Pete Cross wrote:
1. it amazes me how the digital era of audio appears to be going, I saw a
report recently of a new Logitech speaker that was getting rave reviews
saying "it has no distortion at all" reading the reviews it became clear
they where refering to be able to play at max vol with an ipod as the input
without it distorting....a look at logitech.com found me the spec 20wrms @
10%thd ! sheesh! that got me wondering too about the mp3 source, compared to
the orig CD wav what wouth the thd of the mp3 work out at ? can it be
measured ?

I'm not a bit surprised that a cheap computer 'speaker would be reviewed
as having no distortion. I assume the review was in a computer journal
or General Public publication, and not an engineering journal. Reviews
of technical products by non-technical journalists often make these
stupid statements.

MP3 sources generally don't have any worse THD figures than non
data-reduced formats for one reason:- THD is normally measured by
injecting a single tone into the Device Under Test, then measuring the
output with the tone removed by a very deep and tight notch filter.
Everything else is deemed to be distortion (plus noise). MP3 encoding
with a single tone is very good, and won't show any artefacts, as the
entire bit budget can go towards coding that tone perfectly. Things get
a little more difficult when encoding multiple tones, as the bit budget
has to be spread over more information. There are tests available that
inject multiple tones, null them all out separately and measure what's
left. These tests, however have shown that the measured result is very
dependant on how many tones, what levels of each tone etc. As far as I
know there's no standard test that will assess distortion for an MP3
encoder/decoder.

As an aside, if you play a rapid sliding tone (20-20k in 2 seconds)at
constant level into an encoder, then decode the result, you will get
something out that looks very strange. There are great holes in the
frequency response, and you would think it was broken. I did this test
for the French Post Office some years ago when they were assessing MP2
ISDN codecs. They had this standard test they wanted to apply to the
codecs even though they knew it was pretty stupid.

S.