[SI-LIST] Re: Routing of 12 GHz diff pairs
- From: "e.sweetman" <e.sweetman@xxxxxxxxx>
- To: <jonpowell@xxxxxxxxxxxx>, <phillip.r.wellington@xxxxxxxxxx>,<emj14@xxxxxxxxxxxx>, <Giovanni.Guasti@xxxxxxxxxx>,<hannappe@xxxxxxxxxxxxxxxxxx>, <buck@xxxxxxx>
- Date: Wed, 2 Apr 2003 10:39:09 -0800
Broadside coupling sounds attractive for a number of reasons but offers lots
of challenges. (read this as problems) You can get reasonable registration
for pairs printed on opposite sides of a core but, if you alternate signal
layers and grounds, you end up with some diff pairs with prepreq between.
-------- GND
prepreg
---- Sig+
core
---- Sig-
prepreg
-------- GND
core
---- Sig+
prepreg
---- Sig-
core
-------- GND
It is, in fact, not pratical to have strong diff pair coupling with
broadside coupled lines unless you want VERY low Zo. To maintain 100 Ohm
differential impedance, one needs either very narrow lines (high loss) or
thick dielectric layers (thick backplanes). The latter leads to problems
with connector transitions (via coupling and resonant stubs) and large pitch
between diff pairs to reduce crosstalk. In practice, broad-coupled lines
are more often more closely coupled to ground than each other but maintain
identical routing geometry (except, of course, at the connector pin fields).
This does reduce the effect of misregistration but you still have the issue
of thick boards.
In short, I feel the same way about broad-coupled lines as I do about
extraterestrial beings. I'm a skeptic but would welcome strong arguments in
favor of either.
Eric Sweetman
----- Original Message -----
From: "Jon Powell" <jonpowell@xxxxxxxxxxxx>
To: <phillip.r.wellington@xxxxxxxxxx>; <emj14@xxxxxxxxxxxx>;
<Giovanni.Guasti@xxxxxxxxxx>; <hannappe@xxxxxxxxxxxxxxxxxx>; <buck@xxxxxxx>
Cc: <si-list@xxxxxxxxxxxxx>
Sent: Wednesday, April 02, 2003 9:35 AM
Subject: [SI-LIST] Re: Routing of 12 GHz diff pairs
>
> More of a question than an answer:
>
> Wouldn't broadside coupling provide significantly more differential
coupling
> that may tend to mitigate the other factors and keep the signal skew
> smaller?
>
> It seems that differential pair coupling on the same layer is more on the
> level of cross-talk than actual differential coupling.
>
> jon
>
>
> -----Original Message-----
> From: si-list-bounce@xxxxxxxxxxxxx [mailto:si-list-bounce@xxxxxxxxxxxxx]On
> Behalf Of phillip.r.wellington@xxxxxxxxxx
> Sent: Wednesday, April 02, 2003 6:59 AM
> To: emj14@xxxxxxxxxxxx; Giovanni.Guasti@xxxxxxxxxx;
> hannappe@xxxxxxxxxxxxxxxxxx; buck@xxxxxxx
> Cc: si-list@xxxxxxxxxxxxx
> Subject: [SI-LIST] Re: Routing of 12 GHz diff pairs
>
>
>
> Hi,
>
> You really have to be careful with broad-side coupled pairs. If the pair
> elements are on layers separated by prepregs you may have some
> mis-registration because of layer float during manufacturing. The
> mis-registration caused an impedance change due to the traces not aligning
> on top of each other. There may be some mis-registration in the printing
and
> etching process as well even if a core is used. There is also a difference
> between the foot of the trace and the top of the trace (trapezoid etch
> factor related) which may affect your impedance prediction, work with your
> target board house.
>
> To help minimize the mis-registration, wider traces may be used to make
the
> mis-registration effects less of a percentage of impedance change. This of
> course makes the dielectric separations wider making routing more
difficult
> and adds layers because of the high line-to-line impedance that you need.
>
> Side-by-side routes eliminate the mis-registration errors. Side-by-side
> traces present routability issues in high density connectors as you need
to
> route 2 traces between pins and clearances causing length differences
> between the differential pair elements. With broad-side routed pairs, the
> lengths can be kept reasonably close through pins to the destination. With
> side-by-side routes it can be managed, it just requires more attention to
> detail.
>
> You have to be very careful what the return path is above and below both
of
> these configurations at this frequency. This includes what is on both
> reference planes. They should be at the same potential (RF - equal noise
> level) so that the pair elements do not have any differential currents
other
> than the intended differential signals. The planes should also be as quiet
> as possible- certainly not near switching supply return paths. Obviously,
> you would not want to route other traces adjacent to the pair
(differential
> noise coupled into one pair element at a higher level than the other pair
> element). You would not want to route across any plane split (return path
> problems and unplanned delays). It is desired not to route any traces
under
> the pair in the side-by-side configuration.
>
> There are other things to watch for too, not enough time though.
>
> Hope this helps.
>
> Ross
>
> -----Original Message-----
> From: Erik Jacobs [mailto:emj14@xxxxxxxxxxxx]
> Sent: Wednesday, April 02, 2003 7:22 AM
> To: Giovanni.Guasti@xxxxxxxxxx; hannappe@xxxxxxxxxxxxxxxxxx; buck@xxxxxxx
> Cc: Si-List
> Subject: [SI-LIST] Re: Routing of 12 GHz diff pairs
>
>
>
> > Hi Buck,
> > I think the best way to satisfy your customer is to draw the pair on two
> > faced layer.
> > But ... this method has many drawback:
> > 1)it's more difficult to control vertical than horizontal distance
between
> > pairs(due to PCB layers compression). So you could achieve a pair with a
> > wrong impedance.
> > 2)in order to obtain the right impedance, you could require a bigger
> > distance between layers
> > I would prefer to draw the pair on the same layer, making a compromise:
a
> > max difference of 2.5mm (l/10) could be enough?
>
> > > We are working with a customer that is using 12 GHz
> > diff-pair signals.
> > > The customer "requires" that the time of flight across the pair
> > > "must" be equal at all points in time. Can any one shed any light
> > > on how this can be done?
> >
> > The easiest way to achieve this is to use broadside coupled pairs,
> > because there both conductors are identical and routing is therefore
> > much esier.
>
> I'm about to do some 10GHz routing with similar requirements. It seems
that
> if you are using coupled pairs and you made the physical length almost
> identical, wouldn't the electrical length and electrical parameters be
> pretty close? I mean, assuming the signals got onto the PCB the same way
> and assuming that the material was relatively uniform and assuming you had
> good ground planes (or ground wires running under the diff pair traces)...
>
> Also don't you kinda need to assume that there's nothing around the traces
> that will couple INTO them? I.e. if p side of diff couple has coupling to
> some other line but N side has no such coupling? Even if it's only
> capacitive/inductive coupling (non-crosstalk)
>
> -Erik
> Test Engineering
> LeCroy
>
> ------------------------------------------------------------------
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> 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 archives are viewable at:
> //www.freelists.org/archives/si-list
> or at our remote archives:
> http://groups.yahoo.com/group/si-list/messages
> 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 archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
