DeMura & Pixel-Level Measurement

With the LMK 7 camera and LMK DeMURA software, TechnoTeam introduces a new measurement approach to DeMura correction on today's fine-pitch OLED, µLED and Micro-LED displays. It combines APR for sub-pixel emitter registration, SPR for production-speed capture, and the patented µShaking method for Moiré suppression at sampling ratios optimized for production.

The workflow covers flat and slightly curved emitter surfaces alike — from OLED smartphone panels to curved automotive cluster and center information displays (CID) — because APR teaches the imaging geometry of the actual DUT, not an idealised flat one.

At a glance

• APR — pure software; sub-pixel emitter registration on the sensor
• SPR — production-speed variant of APR; one luminance image per test input signal after teaching
• µShaking — optional on the LMK 7; TechnoTeam's patented method (US 12,387,297 B2) for Moiré suppression at low sampling ratios

Talk to a DeMura specialist

LMK during DeMura measurement

What is APR?

Advanced Pixel Registration (APR) is TechnoTeam's software workflow for pixel-level luminance measurement. A dense geometric registration pattern (point cloud), delivered as a dedicated test image, teaches the exact sub-pixel position of every emitter on the sensor — automatically, without manual alignment. That sub-pixel knowledge is what enables pixel-level luminance measurement at low sampling ratios, without having to match the camera grid to the display grid by optics alone.

APR is pure software: independent of geometric LMK calibration, and technically independent of µShaking. Where the native sampling ratio allows, APR alone is enough; denser panels and tighter working distances benefit from the optional µShaking stage.

Single Shot Pixel Registration (SPR) — the production-speed variant

On a production line the same display type is measured thousands of times, ILMD imaging conditions are stable, and each new DUT only shifts or rotates slightly. Single Shot Pixel Registration (SPR) exploits this stability: after a one-time teaching phase on the first panel of a batch, each subsequent DUT is registered and measured with one image capture per test input signal — the registration pattern is not re-taught for every panel.

Validation against full APR shows Q ≥ 0.98 correlation between the SPR luminance map and the full-APR map, measured across shift, rotation and combined misalignments typical for a production control environment. SPR loses no luminance accuracy without requiring the registration capture on every DUT.

Throughput: APR vs. segmented capture

Per DUT, APR needs one registration image (three for RGB panels — one per colour channel) plus one luminance image per test input signal. Segmented capture, by contrast, needs n² images per test input signal (where n is the row/column stride). After teaching, SPR drops the per-DUT cost to one luminance image per test input signal.

APR + µShaking — Moiré-proof accuracy at low sampling ratios

When the emitter grid and the sensor grid interact at unfavourable sampling ratios, Moiré artefacts become the dominant error source. The prior state of the art suppresses Moiré either by increasing the sampling ratio (high-magnification optics, larger working distance) or by defocusing — both come at a cost: larger setups, slower cycle times, or loss of sub-pixel resolution.

The optional µShaking sensor stage on the LMK 7 — also known as Phase Compensation DeMoiré — decouples the emitter and sensor grids between captures, suppressing both high- and low-frequency Moiré without losing resolution.

Left column: the same panel captured three ways — focused (Moiré clearly visible), defocused (Moiré suppressed, but fine detail smeared), µShaking (focus with phase compensation — Moiré suppressed, detail preserved). Right column: two quantitative verifications along the lines annotated in the captures. Top graph — normalised luminance across a sharp edge on line 1 — focused and µShaking track each other, the defocused curve flattens the edge. Bottom graph — luminance in cd/m² along line 2 in a nominally uniform area — the focused measurement shows the Moiré oscillation, the defocused measurement suppresses Moiré but also washes out the fine structure, and µShaking returns a flat profile at the expected luminance

With µShaking, APR works at approximately 2–3 camera pixels per display pixel, instead of the ~10 cpx/dpx typically needed without — enabling smaller ROIs and denser panels at the same working distance.

Patented correction-data-from-µShaking workflow

The method of deriving correction data from µShaking measurements — so that the subsequent routine measurement runs without shaking — is covered by TechnoTeam's international patent family (e.g. US 12,387,297 B2). The panel is characterised once via µShaking; the derived correction data then lets the line camera operate shake-free in routine production — a key enabler for DeMura on lines with permanent vibrations.

The DeMura production workflow

On the production line, TechnoTeam's DeMura correction workflow runs in three stages:

1. Teach (first panel of a batch only) — APR teaches the sub-pixel emitter position on the first panel. Where Moiré or vibration is a concern, µShaking is run on this first panel too, so the patented correction data can be derived.
2. Acquire (every DUT) — each subsequent DUT is captured with one image per test input signal. Multiple grey levels require one image per grey level per DUT. µShaking does not need to be re-run — the derived correction data takes over.
3. Generate — the LMK DeMURA software computes per-emitter correction coefficients from the acquired luminance data and exports the correction data files.

The workflow runs on each LMK and integrates into the line via TCP/IP. For lab-side APR stitching and characterisation, the same camera runs under LabSoft.

FAQ - DeMura, APR and µShaking

RELEVANT PUBLICATIONS