Withdrawn Hisense evidence, 2026-09-22: the user explicitly rejects the earlier video-derived response and scan readings below. The numerical Hisense examples and resulting claims about what its panel can do must not be used for calibration. This page is a historical record, not a current setup recipe. Use the exact model’s manufacturer specifications; see the current U6 Pro investigation.
Audit correction, 2026-09-14: the numerical panel response/scan values and exclusive attribution to the panel below are historical hypotheses, not independently established optical measurements. Successful USB transfers and stable DXGI statistics cannot rule out lens latency, emitter phase error, panel buffering or camera artifacts. The formula is a simplified sample-and-hold bound; it does not describe all BFI/strobed cases. Use PANEL-EXPERIMENTS.md for the current evidence and procedure.
Written 2026-09-13 after fullscreen output failed to fuse on both the Hisense U6 Pro and the Samsung G80SD while the embedded preview window fused on both. The session log for those fullscreen runs shows the presenter locked in direct flip at the measured refresh with single-digit misses and zero USB errors, so the emitter command stream is not the difference. The emitter only ever sees eye commands timed to the predicted vblank; it cannot tell a window from fullscreen.
A panel rewrites its rows top to bottom over most of each refresh. Rows that have finished show the new eye; rows below the scan line still show the old one. The shutter can only be clean while every lit row shows the same eye:
window = hold - response - scan * height
hold: how long the panel keeps receiving the same eye image. One refresh for
Left/Right and Left/Black/Right/Black, two refreshes for Left/Left/Right/Right.response: how long a rewritten row takes to settle. About 0.2 ms on the OLED;
3 to 5 ms on the Hisense VA panel (measured 2026-09-13 with a 240 fps camera).scan: how long the active rows take to be rewritten. The signal timing gives
active lines / total lines x refresh period (about 8.1 ms of 8.33 ms for a
standard 4K 120 Hz mode). The Hisense re-times its panel and measured 6.0 ms at
144 Hz; the app accepts a measured value under Advanced > Panel.height: the fraction of the screen height that carries the image.| Display, mode | hold | response | scan | Whole screen | 40 % band |
|---|---|---|---|---|---|
| G80SD 4K 120 Hz, Left/Right | 8.33 | 0.2 | 8.1 | 0.0 ms | 4.9 ms |
| Hisense 4K 120 Hz, Left/Right | 8.33 | 3.0 | 6.0 | none | 2.9 ms |
| Hisense 4K 144 Hz, Left/Right | 6.94 | 3.0 | 6.0 | none | 1.5 ms |
| G80SD 4K 240 Hz, Left/Left/Right/Right | 8.33 | 0.2 | 4.05 | 4.1 ms | 6.5 ms |
The 1.6 to 1.8 ms shutters in the saved profiles need a window at least that long. On both displays, at 120 Hz with the whole screen lit, there is none. The embedded preview covered about a quarter of the screen height and therefore had a window of several milliseconds. That is the entire difference.
Stereo area (Live tab, under Image alignment; profile fields band and
band_center). The image is scaled about a chosen center and presented in a
band of the chosen height; every row and column outside is exactly black. Black
rows cannot leak into the other eye, so a band behaves like the small window that
already fused, on the real fullscreen path with direct flip and HDR. The same
band is sent to the game hook through shared memory and applied to both eye
blits. In fullscreen, PageUp/PageDown change the height and Home/End move it.
Panel timing. Displays report their active and total line counts from the signal timing; the app derives the scan time and shows the settled window for the whole screen and for the current area. Since 2026-09-13 (night) a numeric model follows every row of the area through the cycle and predicts the other-eye leakage of the current phase for the top, center and bottom. Fit area to shutter picks the largest band whose best phase leaks at most 1 %. Suggest phase picks the least-leakage phase of that model, converted to the emitter’s origin with the measured scan start and USB latency; lens latency is not known, so sweep from there. Advanced > Panel holds the response time, an optional measured scan time, the illumination type and the strobe pulse.
The self-test renders both eye frames with a 50 % band and requires every pixel outside the band to be exactly zero in SDR and HDR, and the pattern to appear inside it.
Hisense at 120 Hz. The panel’s 3 to 5 ms response leaves no whole-screen window at any 4K rate it offers. Use the stereo area; a 40 to 50 % band leaves 2 to 3 ms for the shutter if the 6 ms scan holds at 120 Hz (a 5 s bare-TV 240 fps clip of the Both eye targets pattern would confirm the scan time). Any TV-side black-frame-insertion or “motion clearness” setting is worth a try because it blanks the panel between frames; the app cannot control it.
What software cannot do: make a panel scan faster, settle faster, or show one eye on rows it has not yet rewritten. The stereo area trades screen height for a clean window; the 240 Hz and custom-mode routes trade nothing but require the display to cooperate.