Sony Patents a Detection System That Spots Changes by Counting Light Pulses in Time Slots
Most cameras capture everything and sort it out later. Sony's new patent describes a sensor circuit that only triggers when something in a scene actually changes, by counting individual light particles arriving in precise time windows.
What Sony's photon-counting event sensor actually does
Imagine you're trying to catch a hummingbird on camera. A normal camera films every quiet millisecond and then you scrub through hours of footage. A sensor built on this patent would stay silent until the hummingbird actually moves into frame, because it notices changes in light, not just light itself.
Here's the idea: the circuit counts individual photons (particles of light) hitting a tiny detector during a short time window. It then compares that count to the count from the last time something interesting happened. If the two counts differ by enough, it flags that as an event, meaning something in the scene changed.
This is called an event-based sensor, and it's a different philosophy from traditional cameras. Instead of delivering a flood of image frames every second, it delivers only the moments that matter, which can make downstream processing far lighter and faster.
… associate the photon detection signal with a photon detection time period; and detect an event if, within a predetermined integration time period, a predetermined difference is detected between a number of associations of the photon detection signal with a photon detection time period and a number of associations of a previous event.
Translation: It flags a change when the count of timed light pulses shifts compared to the last recorded event.
How the circuit bins photons and compares burst counts
The patent describes circuitry built around a single-photon detection unit, a pixel-level sensor capable of registering individual photons. When a photon arrives, the circuit logs the detection and assigns it to a specific photon detection time period, essentially a short named slot in time.
Over a predetermined integration time period (a configurable window), the circuit accumulates a count of how many time-slot associations occurred. It then compares that count against the count stored from a previous event, which is the last moment the sensor decided something significant happened. If the difference between those two counts crosses a set threshold, a new event is declared.
The key components the claim covers include:
- Obtaining a photon detection signal from a photo detection unit
- Associating each signal with a specific time period (binning it)
- Counting associations within an integration window
- Comparing that count to the count at the last recorded event
- Declaring an event when the difference is large enough
This is fundamentally a change-detection architecture rather than a frame-capture architecture. Instead of delivering a steady stream of full images, the circuit stays quiet and only speaks up when the photon arrival rate shifts meaningfully.
What this means for cameras that react instead of record
Event-based sensors are a serious area of interest for robotics, autonomous vehicles, and high-speed industrial vision, because they can react to changes in microseconds while ignoring static parts of a scene. A camera that only fires when something moves uses far less power and generates far less data than one recording everything at 60 or 120 frames per second.
Sony Semiconductor Solutions' steady investment in sensor architectures shows up here in the single-photon layer: building the event logic directly around photon-level detection, rather than analog pixel voltage, could allow this approach to work in very low-light conditions where traditional event cameras struggle. For your everyday camera or phone, the practical effect would be faster, more efficient scene-change detection without draining the battery on pixels that didn't change.
This is the 37th Sony filing in our chip coverage we've tracked since May, adding to work like one on self-correcting LiDAR and one on AI chip memory.
Claim 1 is written broadly. It covers any circuitry that (a) gets a photon detection signal, (b) bins it into a time period, (c) counts those bins over an integration window, and (d) fires an event when that count differs enough from the count at the last event. There is no restriction to a particular sensor type, integration-time length, or threshold value. That scope is wide.
In practice, a granted claim this broad could sit across a large portion of photon-counting event-sensor designs, at least as written. The prior art in event-based vision sensors is well established, with academic work and commercial sensors from companies like Prophesee going back over a decade, so the patent office will likely push back on breadth during examination. The real question is how much the claim survives prosecution narrowing.
From a reader's standpoint, the underlying technology is genuinely useful and not trivial engineering. But the patent as filed is ambitious enough in scope that its commercial life will probably be decided in the back-and-forth with examiners, not in the claims as they stand today.
There are more where this came from
We read every patent application Big Tech publishes and send you the ones worth knowing. Plain English, free, every week.
The drawings
11 drawing sheets from US 2026/0270572 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →