SSIM/PSNR figures for surround audio? DTS vs AC3

[robh]

Is there an equivalent of SSIM or PSNR for surround sound audio? This thread got me thinking about that. https://goughlui.com/2016/08/27/video-compression-testing-x264-vs-x265-crf-in-handbrake-0-10-5/ . I may have seen something in the past for MP3 and AAC vs lossless {WAV, ALAC or FLAC} and their various bitrates, but in this case I'm more interested in the common surround formats on DVD and Blu-ray.

It has been said that 1.5mb DTS is better quality than 640kb AC3, but how much better? When a DVD or Blu-ray only offers one, there is no way to compare. But now with lossless audio discs and software that can convert it to DTS or AC3, is there a way to finally compare just how much difference there is versus a pure lossless standard? Likewise, I'd love to know the same for EAC3.

I'm not interesting in double-blind tests, subjectivity, or anything in the analog space. I was wondering if there was a computer program, existing analysis charts, or other methods where the differences can be compared in the digital domain exactly. ie. If a lossless 5.1 audio is 100%, then is 1.5mbps DTS at 95% ? etc.

 

[DVDdoug]

I don't know but I have a shelf-full of DVD concert videos (Only one or two Blu-Rays) and the ones with Dolby Surround or DTS are some of the best sounding, most enjoyable, music I own!

Some of the DVDs have LPCM stereo (lossless) and Dolby Surround, and I'll choose the (lossy) surround sound every time!

I haven't noticed a difference between DTS and AC3 when the discs have both, but the rumor is that sometimes they are different mixes and the DTS may be mixed louder. But I'll usually choose the DTS because it's "supposed" to be better.

I'm not interesting in double-blind tests,

That's the ONLY useful way of comparing lossy compression. That's how psychoacoustic compression is developed & optimized. But with "shiny discs" that's not easy because we don't know if the different tracks are the same mix or the same master. You need a lossy original that you can compress yourself so you know the only difference is the compression. That's not hard with music if you have a CD or other lossless original.

Often with music there is no audible difference in a blind ABX test?, and if you can't reliably hear a difference we can't say the lossless or a higher bitrate is "better". It depends on the program material (some audio is easier to compress), and of course the listener's ability to hear compression artifacts. It doesn't take high-end system to reveal compression artifacts but headphones and/or a quiet listening environment helps.

People often look at spectrum of an MP3 and see the high-frequency loss but we rarely hear that. It's actually easier to get a "pretty graph" than to get good sound. You can tweak the LAME MP3 encoder to keep more high frequencies but that can make it sound worse because lossy compression is going to throw-away information and it's usually best to let it choose what to throw away. If we hear a compression artifact (especially higher bitrates) it's usually something else, often "pre-echo" which is difficult (or impossible) to measure. In normal program material, those highest frequencies are masked by slightly-lower frequencies anyway, and the main way psychoacoustic encoding/compression works is by throwing-away sounds that are masked by other more-dominate sounds, similar in frequency.

 

[robh]

You need a lossy original that you can compress yourself so you know the only difference is the compression. That's not hard with music if you have a CD or other lossless original.

I posted this question because we do have access to lossless originals in modern Blu-rays, and software (ffmpeg) that can compress that same file to either DTS or AC3 within their specs.

 

[MaxwellsEq]

Is there an equivalent of SSIM or PSNR for surround sound audio? This thread got me thinking about that. https://goughlui.com/2016/08/27/video-compression-testing-x264-vs-x265-crf-in-handbrake-0-10-5/ . I may have seen something in the past for MP3 and AAC vs lossless {WAV, ALAC or FLAC} and their various bitrates, but in this case I'm more interested in the common surround formats on DVD and Blu-ray.

It has been said that 1.5mb DTS is better quality than 640kb AC3, but how much better? When a DVD or Blu-ray only offers one, there is no way to compare. But now with lossless audio discs and software that can convert it to DTS or AC3, is there a way to finally compare just how much difference there is versus a pure lossless standard? Likewise, I'd love to know the same for EAC3.

I'm not interesting in double-blind tests, subjectivity, or anything in the analog space. I was wondering if there was a computer program, existing analysis charts, or other methods where the differences can be compared in the digital domain exactly. ie. If a lossless 5.1 audio is 100%, then is 1.5mbps DTS at 95% ? etc.

This is an excellent question. In terms of testing, you could encode an uncompressed piece of music in the two schemes then decode it and use the comparison tools such as Deltawave by @pkane to measure the differences.

It's probably trickier in audio than video because the ear/brain are much less easily fooled than the eye/brain. Even the best compression methods in video produce obvious artefacts which can be seen, but the brain ignores. Whereas with audio, the brain is quite perceptive and so compression artefacts need to be quite small.

 

[robh]

This is an excellent question. In terms of testing, you could encode an uncompressed piece of music in the two schemes then decode it and use the comparison tools such as Deltawave by @pkane to measure the differences.

Yes, there should be software available that will take the raw .ac3 or .dts file and convert it back to .wav (for example) then compare those .wav files against the lossless standard. I'm honestly surprised it hasn't been done, but my Google-fu is failing me.

 

[robh]

Bumping this to grab more eyeballs. Looking for "PSNR" type of software for audio file comparisons in the digital domain.