Browse Prior Art Database

Using a Single Source for All Clock Generation

IP.com Disclosure Number: IPCOM000114907D
Original Publication Date: 1995-Feb-01
Included in the Prior Art Database: 2005-Mar-30
Document File: 2 page(s) / 58K

Publishing Venue

IBM

Related People

Johnson, DWJ: AUTHOR

Abstract

A method for minimizing the amount of logic necessary for clock generation in a multiprocessor system is described.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 54% of the total text.

Using a Single Source for All Clock Generation

      A method for minimizing the amount of logic necessary for clock
generation in a multiprocessor system is described.

      Complex systems require a multitude of logic clocks.  In a
hypothetical multiprocessor system, each processor complex requires
four different frequencies, and the shared memory card requires five.
With an eight processor system, this would then be at least 37
oscillators.  This leads to a high cost, increased board space
requirements, and higher failure rates.

      Oscillators are electro-mechanical, relying on vibrating
crystals to produce the desired frequency.  They are one of the
higher failure rate components in the system.  Having more of them
results in a higher than allowable failure rate.

      In addition, something needs to be done to make the system
synchronous to align the clocks between processors, and so that each
processor's Time-Of-Day (TOD) clock remains in synch with the others.
It might not be desirable if each processor kept a totally
independent time in relation to the other processors in the system.

      A single oscillator is used to form a time base for the system.
Drivers are used to ensure signal quality is maintained, and so that
each load is independently driven by its own driver.  This time base
can be used directly for one of the frequencies on the shared memory
card, and distributed to each processor for one of theirs.  A simple
flip-flop...