From: David Instone ([email protected])
Date: Tue Apr 24 2001 - 02:55:48 PDT
Thanks for your comments. My question was to some extent
hypothetical, but based on a situation I could see arising. I've
inserted a few responses below.
Ingraham, Andrew wrote:
> > Source (series) termination of transmission lines to absorb
> > reflections is well known. However, how long does the transmission line
> > have to be to be recognised ,by the signal, as a transmission line?
> There is no minimum length. A t-line of any length is a t-line. When it is
> very short, we might not bother to model it as a t-line (purely for the sake
> of mathematical simplicity); but at any length, it pretty much has to obey
> the Telegrapher's equations, Maxwell's equations, etc.
I have a hard time visualising that in practice. When I TDR lines I
know that any discontinuities which last less than the TDR risetime do
not appear with their full value. So yes, to a zero risetime signal an
infinitely short line would appear as a line, but for signals with real
risetimes I think the line has to have a minimum length depending on the
> But what I think you are asking is, can you somehow magically "fix" an
> impedance mismatch problem over here by adding or lengthening a trace over
> there until it behaves like a transmission line? If so, then I think you
> misunderstand. (Unless your problem is narrowband, i.e., RF; as opposed to
> logic signals.)
No, not a "fix" but a means to reduce re reflections. and yes the
application is Gbaud signalling.
> When the driver's impedance matches the t-line connected to it, all it does
> is eliminate reflections that might otherwise happen RIGHT THERE at the
> driver/t-line boundary. It does nothing for the impedance mismatches
> elsewhere, i.e., between the connector and the t-line that comes before it,
> or the one after it.
> The reflection from the 40 ohm / 50 ohm interface will move towards the
> source. Upon reaching the 40 ohm / 75 ohm interface, it will reflect again,
> causing a forward-moving wave. It doesn't matter whether that "75 ohms" is
> just the driver, or the driver plus any arbitrary length of transmission
> line. (Well, it might, as a second- or third-order effect, partly because
> the driver's actual impedance is probably nonlinear and/or time-varying,
> etc. But for basic considerations, no difference.)
True but while some of the energy is reflected back towards the load the
greater part carries on towards the driver and if the driver is source
terminated it ends there and doesn't get re reflected so I get just one
ring and not a whole sequence. Now my thinking is - if the driver is
hard on the connector it's just 75 ohms trying to terminate 40 ohms,
doesn't work and in any case putting the driver hard on the connector is
generally not possible. If the driver is some way from the connector
then it needs a t-line to the connector and if that t-line matches the
driver impedance then anything that gets back to the driver stops
there. Saves the shunt resistor and it's imperfections
> So don't bother lengthening the trace; you will just get more delay.
> I wonder if you would have better luck with a 50 ohm driver, to match the
> other trace, and accept the (usually short) impedance discontinuity through
> the connector. If the driver is fixed at 75 ohms, you might try adding a
> shunt resistor there to bring the combined impedance down. Or maybe not.
> It might not actually help; but there are various things you can *try*
In theory could work but if the series R in the package is a fair way
from the package 'pin', and in a large BGA or a GBIC that could be 1/2
inch or more, then there is likely to be a section of t-line in the
package that's 1/2 inch long, so putting a shunt resistor at the package
pin might make things worse, the shunt resistor really needs to be hard
on the series resistor.
Dave Instone. Compliance Engineer Storage Systems Development, MP24/22 Xyratex, Langstone Rd., Havant, Hampshire, P09 1SA, UK. Tel: +44 (0)23-92-496862 (direct line) Fax: +44 (0)23-92-496014 http://www.xyratex.com Tel: +44 (0)23-92-496000
**** 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 ****
This archive was generated by hypermail 2b29 : Thu Jun 21 2001 - 10:11:42 PDT