Re: [SI-LIST] : Re: Guard banding

Robert Voelker (rhv@mamacita-178)
Wed, 12 Nov 1997 15:59:48 -0800 (PST)

Hi Jon,

If you know beforehand that not all of the bits will switch
simultaneously, the transposed line bus is effective in reducing
nearest-neighbor crosstalk compared to the parallel line bus.
The fewer the conductors switching simultaneously, the better the
effectiveness of a transposed line bus at reducing this crosstalk.

You are correct in stating that there is no crosstalk reduction
advantage to a transposed bus compared to a parallel line bus
in the worst-case situation where all lines but the quiet line
are switching. However, for the on-chip bus described in the paper,
you still have the benefit of reduced skew in the signal propagation
delay.

As a best-case example, if the data being sent down the bus is a
Gray code, then only one bit will switch at a time. Another example
would be an on-chip control signal bus in which a limited number of
bits would transition simultaneously.

As discussed in the paper, the transposed line bus has chip area
and fabrication complexity advantages compared to several other
crosstalk reduction techniques.

Bob
--------------------------------------------------------------------
Robert H. Voelker, Ph.D. Sun Microelectronics Div.
Signal Integrity Specialist Sun Microsystems, Inc.
Phone: (408) 328-8836 M/S USUN03-312
FAX: (408) 774-8769 901 San Antonio Road
Sun internal phone: x48836 Palo Alto, CA 94303
EMAIL: robert.voelker@eng.sun.com USA


> Date: Wed, 12 Nov 1997 13:29:42 -0800
> From: Jon Powell <jonp@pacbell.net>
> MIME-Version: 1.0
> To: Robert Voelker <rhv@mamacita-178.Eng.Sun.COM>
> CC: si-list@silab.Eng.Sun.COM, robert.voelker@Eng
> Subject: Re: [SI-LIST] : Re: Guard banding
> Content-Transfer-Encoding: 7bit
>
> It seems that transposing bus signals would only reduce average
> crosstalk and not worstcase crosstalk (ie. when all signal are switching
> at the same time). If this is true, then what does this buy you?
>
> jon
>
>