Return loss measures impedance match quality at the transformer port. Here's the complete technical reference — formula, VNA method, physical causes, and PCB layout contribution.
Definition and Formula
pythonimport math

def reflection_coefficient(Z_load, Z_source=100):
"""
Z_load: impedance presented by transformer at measurement port (Ω)
Z_source: system characteristic impedance (100Ω for Ethernet)
Returns: reflection coefficient magnitude |Γ|
"""
gamma = abs((Z_load - Z_source) / (Z_load + Z_source))
return gamma

def return_loss_dB(gamma):
"""Convert reflection coefficient to return loss in dB"""
if gamma == 0:
return float('inf')
return -20 * math.log10(gamma)

def vswr(gamma):
"""Convert reflection coefficient to VSWR"""
return (1 + gamma) / (1 - gamma)

Reference table

print(f"{'RL (dB)':>10} | {'|Γ|':>8} | {'VSWR':>8} | {'Refl. Power':>12}")
print("-" * 50)
for rl in [10, 16, 20, 26, 30]:
g = 10 ** (-rl / 20)
print(f"{rl:>10} | {g:>8.4f} | {vswr(g):>8.3f} | {g**2*100:>10.2f}%")

Output:

RL (dB) | |Γ| | VSWR | Refl. Power

────────────────────────────────────────────

10 | 0.3162 | 1.925 | 10.00%

16 | 0.1585 | 1.377 | 2.51% ← IEEE 802.3 minimum

20 | 0.1000 | 1.222 | 1.00%

26 | 0.0501 | 1.106 | 0.25%

30 | 0.0316 | 1.065 | 0.10%

IEEE 802.3 Return Loss Requirements
StandardMin. Return LossFrequency Range100BASE-TX> 16 dB1MHz – 100MHz1000BASE-T> 16 dB (per pair)1MHz – 100MHz
Typical quality parts: 18–25 dB across passband.
VNA S11 Measurement Setup
VNA
Port 1
│
┌─────▼──────┐
│ Transformer │
│ (Primary) │
│ │
│ (Secondary)├──[100Ω]── GND
└─────────────┘

Step 1: Calibrate VNA at Port 1 reference plane (SOLT or TRL cal)
Step 2: Connect Port 1 to transformer primary with matched fixture
Step 3: Terminate secondary with 100Ω ±1% to chassis GND
Step 4: Sweep 100kHz to 200MHz, read S11 magnitude in dB
Step 5: Repeat with ports swapped (Port 1 to secondary, 100Ω on primary)

S11 reading = Return Loss at each frequency

⚠️ VNA calibration quality directly determines measurement accuracy
Uncalibrated VNA: results are meaningless for return loss
Poor fixture: adds 3–5 dB of apparent return loss degradation
Physical Causes of Degraded Return Loss
Cause Frequency Range Impedance Effect
──────────────────────────────────────────────────────────────────
Magnetizing inductance < 1 MHz Low ZL at low freq
Turns ratio deviation All frequencies Scales impedance by N²
Leakage inductance > 30 MHz Adds series Z
Inter-winding cap > 100 MHz Reduces parallel Z
PCB trace impedance All frequencies Adds in series with transformer

Turns ratio deviation example:
Ideal: N = 1.000 → Z_sec = 100Ω → RL = ∞
Error: N = 1.010 → Z_sec = 1.01² × 100 = 102Ω
Γ = (102-100)/(102+100) = 0.0099
RL = -20 × log10(0.0099) = 40.1 dB ← 1% turns error → 40dB RL
(Turns ratio deviation is rarely the dominant cause at high frequencies)
PCB Layout Contribution to System Return Loss
System RL = f(transformer RL, trace impedance, via parasitics, pad geometry)

Typical PCB contributions (100MHz):
50mm of 85Ω trace (vs 100Ω target): adds ~3–5 dB degradation
2 vias in signal path (1pF each): adds ~1–2 dB at 100MHz
Mismatched pad geometry: adds ~0.5–2 dB

Example:
Transformer intrinsic RL: 22 dB
PCB layout contribution: -5 dB
Measured system RL: 17 dB (still above 16 dB, but marginal)

→ Root cause of many "the part seems bad" return loss failures is PCB layout
Return Loss vs Insertion Loss Comparison
Parameter S-param Measures Limit (802.3)
──────────────────────────────────────────────────────────────
Return loss S11 Reflected at input > 16 dB
Insertion loss S21 Transmitted to output < 1.0 dB

Independence: IL and RL are NOT complementary
Can have: good S21 (low IL) + poor S11 (low RL)
Both must independently meet specification
Source
Voohu Technology (www.voohuele.com) — network transformers with full S11/S21 characterization. MOQ 50pcs, DHL 3–5 days.