Micro OLED Brightness at the Eye | Panel Nits, Optical Loss and Waveguide Efficiency

Panel luminance is the starting value, not the product result. What reaches the eye is that value reduced by duty cycle, angular emission, the optical path and pupil sampling, then measured at a defined eye point. Treating panel nits as eye nits overstates a near-eye display by a large factor, and the error grows with every optical element added.

Nits are luminance

A nit is one candela per square metre, a unit of luminance: light leaving a surface in a stated direction. It is not illuminance in lux, and it is not luminous intensity in candela. Panel cd/m² and room lux describe different quantities and cannot be divided into a meaningful ratio.

A datasheet value also only holds for its test conditions. Pattern size, colour, frame rate, pulse width, drive current, case temperature, viewing angle and meter bandwidth all move the number. Record those conditions beside every brightness figure, or the figure cannot be reused in a second design.

Duty cycle sets the average

Micro OLED panels are often driven with short emission pulses inside each frame. For a rectangular pulse with stable output, the time average is the on-state luminance multiplied by the duty cycle. A panel driven at a stated on-state level for one fifth of the frame therefore delivers roughly one fifth of that value as a time average.

Real output departs from that ideal because of current limiting, voltage drop, efficiency roll-off, thermal control and automatic brightness limiting. Confirm whether a published figure is peak on-state or time-averaged, and at what duty.

Average picture level matters too. A small white symbol on a black field can hold a higher local luminance than a full white field, because the current budget is shared across fewer pixels. Test a small patch, a partial field and full white before trusting a single number.

Losses along the optical path

Emission from an OLED stack spreads over a wide cone, so only part of it enters a small-aperture collimator.^[2] Each surface after that can reflect, absorb, scatter or clip light. An uncoated air-to-glass interface at normal incidence reflects around four percent for typical glass, and the loss compounds with the number of surfaces. Coatings reduce that figure, but only across their designed band and angle range.

Polarisation is often the largest single term. An ideal linear polariser passes no more than half of unpolarised input, and a real polariser plus retarder stack passes less. Where the panel output is already partly polarised the measured result can be better or worse than the simple half-power case, so the assembled stack has to be measured rather than modelled from a catalogue.

Do not model a birdbath combiner and a waveguide with the same chain. They contain different elements at different angles with different loss terms, and mixing them double-counts.

Waveguides are not neutral

A waveguide couples light in, carries it by total internal reflection, expands the pupil and couples light out towards the eye. Each step depends on wavelength, polarisation, incidence angle and position, so a single efficiency figure describes one geometry at best.

Output grating design creates a direct trade-off. Extracting more light near the input edge brightens that region and leaves less for the far edge, so efficiency and uniformity pull against each other. That is why a waveguide is specified by a map across the eye box rather than by one percentage.

Colour divides further. Red, green and blue usually see different coupling and propagation losses, so an equal white panel can arrive at the eye with a colour cast. Report each primary separately, and note that correcting the cast in firmware raises drive current on the weakest channel, which shortens emitter life.

What the eye receives

The defensible number is measured virtual-image luminance at a stated eye point, aperture, field position, colour, pattern and temperature. An end-to-end ratio, eye-point luminance divided by panel luminance, can then be applied to a new panel value, provided both measurements used matched timing and a known pattern.

Pupil geometry is part of that measurement, not a multiplier. A smaller pupil admits less total light but does not reduce the luminance of an extended image. In a real near-eye system a displaced pupil clips part of the ray bundle instead, producing vignetting, field loss and colour shift. Scan a grid of eye positions and report minimum, maximum and mean rather than one centre reading.

Ambient contrast outdoors

Outdoor usability is a contrast problem. For a see-through display the eye receives the virtual image plus the real scene transmitted through the combiner, so the background sets the floor against which the image has to be read. Comparing display nits with ambient lux compares different quantities and proves nothing.

Converting lux to scene luminance needs a reflectance and a geometry assumption: a diffuse surface under a given illuminance produces a luminance of roughly illuminance times reflectance divided by pi. Under strong daylight the transmitted background can exceed the virtual image by a wide margin, and no practical increase in panel drive recovers the contrast.

Dimming the see-through path helps contrast, but it also reduces the view of the real world, which is a safety decision as much as an optical one. Both paths have to be measured separately, because a coating rarely attenuates ambient and display light by the same amount.

Measure at the eye

Put the entrance pupil of the instrument at the defined eye point and focus on the virtual image, not on the cover surface. State the aperture diameter, because a small aperture samples part of the exit pupil while a large one averages it. Move the fixture in small steps; a millimetre of misalignment changes clipping and the sampled rays, not the luminance of the image itself.

Run a repeatable sequence: dark-level check, fixture mounting, nominal eye relief, focus, aperture selection, warm-up, then white, black, grey and each primary, followed by an eye-box scan and a repeat after thermal stabilisation. Report the instrument, calibration date and uncertainty, since alignment commonly dominates the error budget.^[1]

For product work, the useful output is not one number but a small set: panel luminance with its drive condition, eye-point luminance at the nominal position, the end-to-end ratio, an eye-box map, white point and black level, plus the ambient contrast measured against a controlled background. Panels for near-eye systems are usually chosen from a micro OLED range where those values can be requested for evaluation, and the optical architecture itself is usually discussed alongside the AR and VR display options available for the same project.

Panel nits are the source term, not the answer. Publish eye-point luminance with its eye position, aperture, pattern and temperature, and treat every optical efficiency figure as a property of one specific design.

References

1. ISO 9241-307:2008 — Ergonomics of human-system interaction — Part 307: Analysis and compliance test methods for electronic visual displays. International Organization for Standardization. https://www.iso.org/standard/40102.html
2. Hilsum, C. (2010). Flat-panel electronic displays: a triumph of physics, chemistry and engineering. Philosophical Transactions of the Royal Society A, 368(1914), 1027–1082. https://doi.org/10.1098/rsta.2009.0247