What is the role of silicon CMOS in millimeter wave applications and when can it replace III-V technologies?
Si CMOS at mmWave
The CMOS revolution at mmWave is driven by economics: the superior performance of III-V is traded for the massively lower cost and higher integration of CMOS in high-volume applications.
- 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
- Interface compatibility: verify impedance, connector type, and mechanical form factor match the system architecture
Frequently Asked Questions
Will CMOS completely replace III-V?
No. CMOS will dominate in high-volume, cost-sensitive applications where integration is paramount: consumer 5G, automotive radar, and Wi-Fi. III-V will remain essential for: high-power applications (> 5 W): GaN is the only technology with adequate power density. CMOS cannot efficiently generate > 1 W per die at mmWave. Ultra-low noise: GaAs and InP will maintain their NF advantage (fundamental material properties, not a scaling issue). Frequencies above 200 GHz: InP is the only technology with useful gain. Military and space: where absolute performance matters more than cost. The future is heterogeneous integration (chiplet approach): CMOS for digital and beamforming control, SiGe for the transceiver, and GaN/GaAs for the PA and LNA. These die are co-packaged in an advanced SiP (system-in-package).
How does CMOS PA efficiency compare to GaAs?
At 28 GHz: CMOS PA (28 nm): PAE = 10-18% at P_sat. SiGe PA: PAE = 15-25% at P_sat. GaAs HBT PA: PAE = 25-35% at P_sat. GaN HEMT PA: PAE = 25-40% at P_sat. The gap: CMOS PAE is approximately half that of GaAs/GaN. For a 5G UE with 4 PA elements at +12 dBm each: CMOS PA total DC power: 4 × 16mW / 0.15 = 427 mW. GaAs PA total DC power: 4 × 16mW / 0.30 = 213 mW. The CMOS PA uses 2× more DC power for the same RF output. This means: shorter battery life, more heat, and potentially thermal throttling. The trade-off: CMOS saves $2-$5 per die (vs a separate GaAs PA die). The cost saving is worth it at volumes > 10 million units/year (the total cost saving >> the additional battery/thermal cost).
What about RF-SOI for mmWave?
RF-SOI (Radio Frequency Silicon-on-Insulator): a technology where MOSFET transistors are fabricated on a thin silicon layer above a buried oxide (BOX). Advantages for mmWave switches: the BOX isolates the transistor from the lossy Si substrate, enabling high-Q passives and excellent switch isolation, and low parasitic capacitance for fast switching. Performance: switch insertion loss < 1 dB at 28 GHz, isolation > 25 dB, and power handling > +30 dBm (using stacked-FET topology). RF-SOI is the dominant technology for mmWave switches and tunable components (tunable filters, antenna tuners). Limitations: the SOI transistor fT is typically lower than bulk CMOS at the same node (the thin body reduces the current drive). PA and LNA performance is inferior to SiGe and III-V. Best use: RF-SOI for switches + SiGe or III-V for PA/LNA = optimal system partitioning.