The Image Darkens by the Third Shift: Separate the Lens From the Cell

An image that darkens through the day is a measurement question before it is an optics question, and the answer is a log, not a theory. The exposure value climbing through three shifts can come from the light source, the lens, the camera, the target, or the cell itself, and the only way to separate them is to fix what you can and watch the rest. Fix the camera settings and watch the light with a meter or a reference target in frame; if the light is dropping, the lens is innocent. Then watch the lens. Before the next order, ask the lens maker for the measured transmission of your part number at your waveband, on production samples, and for the storage and cleaning conditions that keep that number true, because the number is what the shift log is compared against.

Five Suspects, One Symptom

The light source is the first suspect and the one owners most often skip, because it was the newest thing in the cell and sounds innocent. LED output drops as the emitter heats through a shift, and the drop is often invisible until the exposure is compared across the day; a dirty diffuser, a drifting driver, or an aging array add to it. The camera is the second: sensor behaviour shifts with temperature, and a camera whose internal temperature climbs as the cabinet heats changes its response before anyone touches the lens. The lens is the third, and the honest statement is that a clean lens’s transmission is among the most stable things in the cell; when it changes, it changes for a reason, condensation in a temperature cycle, contamination inside the barrel from a compromised seal, or a coating being slowly abraded by cleaning. The target and the cell are the fourth and fifth: a target that changes reflectivity with temperature, and a weld cell whose own ambient light shifts through a hot afternoon. The physics of what transmission actually measures is set out under transmittance, and the reason it matters here is that the number is a curve, not a percentage, so a lens can lose more at one waveband than at another and read as a colour shift rather than a brightness one.

Order matters because the cheap checks are the ones that eliminate the expensive suspects. A reference target taped in view and photographed hourly at fixed camera settings turns the shift log into a chart within a day; a light meter reading at the same moment splits the chart into light-side and imaging-side halves. Only when the light and the target are ruled out does the lens deserve a bench, and only then does a call to the maker become the right move rather than a deflection.

One more suspect deserves naming because it hides in plain sight: the aperture. An electronically controlled iris that is drifting, or a lens whose external control lever is being nudged by a cable or a cover through the day, changes exposure with no change in transmittance anywhere. The check is trivial, lock the aperture by whatever means the lens provides and see whether the drift survives, and it belongs in the first hour of the investigation, before anything is pulled off the machine. In the same family sits focus, a lens slowly losing its focus from vibration changes apparent contrast and can read as darkening on an autofocusing system that compensates, which is why the reference target photograph should be captured with focus locked too.

What the Lens Maker Should Hand Over

Three documents turn the lens from a suspect into a datum. The measured transmission curve for the actual part number, at the waveband the cell uses, from production samples rather than from a generic catalogue page, because the coating stack is where transmission lives and stacks change quietly. The temperature behaviour at the cell’s operating range, since a lens that spends its day at fifty degrees in a weld cell is not being used at the conditions the datasheet assumed. And the conditions that preserve the number: the cleaning specification, the storage conditions, and anything about the barrel’s sealing that decides whether contamination can reach internal surfaces. The cleaning discipline itself belongs to the same document, since the wrong solvent does more transmission damage in a month than the right one does in a decade, and the general principle is the same as any coated element handled in service, described in the coated lens families the line runs on.

The temperature document deserves a sentence of its own, because it is the one the datasheets usually omit. Transmission curves are traditionally measured at one comfortable temperature, and a cell that runs hot is not at that temperature. The useful form of the answer is not a single number but a statement of how transmission moves across the cell’s actual range, from a cold morning start to a hot afternoon, so the shift log’s shape can be matched against a predicted curve rather than against surprise. Ask for the range the maker tested, and if the answer is a laboratory band far from the cell’s reality, that gap is itself worth knowing before the log is written.

The part that the maker cannot supply, the buyer supplies: the as-installed datum. A reference image of a stable target, captured on installation day at recorded camera settings and light readings, is the baseline every later comparison stands on. Without it, the dispute is a matter of memory; with it, a shift is a measurement, and the question of who owns it becomes a matter of arithmetic. The discipline of recording what was installed is the same one that applies to alignment and calibration data, and the wider habit is described in the custom optics programme where reference data travels with the unit.

The Shift Log That Answers It

The log is deliberately small: time, camera exposure value, light meter reading, cabinet temperature, and a note of any intervention. One line per hour for a shift or two tells you the shape of the drift, and the shape is the diagnosis. A drop that follows cabinet temperature and recovers overnight points at the imaging chain’s thermal response, camera included. A drop that does not recover after a cool night points at a permanent change, contamination, abrasion, or a source that has actually aged. A drop that resets when the lens is cleaned points at surface film on the optics. And a drop that appears in the reference target but not in the light meter reading localises the change to the imaging side, which is the moment the parts conversation can start, because the log already says the light was constant.

Two habits make the log trustworthy. Photograph the reference target in the same light as the measurement and never clean the optics as part of taking the reading, because a log entry taken after an unrecorded wipe is the most misleading line in the file. And archive the log with the maintenance record, since the same file that settles this question settles the next one, and in a three-shift cell the next one will arrive.

When the Answer Is the Lens

If the log localises the drift to the optics, the lens maker’s data is what decides the disposition. A transmission measured below the production curve by more than the documented tolerance, with the surface clean and the barrel sealed, is a part that has degraded in service, and the conversation is about why: an environment the specification did not cover, a cleaning chemistry the maker never approved, or a batch question that the maker should own. A lens whose transmission still matches its curve, in a cell whose light has actually fallen, has just been cleared, and the finding belongs to the lighting supplier instead. Both outcomes are wins, and both require the curve, which is why it belongs in the crate rather than in an email chain.

A third outcome is worth planning for, because it is the most common one: the log shows a modest drift that is real, localised to the imaging side, and smaller than anything worth replacing. That is the situation where the right answer is documentation rather than action, a note in the maintenance file that the cell’s exposure now sits a few steps higher than at installation, with the log as evidence, and a periodic check that the drift has stopped. Inspection programs run on thresholds, and a shift smaller than the threshold needs a record, not a purchase order, which is the cheapest and most professional way to close a small mystery.

Lines for the Next Order

Five lines carry the whole fix. The measured transmission curve per part number at the working waveband, with temperature behaviour stated. The storage, cleaning and sealing conditions that protect the number. A reference target and an as-installed baseline image with settings recorded. A shift log format agreed with the line’s own maintenance record. And a stated tolerance for transmission change that separates a degraded part from a lighting problem without an argument. With those lines in place, an image that darkens through the day is a chart with a cause, and the cause is found in hours rather than in a season of replacing parts that were never at fault.