RE: [SI-LIST] : No subject given

West, Todd ([email protected])
Thu, 5 Nov 1998 18:45:18 -0800


I won't claim to be a layout wizard (mostly, I know enough to get myself in
trouble), but perhaps I can shed a bit light on the matter.

Probably the most important thing to realize is that USB is actually *not*
differential signaling---both of the data lines have single-ended drives to
ground, rather than the single drive loop of true differential systems such
as LVDS. As a result, USB is perhaps better called faux differential. The
voltage swings on the lines still look like NRZI differential signaling, but
it's possible to have skew between the data lines from differences in the
driver turn-on times that simply can't exist in a truly differential signal.

So, as far as USB is concered, both the single-ended and differential
impedance matters. However, USB's edge rates are relatively slow (7--26ns
in practice) and well controlled, so the bus is really quite robust in the
face of impedance glitches (which is a good thing, since many USB developers
are new to transmission lines).

As to stripline trace width and spacing, I use the single-ended formulas
found in the back of _High-Speed Digital Design_ and the differential
formulas for side-by-side traces above a ground plane published by National
Semiconductor in their LVDS design guide (see

What I have yet to determine to my satisfaction is the range of trace widths
and geometries over which these formulas are valid---how wide, how far above
ground, how heavily buried by soldermask, etc. were the test traces used to
come up with them? Anyone know for sure? I assume they'll work pretty well
for a typical SPGS four layer board stackup with the controlled impedance
signals routed next to the ground plane on a 6--8 mil dielectric, but some
USB devices are simple enough they can actually use two layer boards. A
straightforward application of the formulas there comes up with 130mil wide
traces 700mils apart for the usual 62mil FR4 thickness!

Todd West
[email protected]

-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Thursday, November 05, 1998 5:31 PM
To: [email protected]; [email protected]
Subject: [SI-LIST] : No subject given

Hi all SI experts,

Does somebody know how to determine the trace width and trace spacing
for differential signals such as USB or LVDS?

For example, the USB signal, Intel recommands 45 ohms for single
trace impedance and 90 ohms for differential impedance. I find 12mils
for trace width and 32mils for space for my pcb stackup will meet the
spec. Also I find 6 mils width and 8 mils spacing will also meet 90
ohms differetial impedance but 58 ohms for single trace.

I believe that all I need to take care is 90 ohms differential
impedance instead of 45 ohms for single trace. Am I right?

What will be different between them (12x32 and 6x8)?

I think that the spacing of 32 mils is too big because the
differential pair shall be routed closely to cancel common mode noise.

Are there any disadvantages for high single trace impedance but same
diffential impedance?

Can we have long stub topology, such as "Y", for the differential
signal routing? Shall the reflection on the pair be cancelled because
of common mode cancellation?

Thank you for your reply in advance.

John Lin
CAE Manager at Arima computer corp.
Email: [email protected]

**** To unsubscribe from si-list: send e-mail to
[email protected] In the BODY of message put: UNSUBSCRIBE
si-list, for more help, put HELP. si-list archives are accessible at ****

**** To unsubscribe from si-list: send e-mail to [email protected] In the BODY of message put: UNSUBSCRIBE si-list, for more help, put HELP. si-list archives are accessible at ****