NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/arch/10667

Internet Message Format | 1992-11-12 | 2.4 KB

Path: sparky!uunet!ukma!usenet.ins.cwru.edu!agate!stanford.edu!unix!clipper!arnold
From: [e-mail] (Roger Arnold)
Newsgroups: comp.arch
Subject: Re: fiber/silicon and Thz clocks?
Summary: use frequency multiplexing
Message-ID: <[e-mail]>
Date: 12 Nov 92 01:58:46 GMT
References: <[e-mail]>
Organization: Intergraph Advanced Processor Division - Palo Alto, CA
Lines: 44

In article <[e-mail]>, [e-mail] (Nick Janow) writes:
> [e-mail] (Craig Partridge) writes:
>
> > For example, 1 terabit per second, on an 128-bit wide bit path is about a
> > 0.1 ns clock cycle, which is only a modest improvement over what DEC
> > reputedly hopes to achieve for Alpha clock cycles in the future.  Now going
> > serial to parallel would presumably require some pico-second logic
> > somewhere, but maybe that's possible.
>
> The serial/parallel conversion might be handled with innovative techniques,
> taking advantage of physics.  What is the physical distance between optical
> pulses travelling through silicon (or some similar material)?  If that
> distance is small enough, it might be possible to do a "flash" conversion,
> with sensor elements in a linear optical path, each detecting whether there
> was a positive optical pulse within it at the time of a central clock pulse.

About .2 mm, if it worked that way.  But it doesn't.  What you'd actually
have would be 128 light frequency channels carried in one fiber, each
channel modulated at 10 GHz.  Or 256 channels modulated at 5 GHz, or
whatever works.  Heavy duty frequency multiplexing, at any rate.  The
receiving end has a corresponding number of tuned receivers, one per
frequency channel.

> Likewise, a linear array of laser elements could be pumped or not, according
> to a digital word input.  A word-sized string of optical pulses could be sent
> at once.
>
> It needn't even require linear arrays.  Non-linear optical techniques could
> stretch or shrink optical paths, or beam splitters, etc, could find a use.
> Brute force "one optical emitter and detector plus shift registers"
> techniques shouldn't be necessary.
>
> Hmmm, should I patent this idea?  :)

With a lot of the things that are given patents these days, sure, why
not?  A little more clutter isn't likely to make much difference.

> --
>
> [e-mail]

--
Roger Arnold
[e-mail]