The Problem Sony Spent a Decade Solving

04.5 §1Why Sensors Lie About Time

Every digital sensor captures light across a rectangular array of photosites. On most sensors — CMOS, BSI, stacked or otherwise — that array is not read simultaneously. Charge is transferred row by row, from the top of the frame to the bottom, with each row exposed at a slightly different moment in time. The process takes anywhere from a few milliseconds to tens of milliseconds depending on sensor architecture and readout speed.

Under static conditions that delay is invisible. Move a fast subject across the frame, or pan quickly, and the rows at the top of the image depict the world at a measurably earlier instant than the rows at the bottom. The result is rolling shutter: vertical elements lean, spinning propellers become curved, and fast strobes pulse unevenly across the frame. The artefact is not a processing error; it is an accurate record of something that was never in the same position at the same time.

Film cameras do not produce this artefact because a mechanical shutter exposes the entire frame simultaneously — a true global exposure. Early CCD sensors also read out globally. When CMOS displaced CCD as the dominant architecture — driven by lower power consumption, on-chip signal processing and easier integration — the global shutter property was largely abandoned. CCD's charge transfer mechanism made simultaneous readout straightforward; CMOS's column-parallel architecture made row-sequential readout efficient and cheap, and global shutter CMOS required additional circuitry that reduced fill factor, increased noise and complicated fabrication.

04.5 §2The Engineering Response

Manufacturers addressed rolling shutter incrementally rather than structurally. Faster readout circuits narrowed the inter-row delay. Stacked sensor designs, which place processing circuitry on a separate bonded silicon layer, accelerated readout further by shortening the signal path between pixel and processor — Sony's own stacked BSI designs, found in bodies across multiple manufacturers' lines, cut readout times substantially compared with earlier planar sensors. Sony's a9 (2017) used a stacked BSI sensor to achieve a readout fast enough to suppress most visible rolling shutter at practical frame rates, but the underlying mechanism remained sequential.

The a9 III, announced in November 2023, changed the architecture rather than accelerating it. Sony fitted a full-frame 24.6-megapixel global shutter CMOS sensor — the first in a full-frame commercial interchangeable-lens camera. Each photosite holds its accumulated charge simultaneously and releases it at the same instant across the entire array. Row-to-row delay is eliminated by design, not reduced by speed. Flash sync extends to 1/80,000 s — a figure that follows logically from global shutter, because a mechanical sync ceiling exists only where a rolling exposure can be caught mid-travel by a closing curtain. There is no mid-travel here.

The engineering cost is real and published. Global shutter CMOS requires a charge storage node at each pixel, which reduces the area available for light collection. Sony rates the a9 III's native ISO range at 250–25600 — a notably higher floor than the stacked a9 series, directly reflecting the smaller effective photosite area. Dynamic range measurements by independent testers have confirmed a penalty at base ISO relative to Sony's stacked non-global-shutter sensors. The solution to rolling shutter trades one set of sensor physics for another.

For the K-mount ecosystem, the question is engineering trajectory rather than current product. Ricoh builds on Sony Semiconductor sensors — the K-3 Mark III's APS-C chip among them — and the pace at which Sony Semiconductor develops global shutter for APS-C formats will eventually determine whether that option reaches mid-market bodies. Global shutter CMOS at APS-C scale has existed in industrial and scientific imaging for years; the barrier has been cost and the fill-factor penalty at usable ISO ranges. Sony's full-frame demonstration in 2023 establishes that the penalty is manageable. It does not yet establish that it is cheap.