At 03:25 PM 10/22/99 -0700, Douglas McKean wrote:
>It's my opinion that the stackup S1/S2/G/P/S3/S4
>is only an extension of a typical 4 layer stackup
>S1/G/P/S2.  I don't think it would buy you anything.
>If you're stuck with a S1/S2/G/P/S3/S4 stackup, then
>you'd have to exercise great care with layout.

Care with layout is advised in either case!

Control of trace impedances may be important; this can be a little easier,
in fact, with xxGPxx than with xGxxPx. The former, for 50 ohm traces, will
imply trace widths on layers 2 and 5 which are a standard production value,
such as 8 mils, with a reasonable prepreg thickness, whereas the latter
gets at little trickier because of the stripline construction, which
implies smaller trace widths; further the latter has more separation
between the power and ground planes and thus less, perhaps much less,
distributed capacitance.

There is no doubt in my mind but that xxGPxx buys a great deal over xGPx.
One gains two inimpeded routing layers, which is *very* important, even
essential, for surface mount design. Those layers will have, perhaps 8 mil
design rules for controlled impedance traces, and the surface traces can be
controlled as well with perhaps 15 mil traces, I forget the value at the
moment, and, of course, it does vary with the prepreg thicknesses
specified. Typically I would set layer bias opposite on the top layer to
the layer immediately underneath it, so cross-talk between layers 1 and 2
will be minimized.

If trace impedance is controlled and is a relatively low value, like 50
ohms, it becomes difficult for noise sources to couple into the traces
(provided one has avoided other design faults).

Splitting the planes is a *excellent* practice for the kind of design under
discussion. But I have seen a designer "split" planes by assigning one
plane to, say digital ground, and another to analog ground, while allowing
the planes to overlap; thus there was strong capacitative coupling between
the planes. To steal and mangle slightly a phrase from a song, I don't know
what he was thinking, he must have been drinking....

[log in to unmask]
Abdulrahman Lomax
P.O. Box 690
El Verano, CA 95433