From: "Saved by Internet Explorer 11" Subject: PCB Stack-Up - Part 2 Date: Thu, 2 Jul 2015 15:06:07 -0700 MIME-Version: 1.0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Content-Location: http://www.hottconsultants.com/techtips/pcb-stack-up-2.html X-MimeOLE: Produced By Microsoft MimeOLE V6.1.7601.17609
=20 =20 =20 = &nbs=
p;
_____________ Sig.
&=
nbsp;
_____________
=
Ground &=
nbsp; &n=
bsp;=20
Figure 1
&=
nbsp;
_____________ Power
&=
nbsp;
_____________ Sig.
Although this configuration is significantly better than a two-layer = board,=20 it has a few, less that ideal characteristics. With respect to the = list of=20 objectives in Part 1, this stack-up only satisfies objective (1). = If the=20 layers are equally spaced, there is a large separation between the = signal layer=20 and the current return plane. There is also a large separation = between the=20 power and ground planes. With a four-layer board we cannot correct = both of=20 these deficiencies at the same time; therefore, we must decide which is = most=20 important to us. As mentioned previously, with normal PCB = construction=20 techniques there is not sufficient inter-plane capacitance between the = adjacent=20 power and ground planes to provide adequate decoupling. The = decoupling,=20 therefore, will have to be taken care of by other means and we should = opt for=20 tight coupling between the signal and the current return plane. = The=20 advantages of tight coupling between the signal (trace) layers and the = current=20 return planes will more than outweigh the disadvantage caused by the = slight loss=20 in interplane capacitance.
Therefore, the simplest way to improve the EMC performance of a = four-layer board is to space the signal layers as close to the planes as = possible (<0.010"), and use a large core (>0.040") between the = power and=20 ground planes as shown in Fig. 2. This has three advantages and = few=20 disadvantages. The signal loop areas are smaller and therefore = produce=20 less differential mode radiation. For the case of 0.005" spacing = (trace=20 layer to plane layer), this can amount to 10 dB or more reduction = in the=20 trace loop radiation compared a stack-up with equally spaced = layers. =20 Secondly, the tight coupling between the signal trace and the ground = plane=20 reduces the plane impedance (inductance) hence reducing the common-mode=20 radiation from the cables connected to the board. Thirdly, the = close trace=20 to plane coupling will decrease the crosstalk between traces. For = a fixed=20 trace to trace spacing the crosstalk is proportional to the square of = the trace=20 height. This is one of the simplest, least costly, and most = overlooked=20 method of reducing radiation on a four-layer PCB. With this = configuration=20 we have satisfied both objectives (1) and (2).
&nbs=
p;
_____________ Sig.
&=
nbsp;
_____________ Ground
&nbs=
p;  =
; =
&=
nbsp; &n=
bsp; &nb=
sp;
Figure 2
&=
nbsp;
_____________ Power
&=
nbsp;
_____________ Sig.
What other possibilities are there for a four-layer board =
stack-up?
Well, we could become a little non-conventional and reverse the =
signal=20
layers and the plane layers in Fig. 2, producing the stack-up shown in =
Fig 3a.=20
&nbs=
p;
_____________ Ground.
&=
nbsp;
_____________ Sig.
&nbs=
p;  =
; =
&=
nbsp; &n=
bsp; &nb=
sp;
Figure 3a
&=
nbsp;
_____________ Sig.
&=
nbsp;
_____________ Power
The major advantage of this stack-up is that the planes on the outer = layers=20 provide shielding to the signal traces on the inner layers. The=20 disadvantages are that the ground plane may be cut-up considerably with=20 component mounting pads on a high density PCB. This can be = alleviated somewhat, by reversing the planes and placing the power plane on the = component side, and the ground plane on the other side of the board. = Secondly, some=20 people don't like to have an exposed power plane and thirdly, the buried = signal=20 layers make board rework difficult if not impossible. This = stack-up=20 satisfies objectives (1), (2), and partially satisfies objective (4). =
Two of these three problems can be alleviated with the stack-up shown =
in Fig.=20
3b, where the two outer planes are ground planes and power is routed as =
a trace=20
on the signal planes. The power should be routed as a grid, using =
wide=20
traces, on the signal layers. Two added advantages of this =
configuration=20
are that; (1) the two ground planes produce a much lower ground =
impedance and=20
hence less common-mode cable radiation, and (2) the two ground planes =
can be=20
stitched together around the periphery of the board to enclose all the =
signal=20
traces in a faraday cage. From an EMC point of view this =
configuration, if=20
properly done, is the best stack-up possible with a four-layer =
PCB. Now we=20
have satisfied objectives, (1), (2), (4), and (5) while using only a =
four-layer=20
board.
&nbs=
p;
_____________ Ground.
&=
nbsp;
_____________ Sig./Pwr.
&nbs=
p;  =
; =
&=
nbsp; &n=
bsp; &nb=
sp;
Figure 3b
&=
nbsp;
_____________ Sig./Pwr.
&=
nbsp;
_____________ Ground
A fourth possibility, not commonly used, but one that can be made to =
perform=20
very well, is shown in Fig. 4. This is similar to Fig 2, but =
with=20
the power plane replaced with a ground plane, and power routed as a =
trace on the=20
signal layers.
&nbs=
p;
_____________ Sig./Pwr.
&=
nbsp;
_____________ Ground
&nbs=
p;  =
; =
&=
nbsp; &n=
bsp; &nb=
sp;
Figure 4
&=
nbsp;
_____________ Ground
&=
nbsp;
_____________ Sig./Pwr.
This stack-up overcomes the rework problem mentioned before, and = still provides for the low ground impedance as a result of two ground = planes. The planes however do not provide any shielding. This = configuration satisfies objectives (1), (2), and (5) but not objectives (3) or (4). =
So, as you can see there are more options available, than you might =
have=20
originally thought, for four layer board stack-up. It is possible =
to=20
satisfy four of our five objectives with a four layer PCB. The
configurations of Figures 2, 3b, and 4 all can be made to perform well =
from an=20
EMC point of view.
=A9 2000 Henry W. = Ott &nbs= p;  = ; = Henry Ott Consultants, 48 Baker Road Livingston, = NJ 07039 (973) 992-1793