REAL-TIME CALIBRATION2026

Millisecond-Scale Calibration and Benchmarking of Superconducting Qubits

Malthe A. Marciniak, Rune T. Birke, Johann B. Severin, Fabrizio Berritta, Daniel Kjær, Filip Nilsson, Smitha N. Themadath, Sangeeth Kallatt, James L. Webb, Kristoffer Bentsen, Tonny Madsen, Zhenhai Sun, Svend Krøjer, Christopher W. Warren, Jacob Hastrup, Morten Kjaergaard

An on-FPGA workflow that brings pulse generation, acquisition, inference, optimization, and feed-forward together to calibrate and benchmark superconducting qubits on millisecond timescales.

Read on arXiv

Twelve-qubit device, continuous T1 tracking, and comparison of a 9.8 ms on-FPGA estimate with a roughly 250 ms CPU-offloaded measurement — Figure 1. Rapid parameter estimation on a superconducting quantum processor. (a) Tilted SEM micrograph of the 12-qubit flux-tunable transmon qubit processor. The specific qubit used in this work is highlighted in a white dashed box. (b) Continuous $T_1$ tracking using our on-FPGA control, with roughly 200 $T_1$ estimates performed in 2 seconds of wall-clock time. (c) A $T_1$ measurement (red points) measured with our on-FPGA loop performing both measurement and estimation within a closed-loop taking 9.8 ms of wallclock time. (d) A $T_1$ measurement of the qubit performed with a conventional offloading loop, taking roughly $\sim250\,\mathrm{ms}$ of wall-clock time, and dominated by data transfer, dense sampling and software overhead.Marciniak et al., arXiv:2602.11912 (2026). Original Figure 1 caption. Figure and caption reproduced under CC BY 4.0.

THE WORK

Calibration needs to keep pace with a changing qubit. This work places pulse generation, measurement, analysis, and parameter updates together on the FPGA controller, avoiding the repeated transfers to a host computer that slow conventional calibration loops.

The toolkit combines three-point decay estimation, sparse phase estimation, and on-FPGA optimization. These methods extract useful parameters from a small number of carefully chosen measurements. Nelder–Mead optimization adjusts readout settings, while golden-section search locates spectroscopic peaks. Together, the tools support readout, spectroscopy, pulse-amplitude calibration, coherence measurements, and randomized benchmarking.

A representative T₁ measurement and estimate take 9.8 ms, compared with roughly 250 ms for the illustrated CPU-offloading workflow. Crucially, the T₁ figure includes both acquisition and analysis. The study also demonstrates pulse-amplitude optimization in about 1 ms and Clifford randomized benchmarking in 107 ms. These timings describe the reported experiments; the paper separately examines how sparse sampling trades measurement time against estimation uncertainty.

The methods are then combined into a closed loop that recalibrates and benchmarks a superconducting qubit continuously. Over six hours, it performs more than 74,000 consecutive recalibrations and maintains lower gate errors than the initial-calibration baseline. Correlation analysis shows reduced sensitivity to control-parameter drift, while performance remains linked to the qubit's coherence.

Pelagic orchestrated the experiments, while the on-FPGA tools were implemented through QUA on a Quantum Machines OPX1000. The broader result is a calibration workflow that can run repeatedly as hardware changes. By combining sparse measurements with local analysis and immediate updates, it makes calibration part of the ongoing experiment and creates room to study dynamics that slower procedures can average away.

Explore the full methods, measurements, and discussion in the paper.

Read the full paper