Mounting and Tolerance

Generic engineering material, applicable to any module from any supplier.

For the project-operations side of technical work, this resource is available here.

The mounting aims the camera, and at distance a small angular error becomes a large positional one — which makes mechanical tolerance an imaging specification rather than a fitting detail.

The arithmetic that matters

A one-degree tilt displaces the view by roughly 17 mm per metre of working distance.

At 200 mm that is 3.5 mm, which against a small field of view is a substantial fraction of the frame.

An independent reference for adjacent engineering topics is arXiv.

Which means an assembly tolerance that would be unremarkable for a bracket is a real error for a camera — and the requirement should be stated in the drawing rather than assumed.

Where the field of view is narrow the sensitivity is worst, and it is exactly the case where the coverage margin is smallest.

What to control

Angle, in both axes, which is the term the arithmetic above governs.

Rotation about the optical axis, which tilts the image and matters wherever the processing assumes horizontal or vertical features.

Distance, which depth of field constrains more tightly than people expect.

And repeatability, which is separate from accuracy: a mounting that is consistently 2 mm off can be calibrated out, and one that varies by 2 mm between units cannot.

Stress on the module

A rigid module clamped to a surface that is not flat is a module under bending stress.

Which can shift the sensor relative to the lens — small movements matter at these scales — and which over temperature cycling is a reliability question rather than a one-off alignment error.

Mount to a flat, machined surface where the tolerance matters.

Use three points rather than four where flatness cannot be guaranteed, since three define a plane and four fight each other.

And avoid tightening a module into place as a way of correcting alignment, which stores the error as stress.

Thermal expansion

Materials move, and a mounting that is aligned at 20 degrees may not be at 60.

Which matters where the accuracy requirement is tight and the environment varies — and it compounds with the focus shift that comes from the same cause.

Matching materials between the module carrier and the bracket reduces differential movement.

And where it cannot be avoided, calibrate at operating temperature rather than at whatever the workshop happens to be.

Access

The consideration that arrives at the first service call.

A module mounted where the lens cannot be reached is a module whose window cannot be cleaned — which is a routine maintenance need rather than a fault.

And one behind a permanently fixed panel is one that cannot be replaced without disassembly nobody planned for.

Both are cheap to design in and expensive to retrofit.

Adjustment, and whether to allow it

A mount with adjustment screws is a mount that can be set correctly and can drift.

Where the installation varies — different machines, different sites, field-installed systems — adjustment is necessary and the design question is how to lock it afterwards.

Where the geometry is fixed and repeated, a machined mount with no adjustment is more reliable: nothing to set wrong, nothing to vibrate loose, and nothing to move under vibration.

The worst option is adjustment with no locking, which is common and which converts a one-time alignment into a recurring service task.

Documenting the alignment

Photograph the correct field of view during commissioning.

A reference image showing what the camera should see, stored with the machine documentation, makes a later misalignment obvious in seconds — where without it, somebody compares against memory.

And record the measured distance and angle, so that a replacement module can be set to the same values rather than adjusted by eye.

Replacing a module in the field

Plan for it, because it happens.

A mount that locates the module positively — dowels, a machined recess, a keyed bracket — returns a replacement to the same position without realignment.

One relying on screws through clearance holes does not, and every replacement becomes a commissioning exercise.

The difference is a few millimetres of machining and it is invisible until the first failure in service.

Checking alignment without instruments

A target with a marked centre, placed at the working distance.

If the mark sits where the processing expects it, the alignment is right. If it does not, the offset is readable directly in pixels and converts to millimetres through the same arithmetic as everything else.

Two minutes, and it is a check anybody can run at handover and at service.

Where the module moves deliberately

Pan, tilt or a moving stage.

Then repeatability of the mechanism becomes the specification rather than repeatability of a fixed mount — and the backlash figure is the one to ask for.

A mechanism returning to a nominal position with half a degree of backlash displaces the view by more than most inspection systems tolerate, by the arithmetic at the top of this page.

Which is usually an argument for moving the part rather than the camera, where the choice exists.

The bracket that was designed last

The commonest reason a good optical design underperforms.

A module selected carefully, a lens calculated properly, and a bracket cut from whatever was available — which introduces the angular error the calculation assumed away.

Design the mount alongside the optics, since the two share the same requirement and are usually owned by different people.

And where a bracket already exists, measure what it actually achieves rather than what the drawing says: a reference target in the field of view settles it in minutes.

In one line

State the angular tolerance on the drawing, because nobody assumes a camera bracket needs one.

The short version

• The mounting aims the camera: one degree of tilt displaces the view by roughly 17 mm per metre
• Control angle in both axes, rotation about the optical axis, distance, and repeatability
• Repeatability differs from accuracy: a consistent offset calibrates out and a varying one does not
• A rigid module clamped to a non-flat surface is under stress that can shift the sensor relative to the lens
• Mount to a flat machined surface, use three points where flatness is uncertain, and never tighten to correct alignment
• Design for access: a lens that cannot be reached cannot be cleaned, and a module behind a fixed panel cannot be replaced