Skip to content

P7 FEL 12-bit test files, set 2

Latest

Choose a tag to compare

@djnice djnice released this 08 Oct 11:58
· 1 commit to main since this release

Dolby Vision P7 FEL test files with exactly known 12-bit values. Details are in README_set2.md inside the zip.

  • P7_FEL_12bit_gray2.mkv: 12-bit FEL. Gray ramps; "12 bit" text +2 twelve-bit codes in the EL only, over backgrounds at all four 12-bit positions within a 10-bit step; a smooth 2000→4000 nit gradient.
  • P8_10bit_gray2.mkv: the same BL and DM metadata as profile 8.1, without the EL (the 10-bit version).
  • P7_FEL_12bit_rgb.mkv: red, green and blue bars one 12-bit code apart, plus smooth gradients.
  • IPT 4:2:2 12-bit references in the HDMI tunnel format, CSV files with every value, and the scripts that made the files.

3840x2160, 23.976 fps, 120 s, lossless HEVC. Tested on an S928X (CoreELEC FEL composition, hardware EL decoding): all three play to the end, every frame composed with its EL.

Updated 2026-10-11: the IPT reference files in the first upload were wrong (the raw IPT was taken as limited range when it was encoded, e.g. I 1570 became 1536). Fixed; the test videos did not change.

Updated 2026-10-11: added *_ipt.csv files with the IPT values of the references (I, P, T as in the HDMI tunnel). The other CSV files contain the composed 12-bit Y'CbCr (VDR12), not IPT.

Updated 2026-10-12: the EL now carries the same VUI (BT.2020 / PQ), HDR10 SEI and level 5.1 as the BL, like a real profile 7 disc. Before, tools that split the file into a dual-track .m2ts reported the EL as profile 4 (SDR compatible); now it shows as profile 7.6. The decoded pictures did not change.

Updated 2026-10-21: the IPT reference of the RGB file was wrong in the colours: our IPT tool left out the 2 % LMS crosstalk that the Dolby Vision core applies (the "c2" of IPTPQc2). Gray was not affected. Fixed reference and *_rgb_ipt.csv; the fixed tools (ref_ipt, ipt_preview, Windows builds and source) are in fel_ipt_tools_win64.zip.

Updated 2026-10-22: P/T of the references are now coded as 2048 + 4096 × P, so neutral is exactly 2048 as the players output it (before: neutral 2047.5, gray came out 2047). I is unchanged. Updated references, *_ipt.csv, tools, and the set-1 gray reference (set1_P7_FEL_12bit_gray_test_reference_IPT422p12_lossless.mkv).

Updated 2026-10-23: checked against the reverse-engineered dovi.ko output map: 2 % crosstalk for IPT output (0x28f / 2^15), the same LMS→IPT matrix, chroma offset 2048 and full-range scale 4095. P/T are now 2048 + 4095 × P (gray stays 2048, colours move by at most 1 code). References, *_ipt.csv and tools updated.

Updated 2026-10-24: the gradients are now carried by the EL alone over a flat BL, so they step strictly +1 with no dips at the 10-bit steps (the P8 file has its own BL with the gradient rounded to 10 bit). Band C of set 2 now really has BL 512 with the blocks in the EL (before, the BL carried most of them). New *_reference_YUV420p12_lossless.mkv files: the composed 12-bit Y'CbCr, for tools that work in YUV.

Updated 2026-10-25: new near-black test in the zip, P7_FEL_12bit_dark.mkv (12-bit FEL) and P8_10bit_dark.mkv (10-bit): only 10-bit 64…128 (up to ~0.05 nits), 64 gray bars one 12-bit code apart, smooth gray and red / green / blue gradients, "12 bit" squares at black. IPT and YUV references and CSV files included.