Re: Eye Patterns

Tim Postel (tim@qdt.com)
Mon, 22 Jan 96 18:27:55 PST

Greg Edlund writes:

[*snip*]
>> I'd be willing to bet not many people simulate eye patterns;
>> they probably try to get as close as they can and then
[*snip*]

Raymond Anderson replies:

[*snip*]
> is true that jitter is a key component of an eye pattern as mentioned above
> by Greg, other factors such as inter-symbol interference due to group delay
[*snip*]

It is worth adding to Raymond's comments that transmission lines, especially
those that tend to exhibit underdamped ringing outside of threshold following
a state transition, are a excellent generators of inter-symbol interference on
non-repetitive (i.e. datapath) signal nets:

Ex. : ... 0 0 0 0 0 1 (sequence A)
... 0 0 0 1 0 1 (sequence B)

in sequence B, the last transition (from 0 to 1) occurs with a different
set of initial net conditions than in sequence A owing to prior switching
not present in "A", and could cause either an increase or decrease
in switching time relative to "A".

This extra "jitter" eats into available setup/hold margins available at
(datapath) input registers, and is independant and additive to any
"source jitter" along a given path. We (Quad Design) have customers who have
used XTK/TLC with psuedo-random datapath excitation in order to flesh out
exactly this phenomena, and at least one customer who "rolled his own"
eye chart post-processor...

-Tim Postel (tim@qdt.com -OR- tpostel@viewlogic.com)

Mon Jan 22 18:19:07 PST 1996
c/o
Quad Design Technology, Inc.
1385 Del Norte Road, Camarillo, CA 93010
(805) 988-8250 FAX: (805) 988-8250