Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 Fluke 8/7/84; site fluke.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxj!houxm!vax135!cornell!uw-beaver!microsoft!fluke!fudd
From: fudd@fluke.UUCP (Mark Freeman)
Newsgroups: net.audio
Subject: Re: Controlled compression, etc
Message-ID: <1494@vax2.fluke.UUCP>
Date: Mon, 22-Oct-84 17:43:09 EDT
Article-I.D.: vax2.1494
Posted: Mon Oct 22 17:43:09 1984
Date-Received: Wed, 24-Oct-84 03:57:50 EDT
References: <86@azure.UUCP> <50@vice.UUCP>
Organization: John Fluke Mfg. Co., Everett, WA
Lines: 23

---------

Shaun Simpkins' analysis of the noise problem is not quite right:

> Let's take an example - we have two 16bit
> digital words.  The LSB of both fluctuates randomly (i.e., the noise floor
> of the digitizer is equal to 1 LSB).  We now add them together.  Oops! The
> 2nd LSB is dithering, too.  The noise floor has moved up from 1/2 LSB average
> to 1 LSB average. Do this a few times and you've lost a lot of your
> resolution to noise.

If the dither noise is uncorrelated, then the addition of the noise is
an RMS process, so the noise floor does not grow quite so fast.  Also,
the signal may have to be scaled down after mixing, to keep the signal
from limiting, sending the noise floor back down.

So, how many bits do studio recorders need?


				Mark Freeman
				John Fluke Manufacturing Co., Inc.
{allegra,lbl-csam,microsoft,ssc-vax,sun,telematic,teltone,uw-beaver,
uw-vlsi}!fluke!fudd