I. Introduction
Fluorescence imaging is a mature field with probes available for detecting almost any analyte, from DNA analysis to biological agent detection. One of the common characteristics across these applications are that fluorescent signals are generally weak compared to other imaging applications. Therefore, this application requires low-noise design so that the detected light signal is above the noise floor. Primary sources of noise include thermally generated noise such as reset noise, readout noise, and 1/f noise. Photon shot noise scales with incident illumination power, and will therefore be relatively small at low light levels. Thermally generated dark and leakage currents can also limit detection and may be suppressed using methods such as pinned photodiodes and capacitive feedback transimpedance amplifiers, which are not explicitly examined in this work. Other noise sources include environmental noise such as power supply fluctuations and coupled electromagnetic fields. Low frequency and DC noise sources such as fixed pattern noise can be compensated through correlated double sampling (CDS) or off-line calibration. Active reset techniques have been developed to further decrease reset noise by incorporating explicit feedback into the reset path to ensure a specific reset value [1]–[6]. Although reset noise is considered to be a primary noise source, in practice power supply fluctuations and other environmental noise sources often dominate circuit performance. While environmental noise may be well controlled for many scientific imaging applications, this is not the case for portable system-on-a-chip devices. This presents a driving force for developing strategies to minimize environmental noise. Differential sensor structures have long been recognized for increased environmental noise immunity (which is often correlated along multiple readout paths), at the expense of increasing uncorrelated noise (such as thermally generated reset noise) [7], [8]. In this work we present a noise model, analysis, and experimental verification for a fully differential fluorescence sensor. Similar results were reported by Tian et al. for a single ended sensor [9], and Phillipp et al. for a linear mode sensor [10]. Section II describes the design and operation of the sensor. Section III shows the theoretical analysis of reset and readout noise in the system. Section IV describes the experimental setup and results. This work is summarized in Section V. Differential sensor and readout chain Single-ended sensor and readout chain