The Set Screw Moves the Lens: Locking S-Mount Optics Without Decentring

When a single radial set screw is turned against a threaded M12 barrel, it does not lock the lens in the position you set. It pushes the barrel sideways until the thread clearance is taken up on the opposite side, and it usually adds a small tilt, because the force is applied at one point along a barrel that is only supported by its thread. The fix is not a tighter screw. It is a different locking arrangement, a holder with a face the lock can bear on, and a measured before-and-after position so that the shift is a number rather than a surprise. Ask the maker for the locking method, the holder datum and the post-lock centring figure, in writing.

When a project calls for this format, the Lensmind M12 lens series groups the focal lengths and locking options we ship in volume.

Why a Radial Set Screw Decentres the Barrel

An M12 S-mount is a threaded joint, and threaded joints have clearance by design. The male thread on the lens and the female thread in the holder have to assemble without seizing, so there is a small radial gap between the flanks and the crests. When the set screw presses on one side of the barrel, that gap is consumed in one direction: the barrel slides until the thread flanks contact on the far side, and the optical axis moves with it. The same contact point also creates a moment, so the barrel tilts slightly in the holder. Both effects happen at the moment of locking, which is exactly when you thought the alignment was finished.

This is why every station behaves differently. The shift depends on the direction of the set screw relative to the sensor, on how much thread clearance that particular holder and that particular lens happen to have, and on how hard the operator tightened. Three uncontrolled variables produce a distribution of results, and the distribution is what you see when one station is sharp and the next has a soft corner.

A Locking Hierarchy You Can Specify

Rank the locking methods by how symmetric the clamping force is, and specify the best one the holder can accept. An axial lock ring, or jam nut, tightens against a square shoulder on the holder and clamps the barrel along the thread axis. The force is distributed around the circumference, so the barrel does not need to move sideways to lock, and axial position is held by friction across the full face rather than by one point. This is the reference solution for an S-mount that has to hold centring, and a lock ring is a normal accessory in this format.

A split collet or a slotted clamp collar is the next best option. It squeezes the barrel symmetrically over a length of the thread, and it can be designed to close without dragging the barrel off axis. Where neither is available, three set screws at 120 degrees, tightened in sequence to a controlled and equal torque, cancel most of the directional push, but only if the torque is actually controlled. A single set screw with thread adhesive is the weakest arrangement for centring: the adhesive sets the position the screw pushed the barrel into, and it sets it permanently.

The Holder Datum Is Part of the Lock

A lock ring only works if it has something flat and square to bear on. The holder’s front face should be machined perpendicular to the thread axis within a stated tolerance, and it should be wide enough that the ring seats on a full annulus rather than on a chamfer. If the face is rough, or if the ring bottoms on a thread relief, the clamp load goes into distorting the ring and the barrel instead of holding them, and you have reproduced the set screw problem with a more expensive part.

Ask for the holder drawing with the datum defined: which face or bore establishes the axis, what the perpendicularity and runout are, and what thread class the female thread is made to. This is the same subject as a thread tolerance dispute between a lens and a body, seen from the holder side. If the maker supplies lens and holder together, the two threads can be matched in production, and that matching is worth asking about explicitly, because a lens that binds in one holder and locks cleanly in another is telling you the two parts were made to different gauges.

Before and After: The Measurement That Settles It

The only convincing evidence of a good lock is a centration measurement taken before locking and after locking, on the same unit, with the same method. On a bench, that means a collimated target or a centring instrument that reads the optical axis relative to the holder datum, with the lens focused, then locked, then measured again. On a station, it means a target that produces a measurable feature at a known image height, before and after the lock, with the camera fixed.

Ask the maker for that figure as a specification, not as an anecdote: the maximum post-lock axis shift in micrometres, or in pixels at a named image height, plus the tilt if they can measure it. A number like that is useful to a supplier because it is testable, and it is useful to you because it turns a complaint into an acceptance criterion. If the maker cannot measure centring at all, that is information too: it tells you the locking method was never verified.

Station-to-Station and Batch Variation

A single good unit is not a process. Once the lock method is fixed, measure the post-lock shift across the stations you are building, and keep the spread rather than the average. A tight spread around a small shift means the method is under control and the remaining error can be compensated in the station recipe. A wide spread means one of the variables is still loose: operator torque, holder face quality, or the thread pairing between a particular lens batch and a particular holder batch.

Batch variation is where the maker’s production data earns its place. A lens family that is produced with 1 μm-class active alignment and shipped with batch test documentation covering parameters such as MTF, distortion, back focus and depth of field is a family whose optical centring is controlled at the factory. What that documentation does not tell you is how the lens behaves when your holder clamps it, because that interaction happens at your station. The two facts are complementary, not substitutes.

Retrofitting Stations You Already Built

For stations already in production, the sequence is inspect, change, re-measure, record. Inspect the holder face for a chamfer or a burr that prevents a ring from seating, and inspect the set screw for a point that has deformed the barrel thread. Change the locking method for one station first, measure the before-and-after shift, and only then roll the change out; a retrofit that is applied everywhere at once destroys the comparison that would prove it worked.

Re-align after the retrofit, because a station that was compensated around a set-screw shift is now compensated around nothing. Then record the new lock method, the torque, the measured post-lock shift and the date in the station file, so that the next person who services the station knows what the lock is supposed to be. Without that line, the next technician will find a plain set screw in a drawer and put it back.

Budget the retrofit as a diagnostic exercise rather than a parts swap. The replacement ring is inexpensive compared with the cost of a day of production, and the value of the exercise is in the first station’s measurements: the before-and-after shift tells you how much of your previous alignment error was the lock and how much was the holder. If the shift barely changes after the ring is fitted, the holder face or the thread pairing is the dominant problem, and buying rings for every station would have hidden it rather than fixed it.

The Purchase Specification Lines

Write the requirement into the purchase order rather than into a conversation. The lines that do the work are: locking method (axial lock ring on a square shoulder preferred, collet or symmetric clamp acceptable, single radial set screw not acceptable for a station that requires centring); holder datum with perpendicularity and runout stated; thread class and, where the maker supplies both parts, the matching between lens and holder; the maximum post-lock axis shift with the measurement method named; and the torque to be applied, with the tool that controls it.

Add a line for the anti-rotation feature if the image orientation matters, because a lock ring holds axial position and focus but does not stop the lens rotating. Add one more for the spare rings and the torque tool, so that a station can be serviced without improvising. The mount decision itself, and the reasons a board-level camera usually ends up on an M12 thread, are covered in the M12 lens selection guide; the vibration case for a locked thread is covered in M12 focus lock under vibration. This article is the third piece: how to lock without moving the axis.

None of this requires a new lens design. It requires that the locking arrangement be treated as part of the optical system rather than as a fastener, and that the shift be measured once so that the specification, the holder and the station recipe all point at the same number.

Written by Baker Zhao, Founder at Lensmind Optics.