Browse Prior Art Database

Statistical Method of Noise Estimation in a Synchronous System

IP.com Disclosure Number: IPCOM000112866D
Original Publication Date: 1994-Jun-01
Included in the Prior Art Database: 2005-Mar-27

Publishing Venue

IBM

Related People

Cook, J: AUTHOR [+5]

Abstract

Disclosed is a process to analyze coupled noise within a synchronous digital machine. Former methods simply summed noise components or used root mean square techniques to estimate the total cross-talk on a quiet line. A method that considers the spatial relationship of the noise generating components as well as the timing of the incident pulses is discussed. Consideration is given to factors such as path delay, reflected noise, driver slew rate, noise pulse width, termination, net topology, noise from other sources and the probabilistic nature of the time when a signal is launched as well as other parameters. The result is a probability of achieving a noise level on the circuit and a noise versus time envelope. The method is more accurate than previous techniques due to timing and statistical considerations.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 12% of the total text.

Statistical Method of Noise Estimation in a Synchronous System

      Disclosed is a process to analyze coupled noise within a
synchronous digital machine.  Former methods simply summed noise
components or used root mean square techniques to estimate the total
cross-talk on a quiet line.  A method that considers the spatial
relationship of the noise generating components as well as the timing
of the incident pulses is discussed.  Consideration is given to
factors such as path delay, reflected noise, driver slew rate, noise
pulse width, termination, net topology, noise from other sources and
the probabilistic nature of the time when a signal is launched as
well as other parameters.  The result is a probability of achieving a
noise level on the circuit and a noise versus time envelope.  The
method is more accurate than previous techniques due to timing and
statistical considerations.  Treatment of the resultant output is
discussed and compared to simpler deterministic methods.  Failure
criterion, critical time periods and some sources of errors are also
considered.

      Coupled noise often attracts our interest as a problem
encountered during the wire layout phase of package design.  Intent
on maintaining the noise on a wired path at a value less than the
receiver noise margin, methods for estimating the total coupled noise
are devised.  Often a worst case scenario has been considered and
analyzed.  The result is used to determine what constraints must be
placed on physical wire placement in order to control coupled noise.
Those constraints are then used during wire layout.

      Another method that has been used has been to tabulate the
magnitude of the coupled noise pulse that could be induced by each
wire section adjacent to the network of interest and simply sum up
all these values.  The sum is then adjusted to account for just how
much of the individual pulses might actually superimpose.  Factors
such as .5, .8, etc. have been used to adjust the sum.

      Methods that utilized RMS addition have also been employed.
The magnitudes of the individual noise pulses are combined as if they
were statistical variables having a mean of zero and a standard
deviation equal to some portion noise pulse's magnitude.

      Frequently the result of these methods has included burdensome
wire layout constraints, manpower dedicated to reducing wire density
on a package and additional turn around time in the wiring design
phase.  These operations may be viewed as costly penalties especially
when considering that they may not be necessary since the prediction
methods are not very accurate.

      More accurate approaches that can be employed include
constructing the noise versus time envelope of the pulses coupled
into the net under consideration and additionally a calculation of
the probability that noise pulses superimpose on a net.  The envelope
approach combines all possible superpositions of pulses or a better
de...