11

• Jason R. Miller, Gustavo Blando, • Istvan Novak• Sun Microsystems

CROSSTALK DUE TO PERIODIC PLANE CUTOUTS

Outline

1 – Introduction2 – Crosstalk Theory3 – Measurement4 – Simulation correlation5 – Parameterized field simulations6 – Conclusions

2

3

Introduction• Multilayer pcbs have dense trace routes, forcing

routing through regions containing packages, connectors, and/or multi-pin sockets

• In these regions the trace may periodically encounter regular patterns of via barrels, antipad cutouts, BGA solder balls, etc.

Conn. pin-field

BGA pin-field

4

Introduction• Single trace: Periodic discontinuities can line up

these small reflections, increasing the slope of the loss curve (S12) by the presence of resonance dips [1].

• Coupled traces: Traces can get routed near to or over cutouts and thermal reliefs, due to misregistration and manufacturing tolerances, allowing for layer to layer coupling.

1. Gustavo Blando, et al., Attenuation in PCB Traces due to Periodic Discontinuities, Proceedings of DesignCon 2006, Santa Clara CA

5

Introduction• Trace routed through a pin field,

such as a BGA with antipads• Identical trace routed below the

immediate plane layer

+/-3 mils

+/-5 mils

Range

-1.315 mils (-33%)Same core misregistration

-3.315 mils (-82%)Different core misregistration

Antipad edge to trace edge separation

Example: 4 mil line width, BGA (1 mm, 39.37 mil), 30 mil antipad

w

6

Crosstalk Theory• Crosstalk is caused by the coupling of energy from one trace to

another

R

Active line Z0 R

R

Victim line Z0

R

Port 1 Port 3

Port 2 Port 4

AC

• Coupled energy on port 2 is called the near-end crosstalk (NEXT). NEXT is proportional to the sum of the capacitive and inductive couplings. It appears on both microstrips and stripline traces

• Coupled energy on port 4 is called the far-end crosstalk (FEXT) and is proportional to the difference in the capacitive and inductive coupling. FEXT is zero for homogenous structures such as a stripline

7

Crosstalk Theory• Pair of 10-inch, edge-coupled microstrip traces

spaced 3 mils apart over solid ground, where the traces are on the same side of the plane

• At low frequencies, both S12 and S14 have rising slopes

• The first maximum of the S12 profile occurs at λ/4 of the coupled line length. The subsequent peaks occur at the odd harmonics of λ/4

• S14 profile reaches full coupling at frequencies where the time-of-flight difference of the even and odd modes equals half of the period

• Subsequent S14 peaks occur at odd harmonics of the fundamental

• S13 profile has corresponding minima at the frequencies where the far-end coupling peaks

• S14 is zero for stripline

108 109 1010-50

-40

-30

-20

-10

0

Frequency [Hz]

[dB]

S14 S12

S13

8

Measurements• A pair of mirrored, serpentine traces

on either side of a perforated plane at three different distances from the cutout

• Unit cell, repeated six times (2.9 inch)

• Board milled down to contact horseshoe ground pads and signal

• Differential 500 um GSSG (Ground-Signal-Signal-Ground) picoprobes with SOLT calibration

• Four-port S-parameters were captured with an Agilent N4230A VNA from 300kHz to 20GHz

1 2 3

474mil

9

Measurements

108 109 1010-100

-80

-60

-40

-20

0

Frequency [Hz]

S12

[dB

]

3

21

108 109 1010-80

-60

-40

-20

0

Frequency [Hz]

S14

[dB

]1

23

10

Measurements

0.5 1 1.5 2x 1010

-50

-45

-40

-35

-30

-25

-20

-15

-10

-5

X: 4.878e+008Y: -23.72

Frequency [Hz]

S12

[dB

]

X: 5.775e+009Y: -12.57

λ/4

oddharmonics

λ/2harmonics

• Measured S12 for trace 3• First maximum occurs at

λ/4 of the total line length (2.9 inch)

• Subsequent peaks occur at the odd harmonics of λ/4

• Periodic cutouts causes a series of peaks with a fundamental corresponding to λ/2 of the unit cell length, here 474 mils

• Subsequent peaks are even harmonics of λ/2.

11

Measurements

• Relative magnitude of the λ/2 series (first peak) to the λ/4 series, for S12 is:

Trace 3: 8.4 dB (40%) Trace 2: 10 dB (21%)

0 0.5 1 1.5 2x 1010

-80

-60

-40

-20

0

Frequency [Hz]

S12

[dB]

3

2

8.4 dB+40%

10 dB+21%

12

Measurement Correlation• Trace 2 & 3 (2.9 inch) simulated

using Ansoft HFSS• Simplifications:

> Thermal relief pattern was not included

> Vias were placed outside of the pin field

> Only the antipads nearest to the trace were included

• Geometrical simplifications limited our range of simulation accuracy to approximately 10 GHz.

• A wideband causal model was used to capture the frequency dependence of the dielectric material

13

Correlation

108 109 1010-80

-70

-60

-50

-40

-30

Frequency [Hz]

S12

[dB]

HFSS

measured

108 109 1010-70

-60

-50

-40

-30

Frequency [Hz]

S14

[dB

]HFSS

measured

Trace 2 S12 and S14

14

CorrelationTrace 3 S12 and S14

108 109 1010-50

-40

-30

-20

-10

0

Frequency [Hz]

S12

[dB

]

HFSS

measurement

108 109 1010-50

-40

-30

-20

-10

0

Frequency [Hz]

S14

[dB

]

HFSS

measured

15

Parameterization

• 4-mil wide traces and 4 + 4 mil thick dielectrics

• Single cutout with a thermal relief pattern located next to the trace, centered

• Unit cell length is 500 mils. 20 unit cells concatenated in MATLAB for a total length of 10 inches

• Trace to cutout perimeter distance (s) was swept from 6 mils to -8 mils

• Positive s value is over the cutout• s = 0 means trace routes along the edge

Parameterizing the Spacing between Trace and Cutout

500 mils

16

Parameterization

• First maximum of the S12 profile occurs at λ/4 of the total line length (10 inch)

• Subsequent peaks occur at the odd harmonics of λ/4

• λ/2 series of peaks due to periodic coupling start at 5.45 GHz, corresponding to λ/2 of the unit cell length (500 mils)

• Subsequent peaks are at even harmonics of the fundamental

Parameterizing the Spacing between Trace and Cutout

108 109 1010 1011-140

-120

-100

-80

-60

-40

-20

0

Frequency [Hz]

S12

[dB]

S = 6S = 4S = 2

S = 0

S = -2

S = -8

17

Parameterization

• S14 profile also shows a series of peaks and dips starting at the λ/2 frequency (5.45 GHz)

Parameterizing the Spacing between Trace and Cutout

108 109 1010 1011-140

-120

-100

-80

-60

-40

-20

Frequency [Hz]

S14

[dB

]

s = 6s = 4s = 2s = 0s = -2

s = -8

18

Parameterization

• S12 magnitude of the first λ/2 peak at 5.45 GHz as a function of the spacing, s

• S12 coupling increases at approximately 5-6 dB/mil as the trace moves closer to the cutout edge

• With the trace 4 mils over the cutout, the coupling starts to saturate

Parameterizing the Spacing between Trace and Cutout

-8 -6 -4 -2 0 2 4 6-100

-90

-80

-70

-60

-50

-40

-30

-20

6.6144 dB/mil

5.6105 dB/mil

5.2188 dB/mil

3.2053 dB/mil

2.1369 dB/mil

s [mil]

S12

[dB

]

19

ParameterizationParameterizing the Spacing between Trace and Cutout

0 0.5 1 1.5 2x 109

-140

-120

-100

-80

-60

-40

Frequency [Hz]

S12

[dB

]

-5 0 5-120

-100

-80

-60

-40

6.1326 dB/mil

5.7629 dB/mil

4.5929 dB/mil3.5544 dB/mil

2.2548 dB/mil

s [mil]

S12

[dB

]

20

ParameterizationParameterizing the Spacing between Trace and Cutout

-8 -6 -4 -2 0 2 4 620

22

24

26

28

30

s [mil]

S12

incr

ease

[dB

]

• Magnitude difference between the λ/4 peak envelope and first λ/2 peak

• > 24 dB increase in the S12 coupling across range of separations

21

ParameterizationParameterizing the Number of Cutouts

• Spacing fixed at zero• 500-mil unit cell

concatenated using 5, 10, 20 and 30 unit cells

• With 5 unit cells, the λ/4 fundamental is pushed to a center frequency four times higher

• λ/2 fundamental remains the same because the unit-cell length dictates the location of the first λ/2 peak

108 109 1010-100

-90

-80

-70

-60

-50

-40

Frequency [Hz]

S12

[dB]

5 unit cells20 unit cells

22

ParameterizationParameterizing the Number of Cutouts

• Magnitude of the first λ/2 peak as a function of number of cells

• Saturation occurs because eventually we get a uniform structure and the per-unit-cell characteristics become stable

• Same characteristic observed on loss of periodic discontinuities

0 10 20 30 40-48

-46

-44

-42

Number of unit cells [-]

S12

[dB

]

23

Conclusions• Measured and simulated data of traces on opposite sides of a

reference plane, passing close to or over periodic plane cutoutsexhibit non-zero high-frequency crosstalk.

• The near-end crosstalk due to periodic apertures show two distinct signatures: > Quarter-wave peaks determined by the full coupled length. This was below 1%,

even over the voids.> Half-wave peaks determined by the length of between two cutouts. These are

1% with edge aligned and 10% when the trace passes over the void.

• The half-wave series of peaks depend on the number of periodic voids, and it shows saturation with large number of cutouts.

• Far-end crosstalk due to vertical coupling through plane voids is also non zero, even if the cross sections above and below the perforated plane form homogeneous striplines.