CCTV pipe inspection limits: minimum diameter and bends

Aged cast-iron and steel lines under a building usually offer one access point and one chance to choose the camera correctly. Three numbers decide whether a CCTV survey can proceed: the bore the camera enters, the direction changes it must pass, and the pixel scale that lands on the pipe wall. This page gives you all three — a diameter ladder for crawler and push cameras, a geometric check for bends and laterals, and a worked resolution example with the arithmetic shown. For lines that cannot be taken out of service, the guide on camera inspection of a pipe under pressure covers the live-main case instead.

The diameter ladder: 75 mm, 100 mm and 800 mm

Four bands cover nearly every line you will be asked to quote. A standard axial-viewing crawler works reliably from about 100 mm internal diameter upward. In the DN75–DN100 band, compact crawlers and push rigs overlap, and the run's bends decide. Below roughly 75 mm, a push camera with a small fixed forward-view head is the practical tool. Above roughly 800 mm, a person can physically enter the pipe — which is exactly why every smaller bore is inspected through a camera on a rod or a wheeled carrier.

Quote against the measured clear bore, never the nominal size on the drawing. Decades of tuberculation and scale can close a nominally DN100 cast-iron bore to under 80 mm, and one encrusted joint can stop a run the clean barrel would allow. Measure the smallest opening — at a service tap, a misaligned joint or the invert — and keep about 10% clearance between the head diameter and that opening when you select the carrier.

Why crawlers stop near DN100 and push rods keep going

A crawler carries its own drive wheels, a self-leveling head and a winch for recovery. All of that needs volume: wheels need contact on a curved wall, the head needs clearance through hub-and-spigot joints and small offsets, and the smallest standard carriers are built around a DN100 bore. Below that, traction fades, joints snag heads, and recovering a stuck carrier becomes an excavation job.

A push camera trades all of that for simplicity: a short rigid head on a flexible rod, pushed from a drum. The head passes elbows that would trap a crawler, but friction grows with every bend, and typical reels hold 50 m to 150 m of rod. In that DN75–DN100 band, decide by the run: a straight or gently graded length suits the push rig; several elbows, heavy scale or branch mapping justify the compact crawler.

Decision, then: under 75 mm, plan one push pass that covers the main run only. From DN100 upward, plan the crawler and add the heads the junctions require.

What a bend does: two separate limits

A change of direction does two different things to an axial camera, and both need checking.

The first limit is optical. Line of sight is straight, so the view ends at the first tangent point of any bend, and everything beyond stays invisible until the head itself travels there. A 90° elbow hides the whole downstream barrel from a stationary head, and no lamp repairs that.

The second limit is mechanical. A rigid head of outside diameter d and length L passes a sharp right-angle corner only when it is short enough for the clear channel. The flat-channel corner model gives the conservative bound:

Real fittings are swept rather than mitred: a long-radius elbow curves with a radius of about 1.5 × D, a short-radius one about 1 × D, and both pass heads longer than the sharp-corner bound. Worked check for a 100 mm bore with a 40 mm head: L_max ≈ 2.8 × (100 − 40) = 168 mm. Round to 170 mm: a head shorter than that passes the worst case; a longer one needs a generous swept radius or a second entry from the far end.

Push heads sit far below the bound at typical lengths of 30 mm to 50 mm, so their cost at each elbow is rod friction, not passage. Continuous sags in buried cast iron behave the same optically — the view stops at each change — but they bend the rod gently instead of trapping it.

Laterals: every junction is its own job

Looking down the main, an axial head sees a branch connection as a dark opening in the wall — a wye joining at about 45° shows an ellipse, a 90° tee a circle. It cannot look up either one. Under a building, where fixture branches meet the horizontal main every few metres, that gap decides what the report can claim.

Three methods cover a junction, and each is priced as its own pass:

1. A steerable or lateral-launch head drives part-way into the branch and inspects it from the main.
2. A rotating side-scan head documents the main's circumference and the condition of the junction itself, but not the branch interior beyond the opening.
3. A second push camera, entered from the branch's own cleanout or fixture connection, inspects the full lateral as a separate segment.

Decision: list every branch on the drawing before quoting, assign one method to each, and state plainly which segments a main-line-only pass leaves unassessed.

Resolution at the wall: the pixel math

Two thresholds drive the camera choice. A feature needs roughly 3 pixels across it to be found reliably, and 5 or more for its width to be measured with confidence — the ordinary sampling rule applied to a crack.

The pinhole relations set the scale:

Geometry then forces the lens. Covering a DN100 bore at the near-ring plane, 50 mm out, needs f ≈ 6.4 × 50 / 100 = 3.2 mm on a 1/2"-class sensor — and the ring itself sits near 90° off the axis, which only a fisheye reaches. A moderate lens sees just the far-wall band, but sees it with far more pixels per millimetre. That trade is the whole lens decision inside the pipe.

Worked example — DN100 line, camera on the centreline, far wall at a working distance equal to the bore. Setup A: 5 MP sensor of the 1/2" class, 6.4 mm wide, 2592 pixels across, 1.5 mm fisheye. Setup B: the same sensor with a 6 mm lens. Setup C: 12 MP sensor of the 1/1.7" class, 7.4 mm wide, 4000 pixels, 1.5 mm fisheye.

The arithmetic. Setup A: FOV = 6.4 × 100 / 1.5 = 427 mm, px/mm = 2592 / 427 = 6.1, so a 0.5 mm crack spans 0.5 × 6.1 = 3 px. Setup B: FOV = 6.4 × 100 / 6 = 107 mm, px/mm = 2592 / 107 = 24, and the crack spans 12 px. Setup C: FOV = 7.4 × 100 / 1.5 = 493 mm, px/mm = 4000 / 493 = 8.1, and the crack spans 4 px.

Decision: survey on the fisheye to see the whole bore, then re-inspect suspect joints at the 6 mm setting to size cracks. If one pass must do both at DN100, a 4000-pixel sensor is the floor that brings a 0.5 mm crack to 4 pixels — enough to find and grade, still short of certified width measurement.

Bore size moves every number. Pixels per millimetre rise as the bore shrinks — setup A reaches 2592 × 1.5 / (6.4 × 75) = 8.1 px/mm in a 75 mm line — and fall as it grows: the same setup drops to 2592 × 1.5 / (6.4 × 300) = 2 px/mm in a 300 mm line, where a 0.5 mm crack is under one pixel wide. Large bores therefore need the narrow-view treatment even for the survey.

Treat the pinhole figures as planning estimates: a real fisheye compresses the image toward its edge and oblique walls stretch pixels, so expect roughly ±20%. Focus is easier than in most machine-vision work — at these focal lengths the depth of field is large, and a fixed head set near half the bore stays sharp from near ring to far wall. If a wide-angle image looks soft anyway, the fault is usually focus stop, and the checks in the wide-angle CS-mount focus check apply; the close-focus calculation at 127 mm shows the working-distance method at similar ranges.

Cast iron and steel under a building: three field realities

First, the bore shrinks. Tuberculation, scale and hard-water deposits pull the clear opening below nominal, so measure before promising, and schedule a descale pass when the head fits with less than the clearance from the ladder.

Second, the surface fights the light. A wet steel or galvanized wall mirrors the lamp straight back into the lens and whites out the image. Diffuse the ring light, add a polarizer if the reflection survives, and keep exposure short enough that dripping water does not smear. The methods in pipe internal surface inspection: controlling glare transfer directly.

Third, CCTV reads the surface only. It maps cracks, open joints, offsets, scale, lining loss and water entry — most of a service-condition report — but it cannot measure remaining wall thickness, and corroded cast iron can look intact while carrying little of its original load. For structural assessment under occupied floors, pair the survey with ultrasonic thickness readings on the segments the video flags.

Parts that fit an inspection head

The survey lens is a fisheye in the 1.5 mm class for a 1/2"-format sensor in an M12 mount — the format most compact head cameras take.

For the detail pass, a varifocal beats opening a sealed head to swap fixed lenses. The LS-CTV-0135 spans 6-15 mm on 2/3" sensors at up to 3 MP in a C-mount with an 8-11 mm image circle. Set it toward 6 mm to crop a high-detail far-wall band, and toward 15 mm when the wall sits farther out in a larger bore.

When one pass must survey and grade at once, pixel count is the lever. The LS-CMR-0166 captures 4000 × 3000 (12 MP) on a 1/1.7" sensor over USB3 at 30 fps. A 4000-pixel fisheye survey lands about 8 px/mm on a DN100 far wall, moving a 0.5 mm crack from 3 to 4 pixels.

For the smallest heads and the simplest start, a compact colour camera with a global shutter fits the same optics: the MV-CU020-90GC is a 2.5 MP colour GigE camera with a 1/1.7" sensor, C-mount and global shutter.

If the deliverable is a pass or fail at each joint rather than recorded video for review, a smart camera runs the inspection on the device — start in the smart cameras category or the camera finder. Whatever the camera, the lens must cover the sensor: the lens sensor format guide explains image-circle matching, and the CCTV lenses category plus its finder hold the wide and varifocal options.

Checks before you commit to a line

1. Measure the smallest clear opening — at a tap, an offset joint or the invert, not the nominal bore. The head plus roughly 10% clearance must pass it.
2. Count direction changes and check rigid head length against 2.8 × (D − d). A miss means a second entry point, not a brighter lamp.
3. List the branches and assign each a method: steerable head, side-scan, or a second camera from the branch's own access.
4. Compute pixels per millimetre at the far wall for your sensor and lens, and hold 3 px across the smallest defect you must find, about 5 px to measure it.
5. Set focus near half the bore and run the head through a dry pipe section of the same diameter before mobilizing.
6. Plan the light for the material: diffusion for wet steel, a polarizer if reflection survives, short exposure under dripping water.
7. Reserve person entry for 800 mm and above, and only inside a permit-controlled confined-space procedure.

When CCTV is not the answer

A silted or blocked line shows the camera nothing until it is cleaned — jet first, camera second, in that order. A line whose wall thickness decides structural safety needs ultrasonic or acoustic thickness methods, with the video used to choose where to measure; no optical view through scale returns that number. And a run whose clear opening is smaller than any head you can source, with no alternate access, is an excavation question rather than a camera question.

Where the geometry passes, build to the numbers: pick the carrier from the ladder, the lens from the pixel math, then start matching parts in the camera configurator, or send us the drawings and we will run the ladder and pixel checks with you. The applications topic page collects more write-ups like this, and industrial lighting covers the lamp side if you are building the head yourself.