Re: [SI-LIST] : crosstalk through resistor networks?

Dennis Tomlinson (det@tellabs.com)
Thu, 20 Aug 1998 15:12:39 -0500

Jory McKinley wrote:
>
> Hello Dennis,
> Sorry for the late repsonse but I noticed one of your replies concerning
> reverse crosstalk. I am/was under the impression that reverse crosstalk
> amplitude is INDEPENDENT of electrical coupled length between two
> conductors. The Kb coefficient is related to physical properties of the
> traces (distance separation and distance from a ground/power reference
> plane), the amplitude of the reverse crosstalk on the victim is
> proportional to the step size of the aggressor and the duration of the
> crosstalk on the victim is 2 times the electrical coupled length (the tp
> below)between the traces.
> This would leave us with:
> Vb(victim) ~= Kb*dVs(aggressor)
>
> Am I missing something?

Everything you have said above is correct if the coupled length is greater
than than what some call the critical length, lc. Referring to "BiCMOS/CMOS
System Design", by Buchanan, pp 118 & 119 (where I got the information for
the first posting):

lc = tr/2t'pd

Where:

tr = risetime, and
t'pd is the propagation constant of the media.

When the coupled length is > lc, your description of the backwards crosstalk
voltage is exactly correct. Additional coupled length will only increase
coupled pulse width, not amplitude.

When the coupled length is <= the critical length, Vb amplitude is a linear
function of coupled length, as given below.

On pages 120 and 121, Buchanan shows three families of curves for Vb
(amplitude) vrs coupled length extracted from simulation runs using
Viewlogic/QDT's XTK tool. They are piecewise linear with a single
discontinuity at l = lc. The graphs show a linear, positive slope
ramp for l <= lc, and are flat for l > lc.

Cheers,

Dennis

>
> Enjoy,
> Jory
>
> At 11:46 AM 8/14/98 -0500, Dennis Tomlinson wrote:
> Andrew Phillips wrote:
> >
> > Hello,
> >
> > I am interested in using some small-package (QSOP) resistor networks for
> > terminating a large number of bus-signals in a design. Signal rise-times
> > range from ~200ps to 1ns or so.
> >
> > I have found some devices that look suitable here:
> > http://www.bourns.com/cat/MNETXX1.htm
> >
> > I am wondering whether such an approach will have crosstalk problems.
> > Has anyone analysed this and can give me some advice?
> >
> > I presume that if traces are suitably separated before and after the
> > resistor package then the crosstalk induced between adjacent signals
> > will be reduced. These signals are very close together for the 3mm or so
> > across the resistor package - is this likely to cause problems?
> >
> > Thanks for any help,
> >
> > Andrew Phillips
> > Supercomputing Systems AG
> > Zurich, Switzerland
> >
>
> Hi Andrew,
>
> I can't give you rigorous results from an exhaustive analysis, but perhaps
> a rule of thumb will suffice.
> Assuming backwards crosstalk dominates, the coupled voltage between two nets
> is given by:
> Vb = Kb*(2tp/tr)*dVs
> Where:
> Vb is the coupled voltage in the backwards direction,
> Kb is the backwards coupling coefficient,
> tp is the propagation time across the coupled length,
> tr is the signal rise time, and
> dVs is the source voltage swing of the aggressor.
>
> The ratio 2tp/tr can be viewed as the percent of the source voltage
> swing available to be coupled from aggressor to victim across the
> coupled length. Assuming about 6 ps/mm propagation constant across
> the 3 mm body width, this ratio takes on values from 18% for 200 ps to
> 3.6% for 1 ns.
> Values of Kb, I can only guess at, but 0.25 is a rather high coupling
> coefficient. Using this value gives Vb in the range of 0.045*dVs for
> 200 ps risetime, down to 0.009*dVs for 1 ns.
>
> Also, for the QSOP package, you have an additional 1.5 mm pin length
> from the seating plane to the body on each side of the package. The
> above does not take pin-to-pin coupling into account.
>
> BTW, is this R-pack used as a series terminator? Do all bus signals
> switch at the same time - with some time lag before stability is
> required? Do the signals all propagate the same direction down
> the bus? If any/all the answers to the above are yes, you could have
> errorless circuit operation with some rather high levels of crosstalk.
>
> Cheers,
>
> Dennis
>

**** 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 ****