From: Muranyi, Arpad (firstname.lastname@example.org)
Date: Thu Dec 23 1999 - 10:46:19 PST
Interesting idea. But I wonder, even if you use capacitive coupling for
the I/O pins of the chip, you still need to connect the chips with some
sort of an interconnect. The question is: if the T-line is a bad thing
because it is a (large) parallel capacitance, what do you replace it
with to turn it into a series capacitance? Your example of the capacitive
pin or pad connection at the chip does not eliminate the conventional
T-line (and its parallel capacitance). How do you go about that?
Driving a large parallel capacitance (T-line) through a small series
capacitance (under the chip) is not going to get much of the signal
across the T-line to the other end...
From: Jon Powell [mailto:email@example.com]
Sent: Wednesday, December 22, 1999 7:29 PM
Subject: Re: [SI-LIST] : Clean sheet of paper
I saw an interesting paper at an IEEE conference (In Monterey, perhaps
EMC?) a couple of years ago. The author conjectured that the problem
people are having is that everything is dependent on charging a large
capacitor (the TLINE) and we have to get away from this.
His proposal. Make all drivers drive through a large series capacitor.
So the only signal sent it a pulse (+ or -) and you re-integrate the
signal at the receivers (through some simple logic he presented).
A couple of good things:
1) You don't need low-resistive contacts for your pads. Just squares of
metal that sit over top of each other and form the capacitors.
2) You don't have to worry about current content.
Some bad things (that he didn't mention).
1) Crosstalk noise probably gets more critical.
2) Nothing is really "clocked" anymore (hey, maybe this goes in the good
Once again, this is not my idea. (but I sort of like it).
regards and Happy Hollidays.
-- Jon Powell Director of HSSD Consulting Services Viewlogic Systems, INC. 805 988 8250
This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:14 PST