A lens assembly that measures diffraction-limited on the bench can still produce an image with corners at 40 % of center brightness. The MTF is fine. The coatings are fine. The problem sits in the mechanical stack: a retaining ring with an inner diameter 0.4 mm tighter than the design called for, a barrel shoulder that clips the marginal ray beyond 18 degrees field, a baffle placed to suppress stray light that also removes a third of the off-axis bundle.
This is vignetting, and in a built system it is a mechanical problem at least as often as an optical one. Sequential lens design codes report vignetting factors, but they only evaluate the aperture stack the designer entered. They do not know about the snap ring, the adhesive fillet, or the spacer that came back from the machine shop 0.2 mm thick.
TracePro traces rays through the imported mechanical solid. The geometry that clips the beam in the model is the geometry that will exist in the assembled unit. This article covers how to set up a vignetting study in TracePro, how to pull relative illumination out of the irradiance map, and how to separate true vignetting from the other mechanisms that darken a field edge.
Vignetting is the loss of ray bundle area at off-axis field points caused by an obstruction other than the aperture stop. The consequence is a drop in irradiance at the image plane that grows with field angle, and a change in the shape of the exit pupil as seen from that field point.
Both consequences matter. The irradiance drop shows up directly as corner darkening, and in a radiometric instrument it becomes a calibration error. A spectroradiometer that vignettes 25 % at the edge of its field returns edge measurements 25 % low unless the falloff is characterized and corrected. In machine vision, a 2:1 corner-to-center ratio forces the exposure to be set for the corners, which drives the center toward saturation and compresses the usable dynamic range.
The pupil shape change is more subtle and often more damaging. When a barrel wall clips one side of the bundle, the surviving rays are no longer centered on the chief ray. The centroid of the point spread function shifts, which introduces a field-dependent distortion error on top of the designed distortion. For a photogrammetry or metrology camera calibrated against a distortion polynomial, an asymmetric vignette that varies unit to unit is a systematic error the calibration cannot absorb.
The clipped bundle also changes the effective f-number at that field point. A system designed at f/2.8 that vignettes to an effective f/4 at the corner has more depth of field there and less light, and the corner MTF changes accordingly. Diffraction-limited performance quoted on axis does not carry to a corner that is running two stops slower.
TracePro treats every solid in the model as a potential obstruction, which makes it useful to keep two categories separate when interpreting results.
Aperture vignetting comes from the optical prescription itself. The clear apertures of the elements, the stop, and any designed field stop combine so that off-axis bundles are trimmed by surfaces the designer intended to be limiting. This behavior is predictable from the sequential model and is usually deliberate, since designers often accept 30 to 50 % vignetting at full field to control oblique spherical aberration and coma without growing the front element.
Mechanical vignetting comes from everything else in the barrel. Retaining rings, spacer inner diameters, bonded joints, lens seat chamfers, filter holders, shutter blades, and baffle knife edges all sit in the beam path and none of them appear in a sequential prescription. In a fast wide-field lens these features frequently become the limiting aperture at high field angles even though the optical clear apertures have margin.
Separating the two in TracePro is straightforward because the ray path is recorded. Trace the field point, then sort the intercepted rays by the object that absorbed them. Rays terminating on a surface named for a lens clear aperture belong to the first category. Rays terminating on a retainer, spacer, or barrel wall belong to the second. The second category is the one worth engineering, because those obstructions can usually be moved or relieved without touching the optical design.
Aperture vignetting is corrected by changing element diameters or redistributing power, which means a new optical design cycle and new tooling. Mechanical vignetting is often corrected by a chamfer, a 0.5 mm increase in a retainer inner diameter, or moving a baffle 3 mm along the axis. The second class of fix costs a drawing revision. Knowing which class you are looking at before you convene a design review saves a great deal of argument.
The workflow below assumes the optomechanical assembly already exists in SOLIDWORKS, STEP, or IGES form. TracePro imports all three, and the import preserves the solid geometry rather than a tessellated shell approximation, so a 0.2 mm chamfer stays a 0.2 mm chamfer.
Import the full assembly, including barrel, retainers, and spacers, not just the lens elements. Assign the optical elements their catalog glass and coating data. Assign every mechanical part an absorbing surface property first, even if the real part is anodized aluminum with 5 % diffuse reflectance. Running the first pass with fully absorbing mechanics isolates pure geometric obstruction, which is what a vignetting study is measuring. Scatter from those same surfaces is a separate analysis and mixing the two makes the irradiance map harder to read.
Vignetting is a field-angle phenomenon, so the source needs to sweep field, not fill the aperture from one direction. Use a grid source configured as a collimated beam that overfills the front element by roughly 10 %, then rotate it through the field angles of interest. A typical sweep for a 40 degree full-field lens runs 0, 5, 10, 15, and 20 degrees. Trace each angle as a separate simulation and keep the ray count identical across the set, since relative illumination is computed by comparing flux between runs.
One million rays per field angle is enough for a smooth irradiance map on a detector divided into 100 by 100 bins. Push to five million if you intend to read the pupil shape from the incident ray plot rather than just the total flux.
Place an absorbing surface at the image plane sized to the sensor, and turn on irradiance map data collection. Set the bin count so that each bin subtends a region smaller than the feature you care about. For a 1/2.3 inch sensor at 6.2 mm by 4.6 mm, a 124 by 92 bin grid gives 50 micron bins, which resolves the vignette gradient without leaving individual bins statistically noisy at one million rays.
Relative illumination is the ratio of irradiance at a field point to irradiance on axis, expressed as a fraction or %age. TracePro reports total incident flux and the irradiance map for each run, and relative illumination follows directly.
For each field angle run, record the total flux reaching the detector. Divide each off-axis value by the on-axis value. A system delivering 1.00 on axis, 0.94 at 5 degrees, 0.81 at 10 degrees, 0.62 at 15 degrees, and 0.38 at 20 degrees has a relative illumination curve that falls off faster than natural falloff alone would explain, which points to obstruction somewhere past 10 degrees.
The irradiance map adds information the flux total cannot. Look at the shape of the illuminated region on the detector at each angle. Rotationally symmetric falloff that darkens evenly toward the edge is consistent with natural falloff and clear-aperture trimming. A hard-edged crescent or a straight cut across one side of the map means a specific mechanical feature is clipping the bundle, and the orientation of that cut points at which feature.
TracePro also plots incident rays on any selected surface. Displaying the ray intercepts on the last mechanical aperture before the image plane shows the beam footprint against the physical opening. If the footprint at 15 degrees is a 12.4 mm circle and the retainer inner diameter is 11.8 mm, the source of the loss is identified and quantified in one view.
Edge darkening has three common causes and they respond to entirely different fixes, so identifying the mechanism before redesigning anything is worth the extra simulation run. Natural falloff follows the cosine-fourth-power relationship for a flat detector and a distant object, and it is a property of the geometry rather than a defect. At 20 degrees field, cosine to the fourth of 20 degrees is 0.78, so a 22 % drop at the corner is expected before any obstruction is considered. Measuring 0.38 relative illumination at 20 degrees against an expected 0.78 means obstruction accounts for the difference, and roughly half the corner light is being clipped by something physical. Pupil aberration adds a second-order effect. In a fast system the entrance pupil is not a flat disc viewed at an angle, and the projected pupil area at high field can differ from the cosine prediction by several percent. TracePro captures this automatically because it traces real rays through real surfaces, which is one reason the traced relative illumination curve is more trustworthy than an analytic cosine-fourth estimate.
Scatter works in the opposite direction and can mask a vignette. If the mechanical surfaces carry a realistic BSDF rather than the absorbing property used in the first pass, scattered light adds a low-level pedestal across the detector that lifts the corner reading. Re-run with measured BSDF data applied to the barrel and retainers, then compare. A corner that reads 0.38 with absorbing mechanics and 0.44 with scattering mechanics is telling you that 6 % of the corner signal is scattered light rather than image-forming light, which is a contrast problem wearing the costume of a brightness improvement.
Once TracePro has identified which solid is clipping which bundle, the fixes fall into a short list.
Relieve the obstructing feature. If a retainer inner diameter is the limiting aperture at 15 degrees and the lens clear aperture has 0.8 mm of radial margin, opening the retainer by 0.6 mm recovers the bundle without risking the glass. Re-trace to confirm the new limiting surface is the one you intended. Move the obstruction axially. A baffle that clips off-axis bundles frequently works just as well for stray light suppression 4 mm closer to the stop, where the marginal ray height is lower. Trace the stray light case and the vignetting case together before committing, since moving a baffle toward the stop generally reduces its stray light benefit. Chamfer or knife-edge the aperture. A square-shouldered retainer presents both an obstruction and a bright grazing-incidence reflector. A 30 degree chamfer on the beam side recovers a fraction of a millimeter of aperture and removes the specular ledge at the same time.
Accept and calibrate. In a radiometric instrument, a stable and characterized 20 % vignette is often preferable to a mechanical change that risks alignment repeatability. If the relative illumination curve is smooth, rotationally symmetric, and repeatable across the tolerance range, correcting it in software is a legitimate engineering decision. Run a Monte Carlo tolerance set in TracePro across the mechanical dimensions that drive the vignette, and confirm the unit-to-unit spread in corner relative illumination stays inside the correction budget. A spread of plus or minus 3 % is usually correctable. A spread of plus or minus 15 % is not, and that result sends you back to the mechanical fix Vignetting analysis belongs in the optomechanical model, not in the optical prescription, because the features that clip real beams are retainers, spacers, and barrel walls that never appear in a sequential lens file. TracePro imports the assembly as built, traces field angle by field angle, and reports both the flux loss and the specific solid responsible for it.
The workflow is short: import the full assembly, assign absorbing properties to the mechanics, sweep a collimated grid source through field, read relative illumination from the detector flux, and inspect the beam footprint on the last aperture to name the culprit. From there the decision is whether to relieve the feature, move it, or characterize the falloff and correct it downstream.
Contact Lambda Research to request a TracePro demo and see a vignetting study run on your own CAD assembly.