Samsung Patents a Camera Chip That Captures Light in Two Separate Passes
Most camera sensors read all their light-collecting cells at the same moment. Samsung is patenting a design that deliberately staggers that readout, firing a majority group of cells first and a smaller group a moment later, all under a single tiny lens.
What Samsung's two-wave pixel readout actually does
Today's camera image sensors work by having each pixel collect light and then hand that information off all at once. That works fine in bright conditions, but when timing and precision matter, firing every cell simultaneously can cause interference and limit how much creative control the sensor has over what it captures.
Samsung's patent describes a pixel made up of a grid of at least nine light-collecting cells (arranged 3-by-3 or larger), all sharing one tiny lens on top. The trick is that more than half of those cells fire at one moment, and the rest fire at a different moment. Two groups, two timings, one pixel.
The practical benefit is that the sensor can gather more exposure information, reduce certain types of image noise, or enable depth-sensing tricks, because it's pulling light data from the same physical spot in two distinct "snapshots" rather than one. Think of it like taking two rapid photos of the same scene from the exact same position and blending them for a better result.
… P transfer transistors among the first N 2 transfer transistors are configured to be turned on at a time t1 and L transfer transistors among the first N 2 transfer transistors are configured to be turned on at a time t2 different from the time t1 …
Translation: Different groups of microscopic switches turn on at separate moments to capture light in two distinct passes.
How the pixel grid splits into timed transistor groups
The patent describes an image sensor pixel that contains N-squared photodiodes (light-collecting cells) arranged in an N-by-N grid, where N must be an odd number of at least 3. So the smallest version is a 3×3 grid, giving you 9 photodiodes packed beneath a single microlens (the tiny curved glass dome that focuses incoming light onto the pixel).
Each photodiode has its own transfer transistor, which acts like a gate: when it opens, the charge that light has built up in the photodiode flows out to be measured. The patent's key rule is that these transistors are split into two named groups. The first group, containing more than half of all the transistors, opens at time t1. The second group, containing at least 2 transistors, opens at a different time t2.
Critically, the majority group (called P) must be larger than half of N-squared. So in a 3×3 pixel with 9 total transistors, P must be at least 5. The remaining transistors (called L) are at least 2 but always fewer than P. This asymmetry, a big group and a small group, is the heart of the claim.
- N: grid size (odd, 3 or larger)
- P: the majority transistor group, fires first (t1)
- L: the minority transistor group, fires second (t2)
- One microlens sits over the entire N×N pixel
What this means for camera quality in Samsung devices
Camera sensors are one of the most competitive battlegrounds in consumer electronics, and pixel-level architecture decisions shape everything from how well a phone shoots in dim restaurants to whether it can measure depth for portrait blur. A sensor that reads subsets of a pixel at different times can support phase-detection autofocus, high-dynamic-range capture, or noise reduction without requiring a physically larger pixel.
For you as a consumer, this kind of design is what separates a camera that handles a dimly lit birthday cake from one that turns it into a blurry orange smear. Samsung's bet on sensor architecture suggests the company sees on-chip pixel design, not just software processing, as where the next round of camera improvements will come from.
Samsung's 140th filing we've tracked in our camera sensor push since May builds on work like one splitting focus and color pixels and one for a fridge camera.
Claim 1 covers any image sensor where at least nine light-catching elements, arranged in an odd-numbered square grid under a single tiny lens, are fired in two distinct waves, with the first wave required to carry a strict majority of the total. That mathematical floor, more than half must go first, rules out any even split and locks in an asymmetry that the patent treats as the core invention.
The scope is broad enough to reach many grid sizes and future sensor generations, yet specific enough to hold up under challenge. Anyone wanting to avoid it would need to restructure the pixel itself, not just adjust timing.
When and how a sensor reads light shapes how sharp a photo looks in a dim room or when something is moving fast. Owning a precisely defined mathematical rule for that process protects a real performance advantage at the hardware level, long before it shows up in any review.
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
51 drawing sheets from US 2026/0303996 A1 · click any drawing to enlarge
Want this weekly breakdown for a company we don't cover? Patentlyze Pro →
Be the first to weigh in