Fast and high-fidelity readout of transmon qubits in scalable QPU architecture

The speed and fidelity of single- and two-qubit gates of superconducting quantum processing units (QPU) have been the subject of optimization but most multi qubit devices only feature modest readout performance in comparison. The high-fidelity dispersive readout with low readout crosstalk requires individual Purcell filters for readout of each qubit. However, the Purcell filters interfere with each other and are sensitive to the QPU package impedance due to the strong coupling to the feedline. These effects make it hard to achieve targeted detuning between the readout resonators and Purcell filters. We present a microwave circuit with reduced sensitivity to loading, and a computationally efficient numerical optimization method for the readout structure parameters. Our approach is validated by a collection of measured readout circuit spectrums with targeted properties and a state-of-the-art qubit readout fidelity.