A wedged PVB interlayer is the bonding film of a laminated windshield, made deliberately thicker at the top than at the bottom so that the inner and outer glass surfaces are no longer parallel and the two reflections of a head-up display (HUD) land on top of each other. Standard laminated glass uses a PVB sheet of uniform thickness, so the two reflections stay separate and the driver sees a ghost image. Wedged PVB interlayer design is not finished when the nominal wedge angle merges the images at one point, though. It is finished when ray-tracing and bench measurements show the ghost stays below the agreed limit at every wedge angle the cutting, relaxation and lamination steps can actually deliver, across the eyebox.
Why a HUD windshield shows two images without a wedge
The HUD projector’s light reflects off both the inner and the outer glass surface, and each surface sends it back to the driver at a slightly different angle. With parallel surfaces the driver sees a primary image plus an offset, usually weaker, secondary image: the ghost. A 1990 patent on wedged windshields notes that the ghost interferes badly with viewing and that neither image alone may be bright enough against the ambient background.
The size of the problem is easy to underestimate. In a worked case published by Lumetrics (Marcus and Ignatovich), a flat windshield 4.9 mm thick at 65° incidence produces a ghost separation of 0.0643°, roughly four times the eye’s angular resolution, which puts the two virtual images about 3 mm apart at 2 m. The 1990 patent shows that separation grows with windshield thickness and with shorter image and viewer distances, and depends on viewing angle and refractive index.
The usual fix is a small wedge angle between the inner and outer glass, made by varying the thickness of the PVB interlayer. The tapered film comes from extruding PVB through an adjustable, wedged slot, or from stretching a film of constant thickness. AGC, Eastman, Kuraray, Saint-Gobain Sekurit, Sekisui and others make wedge interlayers or windshields that use them. A correctly merged image is also brighter: the 1990 patent reports nearly 40% more brightness than the brighter of the two separate images.
How the wedge angle is chosen to merge the primary and secondary reflections
There is no single correct wedge angle for all windshields. It follows from the windshield thickness, the glass refractive index, the local bending radius in the HUD field, the angle of incidence, the observer’s position and the position of the HUD image source. The 1990 patent’s analysis found that both ghost separation and the wedge needed to cancel it increase with thickness, with a shorter image-source distance, with a flatter vertical curvature and with a smaller incidence angle between 75° and 60°.
For the Lumetrics geometry (2,000 mm projector path, 750 mm viewer distance, 4.9 mm glass, 65° incidence, n = 1.518), the best-fit wedge is 0.013°, which brings the separation below the eye’s resolution. Converted to milliradians, 0.013° × π/180 ≈ 0.227 mrad. That sits inside the ranges suppliers quote: approximately 0.5 mrad as a typical value in a Saint-Gobain HUD patent, 0.05 to 2 mrad as that patent’s broad claimed range, and 0.15 to 0.70 mrad as Kuraray’s standard product range for its wedge interlayer.
One wedge angle only fully merges the reflections at a single point (or a line) in the HUD field. For that reason, patents on curved windshields specify a wedge angle that changes continuously across the HUD field. In a Saint-Gobain patent example built on a real windshield and installation position, the angle was calculated at 11 positions over a field about 225 mm tall and falls linearly from 0.482 mrad at the bottom to 0.370 mrad at the top. Lumetrics notes that for most current designs the optimal wedge decreases with distance from the bottom of the windshield, which also covers a range of driver heights.
Two side effects belong in the design brief from day one. A wedge profile tuned to cancel ghosts in reflection can amplify double images seen through the glass in transmission, an effect a Saint-Gobain patent ties to the windshield’s curvature profile. Sekisui also warns that a wedge profile set only for the HUD can create double images in an ADAS camera mounted behind the windscreen. Ray-trace programs such as Code V were proposed for solving the angle math as early as the 1990 patent.
Proving ghost suppression across every wedge angle the line can produce
Our judgement is that ghost suppression requires pairing a ray-trace of the full wedge-angle and coating stack, run across the production tolerance band rather than at the nominal angle, with a standardised bench test on real glass. Once the bench has been shown to agree with the simulation, the qualification protocol built on that pair is meant to allow sign-off without a full vehicle test for every variant.
TechnoLynx can build that coupled pipeline for a glazing or HUD team: implementing the ray-tracing simulation (Zemax or an equivalent) and the wedge-angle optimisation, running it across the full wedge-angle and PVB tolerance range from production stack-up data, correlating its predictions with sunlight and night bench results, and delivering a qualification protocol to sign designs off against. TechnoLynx’s first step with manufacturing and automotive teams is to frame the ghost problem as one coupled simulation, tolerance stack-up and bench-validation problem instead of three separate tasks.
Is there an SAE J1757-2 threshold for ghost separation?
No. SAE states that J1757-2 (issued November 2018) gives measurement methods for HUD optical performance and that setting threshold values for automotive compliance is outside its scope, though it presents some recommended values for reference. A Radiant Vision Systems white paper describes the standard as requiring minimum luminance thresholds. We treat SAE’s own scope statement as the stronger source, so whatever limit applies to ghost separation has to be agreed with the OEM.
Lumetrics’ 0.0643° case is about four times the eye’s resolution. Agree one test method and one threshold definition with the OEM, so that design validation and production QA are never judged against different numbers.
Where the production wedge angle drifts from nominal
The nominal angle does not always survive processing unchanged. Lumetrics measured 0.588 ± 0.011 mrad on a production acoustic windshield, the spread being the standard deviation of local slopes across that one windshield. A 2022 study in Manufacturing Technology followed a wedge PVB roll specified at 0.33 ± 0.03 mrad: 50 stripes averaged 0.333 mrad with a standard deviation of 0.011 mrad, and all 80 sheets cut from the roll stayed inside tolerance.
Relaxation after cutting also moved the wedge. In the same study, the wedge angle of the top five sheets on a pallet fell on average by 0.033 mrad, from 0.343 to 0.310 mrad, mostly in the first two hours, and stopped changing after 48 hours. The authors recommend 48 hours of relaxation, no more than 40 sheets per pallet, and allowing for relaxation when the roll’s wedge is specified.
Bending stretches the panes locally and changes their thickness by a few micrometres where they were heated more, the interlayer stretches slightly during bonding, and both change the wedge progression. Saint-Gobain’s patent warns that even small geometric deviations have a significant effect on HUD optics, and that the wedge profile can usually still be adjusted during the vehicle prototype phase. Where a wedge layer is cut from a larger universal sheet so that the right angle sits at the HUD centre, the cutting position decides which angle lands there. Eastman says its Saflex Horizon LVID interlayer’s improved thickness uniformity keeps the wedge closer to target across the HUD zone, which is the supplier’s claim rather than an independent measurement.
Worked example: putting the wedge tolerance and the ghost budget in the same units
This example uses only the published figures above. It shows the comparison the simulation has to make, not a result for any real windshield.
- Ghost budget: the separation limit agreed with the OEM; J1757-2 sets none.
- Wedge variation, roll as cut: ±0.03 mrad specification, 0.011 mrad standard deviation measured.
- Wedge variation, relaxation: an average 0.033 mrad drop.
- Missing link: how many milliradians of ghost separation each milliradian of wedge error produces at this rake angle, curvature and eye position, which depends on the specific windshield and HUD parameters.
- Decision: multiply the worst-case wedge error by that sensitivity, add the residual separation of the nominal design at the eyebox corners, and compare with the budget.
The missing link is why a nominal-only simulation cannot close this comparison. Wedge error is in the wedge’s units, the limit is in ghost-separation units, and the conversion between them depends on the specific glass and installation.
Covering the eyebox and rake angle before tooling glass
Ghost separation changes as the driver moves within the eyebox, which the 1990 patent attributes to PVB thickness non-uniformity and to the virtual image shifting position. Sekisui notes that the right wedge differs by body type and windshield mounting angle, and that AR HUDs can show double images when the driver’s viewpoint height changes. Larger HUD fields increase ghosting at the top and bottom of the field. Eastman describes AR and large field-of-view systems needing virtual image distance extended from 2 to 3 m out to 4 to 7.5 m.
In the simulation, the outer glass surface has to be modelled to compute the ghost; the Ansys Zemax and Speos HUD workflow computes the ghost optical path and reports ghost results. We would sweep the tolerance band, the rake angle and a grid of eyebox points together, because a design can pass at the nominal eye position and fail at a corner only when the wedge sits at a band edge. The results are only as good as the refractive-index, coating and PVB thickness models fed in, so those inputs need checking against measured samples before the sweep is trusted.
A bench test that can stand in for vehicle testing
J1757-2 covers indoor measurements of the virtual image with simulated outdoor lighting, which is the basis for a sunlight and night bench. Under the standard, ghosting is measured as point deviation from a target virtual image. Radiant argues that spot meters cannot characterise ghosting or distortion because they capture no two-dimensional image, so an imaging photometer is needed. The standard also requires at least three measurements at different locations on the paravan to establish virtual image distance. Software that registers the measured image against the target, such as Radiant’s RADA, can automate the ghost comparison. Under the standard, contrast is measured separately, from luminance on alternating checkerboard patterns in ambient light.
We treat simulation-to-bench agreement as unproven until it is measured on the programme’s own glass. Correlation therefore has to be demonstrated for each programme, and we would set it up in this order:
- Pick bench samples that span the tolerance band, including glass near both wedge limits, not only nominal parts.
- Measure each sample’s actual wedge profile by position (interferometric or contact thickness gauging) before the optical test.
- Run the simulation with each sample’s measured wedge, not the drawing value.
- Measure ghost separation on the bench at the same eyebox points the simulation used.
- Compare predicted and measured separation point by point, and only use the protocol as a sign-off gate once the residuals are small against the OEM budget across the whole band.
Wedged PVB, reflective combiner film or holographic combiner?
| Approach | How it handles the ghost | What the sources support | Watch for |
|---|---|---|---|
| Wedged PVB interlayer | Angles the two surfaces so the reflections coincide | Several manufacturers produce it commercially | Exact only at one point or line; transmission double images; ADAS camera zone |
| 3M Windshield Combiner Film (reflective polariser in PVB) | Selectively reflects P-polarised light | 3M datasheet: Rp 25 to 33% at 60° incidence, Tvis ≥ 78% at normal incidence, target incidence about 55° to 65° | Described by 3M as under development; its claim to remove wedge geometry limits for AR is the vendor’s own |
| Holographic combiner | Not covered by the sources reviewed | None | Needs its own evidence before comparison |
Our reading is that wedged PVB fits when the incidence angle, eyebox and virtual image distance stay within what a wedge profile can cover, while a polarising film becomes worth evaluating when an AR or long-distance HUD pushes geometry beyond that, provided the film’s figures are confirmed on the bench rather than taken from the datasheet.
Setting the wedge tolerance from lamination process capability
The wedge tolerance should be derived from physics, not convention: feed the line’s measured wedge-angle variation into the ghost-separation model, read off the resulting distribution of ghost separation, and set the tolerance that keeps separation under the OEM limit at the required confidence. If current capability, run through the model, already keeps ghost separation inside the spec, keep the tolerance and add monitoring.
The correlation evidence still missing from the literature
The patents tie the wedge to thickness, curvature and incidence angle, and the relaxation data shows an average drop about as large as the roll’s ±0.03 mrad tolerance. Until a programme has measured simulation-to-bench agreement on its own glass, a simulation-only sign-off is a prediction, not a validation.
FAQ
What is the process of glass lamination?
A conventional windshield is made by placing a PVB layer between two glass panes and laminating the assembly in an autoclave. For a HUD windshield, the wedge-shaped thickness profile of that PVB layer carries over to the finished laminate when the panes are bonded.
Which optical design software is used for HUD ghost analysis?
A 1990 wedged-windshield patent proposed solving for the wedge with a ray-trace program such as Code V. A published HUD ghost workflow computes the ghost optical path, needing the outer windshield surface only for the ghost.
