How do I simulate the performance of a PCB via transition at millimeter wave frequencies?
mmWave Via Transition Simulation
Via transitions are one of the most critical structures in mmWave PCB design. A poorly designed via can add 1-3 dB of loss per transition at 60 GHz, severely degrading system performance. 3D electromagnetic simulation is essential for optimizing the via geometry.
Technical Considerations
The key parameters affecting mmWave via performance: (1) Via barrel diameter: smaller diameter reduces parasitic capacitance and inductance. Standard PCB vias: 200-300 μm. Laser-drilled microvias: 75-150 μm (better performance above 40 GHz but limited depth). (2) Anti-pad diameter: the clearance in the ground plane around the signal via. Controls the impedance of the via section: Z_via ≈ (60/sqrt(epsilon_r)) × ln(D_antipad/D_via). For Z_via = 50 ohms: D_antipad/D_via ≈ 2.3 for epsilon_r = 3.5. (3) Signal pad diameter: the landing pad creates a parasitic shunt capacitance (C_pad ≈ epsilon_0 × epsilon_r × pi × (D_pad^2 - D_via^2) / (4 × h_dielectric)). Minimizing pad size reduces this capacitance. For mmWave: use pad-less vias (via directly in the trace, no landing pad) when the fabrication process allows. (4) Ground via fence: ground vias surrounding the signal via create a coaxial environment that contains the fields and prevents radiation. Optimal spacing: 200-400 μm center-to-center, with at least 6-8 ground vias surrounding each signal via. Closer spacing improves isolation but increases PCB cost (more drill hits). (5) Back-drill: for non-functional via stubs (the portion of a through-via extending beyond the signal layer): the stub creates a resonance at f_resonance = c/(4×L_stub×sqrt(epsilon_r)). For a 1.5 mm stub: resonance at ~28 GHz. Back-drilling removes the stub, eliminating the resonance. Back-drill tolerance: ±100-150 μm, which must be included in the simulation model.
Performance Analysis
A well-optimized via transition at 60 GHz should achieve: S21 (insertion loss): -0.2 to -0.5 dB per transition. S11 (return loss): < -15 dB across 57-66 GHz (802.11ad/ay band). Common issues revealed by simulation: (1) Return loss peak at a specific frequency: indicates an impedance discontinuity (via section impedance does not match 50 ohms). Fix: adjust anti-pad diameter. (2) Sharp notch in S21: indicates a resonance from the via stub or a cavity mode in the ground via fence. Fix: back-drill the stub or add more ground vias. (3) Broadband excess loss: indicates radiation leakage through the ground via fence. Fix: add more ground vias, reduce spacing. (4) Mode conversion: energy transfers from the desired mode (e.g., microstrip quasi-TEM) to a parallel-plate mode between ground planes. Fix: add more ground stitching vias around the signal via to block the parallel-plate mode.
- Performance verification: confirm specifications against the application requirements before finalizing the design
- Environmental factors: temperature range, humidity, and vibration affect long-term reliability and parameter drift
- Cost vs. performance: evaluate whether the application demands premium components or standard commercial grades
Design Guidelines
Validate simulated via performance against measurement using a back-to-back test structure: fabricate two via transitions connected by a short transmission line (create a microstrip-to-stripline-to-microstrip back-to-back transition). Measure S-parameters with a calibrated VNA using GSSG wafer probes or end-launch connectors. De-embed the connectors/probes and connecting lines. Divide the total insertion loss by 2 to get the per-transition loss. The simulated and measured results should agree within ±0.2 dB for insertion loss and ±3 dB for return loss up to 60 GHz, assuming accurate material properties and fabricated dimensions are used in the simulation model.
Frequently Asked Questions
When should I use a microvia vs a through-via?
Microvias (laser-drilled, 75-150 μm diameter, depth limited to 1-2 layers): use when operating above 40 GHz, or when layer transitions need minimum parasitic capacitance and inductance. Microvias have no stub (they only span 1-2 layers), eliminating stub resonance. Through-vias (mechanically drilled, 200-350 μm diameter, full board depth): use for layer transitions below 30 GHz, or when connecting layers far apart in the stackup. Back-drilling is required for any through-via at mmWave frequencies to remove the non-functional stub. Some mmWave designs use stacked microvias (V1 stacked on V2) to span multiple layers without the stub problem of through-vias. This requires a sequential lamination process (higher PCB cost, +30-50% vs standard process).
How many ground vias do I need around a signal via?
Minimum: 4 ground vias (one on each side), forming a square coaxial arrangement. Recommended: 6-8 ground vias, placed at equal angular spacing around the signal via at a distance of 200-400 μm (center-to-center). At 60 GHz: the ground via spacing along the signal propagation direction should be < lambda/10 = 2.9 mm / 10 ≈ 290 μm. However, the circumferential spacing (around the signal via) should also be < lambda/10 to prevent radiation leakage. For a ring of 8 ground vias at 400 μm radius: circumferential spacing ≈ 2×pi×400/8 = 314 μm, which is adequate at 60 GHz. For frequencies above 80 GHz: use 10-12 ground vias or reduce the ring diameter to maintain spacing < 200 μm.
What PCB material should I use for mmWave via transitions?
Low-loss, tightly-controlled-thickness materials: Rogers 3003 (εr=3.0, tanδ=0.0013): excellent for mmWave. Well-characterized to 77 GHz. Low moisture absorption. Rogers RO4835T (εr=3.48, tanδ=0.0037): lower cost than 3003, adequate loss for short via transitions. Available in thin cores (2-4 mil) for compact stackups. Megtron 7 (εr=3.37, tanδ=0.002): Panasonic material, widely available, good mmWave performance. Common in automotive radar (77 GHz) PCBs. Isola Astra MT77 (εr=3.0, tanδ=0.0017): designed for mmWave applications, competitive with Rogers. Material selection affects via impedance (through εr) and loss (through tanδ). At 60 GHz: of 0.3 dB total via transition loss, approximately 0.1 dB is conductor loss and 0.05 dB is dielectric loss, with 0.15 dB from radiation and mismatch. Choosing a lower-tanδ material saves only 0.02-0.03 dB per via; the dominant improvement comes from optimizing the via geometry.