v1.3.1

[1.3.1] - 2026-08-29

Added

• Manual exposure overrides (GL128): Scanner.scan(single_pass_exposure=..., me_short_exposure=..., me_long_exposure=...) send an exact REG_EXPOSURE for the retained single-pass, ME short, or ME long pass, bypassing adaptive selection, PPI clamp, and me_hardware_max_exposure. me_long_exposure takes precedence over me_exposure_mode. Values are validated for the 24-bit register range (1–0xFFFFFF); out-of-range values raise ValueError. All three default to None (unchanged behavior) and never apply to the discarded GL128 priming pass. Scan Lab exposes matching Manual exposure overrides fields.
• GL128 priming-pass override: Scanner.scan(gl128_prime=False) skips the discarded first-scan AGOHOME-park pass for that call (debug/testing only — expect ~30 px of first-scan position drift). Skipping does not mark the scanner as primed, so a later call without the override still primes normally. ScanStatus / on_status gains "prime_skipped". Defaults to True (unchanged behavior). Scan Lab Disable priming pass (debug) applies to Prescan and Scan.

Fixed

• Cropped 8200i SE ENDPIXEL dummy suffix: STR/END are computed in native 7200 dpi clocks (session 03 origin STR 242 / END 10610; same crop keeps STR/END at 1800 and 3600). The last ~24 output columns at 1800 (96 native clocks; image left after mirror_x) are a per-line USB dummy suffix plus the invert-white transition — shrinking ENDPIXEL does not remove them. Decode drops that fixed width after shading; USB line_bytes stays locked to the programmed span. Heuristic post-decode edge trim is not used (pcap 7200 URB-wider-than-span trim stays). IR column flatten still uses inset edge levels and a gain cap so residual edge samples cannot blow up to ~56k. Pass registration fills shift borders from interior columns instead of replicating padding (which widened the IR bright band and caused NegPy ICE over-repaint). Scan Lab surfaces IR align shift in the status bar / IR caption.
• OpticFilm 8100 (V2) geometry/timing: five constants inherited from the 8200i SE without V2 capture proof are now V2-specific — feed_to_scan_steps 13128 (was 13704; fixes ~1 mm clipped at the top of full-frame scans), 7200 dpi LPERIOD 16035 (was 15963), 7200 white-shading strip dummy clock 0x10 (was 0x17), max_image_lincnt_by_feed2 {13128: 29012} (was SE preview entry 4836), and ladder_feed2_steps 13128 (was SE 13560).
• GL128 consecutive-scan AFE strip timeout: after an image pass, stationary AFE START could hang for 5s at status 0xcd (home, buffer empty, lamp on, motor still enabled). Scan Lab reported AFE strip: no stationary data ready within 5s (last status=0xcd) — including 1800→3600 with the same crop (DPI cache miss remesures calib). Stationary acquires now abort START, confirm SCAN/motor-off from hardware (park leaves 0x02 armed), program session-03 dark-strip clocks, write SCAN absolutely, and retry START once. Colour shading remesure reuses existing AFE codes when the last search was colour.

Changed

• GL128 AFE zero-collapse recovery logs now name which offset/gain components collapsed, which fallback was used, and warn when image colour may be degraded.

Contributors

• @TobbyTravel for OpticFilm 8100 (V2) capture-derived geometry/timing fixes and for GL128 manual exposure overrides, priming-pass skip, and improved AFE zero-collapse diagnostics.