Re: [SI-LIST] : Definition of Hi-Speed (UC)

About this list Date view Thread view Subject view Author view

From: Donald Telian ([email protected])
Date: Tue Nov 02 1999 - 06:21:22 PST


Lum Wee,

That's not a stupid question. The answers you've already received prove
that many others have puzzled over it too.

After a number of years of trying to explain what "High-Speed" really is,
my favorite explanation has become:

"A net can be considered 'high-speed' when you have to do something other
than simply connect it."

There it is. What do you think?

This explanation naturally carries you in to investigating things like
connection length, topology, impedance, material, driver/receiver
technology, system timings, spacing from other signals, and overall
characterization. As such, "high-speed" nets carry with them other
requirements aside from netlist connectivity.

In retrospect, as applications demanded it and technology allowed it
engineers began to increase frequencies and reduce edge speeds. As that
happened, *digital* engineers had to begin "planning" their connectivity.
(Note that engineers in other disciplines had been handling these kinds of
frequencies for quite some time.) As they found the need to study/plan
certain connections, they called those that needed the attention
"high-speed" signals. So I think that's why the above definition servies
us well - high-speed is that side of digital connectivity we have to be
more careful about.

I do not mean to say that you should analyze *every* net - that's where
your skill comes in. Pick an important net and demonstrate to your manager
what will happen if it's routed in an arbitrary fashion.
 
The fact is that there are many electrical connections that must be
analyzed carefully to be made correctly. Experience has shown that this is
true not just on PCBs, but also in ICs, Cables, etc. But unfortunately
most people have to be stung before applying any mosquito repellent. Many
times, production line systems have begun failing because someone made a
change in the way a net was routed or in a component supplier. So what
happened? Did the net suddenly *become* "high-speed". Actually, it was
all along. It's just that no one ever realized it and attached its
connectivity/operating requirements to the design database.

You can spend a lot of time describing high-speed in terms of rise times,
round trips, crosstalk, terminations, daisy-chains, impedances,
reflections, and the like. I've talked myself hoarse doing this.
Unfortunately, many engineers just aren't interested, especially when
they've spent too many years just connecting things together without
thinking about it or modeling anything. At the PCB level, integration and
increasing speed has (interestingly) changed the focus of the savvy
engineer from "what connection must be made?" to "how must the connection
be made?". I like to view "high-speed" in terms of what it
suggests/requires to be different in the overall design process.

Having watched the typical PC Motherboard go through the transition to
'high-speed', I did see 33MHz become the generally accepted point at which
analysis had to be done. But "the experts" will always tell you that MHz
isn't the issue...

Also, I published a paper at DesignCon '97 that articulated the many facets
of High-Speed Design. It was voted "best paper" at the high-performance
conference, and later used by a number of engineers to help explain to
their management how/when/where/why to do High-Speed Design. If you (or
anyone else) would like a copy of the PowerPoint file, just let me know
off-line.

Don't give up!

Donald Telian
CADENCE

At 12:54 PM 11/01/1999 +0800, Lum Wee Mei wrote:
>Pardon me for asking this stupid question because I am at a loss of how
>to explain hi-speed to my boss. He thinks that hi-speed is as simple and
>straightforward as resistance = V/I and nothing else. Hi-speed should be
>some circuits that need to operate at xxMHz or more. Anything less than
>xxMhz is not hi-speed.
>
>I would appreciate anyone of you experts out there who can enlighten me
>in a simple and easy to understand definition so that my boss can
>understand.
>
>Regards.
>
>
>
>
>**** To unsubscribe from si-list: send e-mail to
[email protected]. In the BODY of message put: UNSUBSCRIBE
si-list, for more help, put HELP. si-list archives are accessible at
http://www.qsl.net/wb6tpu/si-list ****
>
>
>
Donald Telian
Cadence Design Systems
phone: 408-944-7791
email: [email protected]

**** To unsubscribe from si-list: send e-mail to [email protected]. In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at http://www.qsl.net/wb6tpu/si-list ****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Feb 29 2000 - 11:39:31 PST