RE: [SI-LIST] : Combined single-ended/differential termination pa ckage?

About this list Date view Thread view Subject view Author view

From: Ron Miller ([email protected])
Date: Tue Feb 27 2001 - 15:15:06 PST


-----Original Message-----
From: Larry Miller [mailto:[email protected]]
Sent: Tuesday, February 27, 2001 11:52 AM
To: 'Ron Miller'; 'Eric Goodill'; [email protected]
Subject: RE: [SI-LIST] : Combined single-ended/differential termination
package?

I agree with your statements, which brings up some additional questions
(below). Bear in mind that I am looking at 2.5 Gb/s (3.125 Gbaud signaling
rate) with the "souped up" LVDS made by Vitesse and others and proposed for
the IEEE 802.3 10 Gb Ethernet standard.

-----Original Message-----
From: Ron Miller [mailto:[email protected]]
Sent: Tuesday, February 27, 2001 11:09 AM
To: 'Larry Miller'; 'Eric Goodill'; [email protected]
Subject: RE: [SI-LIST] : Combined single-ended/differential termination
package?

If differential termination is used(with a high Z comparator)
1. common mode currents do not flow from the receive lines
        back to the source through the ground.

        Remember that though the current of a comparator is very
        low it does exist and is not perfectly balanced as in a
        differential amplifier input-offset voltage and current.

        This input-offset voltage/current(with respect to ground)
        can only flow (causing noise in the output)if the signals
        are referenced to ground near the comparator. Floating the
        input does not allow this current to flow.

>>>>> So then what happens? If the current cannot flow, then it would seem
that this would imply a very high common-mode impedance, so that you would
get big voltages from unbalance. Wouldn't the comparator see that?

++++++The voltages that result are the voltages from one end of the line
        at the driving source, and the destination end of the line. Obviously
        with high speed/data rate signals there will be an instantaneous and
        an average voltage difference depending on how it is filtered and
        measured. So, the magnitude is no greater than the normal signals.

2. Common mode voltage will develop between the input of the
        comparator and the driving source if the termination
        resistors are not referencded to ground. However, from
        the point of view of the comparator it cannot be detected.
        So, it does not matter.

>>>>> Hmmm at high speeds comparators don't have that great of common-mode
rejection. Also, in a backplane application this sounds like you are putting
all of the inter-board noise in series with signals-- again, at high
impedance.

+++++++ No, sadly, the comparators, and all semiconductors loose their
        effectiveness at some maximum frequency, and that limits our bandwidth.

        However, where the bandwidth of the components is greater than the
        data rate and all important harmonics, eg X3 or X5 a good eye pattern
        can be obtained without adding noise. Low pass filtering removes the
        stuff above the required bandwidth.

        It does not bother me to put the inter-board common mode noise into an
        incomplete signal path where it cannot flow to the comparator.

>>>>> How about at the source end? With current drive I would think that you
would want a termination. What do you say in this area?

At the source end, on practical drivers, the drive always has an intrinsic
impedance which might even be a pull-up/pull-down or the transistor itself.
That source impedance is your termination.

>From a practical point of view I have always got my best, lowest
differential test results when floating the termination.

>>>> One cannot argue with Reality, regardless of theorizing.

+++++ Thanks for the support. Incidentally, the theory came after testing
and head scratching.

A FURTHER NOTE ON GROUND COUPLING:

Considering single-ended(unbalanced signals) all the ground current
for each circuit must flow back to the source through that ground. Now
visulaize a PCB with many signals and power returns flowing from one part
of that ground to another, flowing across different parts of the board and
in different directions someplaces orthogonal and some places in the same
direction coupling strongly. Well, ground is not ground but a matrix of
tranmission lines all connected together and mixing the low level sensitive
signals with higher current power currents. It is a mess, too complex
to analyze thoroughly on a practical basis. So, differential signals with
care to keep common mode currents from flowing by differentially terminating
is the best we can do to keep the noise/coupling effects down.

Ron Miller

Ron Miller

-----Original Message-----
From: Larry Miller [mailto:[email protected]]
Sent: Monday, February 26, 2001 8:11 PM
To: 'Eric Goodill'; [email protected]
Subject: RE: [SI-LIST] : Combined single-ended/differential termination
package?

---which brings up something that has been bothering me:

the LVDS standard goes with the single 100-ohm line-to-line resistor (which
is REQUIRED for DC biasing) but does not seem to provide any specific line
to ground termination. I know that LVDS was originally intended for cables,
but it is showing up in a lot of 2-3 Gb signals, including serial
backplanes.

It looks to me as though you end up with unterminated (to ground) wires in
the pairs.

What do you do with this?

Seriously.

Larry Miller

-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Eric Goodill
Sent: Monday, February 26, 2001 6:09 PM
To: [email protected]
Subject: Re: [SI-LIST] : Combined single-ended/differential termination
package?

Ron,

Our reason for both resistors is to terminate any common-mode signal
that might be on the lines either from crosstalk or from an unbalanced
driver.

-Eric

Ron Miller wrote:
>
> Hi Eric
>
> Why not just use a single resistor to terminate the differential mode,
> which is the signal you are intrested in.
>
> Tieing the center-tap to ground allows common mode currents to flow
> which usually end up allowing power supply ground currents to get into
> the sensitive signal paths.
>
> -----Original Message-----
> From: Eric Goodill [mailto:[email protected]]
> Sent: Monday, February 26, 2001 3:47 PM
> Subject: [SI-LIST] : Combined single-ended/differential termination
> package?
>
> Hi,
>
> We often use this termination scheme:
>
> SIG+ -------+
> |
> R
> |
> +--C--+
> | |
> R V
> | GND
> SIG- -------+
>
> That is, two resistors in series between the differential signals with a
> center-tapped capacitor to ground. Is there a combined device to sucks
> all these parts into a three-pin package?
>
> Regards, Eric

--
Eric Goodill                    Redback Networks
[email protected]               300 Holger Way
voice: (408) 571-5319           San Jose, CA 95134-1362

**** To unsubscribe from si-list or si-list-digest: send e-mail to [email protected]. 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 [email protected]. 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 [email protected]. 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 - 14:30:58 PDT