Passive Ultra-Broadband Single-Photon Imaging

Optical sensors are used in applications including cameras, lidar, autonomous navigation, free-space optical communications, smartphones, scientific imaging, and defense. In each of these applications, the sensor’s sensitivity, bandwidth, and dynamic range are key to overall performance, determining factors such as a system’s maximum range, resolution, and data transfer speed.

This invention allows imaging a dynamic scene over an extreme range of timescales simultaneously - seconds to picoseconds - and doing so passively, without much light, and without any timing signals from the light source(s) emitting it. The system applies a computational algorithm to the output stream of photon timestamps produced by the detector, recovering ultra-fast variations in the intensity of incident light. Specifically, the invention makes it possible to reconstruct a pixel’s time-varying flux (i.e. light intensity) from a discrete, asynchronous, stochastic, and never-ending stream of photon arrival times, showing for the first time, that:

1. Flux reconstruction is possible even if it varies orders of magnitude faster than the timespan between consecutive photon arrivals, and
2. They can compute its Fourier decomposition directly from the inkling photon stream.

Using it in combination with a free-running SPAD (single-photon avalanche diode) camera, the inventors demonstrate never-before-seen abilities, such as:

1. Observing light in flight passively,
2. Intensity imaging at GHz speeds, and
3. Passively recording video that captures variations at everyday timescales – but can be slowed down by a factor of a billion to show the propagation of light itself.

BENEFITS

The main advantages of this technology are:

• Passive and asynchronous. Compared to existing optical sensors, the approach achieves temporal resolution on the order of picoseconds without requiring synchronization to an active source, enabling passive observation of faint, ultra-fast optical signals that are not synchronized to the detector.
• Leverages existing single-photon detection technology. The system is compatible with existing single-photon detector technologies, including CMOS SPADs, which are highly scalable and have already been deployed in consumer electronics (e.g., the Apple iPhone).
• Opens up new applications. The system enables detection of extremely weak and faint sources; for example, it could be used for detection and localization of lidar systems, self-localization from ambient pulsed light sources, or high-bandwidth optical communications.

Current optical sensors either (1) require synchronization to a controlled light source to measure ultra-fast signals, or (2) they are limited by noise and so do not reach single-photon sensitivity. Optical sensing is especially challenging when photons arrive at the sensor at a rate that is slower than the signals being measured. This occurs in time-of-flight sensing, for example, where picosecond or nanosecond-scale signals must be measured from a handful of photons arriving in millisecond timespans. It also applies to setting where a computer vision system must operate reliably in severely light-starved environments (e.g. near total darkness). To overcome this issue, existing solutions rely on synchronized sensing—they pair a sensor with an ultra-fast (e.g., picosecond) strobe light, plus extra hardware to precisely synchronize the sensor with the light. By capturing the same event under strobed illumination repeatedly—often millions of times—enough photons are collected to recover an ultra-fast signal of interest. However, this paradigm is very limiting and cumbersome, and does not allow for asynchronous, passive measurements —for example, detection of ambient lidar signals, measurement of multiple asynchronous laser sources, or characterization of chaotic broadband optical signals. It is also exclusively limited to measuring ultrafast phenomena, as synchronized sensing cannot measure slowly-varying signals.

While fast photodiodes and avalanche photodiodes offer passive, asynchronous sensing at bandwidths exceeding 10 GHz, they do not reach single photon sensitivity and may fail to detect weak signals at ultra-fast timescales. Thus, there are no existing solutions for ultra-broadband optical sensing with single-photon sensitivity.

APPLICATIONS