Path: utzoo!utgpu!watmath!att!tut.cis.ohio-state.edu!ucbvax!hplabs!hp-pcd!hplsla!tomb
From: tomb@hplsla.HP.COM (Tom Bruhns)
Newsgroups: comp.sys.amiga
Subject: Re: Weighing Platters
Message-ID: <5160051@hplsla.HP.COM>
Date: 14 Aug 89 16:49:14 GMT
References: <625@orange6.qtp.ufl.edu>
Organization: HP Lake Stevens, WA
Lines: 23

sutherla@qtp.ufl.edu (scott sutherland) writes:
>
Lots of other stuff deleted, but this caught my eye:

>6) Let the weight of the platter be ~ 3 oz, or ~ 400 g. (I don't have any
>	idea if this is a reasonable guess for the platter weight.)

It's not a reasonable conversion!  1 oz is about 28 grams.

>
>
>7) Then the weight of the platter without lubricant is 400.0000g and 
>	the weight of the platter with lubricant is 400.0004g. Thus you
>	are trying to measure one part in one million by using the 
>	difference between two large numbers. This is a statistical 
>	nightmare! (Any statisticians care to comment?). If the error
>	in your scale is 0.0002g, you have a 50% error in your 
>	measurement. NOT GOOD.

I don't think it's the measurement that's a problem.  Clearly, you'd
use the same balance for both the "before" and "after" measurements.
Rather, it's making sure that nothing else of similar mass is
transferred on or off the platter between weighings.