From: John Larkin
Subject: Re: What makes coils AC or DC??
Date: Sun, 12 Jan 2003 22:24:41 -0800
Organization: Posted via Supernews, http://www.supernews.com
References: <2DoU9.22597$jM5.firstname.lastname@example.org> <email@example.com> <3E223E27.367C@sneakemail.com>
X-Newsreader: Forte Agent 1.91/32.564
On Sun, 12 Jan 2003 23:18:47 -0500, Mike Monett
>John Larkin wrote:
>> On the other hand, once in a while I have to do stuff that even IEEE
>> double-precision floats won't handle. Those can be an infernal
>> nuisance. Most webby calculators just use the Pentium fp hardware
>> behind the scenes...
>Infinite precision software is easy to program and readily available. If
>you use Flat Real Mode in DOS, you can access all the memory in your
>computer to store the arrays:
>You can also use base-256 to reduce the array size and handle even larger
>numbers. How many digits do you need?
Well, for instance, I did a timing gadget that had a 77.76 MHz clock
(1/2 OC3 rate) but was programmed in integer picoseconds, 48 bits of
same. So to convert the customer's requested time into my internal
format (ie, N clock ticks and M bits of interpolation vernier: 16 bits
of vernier) I had to do a multiply by a scaler that would partition
the result into nice bit clusters, N.M bits. Simple enough (all in
bare-metal 68K assembly), but I couldn't find a calculator anywhere
that would do the math to 48-bit precision.
The problem reduced to multiplying by something like 5.09607936...,
which I did by multiplying by 8 (just a shift) then a fractional
multiply by 0.637..., which factor I couldn't get anything to
Everything I tried missed by too much. In the end, I just tweaked the
scaler manually until it gave me the right answer: it was...
So what I needed was an infinite-resolution calculator that worked in
hex, with decimal point (heximal point?) included.