Abstract
Owing to the fragile nature and susceptibility to various forms of electrical noise, quantum technology, such as quantum dot qubits, requires an ultra-low noise and sub-Kelvin temperature operation. The classical and thermal noise reaching the device generated by the room temperature electronics is bound to create decoherence and loss of information. Noise, beyond a few kHz, can be effectively filtered by an array of low-pass filters operated at various temperature stages. Minimizing noise in the lower frequency bands presents a significant challenge, particularly in experiment-specific voltage sweeps, where the sweep duration is susceptible to low-frequency periodic noise or 1/f noise having long-term correlations. Although the active noise cancellation technique is effective in minimizing the short-term noise in audio frequency ranges, it is ineffective in the lower frequency ranges. This paper addresses this gap by introducing a predictive noise cancellation strategy for improved noise mitigation in short-term voltage sweeps. Our system consists of a home-built microvolt precision, dual digital-to-analog converter (DAC). The dual-DAC strategy allows control and compensation in the milli- and microvolt range with better noise isolation independently. We study the short-term noise profile of our system and construct the anti-noise signal using the Fast Fourier transform with computed phase shifts. After implementing our method, the Mean Absolute Error is reduced by 34.58%. We also compare our technique with the Long Short-Term Memory-based machine learning model and find that it is efficient only for much shorter sweep durations compared to the proposed method.