CCSS: Signal Processing for Single-Photon Detectors

点击翻译按钮获取中文摘要

Light has a fundamental smallest quantity - a photon - that is very far from everyday human experience. For example, the number of photons collected by the camera in a mobile phone to form a typical photograph is in the trillions. Nevertheless, there are some increasingly common devices that rely on measuring light down to the smallest possible amounts. These are devices that perform "single-photon detection" (SPD). SPD is used in combination with pulsed lasers in the 3D imaging systems of self-driving cars, and it is used to enable augmented reality in devices like the iPad Pro. Many types of scientific imaging also use SPD to see individual molecules, track proteins, or determine chemical concentrations through spectroscopy. Devices for SPD have complicated behavior that is often modeled naively. Using more detailed modeling, this project will develop data processing methods to use with SPD to improve various systems.Though SPD is on the verge of everyday use, signal processing for SPD has lagged far behind. Both system design and device design are guided by trade-offs, and those tradeoffs depend greatly on the sophistication of the data processing. Therefore, novel signal processing will not only improve applications that use SPD, it will also influence hardware designs. Devices with single-photon sensitivity suffer a limitation by which each detection event causes a non-zero "dead time" during which the system is unable to register incident particles. Since signal processing methods to mitigate dead time effects are few and not well known, it is customary to carefully avoid dead time effects by operating systems such that photons very rarely arrive during dead times. While this indeed makes the effect of dead time negligible, one may naturally ask whether this is a good practice. Preliminary results suggest that allowing appreciable dead time effects and compensating for them can provide dramatic improvements in lidar. A project focus is to improve and exploit modeling of dead time effects to create the most informative measurements for lidar and other applications. In arrays for SPD, maintaining low crosstalk is a major barrier to increasing fill factor and thus increasing detection efficiency. Another project focus is to model and mitigate crosstalk, starting with a deconvolution approach and progressing to more sophisticated modeling of coupled Poisson processes. With these novel models and methods, the project will develop several imaging innovations.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.