André Saraiva

We show four levels of integration between our silicon spin qubit systems and classical logic millisecond feedback by streaming circuits from a computer to an OPX1000 pulse processing unit (PPU) from Quantum Machines;
microsecond integration between an AI server (NVIDIA’s DGX Quantum) and the PPU;
nanosecond integration leveraging the FPGA processing in the PPU in real time;
quasi-instantaneous integration between charge and spin states to perform conditional logic on spins based on spin-to-charge conversion.

We use these to perform tasks with increasingly stringent requirements in latency. At millisecond scale, feedback systems based on calibration measurements over multiple shots are enabled. At the microsecond integration scale, single-shot feedback is enabled, as well as circuit adaptations based on long mid-circuit measurements. With nanosecond integration, shorter mid-circuit measurements are enabled, which better preserve the idling data qubits. Finally, the spin-to-charge conversion protocol used for readout generates a conditional frequency shift on neighbouring quantum dots. This enables a quantum Z gate conditional on the classical outcome of a measurement of spin parity without needing a sensor. We use these methods to perform complete quantum instrument tomography studies that includes the performance of the mid-circuit measurements and conditional operations.