Re: decoupling/ bypass capacitors at connectors

Larry Smith (ldsmith@lisboa)
Wed, 5 Mar 1997 10:47:56 -0800

Ok, I agree to that. As much as anything, I wanted to stir up some
good discussion on si-list. It has been kind of quiet out there for
a while. Thanks for your contribution.

Larry

> From HANS_MELLBERG@non-hp-santaclara-om4.om.hp.com Wed Mar 5 09:28:29 1997
> From: HANS_MELLBERG@non-hp-santaclara-om4.om.hp.com
> X-Openmail-Hops: 1
> Date: Wed, 5 Mar 97 09:23:45 -0800
> Subject: Re: decoupling/ bypass capacitors at connectors
> Cc: si-list@silab.eng.sun.com
>
> Don't get me wrong! I am an advocate of stitching for general pcb design
> guidelines. My point was for a single transmission line. Unless you are
> dealing with ECL logic then the structure on the rise/fall edge is not
> significant. If you are dealing with ECL then "emiter hang-up" can occur
> which will have your logic designers walk around in circles with
> unexplained propagation delay being over spec! In the 0.5 ns realm, I still
> maintain that stitching will have no effect unless at the source,
> end-of-line or during plane changes.
> Best Regards,
> Hans
>
>
> ______________________________ Reply Separator _________________________________
> Subject: Re: decoupling/ bypass capacitors at connectors
> Author: Non-HP-Larry.smith (Larry.smith@eng.sun.com) at hp-boise,shargw2
> Date: 3/4/97 4:10 PM
>
>
> > For 0.5ns requirements, the stitching has no effect except immediately
> > adjacent to the signal trace or connectors. The displacement current will
> > find the most proximal structure to propagate the image current.
> >
> > Hans Mellberg
> > Consultant
> >
>
> The stitching of ground planes every square inch of the board will have
> significant effect. Imagine a signal on the top layer of the card
> that is referenced to a ground plane immediately below. The signal
> goes down a via, through the board and takes off on a trace referenced
> to another ground plane on the bottom of the board.
>
> The question is, what happens to the return current. If the rise time
> is .5 nSec, the 'length' of the rise time will be 3 inches, assuming 6
> inches/nSec. (If these are exterior traces, the velocity may be closer
> to 9 inches/nSec, making the distance traveled in the rise time 4.5
> inches.)
>
> Suppose there is a ground plane stitch via within 1 inch of the signal
> via (1/3 of a rise time distance). True, there will be an impedance
> discontinuity as signal current must depart from the return current as
> the currents go through the vias. But, the time of flight (1/6 nSec)
> will allow for 3 reflections across the impedance discontinuity during
> the rise time. If we can get 3 reflections during the rise time, the
> impedance discontinuity has minimal effect on the waveform.
>
> If one of the reference planes is power, then decoupling will be
> involved in the current path. The fidelity of the edge will be
> degraded if a 'short' path is not provided for the return current.
> Ground and power plane bounce will occur at via locations if this
> path is not provided.
>
> Larry Smith
> Sun Microsystems
>