arXiv:2610.00783
Real-Time Adaptive Filtering and the Boxcar Limit in Superconducting Qubit Readout
Hans Johnson, Tanay Roy, Leonardo Bove, David van Zanten, Silvia Zorzetti, Jafar Saniie
Abstract
Control and readout systems for superconducting quantum computing support real-time control and high-fidelity qubit-state discrimination. We examine when additional digital processing improves readout fidelity. A common baseline for dispersive readout uses a boxcar averager to compute a uniformly weighted average over a fixed window followed by threshold-based state assignment. We evaluated an adaptive finite-impulse-response (FIR) filter trained in real time using the least-mean-squares (LMS) algorithm on a field-programmable gate array (FPGA) quantum controller. We also analyzed measured readout shots offline to compare the frozen filter with boxcar averaging, and evaluate per-sample weighting plus sequential detection. For the fixed FIR filters tested, summing the filtered samples gives the boxcar sum multiplied by one complex number apart from a small edge correction. We call this the kernel collapse. The multiplication scales the state separation and noise equally without improving state discrimination, and we call this the boxcar limit. The frozen LMS filter reduces trace noise, but shows no resolved improvement in readout fidelity. On a 2D transmon, optimizing the readout window accounted for much of the improvement over full-window integration. We then examined how changing the decision method could reduce calibration effort and decision time. On a higher-fidelity 3D ancilla readout, our modeled template estimation reached the same fidelity target as direct averaging with 89.8% fewer labeled shots per state. Offline sequential replay assigned most shots before the acquisition window ended. These results distinguish waveform denoising from state-discrimination improvement, and motivate direct control of decision weights and assignment time in dispersive readout.
AI Summary
The paper asks when digital processing improves dispersive superconducting-qubit readout beyond boxcar integration, distinguishing waveform denoising from state-discrimination gains. It implements a 64-tap complex LMS adaptive FIR filter on a QICK FPGA controller and compares frozen-filter replay, optimized integration windows, matched filtering, model-based calibration, and sequential detection on measured traces. Its central architectural result is that a frozen, shift-invariant FIR followed by uniform accumulation produces $Y\approx W\sum_n x[n]$, where $W$ is the sum of the taps. With zero initial state, only the final $T-1$ input samples have different effective weights. Nonzero scalar rescaling changes signal separation and noise equally, leaving discrimination SNR unchanged. Coefficient readback, fixed-point modeling, and same-shot replay support this identity. Across four transmon operating points, the frozen filter gives no resolved fidelity improvement despite trace-noise reduction. Optimizing window placement captures much of the available gain; at an $8\,\mu\mathrm{s}$ window, matched filtering adds 0.33 percentage points of contrast fidelity beyond the tuned boxcar. On a higher-fidelity ancilla dataset, ring-up-model template fitting reaches within 0.50 percentage points of reference contrast fidelity using about 30 rather than 297 labeled shots per state. Offline sequential replay reduces mean decision time from 2.60 to $1.80\,\mu\mathrm{s}$ on the primary split, with a fidelity difference consistent with zero. However, one sensitivity split shows a resolved fidelity loss. Calibration savings depend on the allowed template family; sequential performance and feedback latency remain unvalidated in hardware. The boxcar limitation applies specifically to the frozen-FIR-plus-accumulator architecture, not general linear readout.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 16 pages, 12 figures, 2 tables, 27 equations