / ω = 13.78 nH, and a shunt capacitor of 1 / (ω × 115.5) = 1.378 pF.
- High-pass: a series capacitor of 1 / (ω × 86.6) = 1.838 pF, and a shunt inductor of 115.5 / ω = 18.38 nH.
You can check the low-pass result in one step. The series branch 50 + j86.6 ohm is equivalent to a parallel resistance of 50 × (1 + Q²) = 200 ohm in parallel with a reactance of 86.6 × (1 + 1/Q²) = 115.5 ohm. The shunt capacitor cancels that parallel reactance and leaves 200 ohm.
With Q = 1.73, the rule of thumb gives a 3 dB bandwidth of about 1 GHz / 1.73 ≈ 577 MHz. Treat that as an estimate of the match width, not a specification.
Pi and T networks: three parts, Q chosen
A Pi network has two shunt parts and one series part. A T network has two series parts and one shunt part. With three parts the designer has one extra degree of freedom, and it is used to choose the loaded Q.
The textbook way to see it is as two L-networks back to back, sharing a virtual resistance in the middle. For a Pi network the virtual resistance is lower than both ends. For a T network it is higher than both ends. The two ends see a higher Q than the direct ratio requires, because the middle resistance is more extreme than either end.
The practical results:
- The Q cannot go below the value of the plain L-network for the same two resistances.
- Raising Q narrows the response. For the 50 ohm to 200 ohm case, doubling Q from 1.73 to 3.46 would cut the rule-of-thumb bandwidth from about 577 MHz to about 289 MHz.
- A narrower match rejects more out-of-band energy, such as harmonics. It is also more sensitive to part tolerance and loss.
Pi networks are often chosen with shunt capacitors at the ends, which gives a low-pass response with good harmonic rejection. T networks with series capacitors at the ends block DC at both ports.
The quarter-wave transformer
A quarter-wave transformer matches two real resistances with a line a quarter wavelength long at the design frequency. Its impedance is the geometric mean:
For 50 ohm to 200 ohm that is sqrt(50 × 200) = 100 ohm. It uses no lumped parts, so it avoids inductor loss. The cost is length: the line must be a quarter wavelength, and the match is correct only near that frequency. Like the others, its bandwidth is limited, and a larger resistance ratio narrows it.
Which one to pick
- L-network: the default. Fewest parts, lowest loss, and the Q is whatever the ratio demands. Pick low-pass or high-pass by what you need to pass or block.
- Pi or T network: when you need a higher Q than the L-network gives, usually for selectivity or harmonic rejection, and you accept a narrower match.
- Quarter-wave transformer: when a line of the right impedance and length is practical, mostly on boards or structures where distributed elements are already in use.
Real inductors and capacitors have loss and parasitic effects. A network designed with ideal values has to be tuned with the real parts. Measure the result on a network analyzer and adjust. Higher-Q networks need the most care, because a narrow match moves further when a part value shifts.
In plain words · for students and beginners
Matching networks in plain words
An everyday analogy
Think of a gearbox between an engine and a wheel. The engine and the wheel want different speeds, so the gears change one into the other. A matching network does the same for the "size" of the load, which is its resistance. Two gears (an L-network) give you one fixed ratio. A longer gear train (Pi or T) lets you pick how sharply the whole thing is tuned.
Where the analogy breaks down: a gearbox changes torque and speed, but a matching network works only well near one frequency. Move away from it and the match gets worse.
The same example in simple steps
We want to connect a 50 ohm source to a 200 ohm load at 1 GHz.
- Divide the larger by the smaller: 200 / 50 = 4.
- Subtract 1 and take the square root: sqrt(3) = 1.73. This is Q, a number that says how sharp the match is.
- Use it to find the two reactances: 86.6 ohm in series and 115.5 ohm in shunt.
- Turn them into parts at 1 GHz: a 13.78 nH inductor in series and a 1.378 pF capacitor in shunt.
- Or choose the mirror version: a 1.838 pF capacitor in series and an 18.38 nH inductor in shunt. It blocks DC, while the first version passes it.
The chart below shows the same journey as a path of impedance values.
Mistakes beginners make
- Expecting any match to work at every frequency. Q says how wide the match is: roughly the center frequency divided by the bandwidth.
- Forgetting that the high-pass version blocks DC. If a bias current must flow, use the low-pass version.
- Trusting the calculated values blindly. Real parts have loss and extra small effects, so the final values come from tuning on the bench.
- Using a Pi or T network when an L-network is enough. More parts mean more loss and more to tune.
- Mixing up the quarter-wave line impedance with the load. For 50 ohm to 200 ohm it is 100 ohm, the geometric mean, not the average.
Remember
- An L-network has two parts and a fixed Q: Q = sqrt(R_high / R_low - 1), which is 1.73 for 50 ohm to 200 ohm.
- Pi and T networks use three parts so you can choose Q: higher Q means a narrower match.
- A quarter-wave transformer for 50 ohm to 200 ohm uses a 100 ohm line, and ideal part values always need tuning with real parts.
Related pages on RF View
Frequently asked questions
When should I use a Pi or T network instead of an L-network?
Use a Pi or T network when you need a different loaded Q than the L-network gives, usually a higher Q for a narrower response. The third part lets you choose Q, and therefore the bandwidth.
What is the loaded Q of an L-network matching 50 ohm to 200 ohm?
Q = sqrt(200 / 50 - 1) = sqrt(3) = 1.73. The ratio fixes it, so you cannot trade it for bandwidth with only two parts.
What impedance does a quarter-wave transformer need to match 50 ohm to 200 ohm?
The line impedance is sqrt(R1 * R2) = sqrt(50 * 200) = 100 ohm, and the line is a quarter wavelength long at the design frequency.
Why do matching networks need tuning on the bench?
Real inductors and capacitors have loss and parasitic effects. A network calculated with ideal values has to be tuned using the real parts.