Rob Saecker rsaecker at
Sun Mar 17 11:07:53 PST 2002

At 2:48 AM -0500 3/5/02, Chris Hill wrote:

>Bird hearing isn't that different from ours (they don't hear very high

>sounds or very low sounds as well as us, but are comparable in midranges).


>And nobody can hear computer compression algorithms; all you (and birds)

>can hear is compression waves in air, which is a very different thing!


I have some musician friends who would strenously dispute Chris'
statement that nobody can hear computer compression algorithms, but
that's another discussion. Here's what Cornell has to say about
Over the past couple of years we have received many inquiries
regarding the use of MiniDisc recorders for fieldwork. MiniDisc
recorders utilize an audio compression algorithm based on
psychoacoustic principles known as ATRAC (Adaptive TRansform Acoustic
Coding). The input signal is divided into sub-bands that are then
transformed into the frequency domain using a variable block length.
Transform coefficients are grouped into non-uniform bands to reflect
the human auditory system, and then quantized on the basis of dynamic
sensitivity and masking characteristics. ATRAC compresses compact
disc audio (16-bit 44.1 kHz stereo) to approximately 1/5 of the
original data rate with minimal loss in sound quality. This
technology permits 74+ minutes of sound to be recorded on a small
64mm optical disc.

Audio compression distorts sound, but in such a way that humans
normally can't detect. The compression method used in MiniDisc
recorders has frequency and timing constants built in that are
intended to make the compression inaudible to human ears, but which
may make it audible to other species. For instance, humans can't hear
a sound at a certain frequency if there is another, louder sound at a
nearby frequency, so the compression algorithm will leave out the
former sound. We can't hear the difference, but it's possible other
species can -- and any scientific analysis of the compressed sound
would have to work with a distorted version. Additionally,
compression algorithms typically can't track very fast changes in
overall loudness correctly, since humans can't hear such changes very
well. Other species (e.g. many songbirds) do have large loudness
changes over very short times in their vocalizations. If you try to
record such species with a MiniDisc, you'll likely introduce some
distortion. We have tested a popular early-entry portable field
machine that utilized ATRAC 2 and found it to be unacceptable for
natural sound recordings Ñ especially recordings that might be
subjected to scientific analysis. However, recent claims that Sony's
ATRAC 3.5 & 4.5 coding versions are considered by some professionals
to be indistinguishable from, or even better than a CD remains to be
proven. We plan on conducting in-house testing of these newer ATRAC
versions to determine if they are applicable for natural sound
recording. Test results will be posted as soon as they are available.
Digital recorders with data compression.--Recently introduced, the
Mini-Disc recorder outwardly appears that it might have application
for field use. Mini-Disc is a compact digital recorder that stores up
to 74 minutes of audio on optical disc. Like other digital formats,
the MiniDisc offers convenient track numbering along with the rapid
accessing of a compact disc.

A significant drawback to this recording system is that it uses data
compression algorithms to maximize data storage capacity. These
algorithms are based on psychoacoustic research -- the study of how
humans hear and perceive sound. As a result, when bird sounds are
recorded, harmonic content at the upper margin of human hearing can
be lost. Furthermore, the "program material" (i.e., the bird song
plus the background noise) is "reviewed" as a whole. This means that
if, in processing the algorithm, the machine determines that the
background sounds are masking elements of the signal of principal
interest, the "masked" components of the principle signal are not
necessarily recorded. Data compression algorithms may also add
spurious data during the recording process (Fig. 1).

For the present, we cannot recommend Mini-Disc recorders or other
recorders that use data compression, although they do have utility as
a sound storage system with quick access capability for listening

This addresses an issue a bit different than the original question,
which was using minidiscs for playback; if the original source was a
cd, for instance, I presume the problem would be mitigated. But if
Cornell isn't ready to accept recordings made on minidiscs, I'm not

More information about the Tweeters mailing list