Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!peregrine!ccicpg!zardoz!tgate!irsx01!ka3ovk!drilex!axiom!linus!mbunix!eachus
From: eachus@mbunix.mitre.org (Robert Eachus)
Newsgroups: comp.sys.amiga.tech
Subject: Re: TICK and J300
Summary: Use statistical quality control methods
Keywords: Dennison statistical quality control
Message-ID: <54898@linus.UUCP>
Date: 1 Jun 89 14:31:05 GMT
References: <3319@hplabsz.HPL.HP.COM> <6854@cbmvax.UUCP> <286@xdos.UUCP> <213@doctor.Tymnet.COM> <18978@cup.portal.com>
Sender: news@linus.UUCP
Reply-To: eachus@mbunix.mitre.org (Robert I. Eachus)
Distribution: usa
Organization: The MITRE Corporation, Bedford, Mass.
Lines: 32

In article <18978@cup.portal.com> FelineGrace@cup.portal.com (Dana B Bourgeois) writes:
>Joe Smith would like a crystal clock that adjusts itself long-term from
>the power line which makes a good long-term time standard but relatively
>poor short-term one.  I like the idea but I'm not sure conceptually how
>you would design this.  At any instant the power line time interval will
>be long or short.  When do you decide to 'sync' up to it? Or maybe you
>sample the intervals over a week, take the average and set your crystal
>frequency to that?  Or sample over a 30 hour period since the power
>companies supposedly average over a day?  How many samples?  Anybody
>have an idea of how to design this?

    This is a thoroughly studied problem, but from the opposite
direction.  If you are runninng a production line and are concerned
with say the final product weight, you measure the weight of units as
they come off the end of the line.  But how do you know when to make
adjustments?  If you adjust for each unit produced, the variance
increases, so statisticians have done a lot of work on rule for making
adjustments which minimize both the bias and variance.  I think the
original work was done by Dennison in the thirties, but check any good
library under the heading of statistical quality control.

    This problem is slightly different in that you are changing the
yardstick, not the thing being sampled, so I imagine that to use some
of the techniques correctly you will have to correct the remembered
samples to reflect the new clock rate.


					Robert I. Eachus

with STANDARD_DISCLAIMER;
use  STANDARD_DISCLAIMER;
function MESSAGE (TEXT: in CLEVER_IDEAS) return BETTER_IDEAS is...