[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Spread Spectum Clocks

Hi All-

Over the past few months, there have been a number of posts on this list
regarding spread-spectrum clocks and the inability to use PLL based clock
regeneration chips to control clock skew on an adapter card.  While
posts were very informative, I did not see a final consensus on this issue
the archives.  I'm currently working on an adapter project based on Intel's
960RP CPU and am looking for a clock regeneration solution.  Any help would
be greatly appreciated.

The problem is summarized as follows:

The PCI spec allows clocks from 0-33 Mhz.  Some system designers are are
taking advantage of this by using 'spread-spectrum' clocks to reduce EMI
emissions.  In such a design, the clock frequency is continuously varied to
eliminate emission peaks at any one frequency.  Most (all?) PCI clock
regeneration chips use either an analog or digital PLL to regenerate low
copies of the PCI input clock.  All such chips I've seen either can't
guarantee operation over the entire 0-33 Mhz frequency range, and/or can't
adjust fast enough to a spread-spectrum input clock to avoid losing
synchronization.  (I've also run across some related problems specific to
960RP which are inappropriate for this list - please contact me directly
e-mail if anyone would like to discuss.)

So, the question remains - if PLL clock recovery chips can't meet the spec,
how are designers regenerating the PCI clock?  Any advice from I.C. vendors
or other designers would be greatly appreciated.

Thanks in Advance,

-Mike Martin
Millennium Engineering, Inc.

3)  On a more general note, if the PCI spec. allows clocks from 0-33 Mhz,
further if some systems are designed with spread spectrum clocks to reduce
EMI, what methods (i.e. chips) are designers using to rebuffer clocks?  Any
recommendations?  Perhaps a list of clock regeneration chips could be added
to the FAQ.