From:  MitchAlsup <MitchAlsup@aol.com>
Newsgroups: comp.arch
Subject: Re: Use of bidirectional transmission gates/transistors beside memory 
	and FPGAs
Date: Wed, 11 Jul 2007 09:48:49 -0700
Message-ID: <1184172529.386771.4760@n2g2000hse.googlegroups.com>

On Jul 11, 8:26 am, acd <acd4use...@lycos.de> wrote:

> Are there other cases where such a bidirectional use of pass
> transistors or transmission gate is common?

T-gates are often found in latches, flip-flops, shift registers--or
maybe I should say that they used to be used often in latches, flip-
flops, and shift registers.

T-gates cannot be used "willy-nilly". Back in the early 1990s, Ross
Technology had a circuit design problem where a flip-flop would
occasionally suffer a 1->0 transition when no input signal was
'logically' applied to the T-gate input part of the flip-flop. A long
wire had coupled the output of a strong driver to this flip-flop. And
it was found that the signal at the end of this long wire had some
ringing. The amount of ringing in the downward direction was enough to
turn on the n-channel device and this allowed the charge stored to be
removed. The amount of ringing on the upwards side was not sufficient
to turn on the p-channel device and alter the stored charge. Thus the
only observed failure mode was the 1->0 transition.

The general rule of thumb for deap sub-micron design is to avoid T-
gates unless you can so constrain the signaling environment (levels,
edge speeds, and timing) that T-gate failures happen no more often
than SRAM read failures. This relegates T-gates to very well controled
arrays and out of the general library of gates and devices.