[SI-LIST] Re: Design rules for estimating crosstalk
- From: "Carrier, Patrick" <Patrick_Carrier@xxxxxxxxxx>
- To: "gwmaichen@xxxxxxxxxxxxxx" <gwmaichen@xxxxxxxxxxxxxx>, "andersson.pehr@xxxxxxxxx" <andersson.pehr@xxxxxxxxx>
- Date: Mon, 23 Apr 2012 20:10:37 +0000
Hi Pehr--
I agree with Wolfgang that you are usually best off using a field solver, but
rules of thumb can be useful too, as long as you understand the assumptions and
limitations used to create them.
For instance, FEXT in stripline is not always zero. That only occurs if you
have a homogenous dielectric. Even in a symmetrical stripline, one dielectric
will be a core and the other a prepreg, so there will be some difference in the
dielectric constants. This effect is most dramatic on a dual stripline, which
usually consists of two smaller dielectrics and a large dielectric between
them. Since dielectric constant tends to increase with larger dielectric
heights, the effect of that difference in dielectric constants (the dielectric
between the trace and reference plane, and the dielectric on the other side of
the trace) will lead to noticeable FEXT in stripline. Of course, the most
nonhomogeneous dielectric setup is microstrip, which is a mixture of a
dielectric and air, which in turn creates the greatest amount of FEXT. It has
to do with the relative amounts of inductive and capacitive crosstalk.
Another rule of thumb is spacing things out according to trace width. Really,
you should be concerned about the height of the trace above the reference
plane. Fields drop off proportionally to the distance of the trace above the
reference plane, so spacing rules should be something like 3H as opposed to 3W.
In the case of a 50-ohm line (as Wolfgang also points out) these two
parameters track each other pretty well, but it is really the height (H) that
you should be concerned about.
I have a presentation which explains both of the concepts above in greater
detail, and with pictures. It also goes into specific details on differential
crosstalk, which was the subject of your original inquiry.
Take a look if you are interested:
http://www.mentor.com/products/pcb-system-design/multimedia/crosstalk-fundamentals-webinar
--Pat
-----Original Message-----
From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx] On
Behalf Of Gwen und Wolfgang Maichen
Sent: Monday, April 23, 2012 9:51 AM
To: andersson.pehr@xxxxxxxxx
Cc: Wolfgang.Maichen@xxxxxxxxxxxx; si-list@xxxxxxxxxxxxx
Subject: [SI-LIST] Re: Design rules for estimating crosstalk
Which equations are you referring to in particular?
The challenge is to get accurate numbers for mutual and self capacitance and
inductance. Only a 2D field solver can give you good numbers for that.
Crosstalk then follows from that through very simple analytical relationships
(which then also correctly predict that FEXT is zero for symmetric striplines).
Is there a specific reason why you absolutely don't want to use a field solver?
There are some free ones out there that can easily deal with these calculations
(e.g. MMTL/TNT, which I have recommended in the past). For specific situations
you can always fit analytical expressions to these results. With a bit of care
such expressions can be quite general, e.g.
express distance in terms of line widths so you are independent of the
particular trace dimensions. Also most paths in high speed signaling tend to be
50 Ohm single ended (or 100 Ohm differential), so one less parameter that you
need to vary.
Finally, unless you want to build a coupler with a specific coupling constant,
most of the time the goal is simply to avoid excessive amounts of crosstak. For
that, rules of thumb are widely known, especially if you assume typical PCB
materials (eps_r arounf 2-5, line impedance 50 Ohms):
- NEXT increases for up to one rise time, the stays at a plateau
- FEXT increases for several meters (assuming typical PCB trace dimensions)
- for NEXT < 1% (with coupled lengths larger than half a rise time), trace
separation > 3-4 line widths for striplines and 4-6 line widths for microstrips.
Wolfgang
On Mon, Apr 23, 2012 at 4:31 PM, Pehr Andersson <andersson.pehr@xxxxxxxxx>wrote:
> Thanks Wolfgang!
> How accurate are the equations that are used for crosstalk analysis
> and that are based on circuit analysis?
> For example with those equations we cannot explain why FEXT = 0 in
> striplines (theoretically) What are other alternatives, I have read
> that modal analysis could be used to accurately model crosstalk, but I
> couldn't find any explanation what it is.
>
> Best regards, Pehr
>
>
>
> 2012/4/23 <Wolfgang.Maichen@xxxxxxxxxxxx>
>
> > Hello Pehr,
> >
> > actually crosstalk for striplines decreases exponentially, i.e.
> > exp(-d/const.). This means beyond a certain distance it drops off
> > much faster than the 1/d^4 you assumed.
> >
> > NEXT increases linearly with coupling length for coupled sections
> > with a round trip delay shorter than the signal rise time (this is
> > why in the
> past
> > people assumed ANY crosstalk - NEXT and FEXT - would increase
> > linearly
> with
> > coupled length, because rise times were cery long). Only for lines
> > longer than that there is the plateau in the NEXT.
> >
> > Regards,
> >
> > Wolfgang
> >
> >
>
>
> ------------------------------------------------------------------
> To unsubscribe from si-list:
> si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
>
> or to administer your membership from a web page, go to:
> //www.freelists.org/webpage/si-list
>
> For help:
> si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
>
>
> List forum is accessible at:
> http://tech.groups.yahoo.com/group/si-list
>
> List archives are viewable at:
> //www.freelists.org/archives/si-list
>
> Old (prior to June 6, 2001) list archives are viewable at:
> http://www.qsl.net/wb6tpu
>
>
>
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List forum is accessible at:
http://tech.groups.yahoo.com/group/si-list
List archives are viewable at:
//www.freelists.org/archives/si-list
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
