PCB power/ground plane edge radiation excited by high-frequency clock

Abstract

This paper describes the PCB power/ground plane edge radiation excited by a high-frequency clock, when it passes through power/ground plane pair by a through-hole signal via. For the analysis, the clock excitation mechanism was invested by TDR-TDT measurement and simulation with balanced TLM and via coupling model. Also, the power/ground plane edge radiation excited by sweeping from 100 MHz clock to 2900 MHz clock were measured. The power/ground plane edge radiation comes from voltage noise in the power/ground plane. From TDR-TDT investigations, we knew that a transition part in time domain makes the voltage noise, which has a strong relation with power/ground plane impedance depending on plane resonances and return current discontinuity of through-hole signal via. Therefore, the consecutive transitions of the clock make larger noise voltage and edge radiation. The PCB power/ground plane edge radiations of a clock depend on the power/ground plane impedance where the clock spectrum is varied. The higher power/ground plane impedance makes the larger power/ground plane edge radiation, when the clock frequency is fixed. In other words, the clock frequency and its harmonics go into the higher power/ground plane impedance range, the power/ground plane edge radiations are increased by maximum 35 dBm in near field measurement.

Full text

Page 1

PCB Power/Ground Plane Edge
excited by High-Frequency
Jun So P & Hyungsoo Kim, and Joungho Kim
Dept. of Electrical Engineering & Computer Science
KAlST (Korea Advanced Institute of Science &
Technology)
Daejeon, Korea
chitoong0eeinfo.kaist.ac.b
shamino0eeinfo.kaist.ac.kr
jounnho0ee.kaist.ac.b
Abstract- This paper describes the PCB powerlgronnd plane edge
radiation excited by high-frequency clock, when it passes through
powerlgronnd plane pair by through-hole signal via. For the
analysis, the clock excitation mechanism was invested by TDR-
TDT measurement and simulation with balanced TLM and Via
coupling model [I]. Also, the powerlgronnd plane edge radiations
excited by sweeping from lOOMHz clock to 2900MHz clock were
measured. The powerlgronnd plane edge radiation comes from
voltage noise in powerlgronnd plane. From TDR-TDT
investigations, we knew that a transition part in time domain
makes the voltage noise, which has a strong relation with
powerlgronnd plane impedanee depending on plane resonances
and return current discontinuity of through-hole signal via 121.
Therefore, the consecutive transitions of the clock make larger
noise voltage and edge radiation. The PCB powerlgronnd plane
edge radiations of a clock depend on the powerlground plane
impedance where the. clock spectrum is. The higher
powerlgronnd plane impedance makes the larger powerlgronnd
plane edge radiation, when the clock frequency is fired. In other
words, the clock frequency and its harmonics go into the higher
powerlgronnd plane impedance range, the powerlgronnd plane
edge radiations are increased by maximum 35 dBm in near field
measurement.
Keywords- Power/Grouad Plane; Edge Radiafiofi; Resonance;
Impednace: Vin; Return Current; Clock: TDR; TDT
1.
INTRODUCTION
The clock is the highest cment signal in PCB. Therefore, a
signal trace including the clock is the most critical radiation
source. So, nowadays PCB is designed to have embedded clock
line (strip line), which gets shielding effect from metal plane
and uses powerlground plane as a return current path.
However, this design method always includes through-hole
signal via, and makes another radiation problem. When the
high-frequency clock is transmitted through the through-hole
signal via, the clock encounters discontinuities at reference
planes l i e a powerlground plane and suffers reflections like
IS1 (inter-symbol interference). It is because the reference
plane is not continuously guiding the high-6equency
electromagnetic waves to support the TEM wave along the
signal trace. In addition to the clock reflection problem, the
Radiation
Clock
Heejae Lee
Computer System Division
Samsung Co.
Suwon, Korea
Hileel090samsunn.com
through-hole signal via excites electromagnetic waves (noise
voltage) in the power/ground plane, since the via is passing
across the power/ground parallel plate waveguide in PCB.
Consequently, the power/ground plane edge radiation occurs
(EMI problem). Accordingly, depending on the edge
termination condition, material and dimensions of the PCB,
standing waves appear with multiple resonance eequencies
inside the poweriground plane. These resonance waves are
responsible for the powerlground plane edge radiation from the
PCB, and are major part of the radiation h m the high-
frequency clock and high-density multi-layer PCB. ,However
the coupling of the signal through the via to the poweriground
resonance have not been thoroughly sNdied yet [3]-151.
In this paper, we have thoroughly investigated 'the, high-
frequency clock coupling to powerlground plane resonance,
and the powedground plane edge radiation with consideration
of clock frequency and its harmonics. To explain the coupling
of the poweriground resonance, we have proposed a
comprehensive 3GHz SPICE-type circuit model including
whole PCB structures, and including through-hole signal via
model in [I]. The PCB power/ground plane edge radiation
peaks are well predicted by the suggested model up to 3GHz.
Even though this model is obtained from 6equency domain
measurement i.e. S-parameters and has a valid limitation to
3GHz, TDR-TDT simulation well predicts TDR-TDT
measurements. From TDR-TDT where only single transition
occurs, we could know the powerlground noise voltage has a
same spectrum as power/ground plane impedance observed
where a through-hole signal via is. Then, we could predict the
clock excitation of noise voltage, which comes from
consecutive transitions. As mentioned previously, since the
powedground plane edge radiation is made by noise voltage,
we also could estimate the edge radiation patterns and obtain
the edge radiation measurement excited by from IOOMHz
clock to 2900MHz clock. These are going to be analyzed by
the simulation and measurements.
11.
SIGNAL COUPLING TO POWEREROUND
In this chapter, test vehicles, brief analysis of frequency
domain, and brief introduction of Balanced TLM and Via
PLANE
0-7803-8443-1104620.00 0 EKE.
197

Page 2

Coupling Model used in this paper. Coupling Mechanism is
well explained by Fig.1.
- I . ? ! .
:~-anu.nC"n.".
-
c-- ..-..;.m ..
1:;
;

j
~
, : ; i
Prn Plan. _ s .
-. ,,, ,I ;
. . .
.,'
R.dh.4S
ThlOY.hA.Ol.
Ll*".l
" I s - ;
'.-'
1 . 1
~~ . . . . . . . . . . . . . . . . . . . . .
~ .................
"I
(bl
~ Tnl-_n*DI.
~~~.
Z I D " . ,
............
~~~~~~~~~~~
...............
Figure I . Poweriground plane edge radiation mechanism caused by return
current discontinuity of through-hole signal via.
A. Test Vehcles (TV)
Four layer PCB is used for three test vehicles (Fig.2).
Substrate is fr-4 (E, = 4.5) and its thicknesses are O.lmm,
l.Omm, and 0.lmm. Fist layer from top is for top microstrip
line, second and thud are for ground plane and power plane,
and fourth is for bottom microstrip line. The powedground
plane are 14cm by 14cm, and the microstrip line is SOR
(176um width). From now on in this paper, each test vehicle is
named as 'TV(#)' and 'TV' means just test vehicle. TVI is
microstrip line without through-hole signal via, TV2 is
microstrip line with center located (7cm. 7cm) through-hole
signal via, and TV3 is that with offset located (3.Scm, 3.5cm)
through-hole signal via. All TVs have 7cm microstrip line.
TSI, "eh151e 2 - N 2
-
-
T ~ # " ~ . o " r w L_
M -
-**,ww.
L -
Figure 2. Tcrt Vehicles: poweriground plane i s I4cm by 1 4 c m WO through.
hole signal via positions wcre investigated.
. ..,
<
1.3 .
2 . 3
1 .
I.,
I,.,
I , . , I
Frqumcy [GHI]
Fsqwncy [GWZI
Figure 3. The- measure3 and tho simulated S-paramten of TV2 are compared
The model used well predicts the meausrement.
Freq"e"N IGH4
F,*IYl"CY
[GHI
Figure 4. The measured and the simulated S-paramten ofTV3 are compared.
The model used wcll predicts the meausrcment.
There is no electrical connection between power and
ground planes. As mentioned, what dealt in this paper mainly
comes from a return current discontinuity of through-hole
signal via due to powedground plane, i.e. two microstrip lines
on different layers have different reference planes. Also, for the
investigation of power/ground plane impedance variation
depending on positions of through-hole signal via, two
different test vehicles (TV2, TV3) are designed. But for TDR-
TDT, they have single DC short pass by single short via.
B. Frequency Domain Analysis
As shown in Fig. 1, power/ground plane impedance acts as
a serial resistance between two microstrip lines. The
powdground plane remnance raises the plane impedance.
Then the S-parameters of TV2 and TV3 have the increased
return loss and the increased insertion loss at the plane
resonance frequencies. These phenomena can be explained by
investing a signal return current path. The return current of
through-hole signal via induces voltage between power and
ground planes. This voltage is a noise voltage, and means that
signal is transferred to planes and that signal losses occur. The
insertion loss is more serious than the return loss. Therefore, in
the view of signal integrity, when the clock 6equency and its
main harmonics meet the plane resonances, the distortion of
signal integrity is most serious. And the increased noise voltage
means the increased powdground plane edge radiation. The
loss patterns are changed in according to the through-hole
signal via positions, which change the plane impedance
patterns. Center location (TV2) of plane suppresses odd
numbered resonance modes, e.g. ( l , O ) / ( O , l ) , (l,l), etc., and
Offset location (TV3) does '2' numbered modes, e.g.
(2,0)/(0,2), (2,2), etc. as shown in Fig. 3.
C.
For prediction of the through-hole signal via effects on
signal integrity and edge radiation of PCB, Balanced TLM
(transmission line modeling) and Via Coupling Model was
developed [2]. A conventional TLM can be applied to just two-
plane system, i.e. single power/ground plane pair. If another
layer is added, there is an ambiguity of reference point.
Moreover, single plane must be treated as two surfaces due to
skin depth effect. So, we could solve these problems with a
balanced manner in transmission line model, which is a base of
TLM method. As show in Fig. 3 and Fig. 4, this modeling
method well predicts the S-parameters of the test vehicles,
which have complex return current path. From now, all
simulations in frequency domain and time domain were done
with the Balanced TLM and Via Coupling Model, and could be
obtained easily and quickly due to Spice type model, i.e.
maximum 20minutes.
Balanced TLM and Via Coupling Model
III.
TDR-TDT ANALYSIS
In chapter 11, the three problems of the through-hole signal
via were the increased insertion loss of microstrip line, the
consequent increased noise voltage in power/ground plane, and
the increased power/ground plane edge radiation. In this
chapter 111, the fust two problems will be invested in TDR-
TDT, which is special time domain response. The TDR-TDT is
the result of a single transition in time domain, and can give a
0-7803-8443-1/04/$20.00 8 IEEE.
I98

Page 3

good understanding of multi-transitions such as clock and
PRBS (pseudo random binary sequence).
A. Band-Stop Filter Effecr as Power/Ground Plane Resonace
Fig. 5 shows simulated S-parameters of 500 microstrip line
including two band-stop filters in the middle of line. Two
band-stop filters are series connected. A band-stop is
composed of inductor and capacitor, which are parallel
connection. One of two band-stop filter has 20nH inductor
and 5OpF capacitor, and the other has 2nH and IOpF. So two
have 160MHz and 1125MHz self resonance 6equencies,
respectively. When Fig. 5 is compared with Fig. 3 and Fig. 4,
we could find the power/ground plane resonance effects to
microstrip line, such as increased return and insertion losses,
are same as a band stop filter.
At this point, we do not overlook the truth t h a t band-stop
filter is energy store device of self resonance frequency's
energy. Therefore, when the step voltage entered into line
meets a band-stop filter, it leaves energy in the filter. And
then, during oscillation of the stored energy, the stored energy
goes out to two microstrip lines with same amount energy
(Fig. 6; 'A' and 'B' are same condition) and appears in TDR-
TDT as fluctuation voltages with self resonance frequencies.
This is shown in Fig. 7. Moreover, since two nodes of band-
stop filter is 180' out of phase, TDR and TDT also shows
180" out of phase. The large cycle is 6.24ns, that is 160MHz,
and small cycle is 889~s. that is 1125MHz.
B. TDR-TDT Simulation
All microstrip lies, through-hole signal via, and
power/ground plane are modeled with the Balance TLM and
Via Coupling Model. For getting the stable TDR-TDT, two
microstrip lies must have the continuous DC return current
path. So, power/gmund plane is connected with 3nH inductor
at the (0.5cm, 7cm), and it generates 104MHz resonance,
which is made of 3nH inductor and 784pF plane capacitance,
in the power/ground plane. At I04MHz, the power/ground
plane impedance is same all over the plane except 3nH
inductor location.
Fig. 8 and Fig. 9 are TDR-TDT responses of TV2 model
and TV3 model, respectively. They show the same trends as
the band-stop filter case. They also show the power/ground
noise voltages at the through-hole signal via and the (13.5cm,
13.5cm). The TDR-TDT fluctuations are smaller than the
power/ground plane noise voltages, which are almost 10% of
source step voltage. If the clock is replaced with step voltage,
the noise voltage will be seriously increased.
From comparison of TDR-TDTs of TV2 and TV3, the 2"'
resonance effect is known. TV2 has higher 2"' resonance
frequency than TV3 as shown in Fig.10 and Fig.11. Since the
Znd, 3". and 4* resonances of TV3 are located closely, the
voltage fluctuation is less periodic. Only TV2 case shows the
2"' resonance 6equency information. As well known, since a
step voltage has larger spectrum at lower frequency, TV3 has
larger voltage fluctuations and longer time existence except
for the I04MHz noise voltage.
ir)c ..... ....... .~
I
0 1
1 s
z ,
Frequency [GHzI
Figure 5. The increased r a m loss and insenion loss at the self resonance
frequencies of two band-stop filters.
Figure 6. TDR-TDT fluctuation mechnism. Band-stop filter is a energy
storage and a energy eminer.
I
*
1o.sm
Time [nsl
z#ws,L,.s~
Figure 1. Band-stop filter effects on TDR-TDT.
resonances appears as vollags flucmtionr
Band-stop filler's
C. TDR-TDT measurement
Through section A and B, we could know the through-hole
signal via effects on TDR-TDT by considering the role of the
powedground plane resonance and impedance. In this section,
the measured TDR-TDT of TV2 and TV3 will be shown.
Firstly, the measured S-parameters of TVl, TV2, and TV3 are
shown in Fig. IO and Fig. 11. The 1' resonance appears at
I21MHz. Since we did not use inductor, but use a short via
(small inductor value), the 1" resonance frequency is shifted to
higher range than that of the model. The measured TDRs and
TDTs of TVs are compared in Fig. 12 and Fig. 13, respectively.
The largest fluctuation comes 60m 121Wz 1" resonance
frequency. For convenient comparison, each TDR-TDT is
plotted in Fig. 14 and Fig. 15. TV2 and TV3 make 180" out of
phase between TDR and TDT. TV3 has lower 6equencies
modulated fluctuations as shown in simulations.
CL7803-&243-1/04/$20.00 8 IEEE.
199

Page 4

rime Ins]
which is composed of Balanced TLM md V i a Coupling Model.
Figurc 8. Simulatsd TDR-TDT and P/G plane noise voltage of TVZ model,
. . . . . . . . . . .
......
Plr Nmre a " I .
PG NOllC 0
w
E % , .
-0 2
50
0
5
3 0
16
20
2s
10
3s
4s
Time Ins]
0 8
Figure 9. Simulated TDR-TDT and PIG plme noise voltage of TV3 model,
which is composed of Balanced TLh4 and V i a Coupling Model.
I
0 5
1 8
*s
Frequency [OH=]
Figure IO. M e a s d SI 1 magnitude of three TVs. Short via of powcrlpund
plane malrcs another resomce BI 121MHz
-
...........
. . . . .
N,
T"z
N a
0
0 5
1.5
2 5
.e
Frequency LGIlzl
Figurc 11. Measured S21 magnitude of three TVs. Resonances make large
amount of insertion loss.
-
.........
. . . . .
N,
Ti.2
N D
YlS FldSP)D"S*
-"
I_
0
6
10
15
20
Tima Ins]
25
i o
31
40 15
50
Figure 12. Measured TDRs of thra TVs. Fkt resonance at IZlMHz makes
voltage fluchlation as expected i n band-stop filter case and simulation.
IV. POWER/GROUND
Until now, the noise voltage excitation in the power/ground
plane is investigated. The noise voltage directly affects the
powedground plane edge radiation. In this paper, we did not
analyze the noise voltage excited by the clock but we could
predict it and confm it by measuring the power/ground plane
edge radiations. The power/ground plane edge radiation is
measured with 2 ways. One is VNA (vector network analyzer),
and the other is SA (spectrum analyzer) with PPG (pulse
pattern generator) as a time domain source for TVs excitation.
At each way, the near field of edge radiation is picked up with .
Lab.-made loop antenna, which has 4cm diameter. This loop
antenna has a weak point to have a low gain under SOOMHZ.
As mentioned in previous chapter, since the powerlground
plane edge radiation is induced by a magnetic current caused
by the noise voltage, the loop antenna vertically located to
power/ground plane can give noise voltage information. Also,
this measured magnetic field by the loop antenna can be
transferred to far field by Huygens' Law.
PLANE EDGE RADIATION
0-7803-8443-1/04/$20.00
0 IEEE.
200

Page 5

I
0 5
1 5
2
2 5
Frequency [GHrl
Figure 17 SA-PPG measlucmcnl ofpowcrlground plane edge radmtiom of
TVI excited by all frequncy clocks, and noise level
A. VNA measurement
The power/ground plane edge radiation of VNA
measurement is the coupling ratio between TV and the loop
antenna. That is, the 50R terminated TV is connected to
forward port of VNA and the loop antenna is connected to
reverse port. Then S21 of S-parameters shows the
powedground plane edge radiation. This VNA measurement
method uses a fi-equency domain sourcing.
In Fig. 16 the noise level is shown, i.e. without TV a t
forward port, and TV1, and they are almost same. llrl has no
through-hole signal via, and then there is no coupling between
power/ground plane and microstrip line. This can be explained
by skin depth, and strengthen the necessity of Balanced TLM.
The power/ground plane edge radiations of three TVs are also
plotted in Fig. 16. As expected, the powerlground plane edge
radiation is in proportion to the power/ground plane impedance
at the through-hole signal via location. The plane impedance
are estimated fiom S-parameters (Fig. 10 and Fig. 11).
I
1 110
0 5
1 5
is
Frequency [GHrl
Figure 18. Measured powerlground plane edge radiation of TV2 excited by
1OOMHz clock. Solid line is SA-PPG meausremenl and dot line is VNA
0 5
1.5
1 5
Frsquenq [GHrl
Figure 19. Measured pawedground plane edge radiation ofTV3 excited by
IWMHz clock. Solid line is SA-PPG meausrement and dash line is VNA
Froqusncy [GHrl
Figurc 20. Measured pomrlground plane edge radiation of TV2 excited by
I20MHz dock which meets resonance frequency 121MHz
3 ~ 5
2%
..
D ...
0 5
Froqvcrry [GHrl
Figure 21. M e a s u r e d pomrlground plane edge radiation ofTV3 excited by
12OMHz clock which meets resonance frequency IZIMHz.
Frsquew IGHzl
Figurc 22. M e a s u r e d pomrlground plane edge radiation ofTV2 excited by
900MHz clock whose 3d h m i c meets anti-resonance and suppressed.
-
T"3 54 uea*urem.nr
Tv3 YNA *C_YremCnt
-30
: .
I !
~ M I I ~ O L I
E -
E
110
1 5
2 5
Frequency [GHrl
Figure 23. Measured pomrlground plane edge radiation of TV3 excited by
900MHz clock. TV3 does not anti-resonance at 27WMlh.
0-7803-8443-1/04/$20.00
0 IEEE.
201