[Table of Contents]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ARSCLIST] MP3

I familiarized myself with MP3 technology through the O'Reilly book "MP3: The Definitive Guide" by Scot Hacker. Unless it's been updated, it's pretty obsolete now, so I would recommend buying it used for a coupla bux vs. new for many more bux. It's somewhat Linux-centric in some ways but the general technical knowledge about how MP3 works hasn't changed, I don't think.

I speak from having years of experience with MP3 and other lossy-compression audio formats, both as a user and producer of files and also editing audio for a proprietary format (Audible) in that format's early years. In fact, I'm one of the first 100 registrants for MusicMatch, which way back when was by far the best Windows MP3 encoder both in sound quality and ease of use. And I've owned and used portable MP3 players since the Rio500 (the original Rio was just too expensive and low-grade to tempt me). To my ears, which I would say are pretty critical because I know what to listen for, MP3 is not similar or same as the original CD audio -- no matter what the encoder or player, they just effect how dissimilar it is -- until you get up into relatively high bitrates for music. 256kbps certainly gets OK for headphone listening and background music and 512kbps starts to get indistinguishable under most conditions from the CD audio. However, in a studio monitoring environment with relatively loud listening levels, even 512kbps is clearly not identical to the source.

In the higher bitrate modes, you don't have problems of digi-swishing in the background or "lisping" sibellence (sp?) like you do with lower bitrates (including almost any streaming bitrate below 128kbps). The problems to listen for are:

-- lack of treble, as if someone had turned down the top frequencies with a graphic equalizer
-- lack of image depth and width, probably due to lack of treble
-- sometimes strange things with loud and fast impulse information
-- some encoders and players seem to also lop off the low bass information under some circumstances, even at surprisingly high bitrates
-- lack of "breath", both literally and figuratively, around acoustic instruments and lack of "air and space" around large ensembles. It's not a terrible loss for headphone or background listening, but if you're in a critical monitoring environment you can tell immediately that the music is less "front and center" or "present."
-- MP3 encoding seems to worsen distortion from "toothpaste" over-compressed CD's. I would imagine it's lossy nature has something to do with this but I'm not sure what.

These effects vary with the type of music you are encoding and how the original source material was recorded and mastered. For instance, I have very good luck with mono LPs crunched to 256kbps. What seems to happen is some of the undesireable LP artifacts are less audible vs, the underlying music. This is using both Sony Soundforge's MP3 encoder and Apple iTunes for Windows.

One thing to note with listening tests is that the listeners probably should be somewhat familiar with the source material. A crappy recording is a crappy recording, no matter what the format. And, listener A might be very clear on what a violin sounds like because she's been to many a concert or maybe plays one herself. Listener B might be very clear that the guitarist is playing a Gibson Les Paul through a Fender Bassman amp but might not know the first thing about how a violin sounds. So even what might seem to be a scientific blind test has all sorts of objectivism built in if humans are used.

One other caveat. For most spoken word material, especially when it's close-mic'd and mono, 128kbps MP3 is more than enough to have perfectly good audibility. Some would argue 64kbps is OK but I'd say bare minimum for avoid digi-swishies around "s" sounds is 96kbps with most human voices.

-- Tom Fine

PS -- back a few years ago, I did extensive listening tests with Windows Media format and found it to be inferior to MP3. They have since improved the format and, for instance, I find 128kbps stream WMA to be much higher quality to 128kbps streaming MP3 -- and more reliable to stream correctly over the Internet at least from the sources I listen to. But WMA is not a universal format, and seems to have been optimized for streaming. I never did similar tests with AAC because I'm heavily invested in MP3 as my lossy format of choice, but 128kbps iTunes purchases are clearly inferior to the original sources, and even include enough digi-swishies to be unlistenable in some cases. Mono 128kbps AAC is much better as far as fidelity to the original source than stereo, which is to be expected from a lossy format. To Apple's credit, they continue to improve the iTunes software and the iPod is far and away the best portable digital music player ever developed (and I've owned and tried many others). I am certainly not one to be singing Apple's praises or joining the Mac-worship cult, but I think they hit a home run with the iPod.

Sorry for the long post but hopefully it's helpful to someone.


----- Original Message ----- From: "George Brock-Nannestad" <pattac@xxxxxxxx>
To: <ARSCLIST@xxxxxxxxxxxxxxxx>
Sent: Wednesday, June 13, 2007 12:51 PM
Subject: Re: [ARSCLIST] MP3


From: Patent Tactics, George Brock-Nannestad

----- sometimes knowledge has a price:

There is a new book out that is essential reading for everybody who has to
deal with data-reduced sound (which is, sadly, everybody - cellphones,
internet, DAB, etc.):

The Perceptual Audio Evaluation: Theory, Method and Application
by Soeren Bech and - N Zacharov,
Hardcover: 462 pages
Publisher: John Wiley and Sons Ltd 2006
Language English
ISBN-10: 0470869232
ISBN-13: 978-0470869239

It is expensive, but well worth it.

Kind regards,


George

--------------------------------------------------

Hello, Steve,

I will tell you a little story about a bad test design and a good test
design.

The bad test started at a lecture I was giving. Part of my standard
lecture shows "what is possible" at a given era by highlighting
high-quality sound from a given era (going back to a 1935 steel tape
copy). Someone asked about MP3 and I had one of the 1980s selections
both in my demo as a WAV file and on my Palm T3 as an MP3. The Palm
sounded way worse than the Sony CD walkman I was using for the rest
of the demo.

When I got home, I took the original file, made a high-quality MP3
within Samplitude, converted it back to WAV and then cut between the
two recordings. I now demo the cut recording and the MP3 is almost
identical to the WAV file.

So, just as with A-D and D-A converters and even CDs themselves which
over time have, for some people, received the reputation of "not
sounding good" for perhaps the wrong reasons. Clearly, here we were
hearing the deficiencies in the Palm T3 audio system as opposed to
the deficiencies in the MP3 format.

I believe that my test is one of the few ways that one can do a test
and remove most of the external variables. I'm passing this story on
as an object lesson and as a caveat to anyone doing a listening test:
make sure that you're really listening to what you think you're
listening to and do NOT make assumptions. I believe that it is almost
impossible to do the test that you describe using A/B hardware
without the hardware differences influencing the rating of the format.

Oh, and I emailed the organizer of the lecture this explanation and
requested she mail it to all attendees. I think she did.

If you want the resultant WAV file I would be happy to share with you.

Cheers,

Richard

At 09:32 AM 2007-06-13, you wrote:
>Has ther been anything published in recent years that addresses
>actual listening comparisons between MP3 and CDs?  I'd prefer they
>have split the panels' sources into those that are acoustical (i.e.
>begin by pushing air)from those which start by exciting
>electrons.  It would be helpful if those doing the reacting were
>identified as professional or casual listeners as well.
>
>I 'm not looking for indivual reactions in print but a designed and
>controlled test.  Anything out there?
>
>Steve Smolian

Richard L. Hess                   email: richard@xxxxxxxxxxxxxxx
Aurora, Ontario, Canada       (905) 713 6733     1-877-TAPE-FIX
Detailed contact information: http://www.richardhess.com/tape/contact.htm
Quality tape transfers -- even from hard-to-play tapes.


[Subject index] [Index for current month] [Table of Contents]