Sensor Area

Sensor area is pixel pitch times pixel count in each direction. A 2048 × 2048 array of 6.5 µm pixels is 13.3 mm square, with an 18.8 mm diagonal.

The instinct is that a larger sensor sees more. It does — up to the field the objective actually forms, and not one micrometre beyond. The objective projects a circle whose diameter is its field number divided by magnification, and a sensor diagonal exceeding that circle records darkness and aberration in its corners. A field number of 25 at 20x gives a 1.25 mm circle at the specimen; the sensor above covers 0.94 mm of that diagonal, so it fits comfortably and uses 57% of the available area.

Both failures are common. Too small and you paid for field you never record; too large and the corners are unusable, which is usually blamed on the objective.

For whole-slide or large-area work, recorded area per frame is the dominant factor in total time, and it enters as an inverse. Double the recorded area and you halve the tile count; a 15 × 15 mm section at 0.665 mm per frame needs about 509 tiles, and at 0.94 mm per frame about 255.

Each tile carries fixed overhead that has nothing to do with exposure: stage settle, autofocus, filter changes. At a second of overhead per tile, that difference alone is more than four minutes of pure waiting per section, before a single photon is collected. Across a study it is the difference between a protocol that runs overnight and one that does not.

A 2048 × 2048 sensor at 16 bits produces 8.4 MB per frame. At 100 frames per second that is 839 MB/s sustained — faster than most single drives will write, and about 3 TB per hour.

This is the constraint that catches people after the camera arrives. A sensor capable of a given frame rate does not mean a system capable of recording at it: the limit moves to the interface, the disk, and eventually the analysis. The honest planning question for a large-sensor, high-speed camera is not whether it can acquire the data but whether you can store it, move it, and open it again.