AR glasses are a wiring problem before they are an optics problem. The electronics have to live in a temple or a bridge that a person will wear for hours, so every conductor crossing the hinge costs weight, stiffness and a flex failure mode. On a 0.96-inch blue monochrome PMOLED the interface choice is therefore part of the mechanical design: both serial options work, but they differ in how much copper the frame has to carry and how the host's timing budget is spent.
Key takeaways
- The bus that crosses a hinge should be the one with the fewest conductors, because every line adds stiffness and a fatigue point.
- SPI offers deterministic timing and a faster full-frame update, at the cost of two or three extra signal lines.
- I2C shares two wires and a bus that may also carry sensors in the frame, which saves copper but adds addressing and pull-up constraints.
- In a near-eye product the display usually shows sparse information, so the average update rate is low and the faster bus is rarely the deciding factor.
- Whichever interface is chosen, the display driver belongs on the panel side of the hinge so the flex carries the low-speed bus rather than the high-speed one.
Glasses frames are unforgiving mechanical systems. A temple has to fold, flex slightly when worn, and survive being put down on a table. Any cable that crosses the hinge is a long-term reliability item, and the fewer conductors it carries, the better the design behaves in the field.
That reframes the interface question. A 0.96-inch PMOLED in blue monochrome draws little current and displays simple information, so the display itself will work with either serial option. What differs is the number of conductors, the electrical environment inside a small frame, and how the host firmware divides its attention.
The sections below compare the two buses for a wearable frame, describe what actually crosses the hinge, and explain how the choice interacts with the sensors and the processor that a pair of AR glasses also has to carry.
What the two serial buses cost inside a glasses frame
SPI separates clock and data and uses a chip select per device, so a display plus one sensor needs at least four signal conductors in addition to power [1]. That is fine on a rigid board and awkward in a temple, where the flex has to bend around the ear and across a hinge.
I2C uses two conductors for every device on the bus, so a display and an inertial sensor can share them [2]. The trade is speed and bus management: the host must handle addresses, and the pull-ups have to be sized for the capacitance of a flex that runs the length of the frame.
Both buses are slow by modern standards, and neither is a burden for a status display. The decision is therefore dominated by conductor count, hinge reliability and whether the processors already needed an I2C bus for other components.
Interface options for a 0.96-inch PMOLED in AR glasses
| Option | Conductors across the hinge | Effect on the frame | Where it fits best |
|---|---|---|---|
| SPI, driver on the panel side | Clock, data, chip select, command line plus power | More fatigue points, but simple firmware | Designs with a flexible bridge and few sensors |
| I2C, shared with sensors | Two signal lines plus power | Fewest conductors and a lighter flex | Frames that already carry an I2C sensor bus [2] |
| Serial bus plus local controller | Low-speed link only, if the panel electronics are local | Best mechanical behaviour, more electronics in the temple | Products that can afford a small driver board near the optics |
| Parallel interface | Not practical in a folded frame | Too many conductors for a hinged product | Bench evaluation only |
Update rates, power and perceived responsiveness
How much data an AR interface really moves
A 128x64 monochrome frame is about a kilobyte, and most updates change a fraction of that. Text prompts, an icon and a battery indicator are the typical content, so the bus spends most of its time idle. That is why a slower shared bus is usually acceptable and why the conductor count should win the argument.
Where the processor budget goes
AR glasses are expected to run optical tracking, connectivity and audio processing. Every function that blocks the main loop competes with those tasks, which favours a display update that can be queued and completed without long critical sections. Keeping the panel work in a small driver task isolates it from the rest of the system [3].
Designing the display path as part of the frame
The best mechanical result usually comes from placing the display driver close to the panel and running a low-speed link back to the main processor. That keeps the flex narrow, reduces the number of conductors that have to survive folding, and moves the high-speed signalling to a short rigid section.
It also changes the thermal picture. A small driver board near the optics adds a heat source close to the face, so the layout should keep it away from the temple contact point and give it a path to the frame. Wearable technology lives or dies on comfort, and warmth is more noticeable than weight in a device worn for hours [4].
Products that use DisplayModule supply for the panel usually settle the question by taking the published interface definition and the outline together, then designing the flex around the module rather than forcing the module into an existing harness. That order avoids the most common wearable mistake, which is discovering the connector orientation after the frame tooling is cut.
Bend a spare flex through the hinge range a few hundred times before the frame design is signed off. The display will work perfectly on the bench and fail after a month of folding if the flex is routed tightly, and the failure appears as a single dead row rather than as an obvious break.
Sources
Frequently asked questions
Is SPI or I2C better for a display in AR glasses?
I2C usually wins on mechanics because it uses the fewest conductors across a hinge, provided the bus has room for another device and the pull-ups suit the flex capacitance. SPI wins on timing determinism and needs no addressing. For a status display at low update rates, the conductor count is the more valuable saving.
How fast does a 0.96-inch PMOLED need to update in a wearable?
Most near-eye information interfaces change a few times per second at most, and a partial update of a few characters is a very small transfer. The bus speed only becomes interesting during initialisation or a full-screen mode change, which are rare events.
Should the display driver be placed near the optics or on the main board?
Near the optics, where the conductor count and the flex stiffness matter most. That choice moves a small amount of heat into the temple, which should be considered alongside comfort, but it makes the mechanical design considerably easier to qualify.
What happens if the display bus fails in a wearable?
Treat it as a recoverable fault: reinitialise the bus, reinitialise the panel and report the event. A wearable that shows nothing but still runs is easier to diagnose than one that resets when the display hangs, and the recovery code is small.