Re: NTLK Questions, questions... (waaay off topic)

From: Michael J. Hussmann (
Date: Tue May 02 2000 - 10:48:33 CDT

THX 1138 ( wrote:

> Even without added
> compression, anything recorded digitally is inherently compressed, at 44
> samples per second,


that's not what is usually called compression, i.e. taking some data,
applying some algorithm, and eventually getting a smaller amount of data
that either can be restored to the original (with lossless compression
algorithms such as LZW) or that preserves the essential qualities
(however defined) of the original, as with MP3 or MD.

What you are talking about is sampling (i.e. you sample the signal at
certain points in time, not continuously) and quantization (you
approximate the sampled value using a finite set of numbers). Sampling
limits the highest frequency that can be recorded, but then, similar
limits exist with analog recordings -- they are just of a different
nature. Quantization introduces distortion, but then again, analog
equipment distorts as well.

> resulting in a square wave.

This is more accurately described as a staircase. Anyway, that's not what
is fed to the amplifier. Prior to the necessary digital-to-analog
conversion, oversampling is used to smooth the signal by introducing
interpolated samples between the real samples. Oversampling can't restore
frequencies that weren't recorded, but it does reduce distortion. After
the conversion, a low-pass filter ensures that the analog output will be
smooth -- no square waves at all. Note that according to the Nyquist
theorem, a sampling frequency f suffices to _exactly_ reproduce
frequencies up to f/2 -- not just some square wave approximation of the

Michael J. Hussmann

Redaktion macmagazin
MACup Verlag GmbH
Leverkusenstrasse 54 VII
D-22761 Hamburg

This NewtonTalk Message brought to you by:

EVOTE.COM, the ESPN of Politics on the
Internet. Visit EVOTE.COM for all the latest
news on Campaign 2000!

Visit today!


Need Subscribe/Unsubscribe Info?


This archive was generated by hypermail 2b29 : Thu Jun 01 2000 - 00:00:07 CDT