2026-09-22. The user reports MAJOR CROSSTALK after the control-fix build. The requested working U6 Pro BFI outcome has not been achieved. The user requests 240 Hz Mini LED/QLED BFI and explicitly rejects the previous video-derived panel readings. No such readings are inputs to this investigation or the BFI control fixes.
The last saved test-build session reports approximately 239.954 Hz, sequence 1 (L / black / R / black), zero gray reset and zero USB errors. Its final requested phase is 858.8 us and shutter is 6980.9 us. These are host settings/counters, not manufacturer panel specifications or measured lens openings. NVIDIA can clamp the requested shutter depending on phase. This log does not identify the cause of the crosstalk or prove the TV’s internal scan cadence. No replacement timing is derived from the old recordings or the generic panel-response model.
The user then reports a dark band when adjusting phase. This observation does not establish a panel response time or uniquely distinguish panel scanning, backlight behavior and shutter timing. Further widening based on the generic model is not a demonstrated fix.
A separate Hisense-240Hz-preload-isolation.ini diagnostic profile is prepared
beside the isolated executable. It uses L/L/R/R, a 500 us manual shutter command
window, 1500 us initial phase, full image area, 1x gain, and no gray reset, black
lift or ghost subtraction. At the existing 239.97 Hz input it submits about 60
new images per eye per second. Those shutter/phase values are search starting
points, not manufacturer panel specifications. This compares preloaded images
against black insertion; it is not BFI or a verified solution. Use nine-row
targets and inspect each lens across the whole screen while sweeping phase.
If a whole-screen visible/clean range exists, record it; a dark image is no pass.
The user identifies the TV as 65U6SF-PRO, 65-inch US Fire TV, Best Buy SKU 6673625. The model is now resolved. Do not use specifications from U6SV, U6Q PRO, U6N, or the non-Pro U6SF.
Reviewed primary manufacturer sources:
The matching retailer SKU lists Game Booster 240, but the inspected Hisense Q&A and setup guide do not supply its resolution/timing table. The app’s existing 2560 x 1440 at approximately 240 Hz log is evidence of the host input mode, not manufacturer documentation of internal panel scanning or pixel transitions. The 144 Hz native specification does not establish that the accepted 240 Hz input is dropped or interpolated; no such claim is made.
Neither inspected manufacturer document supplies gray-to-gray transition curves, black-reset settling, local-dimming latency, or an API to synchronize backlight strobes with shutter glasses. No numerical panel timing or a working BFI preset can be derived from these documents. Lack of a documented backlight API is not proof that a particular TV menu feature is absent.
Local reference copies and rendered pages are in
reports/hisense-65u6sf-pro-manufacturer/. The original setup-guide PDF uses
outlined text, so its port diagram and specification table were read visually.
Existing profile response/rise values remain unverified and are not treated as
manufacturer measurements. Saved user profiles are preserved.
At 240 input refreshes/s, L / black / R / black submits each eye 60 times/s, with 4.167 ms per input refresh. This is sequence arithmetic. It does not establish internal panel scanout, pixel settling, local-dimming response, or lens transparency.
The current changes fix two control-state bugs:
No new Hisense response time, scan time, backlight-control capability or calibrated BFI preset is claimed. Software build/tests cannot establish optical separation. Manufacturer refresh and response specifications alone are not a measured shutter-to-panel timing relationship.
The isolated build is build/u6-bfi-control-fix/bin/Release/VisionRestoration.exe.
The ordinary running app and its saved calibration files are not replaced.
Build targets: VisionRestoration, vision_panel_tests, vision_tests, and
vision_lcd_tests. Regression coverage includes gray reset to black insertion,
preserved OLED timing, switching from a long guarded BFI aperture to manual panel
experiments, and optical-mark invalidation on timing-mode changes.
Validation on 2026-09-22: Release build passed; CTest core,
lcd_temporal_aperture, and panel_experiments passed (3/3). The app’s
--smoke-test --panel-smoke exited successfully with no emitter commands.
No TV/lens result is inferred from those tests.
Selecting Panel > Black reset L B R B in the corrected build remains an experiment, not a verified Hisense setup. Model identification is complete; clean stereo remains unresolved. The pending optical result is whether the 500 us manual L/L/R/R preload comparison removes the dark band and produces visible separation across both lenses and all screen rows. Manufacturer refresh specifications alone cannot answer that question.