What the Sensor Specification Sheet Does Not Say Out Loud

04.1 §1Area, Pitch and the Basic Trade-off

A sensor's format label — APS-C, full-frame, medium format — describes the active imaging area. That area, divided by the pixel count, yields pixel pitch: the centre-to-centre distance between adjacent photosites. Pitch governs how much silicon each photosite occupies and, downstream, how much light it can collect before saturating.

The K-3 Mark III carries a 25.7-megapixel APS-C sensor with a crop factor of 1.5×, giving an active area of roughly 23.3 × 15.5 mm. Divide that area by the pixel count and each photosite sits on a pitch of approximately 3.76 µm. A full-frame body at the same pixel count would spread those same 25.7 million photosites across 36 × 24 mm, stretching pitch to roughly 5.9 µm — a photosite with more than twice the collection area. That is not marketing language; it is elementary geometry, and it is why full-frame sensors retain a light-gathering advantage at matched pixel counts, regardless of in-camera processing.

The specification sheet lists megapixels and format. It does not print pixel pitch. The number that most directly predicts per-pixel light sensitivity is the one manufacturers leave off the box.

04.1 §2What BSI Actually Moves

A conventional front-side illuminated sensor routes wiring — metal interconnect layers — across the face of the silicon, in front of the photodiodes. Those layers block a fraction of incoming light. Back-side illuminated fabrication, or BSI, flips the silicon wafer during manufacture so the photodiodes face the incoming light directly, with wiring behind them. The result is a larger effective fill factor: a greater proportion of each photosite's area actively collects photons.

BSI is a fabrication process, not a larger sensor. It does not change pixel pitch or total area. What it changes is efficiency: more of the light that falls on a given area is captured rather than intercepted by circuitry. For small-pitch sensors — APS-C in particular — where absolute photosite area is already constrained, the efficiency gain is proportionally more significant than it would be on a large-format sensor with naturally generous pitch.

Sony Semiconductor, which supplies imaging sensors to Ricoh and to most of the industry, introduced BSI as standard across its Exmor RS and subsequent lines. The sensor in the K-3 Mark III is a BSI device. Ricoh's specification pages confirm this; the underlying fabrication origin is Sony Semiconductor.

04.1 §3Stacked Architecture and Readout Speed

A stacked sensor adds a second silicon layer bonded beneath the pixel array. The lower layer carries dedicated processing circuitry — DRAM buffers, logic for signal processing — that would otherwise have to sit on the same plane as the photosites. Separating them allows the pixel layer to be optimised for light collection and the logic layer to be optimised for speed.

The practical consequence is readout rate. A non-stacked sensor must read each row sequentially through circuitry that shares real estate with the photosites, setting a ceiling on how fast the entire frame can be scanned. A stacked sensor can read rows faster and process them in parallel on the bonded logic layer. This directly reduces rolling shutter artefact — the skew visible in fast-moving subjects when a sensor scans too slowly to capture all rows at effectively the same instant.

The K-3 Mark III carries a 25.7-megapixel APS-C sensor with a crop factor of 1.5×, giving an active area of roughly 23.3 × 15.5 mm.

The K-3 Mark III uses a BSI sensor without a stacked architecture. Its 12 fps burst rate and electronic shutter rolling shutter behaviour reflect that. Sony's own A9 series and the sensors inside Canon and Nikon's flagship mirrorless bodies use stacked designs; the speed differential between those bodies and the K-3 Mark III in electronic shutter mode is a direct physical consequence of layer count, not processing software.

04.1 §4Reading the Sheet Correctly

Four numbers that the specification sheet always prints — format size, pixel count, BSI or not, stacked or not — define almost everything worth knowing about a sensor's fundamental capability envelope. They do not tell the reader how those numbers interact. Area sets the ceiling for total light. Pitch sets the per-photosite budget. BSI raises efficiency within that budget. Stacking raises readout speed without touching either.

The numbers are geometry. The sheet just declines to draw the diagram.