Digital Signal Processing Reference
In-Depth Information
10.5.1 How Well Do 2D Field Solvers Agree?
The results shown in Figure 10.7 show that simulation and calculation give the
same results when using identical data. The correlation does not prove the field
solver has produced accurate and consistent results even if the trace topology is
known precisely. In fact, the calculation of the mutual capacitance and inductance
requires the field solver to process very small numbers, especially when the spacing
between traces is large. As shown in Table 10.1, numerical imprecision can result in
discrepancies between field solvers. To illustrate this, Table 10.1 shows the worst-
case range of
Kb
for a 5-mil (0.13-mm)-wide, 50
Ω
stripline across various spacing
as reported by multiple 2D field solvers.
Table 10.1 shows that even if the trace topology is known perfectly, the simula-
tion results obtained from different field solvers will not match exactly, especially
when the spacing is large. For instance, the field solvers agree to within 0.5% or
better for spacing of up to two times (10 mils), and the agreement is within 12% for
spacing up to six times (30 mils). It becomes significantly worse for larger distances.
10.6 Guard Traces
Crosstalk can be reduced by placing a grounded trace between a victim and aggres-
sor to create a guard trace [7, 14-16]. When done properly, the aggressor has higher
coupling to the guard trace than to the victim, reducing the voltage induced on the
victim. Guard traces are especially effective with stripline, but they also work well
with microstrip.
One example showing the efficacy of a stripline guard trace is shown in Figure
10.8.
The top view of 4-mil (0.11-mm)-wide 65
stripline traces placed side by side
with a separation of 4 and 12 mils (0.11 mm and 0.32 mm) is shown. The pulse
Ω
1.5V
Aggressor
1.0V
0.5V
0.0V
NEXT
FEXT
-0.5V
0s
1 ns
2 ns
3 ns
Figure 10.7
Simulated microstrip crosstalk results. NEXT is +191 mV (versus +192 calculated). FEXT
is
385mV (versus
388 calculated).
±
±
