Re: [SI-LIST] : Physcially-small far-end LVDS terminations?

About this list Date view Thread view Subject view Author view

From: D. C. Sessions (dc.sessions@vlsi.com)
Date: Tue Dec 21 1999 - 14:38:44 PST


MikonCons@aol.com wrote:
>
> Although many of the comments I have read in this thread have merit, a few of
> the participants in the subject discussion either have lost sight of other
> (perhaps "unknown unknowns") factors that need consideration for a more
> robust system design, or perhaps they have only designed in an electrically
> "clean" environment. With a purely modeling environment, this is a
> particularly easy trap to fall into by overlooking additional noise sources
> and not including them in the model.
>
> The use of two series 50 ohm resistors between the two differential receiver
> inputs, center-tapped with a capacitor to the local (receiver) ground is far
> more tolerant of common-mode noise (CMV) than a simple 100 ohm differential
> load. The arriving CMV is bypassed to the local ground, effectively reducing
> the CMV seen by the receiver by at least 6 dB (based on individual 50 ohm
> lines, loaded by 50 ohm resistors, ac-grounded via the capacitor). Note that
> the CMV signal may come from an externally impinging source as well as a
> poorly designed transmitter that is directly connected. If no ac coupling to
> the local reference is provided, the CMV magnitude could easily exceed the
> tolerance range of the receiver input because the CMV load impedance is
> extremely high (not existent).
>
> The variation in jitter (from imposed CMV) astutely noted by DC Sessions
> would also be reduced (though not eliminated) by the center-tapped
> configuration.

All true, but there's a flip side to the story.
LVDS, for better or worse, is designed to place the burden of
ground potential differences (common-mode shift) on the receiver.
The driver characteristics are only well-behaved over a narrow
range. The output impedance over that range is supposed to be
a loose match to the line impedance, but the common-mode point
is almost always set by common-mode feedback with (IME) much more
attention to stability than high-frequency response.

This makes for a nasty tradeoff. If the primary disturbance is in
ground shift (e.g. a cabled environment) then floating termination
of the kind commonly found in LVDS receivers works well. If the
primary disturbance is in coupled noise (crosstalk, EMI) then
having a victim that floats at high frequencies is ugly and terminating
at both ends is symmetrical but only marginally better (Of course, if
the interference is near a resonant point everything changes.)

Add to this the little detail that complex terminations tend to either
create stubs at the receiver or add $$LOTS$$ to the cost of the system
or both. If anyone gets the impression that I don't like LVDS, it's
only because I'm having to deal with it.

-- 
D. C. Sessions
dc.sessions@vlsi.com

**** To unsubscribe from si-list: send e-mail to majordomo@silab.eng.sun.com. 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 - 11:39:11 PST