Power Waves a and b: The Definition Behind S-Parameters

Power waves are a = (V + Z0·I)/(2√Z0) and b = (V − Z0·I)/(2√Z0). With peak phasors, net power into a port is ½(|a|² − |b|²), so 10 mW incident at |Γ| = 0.5 leaves 7.5 mW delivered.

Engineer’s note: The incident wave a and the reflected or transmitted wave b map directly onto the directional couplers inside a VNA. Once you can picture a and b as physical flows at each port, multi-port systems and coupling between antennas become far easier to reason about and debug.

For practitioners

S-parameters are ratios, but ratios of what? The answer is the power waves a and b. Every S-parameter is a ratio of one wave to another, so if a and b are vague in your head, S11 and S21 are vague too. This page defines the two waves from quantities you can see at a port: voltage, current and a reference impedance. It then shows why the difference of their squared magnitudes is the net power into the port.

The definition from V, I and Z0

Pick a reference plane at a port. Call the total voltage there V and the current flowing into the port I. Both are complex phasors. Choose a real reference impedance Z0, usually 50 Ω. The incident and reflected waves are:

a=V+Z0I2Z0a = \frac{V + Z_0 I}{2\sqrt{Z_0}}
b=V−Z0I2Z0b = \frac{V - Z_0 I}{2\sqrt{Z_0}}

This is a change of variables and nothing more. You can invert it at once:

V=Z0 (a+b),I=a−bZ0V = \sqrt{Z_0}\,(a + b), \qquad I = \frac{a - b}{\sqrt{Z_0}}

So the pair (a, b) holds exactly the same information as the pair (V, I). What changes is the point of view. V and I describe the total state at the port. The waves split that state into the part travelling toward the device and the part coming back. Dividing by √Z0 gives both waves units of √W, so squaring a magnitude gives power with no impedance left over.

The ratio of the two waves at a load is the reflection coefficient. If the port sees an impedance Z = V/I, then:

Γ=ba=Z−Z0Z+Z0\Gamma = \frac{b}{a} = \frac{Z - Z_0}{Z + Z_0}

When Z = Z0, the term V − Z0·I is zero, so b = 0. A matched port reflects nothing, and that is the whole point of choosing Z0.

On a uniform line whose characteristic impedance equals Z0, the total voltage is the sum of a forward wave V+ and a reverse wave V−, and the current is (V+ − V−)/Z0. Put these into the definitions and the cross terms cancel:

a=V+Z0,b=V−Z0a = \frac{V^{+}}{\sqrt{Z_0}}, \qquad b = \frac{V^{-}}{\sqrt{Z_0}}

This is the picture most engineers carry around: a is the forward wave, b is the reverse wave, each scaled by √Z0. The V, I definition is more general, though. It works at a reference plane on a pad, a package lead or a bond wire where no line with impedance Z0 exists. The reference impedance is something you choose. It does not have to be a property of the hardware. Everything here assumes a real Z0. Complex reference impedances need a modified definition, which this page does not cover.

Why |a|² − |b|² is net power

Average power into the port, with peak phasors, is P = ½·Re(V·I*). Now expand the wave magnitudes:

|V + Z0·I|² − |V − Z0·I|² = 4·Z0·Re(V·I*)

Divide by 4·Z0 and you get |a|² − |b|² = Re(V·I*). So:

P=12(∣a∣2−∣b∣2)P = \tfrac{1}{2}\left(|a|^2 - |b|^2\right)

The cross terms between incident and reflected waves cancel exactly. With a real Z0, power splits cleanly into an incident part ½|a|² and a reflected part ½|b|², and the net power is their difference. No interference term is left over. This is why the waves are called power waves.

If your phasors are RMS values, drop the ½: P = |a|² − |b|². Both conventions are common in tools and textbooks. Mixing them gives an error of exactly a factor of 2, which is 3 dB.

Worked example: 50 Ω, V+ = 1 V peak, |Γ| = 0.5

SourceZ0 = 50 Ω linereference planeI+V−incident wavea = (V + Z0·I)/(2√Z0)10 mW (V+ = 1 V peak)reflected waveb = (V − Z0·I)/(2√Z0)2.5 mW (|Γ| = 0.5)Device under testport 1net power½(|a|² − |b|²)= 7.5 mWS-parameters:b1 = S11·a1 + S12·a2
Power waves a and b at one port on a 50 Ω line, showing 10 mW incident, 2.5 mW reflected and 7.5 mW net power delivered to the device. Concept drawing.

Take a 50 Ω line driving one port of a device, with V+ = 1 V peak.

  • Incident wave: a = 1/√50 ≈ 0.1414 √W. Then |a|² = 0.02 W, so incident power is ½·0.02 = 0.01 W = 10 mW. This matches |V+|²/(2·Z0) = 1/100 W.
  • The port has |Γ| = 0.5, so |b| = 0.5·|a| ≈ 0.0707 √W and |b|² = 0.005 W. Reflected power is ½·0.005 W = 2.5 mW, which is the same as |Γ|² × 10 mW.
  • Net power: ½(|a|² − |b|²) = 10 − 2.5 = 7.5 mW.

Now check the result against V and I. For simplicity take Γ = +0.5 (real). Then V− = 0.5 V, the total voltage is V = 1.5 V, and the current is I = (1 − 0.5)/50 = 10 mA. Power from V and I is ½·1.5·0.01 = 7.5 mW, which agrees. The load impedance is 1.5/0.01 = 150 Ω, and (150 − 50)/(150 + 50) = 0.5 confirms the reflection coefficient. The V, I view and the wave view give the same answer, as they must.

A large reflected wave does not by itself mean the device absorbs little power. Net power depends on |a|² − |b|², so always state the incident level. At |Γ| = 0.5, a quarter of the incident power comes back and three quarters is delivered.

From waves to S-parameters

For a multiport, collect the incident waves into a vector a and the reflected waves into a vector b. The scattering matrix is defined by:

b=S a\mathbf{b} = \mathbf{S}\,\mathbf{a}

For two ports, this means b1 = S11·a1 + S12·a2 and b2 = S21·a1 + S22·a2. Each S-parameter is measured with the other port terminated in Z0, so that its incident wave is zero. For example, S11 = b1/a1 with a2 = 0, and S21 = b2/a1 with a2 = 0.

Because the waves are power-normalized, |S21|² is the power delivered to a Z0 load divided by the power incident from a Z0 source. That is the transducer gain when both terminations equal Z0. In the same way, 1 − |S11|² is the fraction of incident power that enters port 1. A lossless network must therefore have a unitary S-matrix, and a passive one can never have |S11| greater than 1. All of these properties follow directly from the |a|² − |b|² result above. For the meaning of each term, see S-Parameters Explained: S11, S21, S12 and S22. For other gain definitions, see Transducer, Available and Operating Power Gain Compared.

In plain words · for students and beginners

Power waves in plain words

An everyday picture

Think of waves rolling into a harbour wall. Some of the wave energy is soaked up by the wall and some bounces back out to sea. The wave going in is "a", and the wave coming back is "b". The energy the wall actually takes is what goes in minus what comes back.

Where the picture breaks down: on a cable the waves also have phase, meaning their timing relative to each other. The two waves add up to a voltage you can measure, but their powers simply subtract. That clean subtraction only works because of how a and b are scaled. The scaling factor is the reference impedance Z0, the "normal" impedance of the system, usually 50 Ω.

The same example, step by step

00.20.40.60.810246810reflection coefficient magnitude |Γ|power (mW)|Γ| = 0.57.5 mW delivered2.5 mW reflectedincident 10 mWreflected =10·|Γ|² mWdelivered =10·(1 − |Γ|²) mW
Incident, reflected and delivered power against |Γ| for 10 mW incident power (V+ = 1 V peak, Z0 = 50 Ω). Computed from the formula.
  1. A 50 Ω cable carries a forward wave of 1 V peak. Its power is 1² ÷ (2 × 50) = 0.01 W, which is 10 mW. That is the incident power.
  2. The device at the end is not a perfect match. Its reflection coefficient Γ, the fraction of the wave's voltage that bounces back, has size 0.5.
  3. Power goes with voltage squared, so the reflected power is 0.5 × 0.5 × 10 mW = 2.5 mW.
  4. The device keeps the rest: 10 − 2.5 = 7.5 mW.

The plot shows this for every value of |Γ|. At 0 nothing comes back and all 10 mW is delivered. At 1 everything comes back and nothing is delivered. At 0.5 you get 2.5 mW back and 7.5 mW delivered.

Mistakes beginners make

  • Using Γ directly for power. Γ = 0.5 does not mean half the power comes back. It is a quarter, because power follows Γ squared.
  • Forgetting the ½. With peak values, power is ½|a|². With RMS values, it is just |a|². Mixing the two makes the answer off by a factor of 2.
  • Thinking b = 0 means no power flows. It means all the incident power is absorbed. That is the best case, not the worst.
  • Treating S-parameters as something separate. S-parameters are just the rules that turn the incoming waves into the outgoing waves, for example b1 = S11·a1 + S12·a2.

Remember

  • a goes into the port and b comes out. Both are built from the voltage, the current and Z0.
  • Net power in = ½(|a|² − |b|²) with peak values, or |a|² − |b|² with RMS values.
  • 10 mW incident at |Γ| = 0.5 gives 2.5 mW reflected and 7.5 mW delivered.

Frequently asked questions

What are the units of the power waves a and b?

They have units of square root of watts (√W). Squaring their magnitude gives power directly. You need a factor of ½ if you use peak phasors and no factor if you use RMS phasors.

How is net power into a port calculated from a and b?

With peak phasors, net power is P = ½(|a|² − |b|²). This is the incident power minus the reflected power, and it equals ½·Re(V·I*) at the reference plane.

Are a and b the same as the forward and reverse voltage waves?

They are scaled versions of them when the reference impedance equals the line's characteristic impedance: a = V+/√Z0 and b = V−/√Z0. The definition from V, I and Z0 works at any reference plane, even where no physical line exists.

How do S-parameters relate to a and b?

S-parameters are the matrix that maps incident waves to reflected waves, b = S·a. At port 1 that gives b1 = S11·a1 + S12·a2.

References and further reading