Closed junh1024 closed 1 year ago
This was experimented with before, around 3adb5b3042218fb0e6314765a596d56a01e6ff47. The Hilbert-transform is implemented in this exact commit, but all results were distorted. Transforming the phases to any direction made it worse.
In the ETSI specification for object E-AC3, section 6.3.2.2 joc_dmx_config_idx, configurations 3-4 aka 5.1 90 & 5.1.2 90, have a 90 degrees phase shift pre-applied to the surround channels on the 5.1/512 core. A proper atmos decode apparently reverses this phase shift before Atmos reconstruction. Configurations with no phase shift should remain unchanged. Your software should mimic this behaviour. I'm not sure in which direction the phase reversal should occur but you can match this with the original. The reference file is in the SMPTE-Dolby order
test reference