Camera Head Knocked Into the Fixture: The Bench Evidence That Decides Reseat or Return

A camera head that still images sharply after a knock into the fixture has not cleared itself. The impact can decenter an element, tilt the optical axis, shift the back focal distance, or stress the mount while the middle of the field stays crisp, so the live picture is a screening observation, not a disposition. Put the lens back in service only when three things hold at once: the mount design allows field removal and refit, the mechanical pass finds no damage, and the unit passes four documented bench checks at your working distance against the test report that shipped with the batch. Any cracked housing, bent mount, one-sided softness, back focus drift, or a sealed or telecentric gauging head changes the answer: it goes back to the maker, and the bench evidence, not the image, is what settles it.

If you are comparing suppliers for this kind of optic, the Lensmind machine vision lens catalog is the hub for our standard models and custom options.

What an impact actually moves inside a camera head

When a technician knocks a camera head into a fixture, the force travels through the barrel into the element stack and the mount. Three kinds of movement follow, and none of them has to blur the center of the picture. An element can decenter or tilt inside its cell. The spacings between groups can compress or open, which shows up as a back focal distance that no longer matches the shipped value. The mount seat itself can deform, which changes where the lens sits relative to the sensor even when the glass is untouched. Center resolution tolerates small amounts of all three; the field edges do not. That is why the classic signs of a decentered lens element are asymmetries: coma that flares to one side, astigmatism that smears tangential and sagittal lines differently, corners that fall off unevenly. These are the named families of optical aberration, and their one-sided signatures are exactly what a centered lens does not produce. A camera lens knocked hard enough to move anything will usually still pass a casual glance at the monitor.

On a gauging cell the sharper question is not whether the picture looks good but whether the numbers still hold. The FA lens line is specified with distortion below 0.05% and 10K+ batch consistency, and that budget exists so a pixel maps to a predictable place on the part. A bent adapter ring or a rotated mount seat can consume that budget without any visible softness. A telecentric head is less forgiving still, because its product is constant magnification across depth: after an impact the magnification can drift with part height while the image stays sharp, and every measured dimension moves with it. A machine vision camera lens impact assessment therefore starts from the numbers the system is specified on, not from how the live image presents.

Start with the mechanical pass, not the optics

Quarantine the head the moment the knock is reported, and photograph the assembly as it sits before anything is loosened. Record the serial number and the batch number from the barrel, because both anchor everything that follows. Then work through the mechanics in order: the mount face for burrs or dents, the threads for full and smooth engagement, the barrel for witness marks that show where something rotated or shifted, a gentle shake for internal rattle, and every ring and screw for a feel that has changed. A stiff or gritty focus ring after an impact is mechanical evidence in its own right, and so is a set screw that has backed out of its hole.

On a C-mount camera the flange distance is 17.526 mm, on a CS-mount camera it is 12.5 mm, and the difference is normally made up by a 5 mm adapter ring. That ring is the cheapest suspect on the bench: swap in a known good ring before you condemn the lens, because a ring bent by the same impact will fail every optical check downstream and point the blame in the wrong place. On an M12 S-mount, the FT/HT thread-locked focus lock exists to keep the setting fixed under vibration; after a knock, verify that the lock is still engaged, that the thread is not stripped, and that the barrel has not backed off its seat. If the mechanical pass finds damage, stop there. The disposition is already decided, and quantifying the damage is the maker’s job, not yours.

The four bench checks that settle the decision

Run every check at the working distance and working aperture the line actually uses, with the camera settings, target, and lighting written down so the run can be repeated after refit. These four bench tests for camera lens damage are enough to settle the reseat question, provided each result is compared against a baseline rather than against a feeling. Where a baseline is missing, the procedure below tells you how to test for subtle lens damage after an impact without shipping a good lens away.

First, image a uniform white target and compare the brightness falloff across the field with the baseline. A new dark crescent, asymmetric shading, or shading that rotates with the lens is evidence of tilt or decentering. This check costs one target and a few minutes, and it catches many impacts that a resolution chart alone will miss.

Second, measure resolution at the center, at mid-field, and at all four corners, and judge side to side symmetry rather than peak numbers. Here the shipped paperwork matters. Every Lensmind production batch ships with test documentation covering the agreed family parameters, typically MTF, distortion, BFL, and DOF, together with the batch number, the test conditions, inspection coverage, and an explicit pass/fail judgment. Treat that document as the baseline and the bench run as a re-test of the same quantities at the named conditions. If the curves on the report have always been a closed book, our guide on how to read an MTF chart for a machine vision lens walks through them. Without a report, borrow a known good identical unit and run the two heads side by side on the same target.

Third, verify best focus at the nominal working distance, then cycle the focus five times, approach from both directions, tighten and retighten the lock, and remount the lens three times. Best focus should return to the same place within the tolerance implied by the shipped BFL value. A focus position that walks after each refit means the seat or the threads deformed, even if every image in the cycle looks sharp.

Fourth, image a precision grid and compare distortion, image tilt, and keystone against the specification. On an FA lens the working tolerance is distortion below 0.05%, so the comparison has to be measured, not eyeballed. Any new keystone, any rotation of the grid, or distortion that grows toward one corner is enough to end the reseat discussion.

When the disposition is the maker, and what to send

The list of hard sends is short and mechanical. Send the head back when any of these hold: visible damage of any kind; any bench check outside the family limits in the shipped report; a housing that is sealed or hermetically packaged, including IP67 or IP68 builds, because the knock may have compromised a seal you cannot inspect; a telecentric gauging head, where magnification behavior needs the maker’s instruments to confirm; or any line where the lens feeds an acceptance or calibration trail that treats an impact as an event that voids the record. Read together, this is when to send a camera lens to the manufacturer for repair instead of reseating it, and when the bench is genuinely split, send it. A lens that is slightly off produces intermittent errors that cost far more to find later than a replacement costs now.

The reason is instrument access, not distrust of your bench. The factory re-verifies a returned lens on Trioptics centration and OptiSurf focus measurement, a laser Fizeau interferometer, reflective centering instruments, a precision air-bearing stage, and collimation systems, and it can rebuild alignment with closed-loop 1 μm-class active alignment where the design demands it. No production floor reproduces that chain, and a sharp target picture is not a substitute for it. If the disposition ends in replacement rather than repair, order against the original specification and batch number: repeat orders carry a 30-day written commitment, so the line is not down for the length of a new project.

Whatever you send, send the evidence with it. The pack that shortens diagnosis to a single pass through the instruments contains the unit serial and batch number, a description of the impact covering what was hit, from which direction, and at what stage of the shift, the quarantine photographs, and the four bench results with their settings. Leave the lens in its as-knocked state: do not loosen a retaining ring or experiment with the focus lock, because the original witness marks and preload are evidence the maker can read.

Write the disposition down like a test report

Whatever the decision, leave a document, not a memory. Mirror the structure of the shipped batch report so the two files can be read side by side: unit identity, batch number, test conditions, the four results, and an explicit pass/fail judgment with the author and the date. Note anything that was loosened and retightened, and which checks were run after refit rather than before. A disposition that cannot be reconstructed later is not a disposition; it is an anecdote.

Fold the record into the as-installed baseline for the station, so the next drift comparison starts from a known state instead of a guessed one; our article on building the as-installed record sets out what that file should contain and who signs it. Close the loop on the floor as well: mount seating checks at preventive maintenance intervals, a focus-lock procedure after every adjustment, and physical protection near fixtures, because the cheapest impact is the one that never reaches the barrel. The bench evidence described above is what settles this incident. The record is what makes the next one, if there is one, a shorter conversation.