Jack,

I found Doug Brooks article from Printed Circuit Design magazine March 1998 p30
"Bypass Capacitors - A conversation with Todd Hubing".  Excellent article BUT it
is written conversationally and you have to dig into it to  make sense of the
comments as there is a lot of back and forth - it isn't tutorial in nature so
you have to have some knowledge of physics or electronics.  Hubing is Assoc Prof
in EE at Univ Mo.  He has done a lot of measurements on boards so he speaks from
both a theoretical as well as a practical point of view.  I found it worthwhile
to re-read the article as I was not quite right - you need both a small spacing
(10 mils) AND bypass caps.

Comments from article:

Bypass caps provide most of the current at the frequencies where they are
effective (up to 1 GHz).

Boards with a power/ground plane spacing of 20 mils or more behave quite
differently.  The inductance of the planes cannot be neglected.  If the spacing
is 10 mils or less then inductance is not a problem (I presume this relates to
sag in voltage from caps to devices due to choke effect).

The critical spacing depends on the size of the board but the <= 10 mil or >=30
mil guideline appears to be appropriate for common board sizes.  Components on a
board with 30 mil or grater spacing may never see the board capacitance (at any
frequency) due to the mutual inductance of the vias between the planes.

The time varying magnetic flux between the planes causes charge to be pulled out
of the nearest coupling cap BEFORE it can be drawn from the planes if the
spacing between the planes is greater than ~30 mils.

The bandwidth of a board increases by the square root of the number of bypass
caps.

On a 10 mil spacing, every time we double the number of caps, we get 40% more
bandwidth.  However as a practical matter, we reach a point of diminishing
returns and added decoupling on boards 10 mils or less power/ground plane rarely
helps above 100 MHz.

He suggests that for >= 10 mil boards, the planes are more effective than
ADDITIONAL caps, as I read the article - you still need bypass caps at the
devices.   He goes on to say that once you have 40 - 50 caps on the board (10
mil) then adding more doesn't have much effect.

Doug then says "well, I've seen 1000 - 2000 caps on a board" and Todd says
"WOW!"
He doesn't believe this is a very good design strategy.  Large numbers of
decoupling caps increase the cost AND decrease the MTBF of a board.  It's more
effective to utilize the interplane capacitance of the board for high frequency
decoupling when you need it.

When planes are 30 mils or greater, closer (cap to device) is far better and in
no case is closer worse.

Basically, the idea is to keep the current loops (caps to devices) as small as
possible.

As for sprinkling caps around the plane - Howard Johnson wrote in his book to do
this but I don't see the value and Hubing is not clear if he does either.

As usual, life isn't all that simple and with Pentiums raising board clocks
higher and higher, the job of a PC board designer is getting more difficult.

I'd like to see a return to ECL but I'm sure I am in the minority.  Yes, they
are current hogs but I was doing 300 MHz phase locked loops in 1972 at COMSAT
Labs and it was a breeze even in prototypes with wire-wrap because the
impedance's were controlled and all signals were differential and terminated.

As Vcc goes down to 1V and lower, life is going to be hell.

Maybe one big BGA chip is the  future?

Regards,

Bob Landman
H&L Instruments




----- Original Message -----
From: Jack Olson <[log in to unmask]>
To: 'Bob Landman' <[log in to unmask]>; DesignerCouncil E-Mail Forum.
<[log in to unmask]>
Sent: July 22, 1999 9:24 PM
Subject: RE: Re: [DC] Stackup assignment for pcb boards.


>
> Thank you for taking the time to respond.
> Boy, it sure is difficult to keep up these days,
> one guy says one thing, another says the opposite.
> So, is there NO advantage to reducing the material between planes?
>
> onward thru the fog,
> Jack
>
> -----Original Message-----
> From: Bob Landman [mailto:[log in to unmask]]
> Subject: Re:      Re: [DC] Stackup assignment for pcb
> boards.
>
> Jack,
>
> Brooks wrote recently about planes acting as caps - they do
> not
>
> Much as we don't like them, every single power pin MUST be
> bypassed by 0.1uF
> caps - if a large chip like a BGA or PLCC has say 4 power
> pins - bypass them ALL
> or pay the piper.
>