Far-End Crosstalk Reduction in Adjacent PCB

Traces employing High/Low-Z Configurations

Mohammad Almalkawi
#1
, Zulfiqar Khan
#2
, Vijay Devabhaktuni
#3
, Charles Bunting
*4

#
EECS Department, University of Toledo, Toledo, OH 43606, USA.

1
mohammad.almalkawi@utoledo.edu
2
zulfiqar.khan@utoledo.edu
3
vijay.devabhaktuni@utoledo.edu
*
Department of Electrical and Computer Engineering, Oklahoma State University,
Stillwater, OK 74078, USA.
4
reverb@okstate.edu

AbstractThis paper introduces a new High/Low-Z trace
configuration to reduce far-end crosstalk in adjacent printed
circuit board (PCB) interconnects. The emphasis is made to
reduce crosstalk without using additional PCB components in the
design. Specifically, we employ stepped impedance elements of
uniform length on the victim traces to suppress high-frequency
electromagnetic interference. The overall design is much simpler
than the usual low-pass filter configurations which are difficult
to implement in the prototype testing. The proposed approach
shows remarkably better results as compared to conventional
intervening guard trace schemes. Further, the proposed
configuration can be combined with the conventional schemes to
achieve even higher reduction in crosstalk.
I. INTRODUCTION
Crosstalk in PCB traces is a well known EMC and signal
integrity problem. A typical scenario involves coupling of a
high-frequency signal (e.g. a digital clock) onto a low
frequency trace (e.g. a power or a ground trace). Since low-
frequency traces are typically long traces routed on multiple
layers on a PCB, high-frequency interference may easily
radiate from multiple and/or little suspected locations on the
printed board. Therefore, it is usually very difficult to
determine the actual location of the radiating area on the board.
Normally, the first line of defence is to mitigate the radiation
source by reducing the current strength, slewing the rise/fall
time of the digital signal (e.g. by adding series resistors or
ferrites [1]) or shortening/straightening the high-frequency
trace. Though such countermeasures may provide a limited
improvement, they may not even be an option for a given
design (e.g. for signal integrity constraints).
Another way to reduce crosstalk is to exploit the inherent
parasitic effects (i.e. mutual capacitance and inductance)
among PCB traces [2-4]. To this end, different configurations
of guard traces have been employed in between adjacent
interconnects [2, 5, 6]. However, routing additional traces may
not be possible in a crowded layout. Furthermore, crosstalk
reduction provided by guard traces is limited and is usually
not sufficient [2, 7]. Therefore, it is highly desirable to explore
new trace configurations that may provide better crosstalk
reduction and can be routed in a crowded layout. One way is
to employ the low-pass filter (LPF) configuration on the
victim traces. A typical LPF employs multiple stepped
impedance elements on the victim trace with much better
performance compared to that achieved by using guard traces
alone. However, since stepped elements have non-uniform
lengths and widths, it is very difficult to implement different
LPF configurations on PCB traces in a prototype testing.
To this end, this paper introduces a new High/Low-Z
victim trace configuration to reduce far-end crosstalk coupling
between two adjacent PCB traces. The far-end crosstalk is of
particular interest due to the significantly stronger coupling as
compared to that at the near-end [8]. The proposed
configuration has LPF characteristics [9] as it employs
stepped impedance elements. However, here elements have
uniform lengths and, therefore, are easier to implement in the
prototype testing in EMC pre-compliance/compliance
chambers. Furthermore, the proposed configuration shows
similar performance to the LPF design and can be combined
with the guard traces for even better results.
The conventional and proposed crosstalk reduction
configurations are given in Section II. Section III compares
these PCB configurations in terms of crosstalk reduction in
adjacent PCB traces, while conclusions are given in Section
IV.
II. CROSSTALK REDUCTION SCHEMES
Fig. 1 illustrates a PCB structure (40 mm x 40 mm) with
two adjacent traces that are spaced 8 mm apart. Both traces
are 3 mm wide and 40 mm long. As illustrated, the PCB
structure is supported by a 1.575 mm thick FR4 substrate.
We assume that one trace (acting as an aggressor) is
carrying digital or high-frequency signals. These signals may
couple onto the adjacent trace (hereby referred as the victim
trace) carrying low-frequency or DC currents. To reduce this
coupling, referred to as crosstalk, various configurations can
be employed. Below, we discuss conventional configurations
as well as our proposed configuration.
A. Guard Trace for Far-End Crosstalk Reduction
As illustrated in Fig. 2, a guard trace is usually introduced
(as a shield) in between the aggressor and the victim trace. For
additional shielding, one may also ground the guard trace with
760 U.S. Government work not protected by U.S. copyright

Fig. 1. PCB structure with two uniform adjacent traces.



Fig. 2. PCB structure with a guard trace between two uniform parallel
traces.
via fences (see Fig. 3). In this paper, we will assume that the
guard trace is 1 mm wide and is terminated at both ends by the
matched impedance (i.e. 87 ). We further assume that the
spacing between each via on the guard trace is 2 mm.
B. Low-Pass Filter (LPF) Configuration
If victim trace is a low-frequency or a DC trace, one may
employ a stepped impedance LPF configuration, i.e. using
elements of high and low impedance alternatively on the trace
(see Fig. 4). Length
Hi
l of the high impedance (
H
Z ) element
can be calculated using the electrical length defined as [9],
H
Hi
Z
LR
l
0
= , (1)
where 2 = , corresponds to the cut-off frequency of
the LPF, and
0
R represents the feed trace characteristic
impedance. Similarly, length
Li
l of the low impedance
element (
L
Z ) can be calculated as,
0
R
CZ
l
L
Li
= , (2)
where L and C are the lumped element values [10]. The
width of elements is chosen corresponding to
impedances
H
Z and
L
Z .
It is worth mentioning that the difference between the two
impedances should be as high as possible for an excellent
filter performance (i.e. sharper cutoff). For higher crosstalk
reduction, LPF configuration can also be employed with guard

Fig. 3. Guard trace with ground vias.



Fig. 4. Stepped impedance configuration with 5 elements.



Fig. 5. Stepped impedance configuration along with a guard trace.



Fig. 6. Stepped impedance configuration with a guard trace grounded by vias.
761
traces (Fig. 5) and grounded guard traces (Fig. 6). However, in
our case, sharper cutoff is not an essential requirement. Fig. 7
illustrates an example of a 5 element LPF design. Here, we
chose
H
Z = 120 ,
L
Z = 20 , and
0
R = 50 . As well, we
selected a cutoff frequency of 4 GHz and 7 dB insertion loss
at 4.4 GHz. The calculated values using (1) and (2) are given
in Table I.

Fig. 7. A 5 element LPF design.
TABLE I
5 ELEMENT LPF DESIGN FOR A 4 GHZ CUT-OFF FREQUENCY.
Element
Value
Impedance
()
Electrical
Length
Length

(mm)
Width

(mm)
C
1
=0.618 Z
L
=20 14.163
0
l
L1
=1.492 w
L
=11
L
2
=1.618 Z
H
=120 38.626
0
l
H2
=4.657 w
H=
0.4
C
3
=2.000 Z
L
=20 45.836
0
l
L3
=4.828 w
L=
11
L
4
=1.618 Z
H
=120 38.626
0
l
H4
=4.657 w
H=
0.4
C
5
=0.618 Z
L
=20 14.163
0
l
L5
=1.492 w
L=
11
C. Proposed High/Low-Z Configuration
The proposed High/Low-Z configuration (Fig. 4) is based
on the LPF design. However, unlike LPF, the proposed
structure employs uniform lengths (
i
l ) for all elements. This
length may be selected to lie within the minimum and the
maximum values calculated for the LPF design. For example,
for the 5 element design given in Table I, the length for an
element in the proposed configuration can be selected between
1.492 mm and 4.828 mm. Choosing a particular length
uniformly for all elements will affect the cutoff frequency and
the return loss of the filter. However, since we are assuming
low-frequency or DC signals on the victim trace, cutoff
frequency and return loss are not the primary concerns. Fixing
the element lengths simplifies the LPF design procedure and
makes it easier to implement it in EMC prototype testing (e.g.
using copper tape). Furthermore, selecting a particular number
of elements is arbitrary and affects the performance of the
proposed configuration in terms of crosstalk. In our
investigation, however, we have selected 3, 5, 7 and 9
elements for comparison with the LPF and guard trace
configurations. The proposed configuration was also
investigated by employing the guard traces (with and without
via fences) between the traces.
III. RESULTS AND DISCUSSIONS
Both conventional and proposed PCB configurations were
simulated using Ansoft HFSS to calculate the far-end crosstalk
(S
41
) over a frequency range from 0 to 6 GHz. Below, we
present a comparison of the calculated results.
First, the crosstalk between adjacent PCB traces (Fig. 1)
using conventional guard trace (Fig. 2) and guard trace with
vias (Fig. 3) was calculated. The results, plotted in Fig. 8,
show that the guard trace alone does not provide any
significant improvement in the crosstalk reduction. Similarly,
using the guard trace with via fences shows a slight but
insignificant improvement in crosstalk reduction.
Frequency (GHz)
0 1 2 3 4 5 6
S
4
1

(
d
B
)
-80
-70
-60
-50
-40
-30
-20
-10
0
Uniform adjacent parallel traces (Fig. 1)
With guard traces (Fig. 2)
With guard traces grounded by vias (Fig. 3)

Fig. 8. Comparison of far-end crosstalk in adjacent PCB traces using guard
trace and guard trace grounded with vias.
Frequency (GHz)
0 1 2 3 4 5 6
-80
-70
-60
-50
-40
-30
-20
-10
0
Uniform adjacent parallel traces (Fig. 1)
3 elements proposed design
5 elements proposed design
7 elements proposed design
9 elements proposed design
S
4
1

(
d
B
)

Fig. 9. Far-end crosstalk in PCB traces using the proposed configuration.
Next the performance of the High/Low-Z configuration
(Fig. 4) employing 3 mm long elements was investigated. As
can be seen in Fig. 9, the proposed design provides significant
reduction in crosstalk as compared to guard trace
configuration (Fig. 8). Moreover, it can be observed that the
crosstalk can be further reduced by increasing the number of
elements in the proposed configuration.
We compared the performance of the proposed scheme with
the conventional LPF design. To this end, Fig. 10 provides a
comparison between a 5 element LPF design and a 5 element
proposed design. Results show relatively similar performance
for both designs in terms of far-end crosstalk. Moreover, the
return loss (S
33
) levels are acceptable for the proposed
configuration. A comparison of the near-end coupling, as
plotted in Fig. 11, shows that the proposed configuration
provides crosstalk improvement similar to the LPF.
A 9 element High/Low-Z configuration along with the
guard traces was also investigated. Fig. 12 shows that adding
762
Frequency (GHz)
0 1 2 3 4 5 6
S
4
1

&

S
3
3

(
d
B
)
-80
-70
-60
-50
-40
-30
-20
-10
0
S41 for the Proposed design
S41 for the LPF
S33 for the proposed design
S33 for the LPF

Fig. 10. Comparison of the proposed configuration with the LPF.
Frequency (GHz)
0 1 2 3 4 5 6
S
3
1

(
d
B
)
-80
-70
-60
-50
-40
-30
-20
-10
0
S31 of the 5 elements LPF (Fig. 4)
S31 of the proposed 5 elements (Fig. 4)

Fig. 11. Comparison of the proposed configuration with the LPF for the
near-end crosstalk.
Frequency (GHz)
0 1 2 3 4 5 6
S
4
1

(
d
B
)
-80
-70
-60
-50
-40
-30
-20
-10
0
9 elements Proposed design (Fig. 4)
With guard traces (Fig. 5)
With guard traces grounded by vias (Fig. 6)

Fig. 12. Proposed configuration combined with guard traces.
guard trace with vias helps, in general, reducing the crosstalk.
However, employing guard traces without ground vias, in fact,
increases the S41 coupling. Finally, 5-element High/Low-Z
configurations with different element lengths were compared.
The results, as plotted in Fig. 13, show that increasing the
Frequency (GHz)
0 1 2 3 4 5 6
S
4
1

(
d
B
)
-80
-70
-60
-50
-40
-30
-20
-10
0
li =1 mm
li =2 mm
li =3 mm
li =4 mm
li =5 mm

Fig. 13. Proposed configuration with different element lengths.

element length shifts the resonances towards the lower
frequency values.
IV. CONCLUSIONS
A new High/Low-Z configuration employing uniform
elements was investigated for reduced far-end crosstalk in
adjacent PCB traces. The performance of the proposed design
was found similar to the LPF design. However, unlike LPF,
the proposed approach is easier to implement in the EMC
prototype testing. The study shows that employing stand-alone
guard traces with ground vias along with the proposed
configuration helps reduce the far-end crosstalk.
REFERENCES
[1] K. Hu, H. Weng, D. Beetner, D. Pommerenke, J. Drewniak, K. Lavery,
and J. Whiles, Application of chip-level EMC in automotive product
design, IEEE EMC Int. Symp., Portland, OR., Aug. 2006, vol. 3, pp.
842-848.
[2] K. Lee, H.-B. Lee, H.-K. Jung, J.-Y. Sim, and H.-J. Park A serpentine
guard trace to reduce the far-end crosstalk voltage and the crosstalk
induced timing jitter of parallel microstrip lines, IEEE Trans.
Advanced Packaging, vol. 31, no. 4, pp. 809-817, Nov. 2008.
[3] B. Eged, F. Mernyei, I. Novak, and P. Bajor, Reduction of far-end
crosstalk on coupled microstrip PCB interconnects, Proc. IEEE
Instrumentation and Measurement Tech., Hamamatsu, Japan, May
1994, vol. 1, pp. 287-290.
[4] S.W. Park, F. Xiao, D.C. Park, and Y. Kami Crosstalk analysis for
two bent lines using circuit model, IEICE Trans. Communications,
vol. E90, no. 2, pp. 323-330, Feb. 2007.
[5] A. Suntives, A. Khajooeizadeh, and R. Abhari, Using via fences for
crosstalk reduction in PCB circuits, IEEE EMC-S Int. Symp., Portland,
OR., Aug. 2006, vol. 1, pp. 34-37.
[6] L. Zhi, W. Qiang, and S. Changsheng, Application of guard traces
with vias in the RF PCB layout, Proc. 3
rd
Int. Symp. Electromagnetic
Compatibility, Beijing, China, May 21-24, 2002, pp. 771-774.
[7] D. Brooks, Signal Integrity Issues and Printed Circuit Board Design,
Upper Saddle River, New Jersey: Prentice Hall, 2003.
[8] A. R. Djordjevic, T. K. Sarkar, and R. F. Harrington, Time-Domain
Response of Multiconductor Transmission Lines, proc. IEEE, vol. 75,
pp. 743-764, June 1987.
[9] D. Pozar, Microwave Engineering, 3
rd
ed. New York: Wiley, 2005, pp.
412-416.
[10] G. Matthaei, L. Young, and E.M. Jones, Microwave Filters Impedance-
Matching Networks, and Coupling Structures, Artech House, 1980s.
763