The great lossless vs. dts 24/96 shootout!

[jimfisheye]

Quick version:
The dts 24/96 format is often criticized for poor sound due to its lossy nature.

The ringer is the dts 24/48 decode. The dts 24/96 format is apparently obscure enough that some decoder apps, media player apps, and hardware disc players only see it as 24/48. There is no reason to be critiquing the core dts 16/48 "legacy" decode right now - it is what it is - so ignoring that here.

Digital files that are identical (same ones and zeros) will sound exactly the same. If you think you hear something different, it's either your perception playing tricks or equipment malfunction.

When you subtract an exact copy of a digital file from itself; just like 1-1=0, the result is all zeros which is perfect silence. The more signal you hear and see in the result of that subtraction, the more different the files are.

There are five 8 second files in the quick sample folder:
loud_sample_LPCM = the lossless master for reference
loud_sample_DTS2496 = the DTS24/96 fully decoded for reference
loud_sample_LPCM-DTS2496 = difference from lossless LPCM and DTS24/96 fully decoded
loud_sample_DTS2496-DTS2448(src96k) = difference from DTS24/96 fully decoded and a lesser DTS24/48 decode
loud_sample_LPCM-LPCM(src48ksrc96k) = difference from a sample rate conversion only on the same source (for reference and sanity check)

The stereo versions are just the front L,R channels of the 5.1 files. Odd to hear isolated but plenty sufficient to quickly hear what's going on.
You can even clearly hear the artifacts (or lack thereof) after converting these sample files to mp3.

Download here -> https://copy.com/DIR8FXWdBhKcXboO
Note the quick folder for short attention spans.


Long version:
There are two 8 second samples from a song.
Then a number of different null (difference) tests run between lossless master and the different dts decodes of them them.
You can knock yourself out comparing everything. I made the obvious difference files. All the files for the different points of reference including originals are there. Prove it to yourself.


One more aside:
You can't instantaneously A/B audio samples at different sample rates without setting up 2 hardware systems. On a single audio system, it would take a moment to switch the hardware. Point being that it really helps to have instant switching to hear something subtle.

The next best thing is to take the file that was downsampled (or a lower sample rate to start) and upsample it to 96k. Now you can instantly A/B between it and the original with no clock change.

Spoiler: You still don't even hear this even though you 1. just made it twice as bad and 2. now you can even instantly A/B between them and REALLY listen for stuff


My conclusions:
I hear nothing from the sample rate only (for reference) difference file. Not without deadly volume increases anyway. I see a little movement on the meters though.

lossless vs. dts 24/96 fully decoded
I don't believe I hear anything A/Bing the files. The difference file has some audible stuff from the cymbals though.

dts 24/96 fully decoded vs. dts 24/96 only 24/48 decoded
I can hear this A/Bing if I'm paying attention. Not night and day or anything. I'm hearing the entire mix in the difference file. Quieter but fully audible.

The lesser 24/48 decode of this is FAR FAR more damaging than a 96k to 48k sample rate conversion.
The lesser 24/48 decode is also FAR FAR more damaging than dts 24/96 is vs. lossless.

If you're hearing something wrong from dts 24/96, then you're hearing the lesser decode.

And if you're hearing anything besides your ears ringing after the piercing high end on the brick wall limited Aqualung release - which is shockingly unlike any other Steve Wilson produced recording I've ever heard - you're doing better than me! This overshadows every last thing mentioned above.

The point?
You may only be a more accurate decoder away from enjoying DTS 24/96 releases. Which still will not help the Aqualung box set! (For the record, the flat stereo transfers are the real thing at least.)

 

[Jim the Oldbie]

Nice work!

My own similar (2-channel, mp3) tests pretty much confirm your conclusions above. I would make only one humble addendum, learned from some older sage back when I was making my own difference files: It's not the amount or level of artifacts (difference) that's most important, but rather how well they are perceptually masked when listening to the actual lossy file. In other words, the smarter the perceptual coding algorithm, the better the output will sound, even with the same amount of (or even more) artifacts in the difference waveform. Having said all that, there does usually seem to be a pretty good (inverse) correlation between the level of the difference sounds and the perceived sound quality, especially within a given encoding environment (like dts).

I've been using FLAC on the big system at home for years now, but at first I was using one of the moderately high-rate LAME VBR mp3 presets for all my portable stuff. After awhile I began to notice a fair amount of crap in the sound of my (not crap) car system. Then I started hearing the same thing when using my portable player to play break music over the PA when I mixed my friends' bands. At first I didn't even think to suspect mp3 artifacts, since the preset I had chosen to encode my collection (can't remember the name, sorry) was deemed pretty much transparent by the HydrogenAudio nerds (no disrespect to those guys, heh). But eventually I realized that I was playing it LOUD in the car, and sometimes at gigs too, if the joint was jumpin'. I now believe that listening so far above "reference" level (whatever that is) was allowing me to hear the low-level mp3 grunge, just like you did.

So I switched to FLAC for the portable stuff for a time. But after some more listening tests, I've since gone back to 320K CBR mp3s on the road, for several reasons: First, I don't hear the grunge (or any other differences) at that bitrate, even when I crank it up. Second, it immediately doubled my tunes capacity, since 320K mp3 takes up roughly half the space of FLAC. Third, those 320K difference files are just wa-ay down there.

-- Jim

 

[jimfisheye]

No argument on qualifying the analysis of any artifacts you hear.

That the gross differences noted are obvious without deeper analysis makes its point though I think.
We have virtual silence vs. swishy swishy swishy vs. clearly hearing the mix.


Yeah, flac is convenient for sure. I use it all the time. Takes half the space and is lossless. Hard to argue with that!
Playing an mp3 on a PA system stands out like playing a youtube quality video on a large screen.

I've never had anything against portable quality (mp3 and all) for convenience either. My only problem is when the master version is never released.

 

[Jim the Oldbie]

...I've never had anything against portable quality (mp3 and all) for convenience either. My only problem is when the master version is never released.

Yes! I worry about this quite a bit. With all the (lossy) streaming & downloading going on these days, I do fear that non-lossy audio release versions are being steadily marginalized. I hope I'm wrong.

-- Jim

 

[user 4725]

For those of us who listen to music via a PC, I can recommend Foobar 2000 and the ABX Comparator plugin for double-blind comparison between two tracks.

https://www.foobar2000.org/components/view/foo_abx

 

[DennisMoore Jr.]

Hi Jim,

Thanks for your detailed & informative post.

Do you know how to find out if one's receiver down samples DTS? I would guess this shouldn't happen with any mid quality or better receiver released within the last 5 years? What do you think?

 

[jimfisheye]

Hi Jim,

Thanks for your detailed & informative post.

Do you know how to find out if one's receiver down samples DTS? I would guess this shouldn't happen with any mid quality or better receiver released within the last 5 years? What do you think?

Actually, from what I've seen combined with the reviews I'm reading on these discs, I kind of expect many receivers to only do the lower quality decode.

I can tell you that the app XBMC Media Center for example, does the lower quality decode and displays DTS 24/96 the whole time.

You have to notice it. And again, this is FAR more apparent than just a sample rate conversion - think of that more as a telltale. You have to notice the 'sheet hanging in front of the speakers', go investigate, and notice the 48k data stream for your first clue.

So... I noticed that. Recorded the output from XBMC. Compared it to a proper decode. Busted! Now I know not to use XBMC for dts 24/96 discs. If something comes along with video + HD audio in that format, I'll rip it to a MKV file first.

The decoder apps (Like DVD Audio Extractor or DVDAExplorer) will show the content as DTS 24/48 which tells you right away they are incompatible.

I have no doubt that there will be receivers out there displaying "DTS 24/96" and only doing a lower quality decode.

It looks for all the world like the intention was to charge more money for the full quality decode in order to sell "pro" machines and apps for a higher price. The hardware (your DA converters) is still sitting right in front of you but if you don't pay the premium, you don't get to fully hear the content on the disc you just purchased. Greedy!

I haven't checked for new software for this for a few years now. (Was hoping that new version of DVD Audio Extractor that they finally got up to speed for many other things would do it, but no.)

Here's what I found and still use:
There was this media player app called ArcSoft something something that sold for around $400 and included the full decoder for DTS 24/96.
This codec ended up getting licensed to some other video app that only ever made a Windows version.

Cheap Mac user that I am, I grabbed the free Windows version of the codec and kludged it into this other free Windows app called AudioMuxer.
The app Wine lets you just run any Windows app directly in OSX. But it's still a Windows app and you still had to kludge around and do a bunch of faux registry additions and deletions in just the right order.

I'll post instructions after getting my old notes in order.

Unless of course someone more on the ball would care to suggest a newer app that just works without all the screwing around.

Sorry, if that's more of a rant than helpful!

 

[Jim the Oldbie]

...Unless of course someone more on the ball would care to suggest a newer app that just works without all the screwing around.

If there is one, I haven't found it either. I'm doing the same as you, with the patched ArcSoft dll hacked onto AudioMuxer. And I'm afraid my notes are nonexistent. I just did a whole lotta Googling until it started working. For those who go this route (ripping to FLAC), you can use MediaInfo or equivalent on your output files to tell you if you're getting it right.

-- Jim

[EDIT] Actually, too much detail on this subject might run afoul of the TOS here anyway?

 

[DennisMoore Jr.]

Actually, from what I've seen combined with the reviews I'm reading on these discs, I kind of expect many receivers to only do the lower quality decode.

I can tell you that the app XBMC Media Center for example, does the lower quality decode and displays DTS 24/96 the whole time.

You have to notice it. And again, this is FAR more apparent than just a sample rate conversion - think of that more as a telltale. You have to notice the 'sheet hanging in front of the speakers', go investigate, and notice the 48k data stream for your first clue.

So... I noticed that. Recorded the output from XBMC. Compared it to a proper decode. Busted! Now I know not to use XBMC for dts 24/96 discs. If something comes along with video + HD audio in that format, I'll rip it to a MKV file first.

The decoder apps (Like DVD Audio Extractor or DVDAExplorer) will show the content as DTS 24/48 which tells you right away they are incompatible.

I have no doubt that there will be receivers out there displaying "DTS 24/96" and only doing a lower quality decode.

It looks for all the world like the intention was to charge more money for the full quality decode in order to sell "pro" machines and apps for a higher price. The hardware (your DA converters) is still sitting right in front of you but if you don't pay the premium, you don't get to fully hear the content on the disc you just purchased. Greedy!

I haven't checked for new software for this for a few years now. (Was hoping that new version of DVD Audio Extractor that they finally got up to speed for many other things would do it, but no.)

Here's what I found and still use:
There was this media player app called ArcSoft something something that sold for around $400 and included the full decoder for DTS 24/96.
This codec ended up getting licensed to some other video app that only ever made a Windows version.

Cheap Mac user that I am, I grabbed the free Windows version of the codec and kludged it into this other free Windows app called AudioMuxer.
The app Wine lets you just run any Windows app directly in OSX. But it's still a Windows app and you still had to kludge around and do a bunch of faux registry additions and deletions in just the right order.

I'll post instructions after getting my old notes in order.

Unless of course someone more on the ball would care to suggest a newer app that just works without all the screwing around.

Sorry, if that's more of a rant than helpful!

Rant? Are you kidding!?!?!?!

Love everything that you wrote.

I will need to check with Onkyo about the decode. Its a 2013 TX NR717. It has an outstanding DAC and Onkyo doesn't sell software so they would have no reason to hold down the sample rate, but you never know, until you find out and then you know!

FWIW, I have always loved and have never had a problem with all the DTS 24/96 music I own. Jethro Tull, The GENESIS Box Sets(though I have the SACD version as well.)...

Thanks again for more fabulous info. Sy Simms from New Jersey said years ago(in a horribly nasal tone):

"An educated consumer is our best customer"! Well that is true about us here as well.

 

[Ge Someone]

I will need to check with Onkyo about the decode. Its a 2013 TX NR717. It has an outstanding DAC and Onkyo doesn't sell software so they would have no reason to hold down the sample rate, but you never know, until you find out and then ...

Although the Onkyo's even back since 2005 include DTS 96/24, be aware that when you engage the Audyssey (room) EQ, the output will be resampled to 48 kHz. This goes for all receivers equipped with Audyssey (Mult)EQ.

Funny isn't it?

 

[neil wilkes]

You will also get 48k downsampled audio under 2 more conditions as well:
1 - If your player is running the DTS-HD Essentials decoder (some players will not output anything at all unless the core is specifically set to 48kHz, although this mainly applies to DTS-HD MA)
2 - if the player is set to 48kHz output - not all players can output at 96kHz

 

[jonmchugh]

Tested this today, Running Kodi on Librelec with SPDIF soundcard, set DTS to passthrough mode, Connected to marantz SR4600 (old) which displays DTS 96/24 on the display. Analogue pre-outs for front channels connected to PC with 192dac and recorded into computer. Analysis with Spec shows frequencys up to 96khz sampling rate equivalent are present and correct.

 

[jonmchugh]

I've been do a lot of experimenting lately. I've found a flaw in DTS-HD's MAS encoder. While the full DTS-HD is totally lossless just like FLAC, the DTS core embedded inside this is pretty poor up when you have used 96khz or 192khz audio as a source.

Its not a problem if you've encoded from a 48khz source but when 96khz is used as the source the encoder just chops it in half without any attempt to gracefully resample the 96khz source to get a 48khz core, the result is a core with lots of resamping artifacts and a bad sound.
My advise is to do a 96khz DTS-HD encode and then use a good resampler like SOXR to make a separate 48khz source and then encode this as DTS standard separately for use on legacy equipment.

If you have any music already in DTS-HD at 96khz and listen to it on core only equipment then decode it to a Wav file, Resample it to 48khz and re-encode it to DTS standard, Much cleaner sound.

This issue does not seem to affect DTS 96/24 however, This encoder makes a nice resample for the 48khz core.

 

[Jim the Oldbie]

...Its not a problem if you've encoded from a 48khz source but when 96khz is used as the source the encoder just chops it in half without any attempt to gracefully resample the 96khz source to get a 48khz core, the result is a core with lots of resamping artifacts and a bad sound.
My advise is to do a 96khz DTS-HD encode and then use a good resampler like SOXR to make a separate 48khz source and then encode this as DTS standard separately for use on legacy equipment...

Interesting...

How were you able to verify this? Were you able to decimate a copy of the 96 KHz stream, and get it to null with the core stream, or something like that?

 

[jonmchugh]

No just running various test files through the encoder and comparing the output with spek.

 

[jonmchugh]

The above image is the DTS core extracted from HD Master Audio 96khz Sweep as attached.

 

[Jim the Oldbie]

No just running various test files through the encoder and comparing the output with spek.

Jeesh - that core ain't pretty. Sure does look like a lot of aliasing up there.
Thanks for posting these.

 

[HomerJAU]

The thing that’s always bothered me about DTS 96/24 encoding for DVD (Not DTS-HDMA 96/24) is that it’s limited to 1440kbps (or whatever the DVD PCM rate is). The original DTS was 48/24, then they added 96/24 as an extension. BUT the bitrate was still limited to same DVD rate so it could play.

So although we got 96kHz we got it at half the bitrate, so basically half the info is missing. So to me, DTS 96/24 is twice as lossy as DTS 48/24. Can DTS 96/24 sound better than DTS 48/24?

 

[Ninecats]

Give me a good mastered dts over a less good mastered lossless album any day.

 

[jonmchugh]

Can DTS 96/24 sound better than DTS 48/24?

NO, It's outside the hearing range of Human Beings, Cats and Dogs might enjoy it though.
Use DTS 48/24 as it gives more bandwidth to stuff you can actually hear.