Borescope field of view and resolvable detail

A borescope sees a cone. The optical head has a fixed angular field of view — typically 50° to 90° for a forward-looking probe — and what that cone covers on the surface depends entirely on how far the tip is from it: W = 2·d·tan(FOV/2). Double the standoff and you double the coverage, which sounds efficient until you realise the detail halves at the same time. Inspecting fast from a distance is the classic way to miss a crack in a bore.

Resolution is set by the image sensor, not the optics, on any modern videoscope. The field of view is divided across a fixed number of pixels, so the ground sample distance is W/N per pixel. A feature needs at least two pixels across it to be distinguished at all — the Nyquist limit — and in practice three or more for a technician to recognise it with confidence against a textured, oily, poorly lit surface. Turning that round gives the maximum standoff at which a specified feature size stays resolvable, which is the number to write into a remote-visual procedure.

Remote visual examination is accepted by ASME V Article 9 in place of direct VT only when the system is demonstrated to resolve at least the equivalent detail. That demonstration is the real acceptance test, and these figures are a planning tool for it: they tell you the standoff at which to run the demonstration and whether the probe you have can possibly meet the requirement before you mobilise.

Treat the results as the best case. Real probes lose resolution towards the edge of the field through barrel distortion and falling MTF, illumination from the tip falls off as the inverse square so the far side of a large bore is dimmer, and side-viewing prisms change the geometry so that the field is no longer symmetric about the target. Contrast matters as much as size — a tight fatigue crack can be invisible at a standoff where a 0.5 mm scratch is obvious.

W = 2 · d · tan(FOV / 2)
ground sample distance = W / N_pixels
smallest resolvable feature = k · W / N_pixels    (k ≥ 2)
d_max = feature × N_pixels / (2 · k · tan(FOV / 2))