Sensor diagonal is the single geometric variable that scales field of view at a fixed working distance and lens focal length: a 1" sensor (16 mm diagonal) yields about 45% more horizontal FOV than a 2/3" sensor (11 mm) and roughly 2.0x the FOV of a 1/2" sensor (8 mm), all under the same optics [S4][S5].
The same four standard format sizes, 1", 2/3", 1/2", and 1/4", still cover most discrete-pixel industrial camera data sheets, with corresponding active areas of 12.7 x 9.6 mm, 8.8 x 6.6 mm, 6.4 x 4.8 mm, and 3.2 x 2.4 mm [S5]. Choosing the wrong format relative to the lens image circle is what produces the classic "tunnel effect": vignetting and black corners that no amount of aperture or illumination can fix [S5].
Why sensor diagonal drives field of view
Field of view in object space equals sensor active dimension times the lens magnification (m = working distance / focal length, for thin-lens geometry); doubling the sensor linear dimension at the same focal length and working distance doubles the FOV [S3]. The "inch" rating on a machine vision sensor is a historical carryover from 1950s Vidicon tubes, where a 1" tube had a 16 mm diagonal photo-sensitive area, and a digital sensor is called 1" only when its diagonal reaches that 16 mm, not because it measures 25.4 mm [S4][S5].
Two sensors rated as the same "format" can still have different active dimensions and aspect ratios (1:2, 3:2, 16:9 are all common), so spec sheets should be read in millimetres, not inches [S4]. A practical sizing rule for in-line vision: target a FOV of about 2A x 2B for a part of size A x B, then back-solve the required horizontal sensor length from the focal length and working distance [S8].
Sensor size, pixel size, and low-light performance
Larger sensors typically carry larger pixels at a given resolution, and pixel area directly drives well capacity and low-light signal-to-noise, which is why industrial 1/2.3" CMOS sensors are commonly limited to about 5 MP while consumer units at the same format ship at 16 MP or higher [S5]. A 5 MP sensor with 5.5-micron pixels will typically outperform a 20 MP sensor with 2-micron pixels in low-light or fast-exposure scenarios, even when both are nominally "high resolution" [S7].
For a fixed resolution, stepping from 1/2" to 2/3" at the same pixel count gives each pixel roughly 1.9x more area, improving dynamic range at the cost of a smaller FOV; stepping from 1/2" to 1" pushes the ratio to about 3.9x per pixel, with the matching FOV scaling up by the same factor [S5]. When pixels shrink below about 2 microns, lens diffraction and pixel angular sensitivity start to dominate, so the lens numerical aperture must be matched to the pixel pitch or resolution gains are wasted [S4].
Matching lens image circle to sensor format
Every industrial lens has a stated "max sensor size" or imaging circle; if the camera sensor diagonal exceeds that, the corners of the frame receive no light, producing vignetting that no aperture or light source can recover [S5]. The safe rule is lens imaging size equal to or greater than camera sensor size, and many 1/1.8" or 2/3" lenses are designed with enough margin to cover 1" sensors with minor corner shading [S5].
For high-vibration lines, even the correctly sized lens-sensor pair can lose calibration: mounting cameras away from vibrating machinery entirely is more effective than any vibration-damping accessory fitted to the machine, because resonance in the lens-sensor stack shows up as pixel-scale blur before the housing shows any movement [S6]. Cable strain, PoE sag, and lens-settling after focus adjustment are the three most common "mystery" FOV shifts in field returns, and each one shows up as a geometric distortion rather than a focus error.
Selection criteria by application type
Area-scan cameras with rectangular sensors dominate part-presence, dimension verification, and label-reading stations, and the sensor format is chosen to deliver enough FOV at the available working distance without forcing a wide-angle lens with its associated barrel distortion [S1][S4]. Line-scan cameras use a single-row sensor and stitch the image as the target moves past, making the sensor's pixel pitch (not its area) the critical FOV lever; the equivalent "sensor length" for a line-scan application is the number of pixels times pixel pitch [S1].
3D vision cameras using structured light, stereo, or time-of-flight add height or depth data on top of the 2D frame, and the 2D sensor still has to be sized for the lateral FOV before the projector baseline is even considered [S1]. For barcode and OCR at typical 100-300 mm working distance, a 1/2" to 2/3" sensor with 5-10 micron pixels covers most 1D/2D codes; for sub-pixel metrology of features under 0.1 mm, designers usually jump to 1" sensors with 3.5-micron pixels or smaller [S4][S5].
Failure modes and engineering constraints
Three failure modes dominate field returns tied to sensor size choice. First, lens image circle undersized for sensor diagonal, producing the "tunnel effect" with hard black corners that no software can fix [S5]. Second, pixel pitch undersized for the lens resolving power, producing crisp-looking but unrepeatable measurements where every micron of vibration changes the measured edge [S4]. Third, FOV undersized for the part plus margin, producing inspections that pass on good parts and fail on parts with normal placement tolerance, which is why the 2A x 2B margin rule is the standard starting point [S8].
Environmental constraints add a fourth failure mode: industrial cameras rated for -20 to 60 degrees C operation still need thermal focus compensation if the sensor package and lens housing have different thermal expansion coefficients, and the focus shift scales with both the focal length and the sensor-to-lens spacing. In dusty or washdown environments, sensor window heating and positive-pressure enclosures are specified before the sensor size decision is finalized, because a larger sensor with a larger window collects more contamination per shift.
Sourcing, standards, and where to look next
The four format ratings (1", 2/3", 1/2", 1/4") remain the de facto sensor-size vocabulary carried over from Vidicon-tube days, and every major vendor's data sheet still publishes the diagonal in millimetres alongside the inch rating, which is the only number that should drive an FOV calculation [S4][S5]. The decision chain runs: required FOV and pixel size on object, then working distance, then focal length, then sensor diagonal to fit the lens image circle, then pixel count to hit the required on-object resolution, with the camera operating temperature and vibration envelope checked last.
Track these two signals before locking a sensor format: pixel pitch versus lens MTF at the chosen aperture (mismatched pairs waste resolution), and the lens image circle versus sensor diagonal in millimetres, never in inches [S4][S5]. For applications where FOV and pixel density must both be high, see how pixel-per-tolerance budgeting works in pixel-per-tolerance budgeting for machine vision measurement.