RE: [SI-LIST] : Attenuation on a PCB trace

About this list Date view Thread view Subject view Author view

From: Jian X. Zheng (jian@zeland.com)
Date: Tue Apr 24 2001 - 08:31:19 PDT


Hi, Michael:

I think your e-mail part of the difference between microstrip and stripline.
However, there are some fundamental thing about the difference is not
clearly explained and I would like to add comments on it:

1. For a stripline, there is only one dielectrics. The wave (or energy) can
propagate parallel to the metallic traces. This is basically the TEM mode.

2. For a microstrip, there are two dielectrics. If the wave propagates
parallel to the metallic trace, the speed of the wave is different in the 2
dielectric media which is well mentioned in your e-mail. The boundary
conditions (continuous tangential E and H) do not allow the wave to
propagate parallel to the metallic trace. In fact, they have to propagate
with an angle to the direction of the metallic trace. They will reflect on
the ground plane and the dielectric surface. At a specific angle, the
sinusoidal wave will reflect back and forth between the ground plane and the
dielectric surface. The final result is that the wave power will propagate
along the metallic trace, indirectly by back and forth between the ground
plane and the dielectric surface. They are in fact the so-called TE or TM
modes in waveguides. In the metallic trace direction, the wave propagation
constant in the trace direction is Kp = K * sin( theta ), where K is the
propagation constant in the dielectrics, and the theta is the incident angle
between the ray of the wave and the normal direction of the ground plane.
The Kp's value will be the same for both the air and the dielectrics to
maintain the boundary conditions.

3. Circuit designers need to concern abou the frequency dependency of Kp. At
different frequency points, the "theta" will be different even you may find
some materials with the the K value independent of frequency. The frequency
dependency of the Kp will cause dispersion and problem in the rise time for
clock signals for wide band (or fast clock-rate) signals.

4. Full-wave simulators can model the frequency dependent properties of
transmission lines very accurately. However, its frequency dependent
results, represented as s-parameters, used to be difficult to apply to
traditional SPICE, which is limited to lumped elements. However, s-parameter
based SPICE simulator will be able to predict the rise time of frequency
dependent circuits precisely.

5. For more information on s-parameter based SPICED simulator, please visit
our web site. I will be giving a presentation on "Full-wave simulation and
S-Parameter SPICE simulation" on the Exhibitor's Presentation of the DAC in
June. Those interested are invited to present. Thank you very much.

-----------------------------------------------------------------------
Jian-X. Zheng, Ph.D
Zeland Software, Inc., 48890 Milmont Drive, 105D, Fremont, CA 94538, U.S.A.
Tel: 510-623-7162, Fax: 510-623-7135, Web: http://www.zeland.com
---------------------------------------------------------------------

> -----Original Message-----
> From: owner-si-list@silab.eng.sun.com
> [mailto:owner-si-list@silab.eng.sun.com]On Behalf Of Degerstrom, Michael
> J.
> Sent: Tuesday, April 24, 2001 7:13 AM
> To: Raymond.Anderson@Sun.COM; si-list@silab.eng.sun.com
> Subject: RE: [SI-LIST] : Attenuation on a PCB trace
>
>
> Ray,
>
> Getting back to Mr. Conn's reply (which I copied below) I think
> a lot of us know of the theory that a signal propagating
> along a microstrip should be quite different when compared
> to a signal that propagates along a stripline. This, as
> is mentioned below, is due to the fact that the microstrip
> signal must propagate through air and the board laminate.
> The theory (also mentioned below) is that the phase response
> vs frequency is not nearly as uniform in microstrip as
> that in stripline. The result is that the edge rates
> in microstrip are not as sharp as that for a stripline.
>
> I would bet that almost all of us simply ignore these
> predicted effects for microstrip. Also, I'm not sure
> that the HSPICE w-element model could predict these
> effects. The older version of the model had frequency
> independant L and C - perhaps the updates to the
> model are more flexible.
>
> My approach for incorporating very high speed signals
> is to route them in microstrip to avoid vias and so that
> I can get a wider line than stripline while maintaining
> a 50 ohm environment. I would like to hear from others
> as to the following:
>
> 1) Choice of microstrip or stripline for very high speed
> applications.
>
> 2) Did microstrip risetime degradation affect your choice
> for (1)?
>
> 3) Has anyone else tried to characterize the microstrip
> risetime degradation effect?
>
> 4) What EM tools are suitable for characterizing the
> microstrip risetime degradation effect?
>
> 5) Can any w-element models model this effect correctly?
>
>
> Thanks,
>
> Mike Degerstrom
>
>
> [Mr. Conn's reply now follows]
>
> Microstrip:
> Surface traces will have at least two major elements of signal
> propagation,
> one in the dielectric and the other above the surface. The latter is the
> fastest. Summing the two signal elements at the termination causes
> considerable risetime degradation for a "digital" signal because of
> different
> phases (and probably magnitude attenuation) of the same frequency
> components.
> The phase relationships of the different harmonics are also changed;
> therefore, an analytically predicted signal may be considerably
> off the mark
>
> from the real signal. In short, I would have no confidence in an
> analytically
> predicted microstrip waveform until proven by correlation with lab
> measurements. I have observed about 2:1 risetime degradation in a 9-inch,
> 50-Ohm microstrip versus the same length of stripline in the same board
> during TDR tests; therefore, I would have little confidence in using a
> 20-inch microstrip interconnection at the bit rate you noted.
>
> [...]
>
> Michael L. Conn
>
> _______________________________________________________________
> Mike Degerstrom Email: degerstrom.michael@mayo.edu
> Mayo Clinic; 200 1st Street SW ; Rochester, MN 55905
> Phone: (507) 538-5462 FAX: (507) 284-9171
> WWW: http://www.mayo.edu/sppdg/sppdg_home_page.html
> _______________________________________________________________
>
>
> > -----Original Message-----
> > From: Ray Anderson [mailto:Raymond.Anderson@Sun.COM]
> > Sent: Monday, April 23, 2001 4:48 PM
> > To: si-list@silab.eng.sun.com
> > Subject: RE: [SI-LIST] : Attenuation on a PCB trace
> >
> >
> >
> > Dave-
> >
> > Yes, our group has modeled the rise time degradation in
> > simulation using the
> > Hspice W-element and have achieved reasonable model to
> > hardware correlation.
> >
> > When using a microstrip model, the fields propagated through
> > the dielectric
> > and through the air are accounted for in the model, so I
> > don't think you
> > really want/need to kluge up a model that utilizes a stripline and an
> > airline to do the simulation.
> >
> > -Ray
> > Sun Microsystems Inc.
> >
> > >
> > >
> > >I interested in whether on not anyone has modeled this
> > degradation of rise
> > >time (see excerpt below) that Mike has mentioned using a
> > simulator. Can it
> > >be done by building two transmission lines in the simulator,
> > one for air
> > >dielectric and one for FR4? If the two parallel
> > transmission line method
> > >works how would one calculated how much impedence to allocate to each
> > >dielectric? Will a field solver tell you this?
> > >
> > >Dave Lorang
> >
> >
> > **** 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
> ****
>

**** 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 : Thu Jun 21 2001 - 10:11:42 PDT