Re: [SI-LIST] : RE: [SI-LIST]: Long bus or star?

D. C. Sessions (dc.sessions@vlsi.com)
Tue, 30 Mar 1999 09:09:11 -0700

Jon Keeble wrote:
>
> >
> >What I realize here is that you will have a reflection that hits one of the
> >devices at the end of the 3" trace. It will reflect all the way back to
> the
> >other 4 devices, which are each 6" away. Because this length is considered
> >"long" it will cause problems, no?
>
> No. If the star is being driven at the hub (this an address bus, right? not
> a bidirectional bus?)
> then the hub is a low impedance, and any returning waves will be mostly
> attenuated, so will pass through to other spokes.
>
> Seems to me that a classic series source terminator of value Zo at the hub,
> one for each spoke, is optimum. It halves the required drive, and turns the
> voltage doubling at the open circuit at the end of each spoke to advantage.
> The reflections witl be fully absorbed back at the hub in the resistors.

It *does* work well, but the big nasty is getting the terminators close enough
to the source. Keep in mind that high ball-count packages are *expensive*
and putting series resistors on them is even more so. Chances are that the
closest you'll get the terminators is 2-3 cm from the driver, which is a
pretty serious stub.

I've used this method before for address drivers and it *does* work, but
the PWB designers sweated long and hard and the result was not popular
with the Operations people who had to build and test it since it pushed
the testability rules and had more process steps than they liked.

-- 
D. C. Sessions
dc.sessions@vlsi.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 ****