Browse Prior Art Database

Global Routing Techniques for Signal Crosstalk Avoidance and Prediction

IP.com Disclosure Number: IPCOM000109169D
Original Publication Date: 1992-Aug-01
Included in the Prior Art Database: 2005-Mar-23
Document File: 7 page(s) / 309K

Publishing Venue

IBM

Related People

LaPotin, DP: AUTHOR

Abstract

Package wiring is often accomplished by dividing the problem into two well-known steps, global routing followed by detailed routing. Disclosed is a practical method for predicting coupled noise during global routing as well as finding global routes which ease the burden on detailed wiring programs to find noise-free solutions.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 24% of the total text.

Global Routing Techniques for Signal Crosstalk Avoidance and Prediction

       Package wiring is often accomplished by dividing the
problem into two well-known steps, global routing followed by
detailed routing.  Disclosed is a practical method for predicting
coupled noise during global routing as well as finding global routes
which ease the burden on detailed wiring programs to find noise-free
solutions.

      A coupled noise model is described for use during the package
global routing step.  Because exact wiring adjacencies are not known,
the model can only capture the expected noise behavior; the model
approximates what would be seen if detailed adjacencies were known.
The approach captures the basic coupled-noise phenomena by employing
a timing model involving driver launch times and receiver sampling
windows.  As a given net is being routed, a probabilistic analysis is
used to estimate the effects of undesired timing coincidences on
overall coupled noise. By considering timing, the model provides the
designer with a more realistic noise analysis than that obtained by
the conservative straightforward approach of analyzing parallelism on
all wires, independent of timing.  The results are beneficial for
predicting the expected noise prior to detail wiring as well as
finding global routes which ease the burden on the detailed wiring
program to find crosstalk-free wiring solutions.

      Consider a 2-point connection (net) modeled as a simple
transmission line terminated at one end by a driver, D, and a
receiver, R, at the other end (multi-point nets are a simple
extension).  Assume that the leading edge of a signal leaves the
driver at time t0.  As shown in Fig. 1, the signal will propagate
down the transmission line (line 1) reaching point a at time t1,
finally reaching the receiver at time t2, where t0 < t1 < t2.  If a
noise pulse (line 2) active at time t1 is injected onto the line at
point a, the noise will also propagate down the transmission line,
and depending on when the receiver samples the signal, the noise
could be interpreted as an undesired distortion to the original
signal.  Time invariance applies to the transmission line model, so
that a noise signal active at time t1 + w, injected at point a, would
reach the receiver at time t2 + w, and could possibly cause an
undesired signal distortion. For the purposes of this model, only
near-end coupling effects are considered.  Thus, the coupling
polarity of line 2 is assumed to be such that the induced noise pulse
on line 1 travels in a direction from driver to receiver.

      For each routed signal, the generation interval, G(l,t), is
defined as the interval when the driver launches the signal on the
transmission line, where the pair (l,t) defines the leading and
trailing edges of the pulse, respectively.  Similarly, the
susceptibility interval, S (l,t), defines an interval when the
receiver is sampling the line and can be affected by a signal present
o...