until I read on and realized that I would have to buy another media player.
Neal Young's venture to deliver quality audio

I would have been interested in re-purchasing some of my music library in uncompressed formats, but I'm not going to buy a new portable media player to do it. Sorry Neal.

Now if the media ends up to be playable in my home system, we may have something.

Which will undoubtedly come with horrendous software and a worse UI. Seen this sort of junk before...

He doesn't even have a good idea to start with:
http://people.xiph.org/~xiphmont/demo/neil-young.html

If you want quality audio, buy CDs that are well mixed/mastered, and rip them yourself to a lossless format your current player does support. And get a better pair of headphones.

I found another post that takes on the bit-depth portion of the hi-res audio debate. I found it quite interesting. http://www.head-fi.org/t/415361/24bit-vs-16bit-the-myth-exploded

True enough. I was assuming he was also going to maybe get permission to remaster from the original masters. However, now I realize this was just my optimism and lazy skimming of the article.

I've also seen that write up. IMO, this is a case where the theory doesn't reflect the reality. I have been listening to quite a bit of HD music lately and have even compared CDs I have ripped at 320Kb mp3 to HD versions. I do notice a difference. Another consideration is that when you get an HD version of an album, many times these are remastered from the original tape.

You're last thought is on the mark. Even CDs that are bundled with the DVD-A version of an album were usually from different masters.

The only way to judge is to take a 24-bit wave, and dither it down to 16-bits and compare those two. I've done that, and can't tell the difference. In fact taking the 24-bit wave, encoding it with Lame using the "extreme" preset, and then turning it back into a 16-bit wave, I can't tell the difference from the original.

We don't need better distribution, we need better mastering.

In fact, here's a test I made for someone on another forum who was claiming to have golden ears...they never replied to me.

http://www.scuzzyeye.com/tests/track-01.wav
http://www.scuzzyeye.com/tests/track-02.wav
http://www.scuzzyeye.com/tests/track-03.wav
http://www.scuzzyeye.com/tests/track-04.wav

One of these files is the original 24-bit version. One has been truncated to 16-bit (not even dithered). One has been converted to a 320 kbps CBR MP3, and back into a 24-bit wave using MADplay. And another an "extreme" VBR MP3, then simply decoded with Lame to a 16-bit wave. They've all been padded back up to 24-bits, so the file sizes are the same.

Tell me which is which.

I gave them a listen and cannot tell a difference over my Grado SR-60 headphones.

Same here. I just send this to a friend who's trying to convince me otherwise.

EDIT: The friend says track 1 is the original. How'd he do? (Feel free to PM me if you don't want to reveal anything here.)

He's wrong, it's the third one.


At least htat's my vote. :-P

Wow, that's a good demo of the sampling Chris.

To me, only track 2 seemed not quite as full as the others, which (1, 3, and 4) were pretty well, no, not pretty well, but in fact entirely indistinguishable to me.

I actually tried to make these worst case, so it'd be easiest to tell them apart. I'll make another challenge that's designed to be difficult this evening.

Oh, and PM me if you do want the answer.

I like that. We've got no consensus on which one is which, and you tell us this is the EASY one. Heh.

Designed to be easy, from the track selection (those high pitched sounds should have been a give-away for poor compression), to the poor depth reduction. I really wanted to be able to hear the difference myself, but alas I couldn't.

That was neat but listening to those made my ears ring.

But which made your ear feel golden before the ringing?

I pick 4 then 3 as the best. Man we are all over he map

OK, the easy test is over. Time for the difficult one. This one is just about bit-depth, not lossy compression.

These files are longer, I wanted a full reference of dynamic ranges, since that's supposed to be what 24-bit gains over 16.

http://www.scuzzyeye.com/tests/track-b1.wav
http://www.scuzzyeye.com/tests/track-b2.wav
http://www.scuzzyeye.com/tests/track-b3.wav
http://www.scuzzyeye.com/tests/track-b4.wav
http://www.scuzzyeye.com/tests/track-b5.wav

Two of the files above are full 24-bit. The other three are dithered down to 16-bit with noise shaping. The 8 least significant bits are 0-filled so the files are all the same size. You challenge is to tell me which two files are 24-bit.

Before you had a 25% chance of guessing right, now it's only 10%. But if 24-bit really is that large of an improvement it shouldn't be difficult.

This has been interesting. We looked at bit depth, what about sampling rate?

BTW, I pick 3 and 5.

If I cut out a 20 second chunk from each track, starting from about 1:25, using Audacity, would it negatively affect the samples at all? I want to be able to easily play the same chunk of music from each track quickly back and forth.

I have another idea. Could you randomly -- and by randomly, I mean assign digits 1 through n to the different versions of the track -- string together 15-second cuts of different tracks so that you're left with a seamless track that smoothly transitions between bit-rates?

The overall goal would be for the listener to identify which 15 second spans are encoded with the highest quality, starting at time index 0:00, 0:15, 0:30, etc. Since you'd assign the parts randomly, you may end up with a few consecutive spans of the same quality. With the chunks only being 15 seconds long, you'd also increase your chances of using all the bit-rates before the end of the track.

Very good, Chris. Although my previous personal experience and study of other test results didn't leave much doubt that both 320 kbps CBR or the "extreme"VBR(around mid 200s kbps)would be audibly transparent(even with about 80% of the audio data discarded), I listened on both speakers and headphones.

I paid special attention to the first 12 seconds before the singing attempt began, particularly the 9th though 12th notes, which have the sort of content which might demonstrate a difference. After several trials on each of the four samples, nothing was heard would support anything beyond a pure guess.

This might be surprising to some who claim to be able to hear a difference even between pieces of connecting wire, but it's the reality of modern audio technology.

I can try something with sampling rate. But while I'm sure of 0-filling LSBs, in order to make the files the same size. I'd have to take a higher sampling rate, and downsample it, and then upsample it again, and that may not be as clear cut.

I'll play around, and see if there are any extremely obvious tell-tale signs that would make the test inaccurate.

No, nothing wrong with doing that. Of course it's also possible to use Audacity to cheat, but you won't do that, right?

As for identifying 15 second segments of one track. I can do that. In fact I was flipping between bit-depths as I was preparing the test. I could identify 8 and 12-bit, but 16-, 20-, and 24-bit all sounded the same to me. I'm just not that sure of how good a test it would be. It's hard to say, "this is 16-bit," without having something to compare against.

Yeah. And as I've said, that first test was really absolutely worse case, and I couldn't tell the difference.

Did you check out the second set of tracks? (You might like the music too.) I'm glad that I had the option for 8- and 12-bit output, or I would have never believed that the dithering software I was using was even doing anything (actually, I can see it doing something on the VU meter at the higher depths, but I can't hear it).

Oh, and a fun fact. I was going to use a track from R.E.M. for this second test, but as I was processing it, my digital clip indicator came on. This is why we can't have nice things, the engineer was given 24-bits and he still slammed it into a brick wall. (Although it's no were near as bad as The Flaming Lips', At War with the Mystics.)

Okay, I listened to the Carmina Burana samples(Thielemann?). Quite dynamic, but nothing that can't be handled easily with 16 bits. The dynamic range on all was about 50dB, from the upper 50s to a max of 106dB on my SPL meter, and sounded identical.

The dithering raised the noise floor up to -80 dB, but that's still well below the audible threshold.

And yeah, it was Thielemann. I actually quite like that disc.

I see what you did there, JohnK. You da man.

Peter doesn't have the audacity to cheat does he?
DOES HE????

I never cheat unless I have something to gain. And then, only if the chance of being caught is slim. And then, only if it's easy. And then, only if I can live with myself. And then, only if I get rid of all the mirrors in the house. And then, only if I buy a lot of sleeping pills first.

Thanks to Murph for starting this intriguing thread, ClubNeon for his sample tracks, and to everyone else that has participated so far. This is a really interesting topic, and one that can easily foster heated debates similar to those discussing the efficacy of different cable dielectrics ad nauseum! *cough* The entire topic of what is and what isn't audible when it comes to digital audio is fraught with complexities and involves serious math, signal processing, and communication theory. To be an expert on this topic is something few can claim, and I'm certainly not one of them. What I do know is that the simple task of taking ClubNeon's sample files and playing them back over headphones plugged into your laptop is probably not going to give you the goods. Why? Simply because most computer systems are not designed for bit-perfect audio playback, which is the first requirement of trying to compare audio files at different sample rates or bit depths. Just because you can play the file does not necessarily mean that it isn't being trans-coded to another format before it gets to your headphones. As an example, if you're casually using iTunes for playing back your music files, there is a slim chance that you are hearing them played back correctly. Then there is the issue, as has already been pointed out, that most commercial material in standard and high-rez formats have been mastered differently, or were created from differing original sources. As a die-hard analogue guy, I am intrigued by the world of high-rez digital, but have never taken serious steps towards educating myself and putting my own ears to work. The first path down this road has been to investigate the idea of home music servers and the various playback and software options that this entails. To say I have repeatedly pulled my hair out is an understatement. The next step has been to talk to people with far more knowledge than myself on the topic and attempt to devise a test that can be performed blind, in a similar manner to our loudspeaker listening tests. Interestingly enough, Ian has also been thinking about this very topic, and we have decided to embark on a research project in the new year which will allow us, and anyone else who's interested, to compare various digital formats in a repeatable manner. To that end we are going to setup a state of the art digital playback system, with the assistance of our friends at Bryston, which will ensure that all of the source material is being passed in a bit-perfect manner. For source material we are going to obtain recordings that were originally created at various resolutions and then down-sample, up-sample, or decimate as required.

Feel free to add your own suggestions and stay tuned!

Andrew

Cool stuff Andrew!! Its nice to see some original research being instigated in a market (hell a society) that has become more about hot air than facts.

I like research.

Go figure.

Great idea, I learned much in this thread myself.

Regarding your comment about files being trans-coded, this definitely is an issue on Windows as the Windows mixer will usually intercept audio playback and muck about with it. I know of two ways to avoid this, either set you AVR to bitstream (ie. no A/D is done, just send the digital stream) or better yet use a great media player such as JRiver Media Center 17. It can (painlessly) use WASAPi - Event Style to bypass the Windows mixer completely and let the output device pull data as required instead of Windows just feeding data dumbly.

If you are going to do any testing on Windows for music playback, JRiver Media Center 17 (18 coming soon) is to my knowledge the best player out there.

That sounds pretty cool, Andrew. Just one more reason to make a pilgrimage to Axiom headquarters.

So are you guys going to be recording a lot of your own material for the sake of the research? It seems like the best way to be sure of what you're working with. You could make your own data bank of samples that theoretically should be prime candidates for showing off higher-resolution formats. Of course, this would mean making a state of the art recording studio to go along with your state of the art digital playback system. Ha ha. Or maybe you have connections to people with an appropriate recording studio already. If you record your own stuff, then you could offer what you ultimately end up with as your demo selections as downloads on your site, for end users to play around with themselves.

Are you going to be researching surround formats as well, or is this strictly stereo, to keep things simple?

There is a great site that has a variety of hi res multichannel lossless master samples, mostly classical. These are the production originals used for SACDs and Bluray, many in the raw DXD format.

I particularly like: "Bridge Over Troubled Water (Simon & Garfunkel) Iver Kleive (church organ)". It's starts out very soft and just gets louder and more powerful as it goes, it scared me the first time I heard it

http://www.2l.no/hires/index.html

That is true, but all my listing tests were done from within an audio production program, that uses ASIO drivers instead of Windows. That does guarantee bit-perfect playback. I still can't hear a difference.

The best comment I've seen on this topic, is that the sampling theorem was not created to explain how digital encoding of an analog wave form works, but the other way around. The theorem existed before it was possible to do the encoding/decoding, and the math is sound. So when electronic technology got to the point where it was possible to test the theorem, the devices were built.

I thought they fixed that with win7?

I heard they've already put the mike in Ian's shower.

A program in Windows 7 can request not to be messed with "exclusive mode", but the majority don't bother.
ASIO drivers completely side step the issue, by not using the Windows mixer at all. But ASIO is usually limited to professional interfaces, and audio production software.

Anyone know how it works on the Mac?

Apple software uses CoreAudio -- which is billed as professional grade. Here are a few quotes from the doc I linked below:

Most Core Audio constructs manipulate audio data in linear pulse-code-modulated (linear PCM) format.... Core Audio generally expects audio data to be in native-endian 32-bit floating-point linear PCM format.

Supported file and data formats

Here are the developer docs.