T-Network Matching Through a Virtual Resistance
A T network matches each termination up to a virtual resistance R_v higher than both, with Q_i = sqrt(R_v/R_i − 1). For 50 to 200 ohm at Q = 5, R_v = 1300 ohm, the series arms are 250 and 469.0 ohm and the middle shunt is about 177.0 ohm.
Engineer’s note: When both the source and load are low impedance, as between transistor stages, a T often beats a Pi. Its series inductors come first, so the chip package's bond-wire inductance or a long PCB trace can be absorbed into the matching network as part of the design.
For practitioners
A two-element L-network has no free parameter. Once the source and load resistances are fixed, its Q is fixed too. If you need a higher Q, for more selectivity or more harmonic rejection, you need a third element. The T network gives you that third element. It works as two L-sections placed back to back, and they meet at a resistance you choose.
This page derives the T-network element values from that idea and works one example from start to finish.
Why a T network needs a virtual resistance
A T network has series elements at both ports and a shunt element in the middle. Cut the middle shunt element into two parallel parts and the T splits into two L-sections. Each one has its series element facing a port and its shunt element facing the middle node.
An L-section with the series element on the resistor side can only transform upward. A series reactance added to R gives R + jX. Converting that to the parallel form gives a parallel resistance of R(1 + Q^2), and that is always larger than R. So both halves of the T raise their terminations to a common resistance at the middle node. This is the virtual resistance R_v. It is not a physical component. It is the resistance you would see if you broke the circuit at the middle node and looked toward either port.
This gives the one hard rule of T design: R_v must be higher than both terminations. If R_v sits between them, one of the two sections would have to transform downward, and a series-first L-section can't do that.
The design equations
Each L-section matches its termination R_i to R_v, so each has its own Q:
The section on the lower-resistance side has the larger ratio, so it has the higher Q. That section sets the loaded Q of the whole network. The usual approach is to choose the design Q for that side and solve for R_v:
Then the reactances follow directly. The series element at each port is
and the shunt reactance each section needs at the middle node is
The two shunt reactances sit at the same node, so they combine in parallel into one physical element:
This parallel formula holds when both shunt reactances are the same type, both capacitive or both inductive.
There is also a lower limit on Q. R_v must exceed R_high, so the design Q must exceed sqrt(R_high/R_low − 1). Below that value you get a single L-section, or no real solution at all.
Worked example: 50 ohm to 200 ohm at Q = 5
Match a 50 ohm source to a 200 ohm load with a design Q of 5, set on the 50 ohm side. The minimum usable Q here is sqrt(200/50 − 1) = sqrt(3) ≈ 1.732, so Q = 5 is valid.
Step 1: virtual resistance. R_v = 50 × (1 + 25) = 50 × 26 = 1300 ohm. That is above both 50 ohm and 200 ohm, as it has to be.
Step 2: Q of the second section. Q_2 = sqrt(1300/200 − 1) = sqrt(6.5 − 1) = sqrt(5.5) = 2.345.
Step 3: series reactances.
- Port 1: X_s1 = 5 × 50 = 250 ohm
- Port 2: X_s2 = 2.345 × 200 = 469.0 ohm
Step 4: shunt reactances at the middle node.
- Section 1: X_p1 = 1300/5 = 260 ohm
- Section 2: X_p2 = 1300/sqrt(5.5) = 554.3 ohm
Step 5: combine the shunts. X_mid = (260 × 554.3)/(260 + 554.3) ≈ 176.99 ohm, or about 177.0 ohm when both are the same type.
You can check section 1 directly. 50 + j250 has Q = 5. In parallel form that is 50 × 26 = 1300 ohm resistive in parallel with 250 × (1 + 1/25) = 260 ohm reactive. A shunt reactance of the opposite sign and 260 ohm magnitude cancels it, which leaves 1300 ohm at the node.
Choosing element types
The derivation only fixes magnitudes. You still pick the signs. A series inductor with a shunt capacitor gives a low-pass section. A series capacitor with a shunt inductor gives a high-pass section. Using two low-pass sections gives series inductors of 250 ohm and 469.0 ohm, plus one shunt capacitor of 177.0 ohm. That is a low-pass T that also attenuates harmonics.
If you mix the two section types, one shunt is inductive and the other capacitive, and they partly cancel. Then the middle element is the parallel combination of +j260 and −j554.3, about 489.7 ohm inductive. That is a different part with a different value. Decide the topology before you combine anything.
To convert to component values at an operating frequency f, use L = X/(2πf) for inductors and C = 1/(2πfX) for capacitors.
Loaded Q and what it costs
The network's loaded Q is about the larger of the two section Qs, here 5. The bandwidth is roughly f0/Q, so a higher Q gives a narrower match. A single L-section for the same 50 to 200 ohm ratio would have Q ≈ 1.732. The T trades bandwidth for selectivity and a free design parameter.
The cost is loss. Circulating reactive current grows with Q, and every component's own resistance dissipates part of that current. A Q of 5 is moderate. Pushing the design Q much higher with real inductors makes insertion loss climb quickly. Pick the lowest Q that meets your filtering or bandwidth target.
In plain words · for students and beginners
T-network matching in plain words
An everyday picture
Imagine you have to move a box from a low shelf (50 ohm) to a middle shelf (200 ohm). A T network doesn't go straight across. It lifts the box up to a high shelf first (the virtual resistance, 1300 ohm) and then brings it down to the middle shelf. A pi network does the opposite. It lowers the box to the floor first and then lifts it up.
Why take the detour? The height of the detour is something you choose. That choice sets how "sharp" the network is: how narrow a band of frequencies it matches and how strongly it rejects other frequencies. That sharpness is called Q (quality factor, the ratio of stored energy to energy passed on each cycle).
The picture breaks down in one place. Lifting a real box costs energy. With ideal parts, the detour costs no power at all. With real parts, a bigger detour means more current circulating through them, so a little more power gets lost as heat.
The same example, step by step
- Pick the sharpness. Choose Q = 5 on the 50 ohm side.
- Find the high shelf. R_v = 50 × (1 + 5 × 5) = 50 × 26 = 1300 ohm. It's above both 50 and 200, so the detour works.
- Find the other side's Q. From 200 up to 1300: Q = sqrt(1300/200 − 1) = sqrt(5.5) ≈ 2.345.
- Series parts at the ports. Multiply Q by the port resistance: 5 × 50 = 250 ohm and 2.345 × 200 = 469.0 ohm. (Reactance is the ohms-like opposition a coil or capacitor gives to AC.)
- Shunt parts in the middle. Divide 1300 by each Q: 260 ohm and 554.3 ohm. Both connect to the same point, so they act in parallel and combine into one part of about 177.0 ohm.
The pi network for the same job goes down to 7.69 ohm instead. It uses the same two Q values, just on opposite sides.
Mistakes beginners make
- Choosing a virtual resistance between the two terminations. Something like 100 ohm can't work, because the 200 ohm side would have to step down and a T half can only step up.
- Adding the shunt reactances like resistors in series. They share a node, so you combine them in parallel.
- Combining a coil with a capacitor as if they were alike. Opposite types partly cancel, and the result is a completely different value.
- Assuming a higher Q is always better. It narrows the bandwidth and raises the loss in real parts.
Remember
- A T network climbs from both ports to a virtual resistance higher than both terminations. A pi network drops to one lower than both.
- Each half has Q = sqrt(R_v/R_i − 1). Set the design Q on the low side: R_v = R_low(1 + Q^2).
- The two middle shunt reactances share a node and combine in parallel, about 177.0 ohm for this example when both are the same type.
Related pages on RF View
Frequently asked questions
How do I choose the virtual resistance for a T matching network?
Pick the design Q on the lower-resistance side and compute R_v = R_low(1 + Q^2). R_v must come out higher than both terminations, otherwise the second L-section has no real solution.
Why does the T network virtual resistance have to be higher than both terminations?
Each half of the T is an L-section with its series element at the port, and that arrangement can only transform a resistance upward. Both terminations therefore have to meet at a resistance above both of them.
How are the two shunt reactances combined in a T network?
They both connect to the same virtual-resistance node, so they combine in parallel. For the 50 to 200 ohm example at Q = 5, 260 ohm and 554.3 ohm of the same type give about 177.0 ohm.
What is the difference between a T and a pi matching network?
A T uses series elements at both ports and a shunt element in the middle, and it steps up to a virtual resistance above both terminations. A pi uses shunt elements at both ports and a series element in the middle, and it steps down to a virtual resistance below both.