Re: [SI-LIST] : USB differential lines.

Norm Ebsary (n.ebsary@computer.org)
Wed, 30 Sep 1998 13:52:02 -0400

Fred,

My intent was to respond to Laurent's question of

"From what I have read in the SI-LIST archives
>> >Zdifferential = 2 * Zsingle ended in the case where there is no coupling
>> >between the traces : the lines have to be separeted by a large
>> >distance, which is stupid for a differential pair where both wire should
>> >be near to each other.
>> >
>> >Any comment on this strange requirement is welcome.
>> >"

and therefore define the differential characteristic impedance as Zo =
120/Er * ln(h/w + ((h/w)**2 -1)**.5) in relation to the distance between
the two differential traces. While you have defined a differential system
which I have no disagreement of, this certainly was not my intent. My intent
was to define for a differential (and not single-ended) pair the
characteristic impedance. The question Laurent asked was "I am designing a
PCB which should use differential lines for USB.", in regards to
"Differential impedance should be 90 Ohms to match the cable impedance if I
have well understund.".

As you have emphasized it is not the practice that differential lines are
routed closely together, but that the characteristic impedance be met for
this differential pair. I agree with you that the differential impedance is
the requirement that must be adhered to, which could be met by routing two
differential lines in a pair closely together as .005"/.005" designs rules
allow, for a Zo of 45 Ohms. The common industry misnomer is to route
differential lines as closely together without regard to their resultant
characteristic impedance. Again I tempted to disagree with you on your
statement "It is not incorrect to route differential signals close to each
other.>It is preferable. " ,but since you have added "The caveat being that
the required ZO be met. " I would have have to agree.

Regards,
Norm Ebsary
Digital Telecommunications Inc.

-----Original Message-----
From: Fred Balistreri <fred@apsimtech.com>
To: Norm Ebsary <n.ebsary@computer.org>
Cc: lsbernard@chez.com <lsbernard@chez.com>; 'Signal Integrity LIST'
<si-list@silab.Eng.Sun.COM>
Date: Wednesday, September 30, 1998 12:49 PM
Subject: Re: [SI-LIST] : USB differential lines.

>Norm Ebsary wrote:
>>
>> Laurent,
>>
>> The differential characteristic impedance is dependent on the center to
>> center distance of the two conductor pair, with single ended
characteristic
>> impedance on the distance from the single trace to ground. While it is a
>> common misnomer in the industry to route two conductors as close as
possible
>> in a differential pair, this is generally incorrect where the
differential
>> characteristic impedance , Zo = 120/Er * ln(h/w + ((h/w)**2 -1)**.5) -
Note
>> that h is the center to center separation of the two traces with w equal
to
>> the trace width. For example the distance between conductors of a
>> differential pair for a w = .005" wide trace with a separation of .005"
>> (h=.010" center to center) and an Er = 3.5 is 45.12 Ohms. So in this
case,
>> keeping two traces as close together as possible with .005"/.005" design
>> rules would yield a 45 Ohm differential characteristic impedance. For a
>> differential characteristic impedance of 90 Ohms, the separation of .005"
>> traces for a differential pair should be approximately .035" center to
>> center or for .005" trace width a separation of .030".
>>
>> Hope this helps,
>> Regards,
>> Norm Ebsary
>> Digital Telecommunications Inc.
>>
>> -----Original Message-----
>> From: Laurent BERNARD <l.s.bernard@usa.net>
>> To: 'Signal Integrity LIST' <si-list@silab.Eng.Sun.COM>
>> Cc: 'Laurent BERNARD' <lsbernard@chez.com>
>> Date: Wednesday, September 30, 1998 8:49 AM
>> Subject: [SI-LIST] : USB differential lines.
>>
>> >Hello SI experts,
>> >
>> >I am designing a PCB which should use differential lines for USB.
>> >
>> >I am a little bit confused by Intel data on the USB. The Single Ended
>> >impedance should be 45 Ohms, the Differential impedance should be
>> >90 Ohms to match the cable impedance if I have well understund.
>> >
>> >>From what I have read in the SI-LIST archives
>> >Zdifferential = 2 * Zsingle ended in the case where there is no coupling
>> >between the traces : the lines have to be separeted by a large
>> >distance, which is stupid for a differential pair where both wire should
>> >be near to each other.
>> >
>> >Any comment on this strange requirement is welcome.
>> >
>> >Thank you very much for your help,
>> >
>> >Best Regards,
>> >
>> >Laurent BERNARD. MicroProcess Ingenierie
>> >Ingenieur d'etudes / Design Engineer.
>> >
>> >Tel : 33 (0)1 47 68 80 80 B.P. 87
>> >Fax : 33 (0)1 47 88 97 85 97 bis, rue de Colombes
>> >Email : mailto:lsbernard@chez.com 92405 COURBEVOIE Cedex
>> >Web : http://www.microprocess.com FRANCE
>> >
>> >
>> >
>> >
>> >**** 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 ****
>> >
>>
>> **** 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
>
>
>It is not incorrect to route differential signals close to each other.
>It is preferable. The caveat being that the required ZO be met. It is
>incorrect to call traces that are not coupled to each other differential
>pairs. One can certainly measure an impedance across such a pair and
>call that an impedance measured differentially. However the pair is NOT
>a electrical differential system. The degree to which a pair can be
>called differential will depend on the coupling. This effects where
>the return current is located and the magnetic flux cancellation. In an
>uncoupled system the return current is not located on the other trace
>but rather somewhere else (hopefully a nearby plane). And there
>certainly is no magnetic flux cancellation between the two. Your
>definition of differential systems seems to be predicated on Zo only.
>
>Best Regards,
>
>--
>Fred Balistreri
>fred@apsimtech.com
>
>http://www.apsimtech.com
>

**** 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 ****