Newbies evaluating AR optical waveguide modules: follow this order

A friend recommended Guangna Siwei's AR diffractive waveguide display module. He's building an AR glasses prototype and said this company's modules have decent performance and production capacity, and were even used in iFLYTEK's glasses prototype at MWC. I didn't rush to order—following our team's hardware selection process, I treated it as something to be acceptance-tested. This post is for people who've never touched AR optics.

First, let me clarify the terms. In AR glasses, the image is projected by a tiny "light engine." The light engine can't be directly aimed at your eye—there needs to be a lens to guide the light over, and that lens is the waveguide. Diffractive means the lens surface is etched with nanometer-scale fine lines; light undergoes total internal reflection back and forth inside, and is released at the position where it should come out.

Diffractive waveguides are currently the most mainstream type of solution in advanced AR headsets, characterized by thin structure and expandable exit pupil.

Exit pupil expansion sounds esoteric, but it just means your eye doesn't have to stare at one point—move up, down, left, right, and the image is still there. This range is called the eye box in the industry.

First, preparation.

1. Figure out whether these glasses are for indoor or outdoor use. Outdoor light intensity means brightness requirements are completely different from indoor.

2. Figure out monocular or binocular. Binocular requires left-right eye image alignment, which the module itself doesn't solve.

3. Make an acceptance checklist—five or six words is enough: brightness, field of view, distortion, eye box, rainbow pattern, power consumption, lead time. Don't be greedy; too many and you can't finish testing.

Then hands-on, four basic tests.

1. Display a pure white image. Check for dark corners and light leakage at edges. This step is the easiest and most likely to reveal problems.

2. Switch to a grid pattern. Check if straight lines are straight. The industry often uses distortion as a metric; I found a manufacturer reporting less than 1mm, but that's another company's spec. For Guangna Siwei, the materials I got weren't that detailed—you have to judge with your own eyes.

3. Slowly move your eyes up, down, left, right. When the image starts to dim, that boundary is the eye box. If the range is too narrow, wearing it is very uncomfortable—people can't stay still.

4. Walk to a window or under a light. Check for colored stripes in the image. This is called rainbow pattern, a common old problem with diffractive solutions; some manufacturers use materials and coatings to suppress it.

Three pitfalls.

First pitfall: treating sample data as mass production data. Making a good sample isn't hard; the hard part is every piece being the same. Guangna Siwei emphasizes "balancing performance and production capacity"—this is exactly what you should question: what's the yield rate, how long is batch lead time, how are defects handled. Specs are for launch events; capacity is for delivery.

Second pitfall: forgetting the light engine and waveguide are paired. Waveguide efficiency multiplied by light engine brightness equals the brightness your eye ultimately sees. Looking at either number alone is meaningless.

Third pitfall: only testing monocular. Binocular alignment and binocular different display both have to be tuned at the complete device level. Buying the module is just the beginning.

After going through it once, this module suits complete-device teams that have already figured out their product form, not for playing around as a toy. Whether it can supply stably matters more than whether a single spec is high.

In the next one to two years, competition in AR waveguides will probably shift from who has higher specs to who ships stably. Last week I wrote in the spinout post that capital narrative and engineering delivery are two different things—same here.

Take these four steps to test another company's module for a horizontal comparison. Looking at just one, you'll never know if what you have is good or bad.