On Tue, Oct 11, 2011 at 11:17:15PM +0200, Atte André Jensen wrote: There are at least two good reasons why such a test is not valid. 1. The difference signal doesn't tell you anything about audible differences between two signals. It's fairly easy to make a linear filter (no compression involved) with a perfectly flat response and that nobody would be able to hear. But when you take the difference between in and out it is 3 dB higher than both. 2. Lossy compression is based on combined temporal and spectral masking - some signals you can't hear in the presence of others. Of course when you take away the masking signal they become apparent... Ciao, -- FA

Hi Atte, if you already mentioned Massenburg, he makes some high quality (and high price tag) analog devices mostly used by mastering engineers. He "should" know his stuff. It happens to be that just today I came across this vid of Bob Katz (a well known mastering engineer), he uses the mp3 artifacts too but with a different purpose. http://www.youtube.com/watch?v=8EgamkLkXW8&feature=related Hope you find it interesting. Moshe On Tue, Oct 11, 2011 at 11:42 PM, Atte André Jensen <atte(a)email.dk> wrote:

On Tue, Oct 11, 2011 at 09:07:33PM +0000, Fons Adriaensen wrote: With suitable hardware. What I mean is, I think a great way of demonstrating the difference between lossy compression and uncompressed audio is to do an A/B test through a consumer device and then do it in the studio. The difference can be striking. I like to think I have decent ears, and I can only very rarely tell the difference once over 192kbps. Though I've also found that bitrate isn't always everything (i.e some audio seems to respond better to a given compression algorithm than other). I wonder what other people's experience has been in that respect. Cheers, S.M.

A/B doesn't remove confirmation bias unless it's randomized. Perhaps that's what you meant (FTR, squishyball will give you casual A/B, randomized A/B, A/B/X and X/X/Y testing). Some very high end ears routinely fall prey to this-- It's not that you can't trust your ears, it's the brain that's gullible. Monty

On Wed, Oct 12, 2011 at 01:34:46AM -0400, Al Thompson wrote: That's not necessarily surprising and has less to do with actual quality of reproduction than with comfort. Many of us who grew up with LPs and tapes had difficulty adapting to digitally reprroduced music (I confess I still get a warm fuzzy feeling with listening to a very slightly worn vinyl record). So, can we really be surprised that people who probably have been listening to MP3 encoded music since the age of ten or younger feel a degree of discomfort or alienness when listening to uncompressed audio? Cheers, S.M.

Also depends on specific encoder. mp3 (and Ogg) specify the decoder. A valid encoder is any encoder that produces a valid bitstream. Encoders vary *widely* in both final quality as well as the nature of their artifacts. Simply saying '192kbps sounds like <foo>' is completely invalid. Monty

Excerpts from S. Massy's message of 2011-10-12 00:23:42 +0200: Was this a double-blind test? Otherwise you don't really need to test at all. The idea of a double-blind test is that you can't fool yourself. With a simple A/B test you might not want to fool yourself but still do it.

I'm fully aware of the effect, a skeptic at heart, and I find that I still fool myself all the time into thinking I hear differences when I don't. I wrote an ABX tool because *I* needed it to use for my own testing. You're on Linux, right? Just grab and build Squishyball: svn co http://svn.xiph.org/trunk/squishyball it's a lightweight, standalone double-blind tester. Hand it a few comparison files, and off you go. It runs in a terminal. It can run in a few modes, the manpage details them. It's useful for non-rigorous testing too. Monty

On 11/10/2011 23:35, S. Massy wrote: But tests should be blind, as Fons suggested. It has always struck me how well you can tell compressed files by listening to applause (especially the kind of applause you get at a talk or classical music concert). My personal experience is that classical and electroacoustic music are much more susceptible to compression artefacts than 'mainstream' music. Lorenzo.

That was Brandenburg IIRC, and an early FhG encoder. Many mp3 encoders still have trouble with that track because of all the HF content. I know I sound like a broken record, but a format is only as good as its encoder, and each has many different encoders. Encoding is NP-hard. Monty

On Wed, Oct 12, 2011 at 7:42 AM, Lorenzo Sutton &lt;lsutton(a)libero.it&gt; wrote: Cymbals---these sound horrible in mp3 encodings. It should be fairly predictable that wide-band sounds have worse encodings. It takes more frequency content and hence higher bitrate to represent these sounds.

Hartmut Noack wrote: My ears are shot (age and rock'n'roll), and my equipment isn't pro level, but I notice a difference between 32-bit WAV recordings and resulting MP3s (LAME's variable bit rate, quality 2). Mostly more high frequencies in the WAV vs the MP3s. But this isn't double-blind testing. How about MP4 - any difference between MP3 and MP4 when it comes to sound? -- David gnome(a)hawaii.rr.com authenticity, honesty, community

On Wed, Oct 12, 2011 at 06:31:26PM +0200, Hartmut Noack wrote: Tonight I'm recording 5 violins made by Guadagnini (about 1 M Euro each), for a listening test about the violins. Could provide some interesting test material for an encoding ABX test... Ciao, -- FA

On Wed, Oct 12, 2011 at 09:09:37PM +0200, Jostein Chr. Andersen wrote: To make this clear, I don't get 1 M Euro for each one I record... :-( It's just the insurance value of the instruments. BTW, 'gaudagnini' means 'little gains' or 'little profits' :-) I'll try to get permission to make some of this available for mp3/ogg testing. Ciao, -- FA

Indeed, 'video' and 'photo-graphs' are also utterly inadequate as compared to the original experience and should be shunned. Indeed, the printed word itself is but a pale shadow of the original unspoken idea. Personally I find life is far richer with all of the above. I never fail to be impressed at the lengths snobbery goes toward rationalizing itself (rhetorical comment, not actually aimed at you Hartmut). On a lighter note: Hmph, you should easily be able to get more wattage out of a proper driver than a bass drum. That's engineering, Sir. (IMHBCO, it's generally not the inadequacy of the recording but the fact that the recoding does not bundle along the original context of the performance. The context carries the greater power). Monty

Am 12.10.2011 23:31, schrieb Monty Montgomery: This is a case of misunderstanding most probably rooted in my poor English: I did NOT say, that recordings are a bad/inferior thing. I just said, that they differ from the experience, the original performance creates. And I said, that recording is an art in its own right. I tried to say, that recordings can be works of art, that cannot be replaced by live performance. The same as films can be works of art, that cannot be replaced by actors on a stage playing live. Recordings can be great, even better than a live-experience. Anyway they should be. While they lack the physical power of a live-performance they can have lots of other qualities that live-performances cannot. I admit, that I am not top-notch when it comes to recording drums. I do not have the gear and not enough experience to do such jobs as good as Albini can. Very well but: many engineers tend to trigger-mic the drum-set and use samples to make drums "sound right". So it seems to be not the most easy part of a rock-band recording. Anyway, I never heard a bass-drum, that really sounded like the punch-samples from the Alesis D4-sets ;-) At the other hand: If a guitar-player stands within hand-reach of a 4x12"-stack playing on 10 and a minute later hears the result on 2x8" the difference is inevitable. Also I do not think, that it is wise to make musicians a little happier by raising the volume of the monitors to make it sound "more real". In the end a recording is made for people, that use middle-class stereos at 50% volume in the best of cases and 3.99-earplugs playing MP3 in the worst. Such equipment cannot produce the same physical effects a 4x12"-stack can produce. One more thing, that only live-performance can do. And a pro for recordings also: in a recording one can fiddle and re-take as long as everything is the way it should be. Live is live: one of the players having his/her bad day, stupids in the audience, a broken monitor or anything else that can happen in real live (but not in a studio) can destroy the experience. Once I saw Neil Young playing Like a hurrican as the last encore and my word is good: he managed it to play about 10 minutes of solo with 2 strings broken. It was great. He was accompanied by the boys of Pearl Jam and they mentioned the situation with awe. Still I dare to believe, that even he would have changed the strings when he would have played it in a studio-session ;-) best regs HZN