How do I design a microwave switch network for routing signals to multiple qubits?
Microwave Switch Networks for Quantum Systems
Switch networks can reduce the cost and complexity of quantum control electronics by enabling time-multiplexed sharing of expensive signal generators across multiple qubits. The trade-off is reduced parallelism: qubits sharing a signal source through a switch cannot be controlled simultaneously.
Frequently Asked Questions
Do production quantum computers use switch networks?
Current production systems (IBM, Google, Amazon Braket) use fully dedicated per-qubit signal chains without switching for maximum parallelism and simplicity. Each qubit has its own DAC channels, signal generation, and cable to the cryostat. For a 100-qubit system: approximately 200-300 DAC channels, 200-300 coaxial cables, and dedicated amplifiers. The cost is $100,000-500,000 for control electronics alone. Switch networks are used only for calibration and characterization routing. As qubit counts grow to 1000+ and beyond, the scaling of dedicated per-qubit electronics becomes a significant cost and complexity barrier, driving research into multiplexed architectures.
What is the maximum switch network size practical for quantum computing?
At room temperature: commercial RF switch matrices up to 64×64 (Keysight, National Instruments) with total insertion loss of 3-6 dB and switching speed of 10-100 μs. Custom switch trees using SP4T switches can scale to 256+ outputs with 5-10 dB total loss. At 4K: limited by cooling power. Each GaAs switch dissipates ~0.5 mW; a 64-switch tree dissipates ~32 mW (2% of 1.5W budget), feasible. At mixing chamber: limited to superconducting switches with zero static dissipation. Current research demonstrations: 4-8 port superconducting switch networks. Scaling to 100+ ports is a major engineering target for the next generation of quantum computers.
How does switch insertion loss affect the qubit?
Switch insertion loss between the DAC and the cryostat input is compensated by increasing DAC output power. However, switch loss at cryogenic stages (below the attenuators) directly reduces the signal reaching the qubit, requiring recalibration of the pi-pulse amplitude. More importantly, if the switch loss changes with temperature, state, or time (aging), it causes amplitude drift that degrades gate fidelity until recalibrated. RF switches with <0.1 dB insertion loss variation over temperature and switching state (achievable with electromechanical and some MEMS switches) are strongly preferred.