RE: [SI-LIST] : Return planes for Gigabit differential signals

About this list Date view Thread view Subject view Author view

From: Larry Miller (ldmiller@nortelnetworks.com)
Date: Thu Nov 30 2000 - 07:53:17 PST


If shunt capacitance doesn't do it, series inductance (easy to come by at
high frequencies!) often does. Also, ferrite beads work as inductors at high
frequencies and lossy inductors at lower ones.

Most transceiver vendors, for example, recommend a series ferrite/shunt cap
decoupling network, or a shunt cap/series ferrite or inductor/shunt cap (pi
network) to decouple their units.

One way to reduce noise in general is to not cause it in the first place!

1) Use differential signals that are closely spaced to keep return (mirror)
currents in the ground planes localized to near the pairs. At any distance
away, the fields and currents cancel. Oh, I see you are doing that....

2) Don't use vias except if you have to, as at connectors. Keep those
impedances matched to avoid noises caused by reflections.

3) Make your routing layer closer to the groundplane than the VCC plane. You
are dealing with an inverse-square law thing here, so you can more tightly
couple the signal to the closer plane quite effectively,

4) Use current-limited drivers if you can (some of the "hot LVDS" starting
to appear.)

5) If you are using actual real PECL, remember that it is VCC-referenced,
not ground-referenced, so you might well be better off using the VCC plane
for your return current mirror. (See Howard Johnson's book, I think, on
this). You do have to be careful here, because optical transceivers, though
operating at PECL levels, are NOT Vcc referenced.

6) I don't think the buried capacitance folks would necessarily agree about
it being "totally ineffective" or even "no help".

Most multi-Gb logic systems have to operate with some noise, which is why
differential architecture helps so much by cancelling. If you really need
low noise, you have to forget about saving layers. SONET systems at OC-192
and above have to use shielding that gives most engineers nightmares in
order to avoid jitter. Cell phones use extra layers as Faraday shields to
get 160 dB isolation between the transmitter and receiver that are as close
together as your ear and your mouth (and the noisy part (transmitter) has an
antenna!) Radar systems enclose sensitive circuits in thick, expensive metal
boxes.

You do what you gotta do.

Larry Miller

> -----Original Message-----
> From: Alex March [SMTP:alx_march@yahoo.com]
> Sent: Thursday, November 30, 2000 4:39 AM
> To: Michael Nudelman
> Cc: si-list@silab.eng.sun.com
> Subject: Re: [SI-LIST] : Return planes for Gigabit differential
> signals
>
> Michael,
>
> Thanks a lot for your answer.
> However, as far as I know, at the maximum
> frequencies associated with such data rates, which
> would be above 3GHz, the discrete decoupling
> capacitors are totally ineffective, and the buried
> capacitance is also of no help.
> So could you comment about the decoupling at those
> frequencies?
>
> Thanks again,
>
> Alex
>
>
> --- Michael Nudelman <mnudelman@tellium.com> wrote:
> > Any well-decoupled power plane is good enough. We
> > use it without
> > noticeable problems.
> >
> > Mike.
> >
> > Alex March wrote:
> >
> > > Hello everybody,
> > >
> > > I would like to ask your opinion about the
> > > following problem: I have a board with some high
> > speed
> > > SerDes at 2.125 Gbps (differential). Since there
> > are
> > > many lines like this, I would like to route some
> > of
> > > them between power planes (2.5V and 3.3V) or
> > between a
> > > GND and a power plane. This would save me some
> > layers.
> > > Now I saw some application notes which recomend
> > to
> > > use only GND planes as reference for these lines,
> > so I
> > > don't know which is the best solution.
> > >
> > > Many thanks,
> > >
> > > Alex
> > >
> > > __________________________________________________
> > > Do You Yahoo!?
> > > Yahoo! Shopping - Thousands of Stores. Millions of
> > Products.
> > > http://shopping.yahoo.com/
> > >
> > > **** To unsubscribe from si-list or
> > si-list-digest: send e-mail to
> > > majordomo@silab.eng.sun.com. In the BODY of
> > message put: UNSUBSCRIBE
> > > si-list or UNSUBSCRIBE si-list-digest, for more
> > help, put HELP.
> > > si-list archives are accessible at
> > http://www.qsl.net/wb6tpu
> > > ****
> >
>
>
> __________________________________________________
> Do You Yahoo!?
> Yahoo! Shopping - Thousands of Stores. Millions of Products.
> http://shopping.yahoo.com/
>
> **** To unsubscribe from si-list or si-list-digest: send e-mail to
> majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
> si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
> si-list archives are accessible at http://www.qsl.net/wb6tpu
> ****
>

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:30:18 PDT