Gaku Imamura | Searcharxiv

arXiv2026

Maximum signal-to-noise ratio enhancement by averaging under a limited measurement time

Averaging through repetitive measurement is a ubiquitous strategy for improving signal-to-noise ratio (SNR) and is commonly assumed to yield a $\sqrt{N}$ enhancement with the number of repetitions $N$. This assumption, however, implicitly requires the signal amplitude to be independent of measurement duration. This condition does not generally hold in dynamical sensing systems with finite response time and a fixed measurement time. We derive a closed-form expression for the SNR enhancement factor by analytically accounting for the competition between statistical noise reduction and dynamical signal attenuation, and demonstrate the existence of a strict upper bound on the SNR enhancement. The enhancement factor is a non-monotonic function of $N$ with a well-defined maximum at an optimal repetition number, beyond which further averaging degrades the SNR. Moreover, below a threshold set by the ratio of measurement time to response time, averaging yields no enhancement at all. These two regimes delimit where the conventional $\sqrt{N}$ law breaks down. Experimental validation using nanomechanical gas sensing, with two receptor-analyte systems deliberately chosen to bracket this enhancement transition, confirms the theoretical predictions. Our results show that measurement time is a finite resource to be optimally partitioned between signal accumulation and averaging, and provide a quantitative guideline for selecting the repetition number in time-constrained sensing such as real-time and repetitive gas or odor detection.

physics.app-ph↗

arXiv2018

Free-hand gas identification based on transfer function ratios without gas flow control

Gas identification is one of the most important functions of gas sensor systems. To identify gas species from sensing signals, however, gas input patterns (e.g. the gas flow sequence) must be controlled or monitored precisely with additional instruments such as pumps or mass flow controllers; otherwise, effective signal features for analysis are difficult to be extracted. Toward a compact and easy-to-use gas sensor system that can identify gas species, it is necessary to overcome such restrictions on gas input patterns. Here we develop a novel gas identification protocol that is applicable to arbitrary gas input patterns without controlling or monitoring any gas flow. By combining the protocol with newly developed MEMS-based sensors (i.e. Membrane-type Surface stress Sensors (MSS)), we have realized the gas identification with the free-hand measurement, in which one can simply hold a small sensor chip near samples. From sensing signals obtained through the free-hand measurement, we have developed machine learning models that can identify not only solvent vapors but also odors of spices and herbs with high accuracies. Since no bulky gas flow control units are required, this protocol will expand the applicability of gas sensors to portable electronics and wearable devices, leading to practical artificial olfaction.

physics.app-ph↗