Digital Signal Processing Reference
In-Depth Information
but to illustrate,
Kb
for a 65
, 5-mil-wide stripline is found from Figure 10.6 to be
0.03 when the victim is a distance equal to three traces away. If the rise time of the
aggressor pulse is small compared to the electrical length of the trace, the crosstalk
voltage is found with (10.2c). For instance, NEXT will be 60 mV if the aggressor
pulse is 2V.
Ω
10.4.3 NEXT Summary
To summarize NEXT behavior:
The NEXT pulse has the same polarity as the aggressor pulse.
•
NEXT is possible for either stripline or microstrip traces.
•
Increasing the aggressor signal swing increases the NEXT pulse amplitude.
•
If the coupled region is long (aggressor rise time is less than twice the coupled
length, where
dt
< 2
tpd
), the NEXT amplitude saturates at a maximum
value.
•
2
tpd
), the NEXT pulse
amplitude depends on the aggressor pulse rise time and will be smaller than
the amplitude of the long line case.
If the coupling occurs over a short distance (
dt
≥
•
10.5
How Closely Do Calculation and Simulation Agree?
Simulation results using the SmartSpice circuit simulator [13] lossy coupled trans-
mission line appear in Figure 10.7 for a pair of 65
Ω
4-mil (0.11-mm)-wide mi-
crostrips separated by 4 mils on FR4.
We see that the simulation predicts a NEXT amplitude of 191 mV, and as
shown in the Problems, (10.2) calculates 192 mV. As expected, the NEXT pulse
has the same polarity and is as wide as the aggressor pulse. In fact, it appears as a
miniature version of the aggressor.
Two FEXT pulses are created because coupling occurs on each edge of the
aggressor pulse. The FEXT pulses are opposite polarity because (as we find in the
Problems) in this case
Kf
is negative. The simulation shows the FEXT pulses to be
385 mV, versus the 388 mV calculated.
The close agreement between the simulation and the equations only occurs
when the transmission line has low loss and when both ends of all the transmis-
sion lines are terminated in resistors having a value equal to
Z
o
. Even with this
excellent correlation, the simulation and calculations are still only estimates of the
amount of crosstalk that will occur in actual hardware. This is because we usually
do not know the true shape or the actual dimensions of the fabricated traces. Nor
do we know the dielectric constant of the actual laminate or the actual solder mask
thickness. These things mean that the topology we specify in a field solver only
approximates the trace configuration that will actually be obtained in production.
