[Help] identifying encoded but unflagged matrixed surround mixes (2.0/4.0) - (5.1/6.1)

[kiyoakiwah]

i see that there's a lot of DVDs/BDs that have matrixed 2.0/5.1 tracks but they aren't flagged so they just show up as standard 2.0/5.1... which shouldn't be the case and it's annoying
is there any way to identify them if they're not labelled on the product? if you have knowledge about the matrix encode you can just force it on the decoder.

this seems to extend to D-VHS AC3 tracks too... probably not the MP2/MP1 stereo tracks though.
i've even seen mentions of LDs that only show stereo, but are actually 2.0 surround sound.
PCM, DTS/DTS-HD MA, AAC-LC and maybe even E-AC3 seem to do this as well. god knows what!

doesn't seem to be a problem for 5 track Cinema DTS as it's indicated in the serial number (65XXX)

[zoidberg]

I guess the simplest way to find out is to force the decoding, it will be pretty obvious if a 2.0 mix is a true Lt-Rt mix as you will either get distinct content in the L-C-R-S channels or you won't (likewise a Surround EX/ES title with it's surround back channel). Usually a film will have either a Dolby Stereo or Ultra Stereo credit during the end credit roll as well.

You will find that with a lot of recent releases, the 2.0 is actually a mix-down of the 5.1 mix. Occasionally it will be the original Lt-Rt but mostly they are mixdowns for soundbars and such.

[swhitt]

Late to this but I went and had a proper look, because the answer turned out to be more annoying than I expected.

For AC-3 there's a flag that's meant to tell you, a two bit dsurmod field in the BSI, only present on 2/0 tracks. I couldn't find anything that surfaces it, ffprobe doesn't seem to have an option for it, so I ended up parsing it out of the syncframes directly. Ran that over my own shelf, 100 AC-3 2/0 tracks across 29 films.

It's mostly just not set. 79 of the 100 say "not indicated". 16 say Dolby Surround, 4 say explicitly not, and 1 is sitting on the reserved value, which shouldn't happen.

The 16 don't hold up that well either. 9 of them are commentary tracks, and I don't think anyone mixed a commentary in Dolby Surround, so that's probably an encoder default rather than anyone describing the content.

So I tried measuring instead, correlation between L and R per second, sampled across four spots in each film.

One thing I had wrong going in. I assumed a real Lt/Rt would go negative on surround heavy moments. It basically doesn't. Across the flagged tracks only about 2% of blocks go negative, and the unflagged ones sit at 1.4%, so I don't think that's usable as a tell.

What does move is the overall level. The tracks flagged explicitly not Dolby Surround average 0.985, so that end behaves. But the flagged ones average 0.833 and the unflagged pile averages 0.846, which is basically the same number. Which might be your answer, the flag doesn't look like it's tracking anything real.

The bottom of the list is the interesting part. My lowest correlation tracks are nearly all old Star Wars laserdisc era stuff, and they're a mix of flagged and not:

Code:

ROTJ Italian 2.0 0.482 not flagged ANH 1993 laserdisc mix 0.623 flagged ANH 1993 stereo remix 0.623 not flagged ANH 1985 laserdisc mix 0.673 flagged ANH 1985 stereo remix 0.692 not flagged

The two 1993 ones come off different projects, the flagged one from Harmy's and the unflagged one from 4K77, and they land on the same number to three decimals, so I'd assume that's the same underlying mix carried twice with the flag set one time out of two. Which is about as clean an example of your problem as I could have asked for.

Caveat I can't get around though. Low correlation only really means a lot of L minus R energy, and wide stereo music does that too without being matrixed at all. Two of my three lowest are isolated scores, which I'd guess are just wide rather than matrixed. So I think this narrows the search rather than settling it.