...

お願い ・製品によっては、お守りいただかないと発煙、発火等に

by user

on
Category: Documents
2

views

Report

Comments

Transcript

お願い ・製品によっては、お守りいただかないと発煙、発火等に
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
C39E.pdf
Jul.20,2010
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
EU RoHS Compliant
•
•
•
All the products in this catalog comply with EU RoHS.
EU RoHS is "the European Directive 2011/65/EU on the Restriction of the Use
of Certain Hazardous Substances in Electrical and Electronic Equipment."
For more details, please refer to our website 'Murata's Approach for EU RoHS'
(http://www.murata.com/info/rohs.html).
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Contents
1. Introduction
1
2. Generation of power supply noise from digital ICs and configuration of decoupling circuits
4
2.1 Mechanism of power source noise generation
4
2.2 Various ways to view the noise and evaluation criteria
5
2.3 Measurement method for insertion loss
7
2.4 Bypass (decoupling) capacitor
8
2.5 Inductors, ferrite beads
9
2.6 Capacitance necessary for a capacitor
3. Noise suppression with a capacitor
11
12
3.1 Frequency characteristics of the capacitor
12
3.2 Influence of the capacitor mounting pattern
13
3.3 Noise path and capacitor mounting position
14
3.4 Influence of peripheral circuit impedance
17
3.5 Parallel connection of capacitors and antiresonance
18
4. Capacitor with improved high-frequency characteristics
24
4.1 Low-ESL capacitor
24
4.2 Low-ESL capacitor lineup
27
4.3 3-terminal capacitor
27
4.4 3-terminal capacitors lineup for power supplies
32
5. Inductors and LC filters
33
5.1 Decoupling circuit using an inductor
33
5.2 Frequency characteristics of inductors
34
5.3 Frequency characteristics of ferrite beads
35
5.4 Characteristics for a combination of capacitors and inductors
38
5.5 LC filter
40
5.6 Considerations when using an inductor for a power supply
43
5.7 Lineup of inductors suitable for power supplies
44
5.8 Lineup of LC filters suitable for power supplies
47
6. Suppressing power supply voltage fluctuation
48
6.1 Relationship between power supply impedance and voltage fluctuation
48
6.2 Voltage fluctuation when a capacitor is present
49
6.3 Suppressing spikes with parallel capacitors
52
6.4 Suppressing spikes with low-ESL capacitors
53
6.5 Voltage fluctuation when the pulse width is wide
53
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
7. Location of a capacitor for suppressing power supply impedance
Jul.20,2010
56
7.1 Power supply impedance relative to an IC
56
7.2 Simple estimation of power supply impedance relative to an IC
57
7.3 Possible range for placing the closest capacitor of an IC
58
7.4 Guideline for the maximum allowable wiring length, lmax
60
8. Configuration of PDN combined with capacitors
64
8.1 Hierarchical positioning of decoupling capacitors
64
8.2 Impedance of PDN
65
8.3 Hierarchical positioning of capacitors
66
8.4 Target impedance on PCB
69
8.5 Bulk capacitor
69
8.6 Board capacitor
70
8.7 Capacitance design of a capacitor
72
8.8 Making a PDN with ultra-low impedance
76
9. Summary
78
References
79
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
1. Introduction
Various capacitors and EMI suppression filters are used for power supplies connected to
digital ICs as shown in Figure 1-1. By forming a decoupling circuit acting as a filter as
shown in Figure 1-2, at the junction connecting an IC’s power source terminal and power
distribution network (PDN), power integrity (PI) can be improved. 1) 2)
Capacitor
IC2
Power
Inductor
IC3
IC1
PCB
Ferrite Bead
Power
Supply Smoothing
Capacitor
IC
Capacitor
3-Terminal
Capacitor
Figure 1-1 An example of noise suppression products used for digital IC power supply
Power Distribution
Network (PDN)
Power
Inductor
Wiring
Decoupling Capacitor (Stand Alone)
IC3
Power Inductor
Decoupling Capacitor (Parallel)
Power
Supply
IC
IC2
Smoothing
Capacitor
Decoupling Circuit (Composite)
IC1
Ferrite
Bead
3-Terminal
Capacitor
Large-Capacitance Capacitor
Figure 1-2 An example of wiring connection for digital IC power supply
As shown in Figure 1-3, this decoupling circuit performs functions such as:
(1) suppressing noise generated by or entering an IC,
(2) providing transient current associated with IC operations and maintaining voltage and
(3) becoming a part of signal path. 3) 4) 5)
When this circuit is not fully functional, the following problems may occur, as shown in
Figure 1-4:
(1) interference with other circuits (such as IC3) or increasing the noise emission of the
equipment due to noise leakage,
(2) intrusion of noise from an external source causing problems with IC operation,
(3) power supply voltage fluctuation, interfering with IC operations, lowering signal
1
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
integrity, and increasing the noise superimposing over signals, and
(4) reduced signal integrity due to insufficient return circuit of the signal current.
Therefore, formation of an appropriate decoupling circuit is important for both noise
suppression and circuit operations.
Power Distribution Network
(PDN)
IC2
Returning
Signal
Current
Signal
Current
Decoupling
Circuit
Power
Supply
Current
Power
Supply
IC
Smoothing
Circuit
Target IC
This diagram focuses only
on operation of decoupling
circuit for this IC.
Noise
Entered
IC3
Working of
decoupling circuit
Main performance
indicators
(1) Noise suppression
Insertion loss
(Permeability
constant/attenuation)
Working of decoupling circuit
(1) Noise suppression
(2) Temporarily supplying current
(3) Forming signal returning path
Signal Line
IC1
Noise
Emission
Receiving IC
Noise generation
Noise receiving IC
Target frequencies
Examples of noise problem and
evaluation criteria
1kHz 1MHz 1GHz
Noise measurement
(Terminal voltage, radiated electric field,
near magnetic field distribution)
Voltage fluctuation, transient voltage
response
(2) Current supply
Impedance
(reflection coefficient)
(3) Signal path
Signal waveform (eye diagram)
Figure 1-3 Working of the power supply filter (decoupling circuit) when focusing on IC1
Noise Level [dBuV/m]
50
40
Increase in power voltage fluctuation
⇒ Decrease in voltage margin
⇒ Increase in signal line noise
⇒ Decrease in signal integrity
30
20
10
0
200
400
600
Frequency [MHz]
800
1000
Increase in noise emission from
the power supply cable
IC2
IC3
IC1
30
Voltage [dBuV]
20
10
0
-10
-20
0
200
400
600
Frequency [MHz]
800
1000
Increase in the noise over the power supply wiring
⇒ Increase in noise emission through other
circuits of the equipment
Figure 1-4 An example of influence by power supply noise
2
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
With a circuit with relatively low clock speed, or a circuit with a large margin against
noise, this decoupling circuit can be easily formed by locating a bypass capacitor,
connecting the power supply to the ground near the power supply terminal. In this manual,
this bypass capacitor is called a decoupling capacitor. However, a more sophisticated
decoupling circuit is needed for ICs with higher clock speed, ICs generating large amounts
of noise, and noise-sensitive ICs.
This manual aims to explain the workings of components for power supplies and how
you may appropriately select components, in order to design this high-performance
decoupling circuit. (However, the smoothing circuit in Figure 1-3 is excluded.)
Typically, performance of a decoupling circuit is assessed mainly with insertion loss from
the viewpoint of noise suppression in Figure 1-3 (1), and with impedance from the
viewpoint of (2) current supply, and (3) signal path formation. Since these two viewpoints
differ, the first half of this manual (Chapters 2 through 5) will focus on noise suppression
performance, and offer explanations based on insertion loss as an indicator. The second
half of this manual (Chapters 6 through 8) will focus on current supply performance, and
offer explanations based on impedance as an indicator.
3
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
2. Generation of power supply noise from digital ICs and configuration of decoupling
circuits
We will first describe the mechanism of power supply noise generation for digital ICs,
configuration of general decoupling circuits for handling such noise, and provide an
overview of the circuit characteristics pertaining to decoupling circuits (power supply
filters) covered in this manual.
2.1 Mechanism of power source noise generation
A simplified model of a C-MOS circuit mainly used for digital ICs is shown in Figure 2-1.
For the purpose of simplicity, the working of the C-MOS transistor on the driver side is
represented as a switch, and gate capacitance of the C-MOS transistor on the receiver side
is represented as a capacitor connected to a ground. With a C-MOS digital IC, by this
switch on the driver side connecting the signal line with either the power supply side
(VDD) or the ground side (GND), the signal output level can be set to be “1” or “0”
respectively.6)
Normally, if the signal level does not change for the power supply of the C-MOS digital
circuit, there is hardly any current flow. However, if charge current at the gate capacitance
(when the signal level switches from “0” to “1”) and discharge current (when the signal
level switches from “1” to “0”) pulsates through the signal line as shown in Figure 2-1,
current consequently flows through the power supply and ground. Further, aside from this
current, a so-called through current flows from the driver power supply to the ground
briefly when the signal switches. Through current also becomes a cause for the pulsating
current flowing through power supplies and grounds.
Since these currents pulsate very acutely, they contain a very wide range of frequency
components causing noise failures when a part of its energy is radiated externally. Also,
since an acute change in current causes variance in power supply voltage due to
inductance of the power supply and the ground patterns, it causes instability in the
operations of peripheral circuits sharing the common power supply. 7) 8)
Therefore, it is necessary to seal the current around ICs (to decouple ICs with peripheral
circuits) so that noise emission can be suppressed and voltage fluctuation would not affect
peripheral circuits. Meanwhile, since fluctuation in power supply voltage would make the
operation of the IC emitting the noise itself unstable, it is necessary to contain
noise-related power supply voltage fluctuation to an acceptbale level. The decoupling
circuit illustrated in Figure 1-3 is used for such a purpose.
Although the model used for Figure 2-1 considers gate capacitance with regard to the
ground for the purpose of simplicity, and charge current and discharge current are
considered to flow through the ground, in actuality gate capacitance occurs with regard to
the power supply as well, necessitating consideration of the case where charge current and
discharge current flow through the power supply.
4
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
VDD
(
)
(
)
Gate
Capacitance
Signal Line
Through
Current
Decoupling
Capacitor
Charge Current
Driver
GND
Jul.20,2010
(
)
Discharge
Current
(
Receiver
)
):Pattern inductance
(
Figure 2-1 Simplified model of a digital IC
Ordinarily, in order to contain the current flowing through the power supply around the
IC, a bypass is formed (an inductor may be combined as described later) by mounting a
capacitor between the power supply and the ground as seen in Figure 2-1. This capacitor is
called a decoupling capacitor. In order to form an effective decoupling circuit it is important
to:
(1) form a bypass that can function under high frequency (using a capacitor with small
impedance),
(2) strictly limit the range current flows through (by placing the capacitor near the IC), and
(3) keep the pattern inductance small (especially between the IC and the capacitor).
An example of capacitor location and power source pattern configuration that take these
points into consideration are shown in Figure 2-2. (With this example, it can be wired in a
single layer under the IC.)
(3) Pattern between IC and capacitor
should be thick and short
(2) Capacitors should be
located near the IC
VDD
VDD
IC
IC
VDD
VDD
Capacitor
Capacitor
GND
GND
GND
(a) VDD enters from the right
(1) Low-ESL capacitors
(such as MLCC) should be chosen
GND
(b) VDD enters from the left
Figure 2-2 An example of decoupling capacitor positioning
Wiring rules for the cases where such an ideal placement is difficult or a higher
performance decoupling circuit is necessary are described in Chapter 3 and later in this
manual.
2.2 Various ways to view the noise and evaluation criteria
There are several measuring methods for evaluating performance of decoupling circuits,
depending on the specific purpose. As described in Figure 1-3 of Chapter 1, decoupling
circuits have three roles, namely (1) noise suppression, (2) temporarily supplying current
5
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
and (3) forming the signal returning path.
Out of these roles, “(1) noise suppression” pertains to the filtering out of noise leaking
from the power supply from the IC and shutting off the noise coming in from an external
source. A measurement taken at the opposite side of the noise source of the decoupling
circuit is used to evaluate this performance as shown in Figure 2-3. In other words, when
questioning the noise leaking out of the IC, measurement is taken on the PDN (point A)
side, and when questioning the noise coming in to ICs from an external source,
measurement is taken at the power supply terminal for the IC (point B).
Measurement criteria include a voltage waveform observed by an oscilloscope and
voltage spectrum measured by a spectrum analyzer. This manual indicates data obtained
in an empirical fashion. Meanwhile, when focusing on comparing decoupling circuit
performances, insertion loss characteristics are used as opposed to measuring voltage and
spectrum in actual circuits. 9) A network analyzer is used to measure insertion loss. Since
measurement conditions are predetermined, results are highly reproducible, making this
method suitable for comparing component performances. Decoupling circuit performances
are compared mainly with these insertion loss characteristics in this manual. The
measurement method of this insertion loss will be described in Section 2.3.
Measurement
Point A
PDN
IC
Decoupling
Circuit
Power
Supply
PDN
Measurement
Point B
Decoupling
Circuit
IC
Power
Supply
Noise
Source A
Noise
Source B
Noise Waveform
(Oscilloscope)
Noise Waveform
(Oscilloscope)
Insertion Loss
(Network Analyzer)
Insertion Loss
(Network Analyzer)
Noise Waveform
(Spectrum Analyzer)
(a) Noise emitting from IC
Noise Waveform
Spectrum Analyzer
(b) Noise entering IC
Figure 2-3 Various measurement criteria for observing noise suppression performance
On the other hand, there are cases where filter performance is evaluated from the
viewpoints of (2) temporarily supplying current and (3) forming signal returning path as
shown in Figure 1-3. In these cases, stabilizing the power supply voltage when IC current
fluctuates or securing the signal integrity are the issues as shown in Figure 2-4. Therefore,
measurement point B’ on the power supply terminal side of the IC and measurement point
C at the signal output terminal are used for measurement locations as indicated in Figure
2-4.
Measurement criteria for point B’, located at the power supply, include noise waveform,
spectrum, and impedance, and for point C, located where the signal is, include jitter and
6
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
spectrum. When focusing on comparing decoupling circuit performances, highly
reproducible impedance characteristics are used. Impedance is the criterion covered
mainly in this manual.
Measurement
Point B’
PDN
Decoupling
Circuit
IC
Measurement
Point C
Current
Fluctuation
Power
Supply
Signal
Current Supply
Jitter
(Oscilloscope)
Impedance
(Network Analyzer, etc.)
Noise Emitted
(Spectrum Analyzer)
Voltage Fluctuation
(Oscilloscope)
Figure 2-4 Measurement to evaluate current supply performance
2.3 Measurement method for insertion loss
Normally, noise filter performance is expressed in terms of insertion loss. 6)
Since
decoupling circuits used for power supplies are a type of noise filter, their noise
suppression performance can be described in terms of their insertion loss.
The measurement circuit for insertion loss is indicated in Figure 2-5. Insertion loss (I.L.)
is described by the effect of a filter mounted on a circuit with 50Ω impedance observed as
the difference in output voltage before and after mounting the filter in dB. The larger the
insertion loss, the more effective the noise suppression is.
Insertion loss may be substituted by the oscillation of transmission coefficient (S21) for S
parameter measured for the 50Ω system. (Note that insertion loss and S21 will have the
opposite positive and negative sign.)
50Ω
50Ω
A(V)
B(V)
50Ω
Filter
A(V)
Insertion Loss =20 log
C(V)
B
(dB)
C
50Ω
(2-1)
Figure 2-5 Insertion Loss Measurement Circuit
7
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
2.4 Bypass (decoupling) capacitor
Next, a basic configuration of a decoupling circuit is introduced. One of its components is
a decoupling capacitor shown in Figure 2-1.
When a decoupling capacitor is used for the power supply terminal of an IC, as shown in
Figure 2-6 (a), a bypass capacitor from the power supply to the ground is formed as a filter,
as seen in Figure 2-6 (b). We will assume that this will be used on a multilayer substrate,
and the ground for the IC and the capacitor will be connected to the ground plane with a
via.
Insertion loss, in this case, would be greater as capacitor impedance becomes less. Since
capacitor impedance decreases inversely in proportion to frequency, this filter becomes a
low pass filter where insertion loss is large at high frequency.
Filter characteristics shown in Figure 2-6, vary depending on the internal impedance of
IC power supply or PDN impedance in an actual circuit. Since impedance must be fixed
when comparing filter performances, it is common to set impedance at 50Ω when taking
the measurement, as shown in Figure 2-5. The filter’s characteristics when mounted on an
actual circuit are estimated from the measurement result at 50Ω.
Internal Impedance
for IC Power Supply
Power
IC (Noise Source) Supply
Filter
Power Supply
Wiring
To PDN
Decoupling
Capacitor
Impedance
of PDN
Noise
Source
Ground
Ground
(a) Decoupling capacitor
(b) Filter circuit
Figure 2-6 Filter circuit with decoupling capacitor
Theoretical value of a bypass capacitor’s insertion loss characteristics measured at 50Ω
is shown in Figure 2-7. As the capacitance of a capacitor becomes larger, or the frequency
becomes higher, the capacitor’s insertion loss increases linearly. This corresponds with the
fact that the capacitor impedance decreases inversely in proportion to frequency, therefore
indicating that capacitors with larger capacitance would basically show a superior noise
suppression effect.
Insertion loss is expressed in dB as shown in figure 2-5. When frequency or capacitance
is increased tenfold, the insertion loss would increase by 20dB.
8
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
100
pF
100
0pF
100
00p
F
20
dB
/de
c.
20dB
100
000
pF
Figure 2-7 Insertion loss characteristics of capacitor (theoretical values)
The actual characteristics of the capacitor’s insertion loss are as indicated in Figure 2-8.
In the high frequency range over 10MHz, insertion loss decreases as frequency increases.
This is considered to be due to the fact that insertion loss is limited by a minute inductance
component (ESL) and resistance component (ESR) contained in the capacitor, as described
later. This indicates that in order to form a decoupling circuit with excellent noise
suppression performance at high frequency, we must use a capacitor with small ESL and
ESR.
0.0
01
0.0
µF
1µ
F
0.1
µF
1µ
F
10
µF
Figure 2-8 Insertion loss characteristics of capacitor (Converted from Data in
Murata Chip S-Parameter & Impedance Library)
2.5 Inductors, ferrite beads
In addition to the bypass capacitor described above, inductors such as ferrite beads can
be inserted serially with the wiring to form a common noise suppression filter. 9) Inductors
are also used for decoupling circuits for power supplies.
However, when only inductors are used for the power supply of an IC, although it may
suppress the noise properly, impedance relative to the power supply terminal becomes high
causing problems for IC operations, or interfering with the signal’s return current, making
it difficult to maintain signal integrity. Therefore, capacitors and inductors are normally
used in combination, placing capacitors in the vicinity of ICs as shown in Figure 2-9 (b) and
(c). (Although inductors are placed on the IC side of high-frequency amplifiers in order to
block specific frequency, for digital ICs, the combinations in Figure 2-9 are generally used.)
9
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
IC
Minimizing impedance
in this range
Jul.20,2010
IC
IC
Minimizing impedance
in this range
Connections to
the ground
should be apart
(c) π-section filter
(b) L-type filter
(a) C-type filter
Minimizing
impedance in
this range
Figure 2-9 Filter configurations used for IC power supplies
(C-type, L-type and π-section)
When capacitors and inductors are combined as seen in Figure 2-9 (b) and (c), the slope
for the characteristics curve of insertion loss can be made more acute compared with using
only a capacitor as in (a). Since the insertion loss in the decay area can be increased
simultaneously this way, it is more advantageous when the noise must be greatly
attenuated. An example of the change in insertion loss when inductors are combined is
shown in Figure 2-10.
0
insertion loss (dB)
(dB)
insertionloss
20
(a
CC
) ML
F
0.1 µ
only
40
60
80
( b)
100
C
MLC
120
F
0.1µ
+ fe
rrite
bea
M
100
0Ω@
d 47
Hz
Hz
00M
F
@1
µ
Ω
1
0
.
0
47
LCC bead
(c) M ferrite 1µF
+
LC C
+M
140
1
10
100
1000
frequency (MHz)
Figure 2-10 An example of insertion loss fluctuation when
an inductor is combined (calculated values)
Since, in Figure 2-9, the IC ground and the ground for the closest capacitor become
return paths for the noise, they should be made as short as possible to minimize impedance.
Also when a π-section filter is formed as shown in Figure 2-9 (c), it is more desirable to use
a ground separate to the capacitor (with a via, etc.) so as to prevent noise from bypassing
the inductor (when conditions of the grounds are not desirable) going through the ground
(the ground plane is shared, however).
The decoupling circuit shown in Figure 2-9 can be applied to noise entering from outside,
in addition to noise emitted by the IC. For example, with circuits exposed to strong high
frequency energy such as those in mobile phones, a decoupling circuit with capacitor and
inductor combination to achieve larger insertion loss is more suitable.
10
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
2.6 Capacitance necessary for a capacitor
When an inductor is used for the decoupling circuit as shown in Figure 2-9 (b) and (c),
the capacitor’s capacitance must be greater than the circuit without an inductor, as shown
in Figure 2-9 (a). When multiple ICs are connected to a power supply, each of the
decoupling capacitors operate in parallel, effectively expanding capacitance, and
impedance relative the power supply terminal is relatively low due to the support given by
the power supply module. In contrast, when an inductor is used, capacitors supporting the
power supply of the corresponding IC are limited to those inside the inductors.
A method of roughly estimating necessary capacitance, when an inductor is mounted as
shown in Figure 2-9 (b), is introduced here as an example. The capacitor’s capacitance
necessary inside the inductor may be established using the following equation:
C ≥
L
2
ZT
(2-2)
In this case, C is capacitance (F) of the capacitor that is necessary, L is inductance (H) of
the inductor and ZT is the power supply impedance necessary for IC: target impedance (Ω).
Although there are many approaches to determining ZT, it can be determined by using
instantaneous transient current ∆I (A) and allowable voltage fluctuation ∆V (V) of an IC.
10)
ZT =
∆V
∆I
(2-3)
Let us estimate the capacitance necessary for the capacitor when using an inductor with
L =1µH (approximately 600Ω@100MHz) for the IC’s decoupling circuit at ∆I=0.1A and
∆V=200mV using the equation above, for example. First, we can derive ZT =2Ω from
equation (2-3). By introducing this value to equation (2-2) we can obtain C ≧0.25µF. In
this case it can be seen that the IC’s decoupling capacitor needs at least 0.25µF
capacitance.
From equation (2-2), we can see that the higher the inductance is, the greater the
necessary capacitance becomes. We can assume that as inductance becomes greater,
impedance will become greater than the lower frequency, expanding the frequency range,
necessitating the lowering of impedance for the capacitor on the low-frequency side.
Since equation (2-2) only gives a rough estimate, capacitance may be insufficient in some
cases when applying this calculation result to an actual circuit. Amply sufficient
capacitance should be chosen when applying this value to circuit design.
11
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
3. Noise suppression with a capacitor
Ideally, the noise suppression effect when using capacitors and inductors in a decoupling
circuit improves as frequency becomes higher. However, in reality, it becomes less effective
in the high-frequency region over 10MHz. Although the cause of becoming less effective
may be due to the characteristics of the capacitor itself, it may also be due to the way the
capacitor is used on the printed circuit. In this chapter, we will describe the causes for
fluctuation in frequency characteristics of capacitors used in a decoupling circuit and how
it may affect noise suppression effect.
3.1 Frequency characteristics of the capacitor
Multilayer ceramic capacitors (MLCC) are widely used as capacitors offering excellent
frequency characteristics. However, they contain small amounts of intrinsic resistance and
inductance causing less than ideal characteristics when used under high frequency. A
typical equivalent circuit of a capacitor is shown in Figure 3-1. 1)
In Figure 3-1, ESR
represents the intrinsic resistance and ESL represents the intrinsic inductance. Because of
these elements, capacitor impedance indicates a V-shaped frequency characteristic, as
shown in Figure 3-2.
Cap
ESL
ESR
Figure 3-1 Equivalent circuit of the capacitor
Capacitive
Inducible
log|Z|
Z =
1
2πf ⋅ Cap
Z = 2πf ⋅ ESL
Z ≈ ESR
Self-resonant
f0 =
1
2π Cap ⋅ ESL
log f
Figure 3-2 Frequency characteristics of the capacitor
The impedance of a capacitor decreases almost linearly, showing characteristics similar
to an ideal capacitor on the left side in low frequency range in Figure 3-2 (marked
“capacitive”). After reaching the minimum value (marked “self-resonant”), it then
increases almost linearly past that point (marked “inducible”). Impedance in the capacitive
zone corresponds with ESR while in the inducible zone corresponds with ESL. Therefore,
in order to use a capacitor with low impedance in the high frequency range, it becomes
12
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
important to choose a capacitor with low ESR and ESL.
Noise suppression effect of a capacitor being used as a bypass capacitor corresponds with
this impedance (The smaller the impedance, the larger the noise suppression effect.).
Therefore, its insertion loss characteristic is charted as a V-shaped frequency
characteristic similarly to Figure 3-2. An example of comparison between MLCC’s
impedance and insertion loss is indicated in Figure 3-3.
In this example, characteristic curves are compared using a 2.0x1.25mm-size (GRM21
series) capacitor for both cases while varying its capacitance. The two curves show
virtually the same curve, and at the frequency where the capacitor’s impedance is almost
25Ω, a cut off (3dB) frequency appears for insertion loss. This is understood to be because
in the insertion loss measurement circuit in Figure 2-5, impedance of the bypass capacitor
becomes significantly large against impedance of the measurement system (50Ω).
By looking at Figure 3-3, we can tell that the characteristic curves are neatly separated
in the capacitive zone, but essentially become one line in the inducible zone. By this we can
assume that MLCC’s ESL is influenced by a factor other than capacitance.
In order to significantly improve the characteristics in this inducible zone, we need a
capacitor with reduced ESL. We will describe such a capacitor in Chapter 4.
MLCC in this manual refers to a regular multilayer ceramic capacitor (without a special
structure for reducing ESL).
1000
0
3dB
0.0
01
0.0
µF
1µ
0.1
F
µF
10
100
10
1
0.1
20
insertion loss (dB)
impedance (ohm)
25Ω
0.0
01
µF
0.0
1µ
F
0.1
µF
1µ
F
10
µF
1µ
F
40
50
10
µF
60
70
0.01
0.001
0.01
30
80
0.1
1
10
100
frequency (MHz)
1000
10000
90
0.01
0.1
(a) Impedance
1
10
100
1000
10000
frequency (MHz)
(b) Insertion loss
Figure 3-3 An example of frequency characteristics of capacitor
(From Murata Chip S-Parameter & Impedance Library)
3.2 Influence of the capacitor mounting pattern
Insertion loss characteristics for the capacitor we have discussed so far pertain to
capacitors mounted on a printed circuit under ideal conditions. These characteristics may
vary when mounted on an actual substrate. For example, when considering connecting a
capacitor to a ground from a power supply wiring as illustrated in Figure 3-4, mounting
pattern and via inductance would be placed linearly to the capacitor. When factoring in the
component inductance (ESL(PCB)) generated by mounting on a substrate, insertion loss
characteristics of the capacitor change, as seen in Figure 3-5 and insertion loss is observed
to decrease in the inducible zone (high frequency range).
When a capacitor is used to suppress high frequency noise, it should be designed with
thick and short wiring so that this mounting inductance (ESL(PCB)) can become small.
13
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
ESL(PCB) must be kept small from the viewpoint of the power supply impedance in addition
to that of the insertion loss (noise suppression effect). Wiring design to suppress
inductance value in the wiring and power supply impedance will be introduced in detail in
Chapter 7.
Equivalent Circuit of Bypass Path
Power Supply Path
Noise
Bypass
Path
Wiring and via for
mounting capacitor
generate
inductance
ESC(PCB)
Capacitor
ESL(cap)
ESL(cap)
ESL(PCB)
Simple Circuit
via
(Connected to ground
layer on the back)
Circuit taking
capacitor’s ESL
(ESL(cap)) into
consideration
Circuit taking ESL
of capacitor and
substrate into
consideration
Figure 3-4 Influence of wiring when mounting a capacitor
Insertion
Insertion Loss
Loss (dB)
0
Characteristics taking inductance of the substrate into consideration
When ESL(PCB)=1nH
20
When ESL(PCB)=2nH
40
Inductance as low as 1nH has a
possibility of reducing the effect of
capacitor by approximately 10dB
60
1005-size 1µF
Characteristics of multilayer ceramic capacitor alone
80
0.1
1
10
Frequency (MHz)
Frequency
(MHz)
100
1000
Figure 3-5 Fluctuation in characteristics of conductor due to mounting
inductance (calculated values)
3.3 Noise path and capacitor mounting position
Mounting Inductance (ESL(PCB)) when the capacitor is mounted may change according to
the noise path and capacitor mounting position. For example, when the capacitor is
positioned over the noise path as shown in Figure 3-6 (a), ESL(PCB) comes from the
capacitor mounting pattern and via, making it relatively small. On the other hand, if the
mounting position is located on the opposite side of the noise path as shown in Figure 3-6
(b), all the wiring from the power supply terminal to the mounting position is included in
ESL(PCB) making ESL(PCB) larger. The effect of the capacitor lessens in the high frequency
region, in this case. We shall call this kind of wiring away from the noise path “branched
wiring”.
An example of calculating fluctuation in insertion loss assuming this branched wiring
being a 10mm-long MSL (Micro Strip Line) is shown in Figure 3-7. In this example,
insertion loss decreases close to 20dB in an over-10MHz frequency range.
When the pattern configuration is complex like a power supply circuit, and there are
multiple power supply terminals emitting noise, it is necessary to place capacitors so that
there is no branched wiring, in consideration of the noise sources and transmission paths
14
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
corresponding with the capacitors.
Power Supply Wiring
Power
Supply
Ground
Noise Path
Noise
Observation
Point
Noise
Bypass
Path
Capacitor
via
(Connected to ground
layer on the back)
IC (Noise Source)
via
(Connected to ground
layer on the back)
Noise
Observation
Point
ESL(PCB)
Preventing noise
bypass
IC (Noise Source)
(a) Layout limiting mounting inductance (Top: Layout Bottom: Schematic Diagram)
Branched Wiring Section
Power Supply Wiring
Noise Path
Noise
Observation
Point
Noise
Bypass
Path
Power Supply
IC (Noise Source)
via
(Connected to
ground layer on
the back)
via
(Connected to ground
layer on the back)
Noise
Observation
Point
ESL(PCB)
Preventing
noise bypass
IC (Noise Source)
(b) Layout with large mounting inductance (Top: Layout Bottom: Schematic Diagram)
Figure 3-6 Relation between noise path and capacitor position
0
ed
nch
bra
insertion
loss (dB)
(dB)
insertionloss
10
ing
wir
ting
oun
rm
pad
lf
itse
o
by
m
or
acit
acit
10m
cap
p
a
g
m
c
erin
1m
µF
sid
e1
ring
con
-siz
ide
tics
608
ons
1
is
c
r
ics
cte
of a
rist
ara
tics
cte
Ch
eris
ara
act
Ch
r
a
Ch
20
30
40
50
60
Wiring configuration: 1mm wide
Assuming MSL to be made with
0.6mm-thick dielectric material
70
80
0.1
1
10
frequency (MHz)
(MHz)
frequency
100
1000
Figure 3-7 An example of fluctuation in insertion loss when branched wiring
is present (calculated values)
Figure 3-8 is an example to confirm the influence of branched wiring shown in Figure 3-7
in an experiment. Noise emitted at the power supply terminal of a digital IC operating at
20MHz is suppressed by mounting 1608-size, 1µF MLCC 6 mm away from the power
supply terminal. Size of the noise can be confirmed with voltage waveform measured with
15
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
an oscilloscope located at 15mm away from the power supply terminal.
Figure 3-8 indicates measurement results when (a) without a capacitor, (b) a capacitor is
mounted on the other side of the noise path (with branched wiring), and (c) a capacitor is
mounted over the noise path (no branched wiring). Compared with (b), voltage fluctuation
(ripple) is less than one third in (c), indicating that existence of branched wiring can
greatly affect noise suppression.
6mm
6mm
Measurement
Point
Wiring
Width: 2mm, Dielectric material thickness: 0.4mm
Measurement
Point
20MHz Operating Frequency
74AC04
15mm
15mm
0.30
0.30
0.20
0.20
0.20
0.10
0.10
0.10
0.00
0.00
-0.10
-0.10
-0.20
-0.20
-0.30
-0.30
-200
-100
0
100
200
time (ns)
(a) Without a capacitor
level (V)
0.30
level (V)
level (V)
Voltage Waveform 100mV 50ns/div(AC coupling)
0.00
-0.10
-0.20
-0.30
-200
-100
0
100
200
time (ns)
(b) With branched wiring
-200
-100
0
100
200
time (ns)
(c) Without branched wiring
Figure 3-8 Confirming power supply noise suppression effect (voltage waveform)
In Figure 3-9, the same circuit as Figure 3-8 is used to evaluate change in the influence
of the power supply with regard to noise emission. A loop antenna emitting noise is
mounted at one end of the power supply wiring, and the noise emitted was measured from
3m away. Horizontal polarization is marked with H and vertical polarization is marked
with V. As with Figure 3-8, measurement results when (a) without a capacitor, (b) a
capacitor is mounted on the other side of the noise path (with branched wiring), and (c) a
capacitor is mounted over the noise path (without branched wiring) are indicated.
Results of Figure 3-9 also shows that the noise suppression effect of (c) without branched
wiring is better than (b) with branched wiring by approximately 10dB in terms of the
frequency showing peak emission. This 10dB difference is understood to represent the
influence of inductance at 6mm-branched wiring section.
Furthermore, while with as seen in Figure 3-8, voltage fluctuation is contained within
approximately 1/5 when comparing (a) and (b) in spite of the presence of branched wiring,
in Figure 3-9, peak emission fluctuates by approximately 8dB (approximately 1/2.5) when
comparing (a) and (b). The reason for this may be that the influence for all frequencies is
multiplied together for voltage fluctuation, while radiation is measured for each frequency,
accentuating the effect of high frequency in this observation. We can see that the influence
of ESL(PCB) such as branched wiring becomes more significant when dealing with
high-frequency noise.
16
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Connecting antenna
6mm
Jul.20,2010
6mm
(Loop antenna with 13cm diameter)
70
60
60
50
40
H
30
V
20
Decrease by approx. 8dB
50
40
H
30
V
200
400
600
800
1000
frequency (MHz)
(a) Without a capacitor
50
40
H
30
V
20
10
0
Decrease by approx. 18dB
60
20
10
70
noise level (dBuV/m)
70
noise level (dBuV/m)
noise level (dBuV/m)
20MHz Operating Frequency
74AC04
10
0
200
400
600
800
1000
frequency (MHz)
(b) With branched wiring
0
200
400
600
800
1000
frequency (MHz)
(c) Without branched wiring
Figure 3-9 Confirming IC power supply noise suppression effect (noise radiation spectrum)
3.4 Influence of peripheral circuit impedance
Insertion loss characteristics for the capacitor described in Section 3.1 show the value
when mounted on a 50Ω system capacitor. Since the actual power supply circuit will be
different from this, we need to compensate for the estimation of capacitor effect, taking
circuit impedance into consideration. One example in Figure 3-10 shows the result of
calculating capacitor characteristic fluctuation, assuming that peripheral circuit
impedance will show a certain resistance value.
As shown in Figure 3-10, insertion loss of a capacitor generally tends to be small in a
circuit with low impedance. Since power supply circuits are considered to be relatively
low-impedance, when designing a filter, we need to figure in the reduction in insertion loss.
Also, in order to improve the capacitor’s effect in such a low-impedance circuit, the use of
an inductor in combination is effective. A decoupling circuit with an inductor added is
introduced in Chapter 5.
17
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
0
Sample
Circuit
10
Output
Impedance
(Normally 50Ω)
Signal
Source
Ω
0.5
20
insertion loss (dB)
Input
Impedance
(Normally 50Ω)
Jul.20,2010
2Ω
30
10
40
Ω
50
50
Measurement
Impedance
(Normally 50Ω)
60
70
Peripheral circuit impedance for insertion loss
Ω
80
0
1
10
frequency (MHz)
100
1000
Figure 3-10 Fluctuation in insertion loss in relation with peripheral circuit impedance
3.5 Parallel connection of capacitors and antiresonance
When the capacitance of a capacitor is insufficient, or target impedance and/or insertion
loss cannot be achieved due to large ESL and ESR, multiple capacitors may be connected
in parallel as seen in Figure 3-11. We must be careful with parallel resonance (called
antiresonance) occurring among capacitors in this case, and as shown in Figure 3-12, to
make impedance of certain frequencies becoming higher than the case with one
capacitor.11) (In the case of Figure 3-11 (b), wiring between capacitors works as an inductor
to increase insertion losses in some frequencies; however, impedance seen from the power
supply terminal tends to become large due to antiresonance.)
Antiresonance is a phenomenon where the self-resonant frequency of two capacitors
differs, and parallel resonance occurs in the frequency region where one capacitor is in the
inducible zone and another capacitor is in the capacitive zone, resulting in increased total
impedance (ideally, it becomes infinitely large). Therefore, insertion loss becomes small for
the frequencies where antiresonance occurs.
IC (Noise
Source)
Ground
Power Supply
Ground
IC (Noise
Source)
Power Supply
Power
Supply
Wiring
Capacitors
(with significantly differing
capacitance)
Ground
(a) Capacitors having significantly different capacitance
Power
Supply
Wiring
Same capacitance, but
wiring is long in between
(b) Capacitors being far apart
Figure 3-11 Examples of capacitor connections likely to have antiresonance
18
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Impedance of parallel circuit
10
ESL1
ESR1
C2
impedance (Ω)
C1
ESL2
Jul.20,2010
Antiresonance
Impedance of C2
MLCC 1.6 x 0.8mm-size 1000pF
ESL=0.6nH, ESR=60mΩ approx.
1
Impedance of C1
MLCC 1.6 x 0.8mm-size 1µF
ESL=0.6nH, ESR=10mΩ approx.
0.1
ESR2
0.01
Selfresonance
of C1
1
Selfresonance
of C2
10
100
1000
frequency(MHz)
Figure 3-12 Parallel resonance of the capacitor (calculated values)
10
impedance (Ω)
ESL1
ESL1
1
C1
ESL2
C2
C2
0.1
0.01
1
10
100
1000
frequency(MHz)
Figure 3-13 Mechanism of the capacitor’s parallel resonance
The following methods, shown in Figure 3-14, would prevent antiresonance by:
(1) inserting resonance suppression components such as ferrite beads between
capacitors,
(2) matching capacitance of capacitors to align self-resonant frequencies, and
(3) containing the capacitance difference less than 10:1 when combining capacitors with
different capacitance.
Method (1) is quite effective for improving insertion loss. However, the effect of lowering
impedance is smaller, as explained in Section 2.6. While complete suppression of
antiresonance is difficult for methods (2) and (3), in practicality, the problem is somewhat
lessened. As shown in Figure 3-14 (d), if the target capacitance is reached with one
high-performance capacitor with low ESL and ESR, it would be ideal since it can eliminate
the problem of antiresonance. This high-performance capacitor will be introduced in
Chapter 4.
19
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Power
IC (Noise Source) Supply
Ferrite Bead
IC (Noise Source) Power Supply
Match capacitances of capacitors
If performance is not enough
increase the number of capacitors
Ground
Ground
IC (Noise Source)
Power
Supply
Wiring
Power
Supply
Wiring
(a) Insert a ferrite bead between capacitors
Jul.20,2010
(b) Match capacitances of capacitors
Power Supply
Power Supply
Power
Supply
Wiring
IC (Noise Source)
Power
Supply
Wiring
Ground
Ground
Make capacitance pitch of capacitors smaller
(At the most less than 10 times)
Ground
Use one high-performance,
large-capacitance capacitor
(c) Make capacitance pitch of capacitors smaller (d) Use one high-performance capacitor
Figure 3-14 Examples of positioning capacitors to suppress antiresonance
Figures 3-15 through 3-17 are examples to confirm the influence of antiresonance in
experiments. The noise observed at the power supply terminal of a digital IC operating at
4MHz is measured with an oscilloscope and a spectrum analyzer. Measurement is taken on
the left and right of the decoupling circuit (left: input side of the noise; right: output side of
the noise).
In this diagram, measurement results of the oscilloscope are located in the center and
measurement results of the spectrum analyzer are located on the outside. Both are
measured using an FET probe. (Spectrum is measured by resolving frequency of the
terminal voltage, as opposed to measuring emissions.)
In Figure 3-15, antiresonance occurring from a combination of two MLCCs is observed.
The first row is the case with only one MLCC, and antiresonance does not occur. The
occurrence of antiresonance is deliberately induced by creating extreme conditions by
applying conditions in Figure 3-11 for the second and third rows. MLCCs at 1µF and
1000pF are connected through wiring. As indicated by the arrows, noise spectrum
increased for some frequencies, and strong ringing is observed in the third row even over
the power supply voltage.
20
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Noise Generation
Observing IC power
supply noise
Switching at 4MHz
(Observing 3V power
supply with AC coupling)
40
30
Connecting to power supply
100
(Reference)
100
50
One MLCC 1µF
Without antiresonance
50
0
-50
60
50
noise level (dBuV)
level (mV)
noise level (dBuV)
50
Decoupling Circuit
(Capacitor)
level (mV)
60
Jul.20,2010
0
-50
20
30
20
-100
-300
10
100
150
200
250
300
-200
-100
0
100
200
-100
-300
300
-200
-100
time (ns)
frequency (MHz)
level (mV)
50
40
30
100
200
10
300
0
50
100
time (ns)
1µF+1000pF
(1mm apart)
50
0
150
200
250
300
frequency (MHz)
When antiresonance is present, noise
increases at its frequency
Two MLCCs
100
60
0
100
-50
60
50
50
noise level (dBuV)
50
level (mV)
0
noise level (dBuV)
40
0
-50
20
40
30
20
-100
-300
10
0
50
100
150
200
250
300
-200
-100
0
100
200
-100
-300
300
-200
-100
time (ns)
frequency (MHz)
0
100
200
300
10
0
50
100
time (ns)
150
200
250
300
frequency (MHz)
50
40
100
Two MLCCs
1µF+Wiring+1000pF
(6mm apart)
0
30
-50
60
50
50
noise level (dBuV)
100
50
level (mV)
60
level (mV)
noise level (dBuV)
Inserting wiring between capacitors changes frequency
and increases influence
0
-50
20
40
30
20
10
0
50
100
150
200
frequency (MHz)
250
300
-100
-300
-200
-100
0
time (ns)
100
200
300
-100
-300
-200
-100
0
100
200
300
10
0
50
time (ns)
100
150
200
frequency (MHz)
250
300
Figure 3-15 Examples of noise observation when antiresonance is present
In contrast, Figure 3-16 indicates results of the experiment suppressing antiresonance.
The first row shows the case with one MLCC as with Figure 3-15. Conditions from Figure
3-14 (b) were applied for the case on the second row to match the capacitance of capacitors
(with two capacitors). Conditions from Figure 3-14 (c) are applied to the third row to make
the capacitance pitch smaller (with four capacitors). In these cases, as the number of
capacitors increases, power supply voltage fluctuation becomes smaller and noise spectrum
decreases as well.
21
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Noise Generation
Decoupling Circuit
(Capacitor)
Connecting to power supply
(Reference)
100
50
One MLCC 1µF
Without antiresonance
50
level (mV)
noise level (dBuV)
40
30
0
-50
60
50
noise level (dBuV)
100
50
level (mV)
60
0
-50
20
30
20
-100
-300
10
100
150
200
frequency (MHz)
250
300
-200
-100
0
100
200
-100
-300
300
-200
-100
time (ns)
Two MLCCs
100
60
level (mV)
50
40
30
0
100
200
10
300
0
0.47µF+0.47µF
(1mm apart)
50
0
50
100
150
200
frequency (MHz)
250
-100
-300
300
50
0
-50
-200
-100
0
100
200
250
300
50
-50
0
100
150
200
frequency (MHz)
60
100
-100
-300
300
40
30
20
20
10
50
time (ns)
noise level (dBuV)
50
level (mV)
0
noise level (dBuV)
40
10
-200
-100
time (ns)
0
100
200
300
0
50
time (ns)
100
150
200
frequency (MHz)
250
300
Output ripple voltage decreases as the
number of capacitors increases
40
30
100
Four MLCCs
50
1µF+0.1µF+
0.01µF+1000pF (1mm apart)
0
level (mV)
level (mV)
noise level (dBuV)
50
-50
60
50
50
noise level (dBuV)
100
60
0
-50
20
40
30
20
10
0
50
100
150
200
frequency (MHz)
250
300
-100
-300
-200
-100
0
100
200
300
time (ns)
Ripple suppression effect is observed on IC power supply terminal side also.
(Effect is insignificant due to the influence of wiring between capacitors)
-100
-300
-200
-100
0
100
200
300
10
0
50
time (ns)
100
150
200
frequency (MHz)
250
300
Frequency characteristics show
unevenness due to differing resonance
frequencies of capacitors
Figure 3-16 Measurement results for combinations with little antiresonance
Figure 3-17 shows the results of the experiment where a low ESL capacitor and a
3-terminal capacitor are used as an example of Figure 3-14 (d). The first row in the
diagram indicates one MLCC as with the earlier example for the sake of comparison.
The second row indicates a case with a low-ESL capacitor. Fluctuations in power supply
voltage become small on the left and right of the decoupling circuit, and the spectrum
becomes smaller because of it also. A low-ESL capacitor is considered to be an especially
effective component for suppressing fluctuation in power supply voltage.
The third row indicates a case with a 3-terminal capacitor. We can see that the voltage
fluctuation and spectrum on the right is especially small. This should indicate excellent
noise suppression characteristics of a 3-terminal capacitor.
These high-performance capacitors are introduced in detail in Chapter 4.
22
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Noise Generation
Decoupling Circuit
(Capacitor)
Connecting to power supply
40
30
100
(Reference)
100
50
One MLCC 1µF
Without antiresonance
50
0
-50
60
50
noise level (dBuV)
level (mV)
noise level (dBuV)
50
level (mV)
60
0
-50
20
30
20
-100
-300
10
100
150
200
250
300
-200
-100
0
100
200
-100
-300
300
-200
-100
time (ns)
frequency (MHz)
Low-ESL capacitor
LW reverse LLL type
1µF
100
60
level (mV)
50
40
30
0
100
200
10
300
0
50
time (ns)
50
0
100
-50
100
150
200
250
300
250
300
250
300
frequency (MHz)
60
50
50
noise level (dBuV)
50
level (mV)
0
noise level (dBuV)
40
0
-50
20
40
30
20
-100
-300
10
0
50
100
150
200
250
300
-200
-100
0
100
200
-100
-300
300
-200
-100
time (ns)
frequency (MHz)
0
100
200
10
300
0
50
time (ns)
100
150
200
frequency (MHz)
Low-ELS capacitor has a relatively large ripple
suppression effect on the IC power supply side
40
30
100
3-terminal capacitor
NFM type 1µF
50
0
-50
60
50
50
noise level (dBuV)
level (mV)
noise level (dBuV)
50
level (mV)
100
60
0
-50
20
40
30
20
10
0
50
100
150
200
frequency (MHz)
250
300
-100
-300
-200
-100
0
time (ns)
100
200
300
-100
-300
-200
-100
0
100
200
300
10
0
50
time (ns)
100
150
200
frequency (MHz)
With 3-terminal capacitor, noise and ripple
on the output side decrease significantly
Figure 3-17 Measurement results when using low-ESL capacitors
23
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
4. Capacitor with improved high-frequency characteristics
In Chapter 3, we described general frequency characteristics when using a capacitor for
a decoupling circuit of IC power supply. In this chapter, we will introduce capacitors with
improved high-frequency characteristics by reducing ESL.
4.1 Low-ESL capacitor
The ESL of an MLCC is understood to be generated by magnetic flux that occurs when
current flows through external and internal electrodes as shown in Figure 4-1. Therefore,
we may alter ESL by changing current path and distribution as a result of the change in
electrode configuration.
An example of a capacitor with ESL reduced through innovation of electrode
configuration is shown in Figure 4-2.
Direction of Current
Internal Electrode
φ
Residual
Inductance
(ESL)
External
Electrode
Current and
magnetic flux flowing
through an electrode
Structure of
an MLCC
Circuit
symbol for a
capacitor
Equivalent circuit
for a capacitor
with ESL
Figure 4-1 Mechanism of ESL generation in an MLCC
(a) LW reverse
type
(b) 8-terminal
type
(c) 10-terminal
type
Reference:
Ordinary MLCC
Figure 4-2 Configurations of low-ESL capacitors
Internal Electrode
Direction of Current
(a) LW reverse type
(b) 8-terminal type
(c) 10-terminal type
Figure 4-3 Structure of low-ESL capacitor (schematic view)
Figure 4-2 (a) indicates a capacitor with reduced inductance due to its wide and short
electrode, called length width reverse capacitor or LW reverse capacitor. As shown in the
internal structure diagram in Figure 4-3 (a), the internal electrode is wider and shorter
compared with common MLCCs.
In Figure 4-2 (b) and (c), multi-terminal capacitors with an increased number of external
24
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
electrodes, where neighboring electrodes have reversed polarities, are indicated. As shown
in the internal structure diagrams in Figure 4-3 (b) and (c), internal electrodes are formed
thick and short, and additionally internal electrode drawers are formed so that they can be
connected to external electrode alternately. (Their external appearances seem similar to
capacitor arrays, but these components’ internal electrode configuration is totally
different.)
By choosing such a structure, mutual inductance occurs between currents where they
flow in opposite directions, cancelling each other’s inductance as shown in Figure 4-4. For
the components where currents flow between neighboring electrodes, the current loop
tends to be extremely small as opposed to the currents flowing in opposing directions.
Furthermore, these inductances are connected in parallel, realizing extremely small ESL
as a total of the components.
M
L
–
+
–
Has an effect for current to flow
toward neighboring electrodes
in addition to opposing
electrodes (Current loop
becomes smaller reducing
inductance)
L
Assuming inductance of each current
is equal (L), inductance per current
flowing in both directions, Lpart
becomes
Lpart =
+
2L − 2M
2
+
–
–
+
8-terminal type
Therefore, as the mutual
inductance M becomes larger,
Lpart becomes smaller.
Figure 4-4 Inductance cancellation effect from mutual inductance
Figure 4-5 indicates an example of comparing impedances of a conventional MLCC and
low-ESL capacitors. All capacitors are 1.6x0.8mm size and 1µF. Impedance decreased to
approximately 1/5 for the LW reverse capacitor and approximately 1/2 of the LW reverse
capacitor for the multi-terminal capacitor in the frequency range over 100MHz. Compared
with the conventional MLCC, the ESL of a multi-terminal capacitor should be less than
1/10.
)
nH
.5
.0
)
x
nH
ro
pp
0.2
(A
x.
o
r
p
CC
Ap
ML
e(
H)
yp
ry
T
a
.1 n
n
e
.0
s
di
r
x
r
ro
O
ve
pp
Re
(A
LW
pe
y
lT
i na
er m
T
8
Figure 4-5 An example of impedance characteristics for a low-ESL capacitor
(From Murata Chip S-Parameter & Impedance Library)
25
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Characteristics shown in Figure 4-5 are those of conversion from S parameter to
impedance, where a capacitor is mounted in on the bypass side of a micro strip line (MSL)
for measurement. Therefore, they represent characteristics specific to the component (and
can be approximated with a lumped constant).
Generally, when mounting a capacitor on a printed circuit, influence by inductances
(ESLPCB) of pattern connected to the capacitor and via, in addition to capacitor’s ESL, is
significant. As shown in the diagram, when a multi-terminal capacitor is mounted on a
substrate, the inductance cancellation effect between currents flowing in opposite
directions next to each other influences currents in the pattern and via as in Figure 4-6,
making the influence of ESLPCB relatively small. Therefore, compared with using MLCCs
with conventional pattern and via, use with patterns and via dedicated for multi-terminal
capacitors would yield a higher impedance improvement effect surpassing the performance
difference indicated in Figure 4-5.
Mounting
Surface
Pattern inductance becomes
smaller due to mutual
inductance between
neighboring wringing
Direction of Current
Ground Layer
Inductance of vias become smaller
due to mutual inductance between
neighboring vias
Power Supply Layer
Figure 4-6 Inductance suppression effect when mounting a multi-terminal capacitor
26
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
4.2 Low-ESL capacitor lineup
An overview of the low-ESL capacitors currently commercialized is shown below. Please
refer to the catalog, etc. for the latest information.
● LW reverse capacitor LLL series
Capacitance range
0.5 x 1.0mm size
:0.1−0.22µF
0.8 x 1.6mm size
:0.0022−2.2µF
1.25 x 2.0mm size
:0.01−2.2µF
1.6 x 3.2mm size
:0.01−10µF
● 8-terminal-type capacitor LLA series
Capacitance range
1.6 x 0.8 mm size
:0.1−2.2µF
2.0 x 1.25 mm size
:0.01−4.7µF
3.2 x 1.6 mm size
:0.1−2.2µF
● 10-terminal-type capacitor LLM series
Capacitance range
2.0 x 1.25 mm size
:0.01−2.2µF
3.2 x 1.6 mm size
:0.1−2.2µF
4.3 3-terminal capacitor
Another method of reducing ESL is the use of 3-terminal capacitors. An example of a
3-terminal capacitor is shown in Figure 4-7. This is a type of feed-through capacitor which
is an MLCC with excellent frequency characteristics, having circuit connection designed to
reduce ESL.
As shown in Figure 4-8, a 3-terminal capacitor is structured with input/output terminals
to draw in the noise path inside the component. Consequently, inductance generating at an
internal electrode branches out three ways to form a T-shaped circuit. When connecting
input/output terminals of a 3-terminal capacitor to the noise path, ESL in the input/output
directions enters the noise path serially, increasing insertion loss (improving the noise
suppression effect). Also, ESL in the bypass direction is only at the ground area, making it
about one-half of an MLCC. The 3-terminal capacitor indicated in Figure 4-7 further
reduces inductance in the ground area by designing it with two ground electrodes on the
left and right sides of the capacitor.
27
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
These innovations contain the ESL of a 3-terminal capacitor in the bypass direction at
about 10 to 20pH, which is less than 1/30 for conventional MLCCs for some models.
Therefore, we can expect a good bypass effect under high frequency over 1GHz.
Insertion losses for an MLCC and a 3-terminal capacitor are compared in Figure 4-9.
They are both 1.6x0.8mm size at 1µF, but 3-terminal capacitor shows approximately 35dB
larger insertion loss in the frequency range over 100MHz.
Ground
Electrode
Input/Output
Electrode
Input/Output
Electrode
Feed-through type
multilayer ceramic capacitor
Ground
Electrode
External Configuration for
NFM18PS
ESL of internal electrode is
divided in three directions,
input, output and ground
Equivalent Circuit
(ESR is omitted)
Internal
impedance of a
power supply
Impedance
of PDN
IC Power Supply
(Noise Source)
Connecting to the circuit
Figure 4-7 An example of a 3-terminal capacitor for power supplies
2-Terminal Capacitor (MLCC)
3-Terminal (Feed-Through) Capacitor
ESL from
patterns
and vias
ESL from
components
An example of via
for grounding
Figure 4-8 Mechanism of ESL reduction using a 3-terminal capacitor
28
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Multilayer Capacitor
1608-size 1µF
Improvement
by over 30dB
3-Terminal Capacitor
1608-size 1µF
NFM18PS105
Figure 4-9 Insertion loss characteristics of a 3-terminal capacitor
In addition to the effect described above, 3-terminal capacitors are characterized by
increasing insertion loss by forming a T-type filter without interfering with current flowing
in the bypass direction since its inductance (ESLPCB) from the circuit pattern and via are
located serially with the noise path where the input/output terminals are mounted, as
shown in Figure 4-10. Although its ESLPCB at the area where the ground terminal is
mounted enters the bypass direction, this can be minimized with multilayer substrates by
connecting with the ground plane with multiple vias with this area directly beneath the
component.2)
Current path from
電源プレーンから
グラウンドプレーンに
power supply plane
至る電流経路
to ground plane
パターンとviaに
Inductance
パターンとviaに
occurs for
インダクタンス①
pattern and
Capacitor
コンデンサ(断面)
Connectionが発生する
via (1) (Cross-Section)
point
to ground
グラウンドプレーン
への接続点
plane
Inductance
occurs
インダクタンス②
が発生する
for
pattern and via
(2)
Inductance occurring
atグラウンド接続部に
ground connection
発生するインダクタンス
③
(3)
3-Terminal part
3端子コンデンサ
Capacitor
入出力部に発生する
Inductance occurring at
インダクタンス②
input/output
part (2)
Inductance
occurring at
入出力部に発生する
input/output
part
インダクタンス①(1)
Ground
グラウンド層
Layer
Ground
グラウンド層
Layer
Power Supply
電源層
Layer
Printed
Circuit
プリント基板(断面)
(Cross-Section)
via
ESL
ESL
基板の
Inductance
of インダクタンス
substrate
(1)①
via
Connection point to
電源プレーンへの接続点
power supply plane
ESR
ESR
Equivalent
コンデンサのcircuit
of等価回路
a capacitor
Power
電源層 Supply
Layer
via
via
電源プレー with
Connecting
ンに接続
power
supply plane
via
via
グラウンドプレー
Connecting with
ンに接続
ground plane
Connecting
with power supply
電源プレーン、もしくは
plane or
wiring going toward IC
ICの電源端子に向かう
power配線に接続
supply terminal
3-Terminal
Capacitor Unit
3端子コンデンサ本体
基板の
Inductance
インダクタンス
of substrate
②
(2)
プリント基板実装に伴うインダクタンスを含んだ
Equivalent
circuit of a capacitor including inductance
コンデンサの等価回路
associated with
mounting on a printed circuit
②
(2)
(1)
①
(3)
③
Inductance from (1) and (2) is
①および②のインダクタンスはノイズの
serial
to the noise path,
経路に直列に入るため、ノイズ除去効
therefore
working to improve
果を高める方向に働く
noise removal effect
③のインダクタンスはノイズ除去効果を
Inductance
from (3) weakens
弱めるので、可能な限り小さくする必要
noise
removal effect, therefore
がある
プリント基板のインダクタンスを含めた等価回路
it should be reduced as much
Equivalent
circuit including printed
as possible
circuit inductance
For MLCC
For 3-terminal capacitor
Figure 4-10 Equivalent circuit of a capacitor including inductance from the substrate
For these reasons, 3-terminal capacitors can achieve larger insertion loss compared with
MLCCs even when they are mounted on a printed circuit. Also, reduction of insertion loss
when mounted on a low-impedance circuit is smaller than MLCCs (due to ESLPCB located
serially with the noise path).
29
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
In Figure 4-11, fluctuation of insertion loss is compared for varying measurement
impedance by taking 1µF capacitors as an example. We assume capacitors are used in a
power supply circuit which tends to be low in impedance, and the impedance of the
measurement system is varied from 50Ω to 5Ω and 0.5Ω. Insertion loss of 3-terminal
capacitors maintains over 30dB at 1GHz for a low-impedance circuit. This suggests that an
inherently large insertion loss of a 3-terminal capacitor, and ESL arranged in T-shaped as
shown in Figure 4-8 is effective in the high-frequency range around 1GHz.
Generally, insertion loss of a bypass
capacitor becomes smaller when used
in a low-impedance circuit
0
0
Measured with
0.5Ω system
Measured with 0.5Ω system
insertion loss (dB)
insertion
(dB)
insertion loss
loss(dB)
insertion
(dB)
Reduction in insertion loss of a 3terminal capacitor is small under
high frequency even when used in
a low-impedance circuit
20
Measured with
5Ω system
40
Measured with
50Ω system
60
80
20
Measured with
5Ω system
40
Measured with
50Ω system
60
80
0.1
1
10
100
1000
frequency (MHz)
(MHz)
frequency
0.1
1
10
frequency (MHz)
(MHz)
frequency
100
(a) For MLCC
(b) For 3-terminal capacitor
1000
Figure 4-11 Change in characteristics when measurement impedance is varied (calculated values)
(assuming impedance of the measurement system to be 50Ω, 5Ω and 0.5Ω)
In Figure 4-12, an example for confirming fluctuation of the capacitor’s noise
suppression effect with varied measurement impedance is indicated through an
experiment. Bypass capacitor operation is observed, in this case, by taking a measurement
of near magnetic field distribution around the capacitor. It visually illustrates the way
noise is being bypassed to the ground with the capacitor, since this near magnetic field is
understood to correspond with the current.
Characteristic impedances of the wiring used in this experiment are (a) approx. 60Ω and
(b) approx. 3Ω. Both ends of the wiring are terminated in order to eliminate the effect of
the reflected wave. The measurement frequency was 100MHz, while the range of
measurement was 40mmx30mm with the capacitor mounted in the center. The diagram
shows that the noise is entering from the right side, and the capacitor’s noise suppression
effect is observed from the intensity of the current leaving from the left side. Intensity of
the current is marked in colors, indicating stronger current as it changes from blue to red.
We could confirm from results of the experiment shown in Figure 4-12 that MLCC
controls the noise relatively well for (a) 60Ω, but its filtrating effect tends to decrease for
(b) 3Ω (current flows through to the left). Meanwhile, the 3-terminal capacitor controlled
the noise well for both (a) and (b). We also found that 3-terminal capacitors have a
tendency for smaller noise diffusion toward the ground compared with MLCCs. This is
assumed to be because the 3-terminal capacitor is connected to the ground with vias
directly under the component.
30
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Since, wide wiring with low characteristic impedance tends to be used for power supply
circuits as in this experiment, a 3-terminal capacitor should be the better choice for noise
suppression.
Without a capacitor
MLCC
3-Terminal Capacitor
1
MSL Cross-Section
0.8
(mm)
(a) When characteristic impedance is approx. 60Ω
Without a capacitor
MLCC
3-Terminal Capacitor
10
MSL Cross-Section
0.2
(mm)
(b) When characteristic impedance is approx. 3Ω
3-Terminal
Capacitor
MLCC
GND
MSL
Noise
Entering
Noise
Entering
Noise
Entering
Ground of a 3-terminal
capacitor is connected to
the back side with five vias
GND
(Entire back surface is GND)
Without a capacitor
MLCC
3-Terminal Capacitor
(C) Reference: Mounted state of a capacitor (wiring thickness is for (b) 3Ω)
Figure 4-12 Change in current distribution around the capacitor
when measurement impedance is varied
31
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
4.4 3-terminal capacitors lineup for power supplies
The lineup of 3-terminal capacitors suitable for IC power supplies is listed below. Please
refer to the catalog, etc. for the latest information.
● 1608 size
High attenuation type NFM18PS series
Capacitance range
+0.2
0.15 –0.1
0.47-1.0µF
0.8±0.1
(2)
(1)
+0.2
0.15 –0.1
(3)
0.2 min.
0.6±0.1
(2)
1.6±0.1
+0.2
0.8 –0.1
0.4±0.1
: Electrode
(in mm)
High capacitance type up to 2.2µF NFM18PC series
Capacitance range
0.1-4.7µF
0.2 min.0.2 min.
0.8±0.1
(2)
(1)
(3)
(2)
0.25±0.1
0.25±0.1 0.8±0.1
0.4±0.1
1.6±0.1
: Electrode
(in mm)
● 2012 size
High capacitance type up to 10µF NFM21PC/PS series
Capacitance range
0.3±0.2
0.4±0.2
0.85±0.1
0.1-10µF
0.6±0.2
2.0±0.2
1.25±0.1
(2)
(1)
(3)
(2)
: Electrode
0.25±0.2
(in mm)
● 3212 size
Rated at 50V NFM3DPC series
Capacitance range
0.7±0.2
0.022µF
1.1±0.3
3.2±0.2
1.25±0.2
(2)
(3)
0.25±0.2
(1)
32
(2)
0.3±0.2
: Electrode
(in mm)
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
5. Inductors and LC filters
When a decoupling capacitor is not sufficient to suppress power supply noise, the
combined use of inductors and LC filters is effective. In this chapter, we will introduce
inductors and LC filters suitable for controlling power supply noise.
Choke coils and ferrite beads are the common inductors used for power supply
decoupling circuits. While ferrite beads are used to control a wide range of frequencies that
are relatively high frequency, choke coils tend to be used to control specific frequencies.
Although ferrite beads are used more for the noise measures described by this manual,
choke coils are also for noise control; therefore they are both introduced in this chapter. We
will also introduce an LC filter for power supplies with a combination of inductors and
capacitors.
5.1 Decoupling circuit using an inductor
A general configuration when forming a decoupling circuit with an inductor added to the
power supply is shown in Figure 5-1. Figure 5-1 (a) shows a decoupling capacitor with an
inductor added, and (b) shows a higher performance π-section filter by adding a capacitor
to (a). Since many capacitors are used with other ICs in power supply wiring, even (a)
could practically become a π-section filter, while the configuration in (b) could suppress
noise with more certainty.
Generally, inductors with larger impedance (and consequently with larger inductance)
would show excellent noise suppression effect (however, there are some considerations
mentioned in Section 5.6).
On the other hand, when an inductor is used as in Figure 5-1, momentary current
necessary for the IC operation will be supplied by the capacitor between the inductor and
the IC. The capacitance necessary for this capacitor becomes larger corresponding with the
inductance as shown in equation 2-2. Therefore, it is not recommended to use excessively
large inductance.
Inductor (Ferrite Bead)
IC (Noise Source) Power Supply
Ground
IC (Noise Source)
Power
Supply
Wiring
Capacitor
Internal
Impedance
(a) When forming a L-type filter
Inductor (Ferrite Bead)
IC (Noise Source)
IC (Noise Source)
Power Supply
Power
Supply
Wiring
Ground
Capacitor
Internal
Impedance
(b) When forming a π-section filter
Figure 5-1 Construction of a power supply filter in combination with an inductor
33
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
5.2 Frequency characteristics of inductors
In Section 5.1, we showed an example of using an inductor with a capacitor. In this
section, we describe characteristics when an inductor is used alone, in order to aid in
understanding its characteristics. Inductors are inserted serially with a noise path as
shown in Figure 5-2. The inductor’s insertion loss characteristics are that of a low-pass
filter as with a bypass capacitor. This is due to the inductor’s impedance increasing
proportionally to the frequency, and the larger the impedance is, the larger the insertion
loss becomes.
0
50Ω
Sample Circuit
1µH
Signal
Source
50Ω
insertion loss (dB)
20
10µH
40
100µH
60
80
100
0.01
0.1
1
10
frequency (MHz)
100
1000
Figure 5-2 Frequency characteristics of the insertion loss measurement circuit and ideal inductor
Actual inductors, much like capacitors, show frequency characteristics. A simplified
equivalent circuit of an inductor and the shape of impedance frequency characteristics are
shown in Figure 5-3.
Z ≈ EPR
L
ESR
EPC
EPR
Capacitive
Inducible
log|Z|
Z = 2πf ⋅ L
Z =
1
2πf ⋅ EPC
Self-resonant
f0 =
1
2π L ⋅ EPC
log f
Figure 5-3 Equivalent circuit of an inductor and its frequency characteristics
As seen in the equivalent circuit illustrated in Figure 5-3, equivalent parallel
capacitance (EPC) and equivalent parallel resistance (EPR) appear in parallel with the
inductor coil (L). Therefore, the impedance of the inductor shows inducibility at a relatively
low frequency, and increases almost linearly. However, it reaches the maximum at a
certain frequency (self-resonant frequency f0) and shows capacitivity and decreases almost
linearly after that.
Impedance for a self-resonant frequency is constrained by EPR, and impedance in the
capacitive region is constrained by EPC. Therefore, in order to realize large impedance
34
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
under high frequency, it becomes important to choose an inductor with small EPC.
Capacitance generated at the winding wire shows in this EPC. Also, the resistance of the
winding wire can be considered with ESR (equivalent serial resistance) aside from these.
An example of comparing impedance and insertion loss as inductance of an inductor is
varied is shown in Figure 5-4. When the measurement is taken with a 50Ω system, at the
frequency where the inductor impedance becomes approximately 100Ω, a cut off (3dB)
frequency appears for the insertion loss. As with the capacitor, it corresponds with the
frequency where inductor impedance becomes a significant size compared with the
impedance of the measurement system.
100µH
10000
0
3dB
10µH
insertion loss (dB)
impedance (ohm)
100000
1µH
1000
100
20
1µH
40
10µH
60
100µH
100Ω
10
0.01
0.1
1
10
100
80
0.01
1000
0.1
1
10
100
1000
frequency (MHz)
frequency (MHz)
Figure 5-4 Comparison of impedances and insertion losses of inductors
(For LQH3C series)
These insertion loss characteristics change, as seen in Figure 5-5, as the impedance of
the measurement system varies. As opposed to the capacitor described in Section 3.4, with
an inductor, insertion loss of an inductor increases as the impedance of the measurement
system decreases. Therefore, when we try to suppress noise for a low-impedance circuit, an
inductor is a more suitable component in general. (However, when used for a power supply,
a low-impedance power supply must be available; therefore an inductor is not used alone,
but used in combination with a capacitor as shown in Figure 5-1.)
0
insertion loss (dB)
50Ω
10Ω
20
2Ω
40
60
80
0.01
0.5Ω
Measurement Impedance
(Normally 50Ω)
0.1
1
10
100
1000
frequency (MHz)
Figure 5-5 Fluctuation in insertion loss in relation with measurement system impedance
(Calculated values, sample circuit LQH3C 10µH)
5.3 Frequency characteristics of ferrite beads
When a ferrite bead is used as an inductor, its characteristics differ slightly from what is
described in Section 5.2.
35
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
5.3.1 Basic structure of a ferrite bead
As shown in Figure 5-6, the basic structure of a ferrite bead consists of a cylindrical
(bead-shaped) ferrite with a lead going through it. (Although recently there have been
ferrite beads with multilayer ferrite with a spiral formed inside, as seen in the right
bottom diagram of Figure 5-6, the basic structure is as described above.) Since magnetic
flux is formed inside ferrite in response to the current going through the lead, it is possible
to gain inductance and impedance relative to the magnetic permeability of ferrite. 12)
This inductance fluctuates depending on the frequency characteristics of ferrite’s
magnetic permeability, and generally is not fixed. Also, impedance generated is strongly
influenced by the magnetic loss of the ferrite. Therefore, characteristics of ferrite beads are
usually expressed in terms of frequency characteristics of impedance, as opposed to
inductance.
An equivalent circuit of a ferrite bead is shown in Figure 5-7, and an example of
impedance frequency characteristics is shown in Figure 5-8.
BL01
Feed-Through
Terminal (Lead)
Ferrite Core
External appearance and product
example for ferrite bead with lead
Magnetic
Flux
Current
BLM
Basic Construction
External appearance and internal construction
example of a chip-type ferrite bead
Figure 5-6 Ferrite bead construction
EPC
L(f)
R(f)
Figure 5-7 Equivalent circuit for a ferrite bead
400
impedance (ohm)
Sample: BLM21PG221SN1
300
|Z|
200
R
100
X
0
1
10
100
1000
frequency (MHz)
Figure 5-8 Frequency characteristics for ferrite bead impedance
36
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
5.3.2 Frequency characteristics of ferrite beads
In Figure 5-8, |Z|, R and X indicate absolute value of impedance, component resistance,
and component reactance respectively. For relatively low frequency under 10MHz,
impedance mainly indicates reactance, but for frequencies over 10MHz, the component
resistance increases. For the frequencies where the component resistance is the majority,
ferrite beads show their propensity to absorb noise by converting it to heat. This resistance
component does not generate with the direct current, and would not affect power supply
(no energy loss occurs for direct current), therefore, is a suitable component for noise
suppression of power supplies.
Also, the impedance in Figure 5-8 shows a peak at a frequency around 300MHz; the
curve is milder than the inductor shown in Figure 5-4. This indicates that while ferrite
beads reduce impedance at high frequency due to EPC as with inductors, the Q value of
resonance becomes smaller due to the component resistance.
On the other hand, the slope of insertion loss characteristic for ferrite beads tends to be
smaller than pure inductors. Figure 5-9 shows insertion loss characteristics of a ferrite
bead shown in Figure 5-8. For a pure inductor, insertion loss characteristics have a slope at
-20dB/dec. shown as the dotted line, while that of a ferrite bead would show a milder slope.
insertion loss (dB)
0
10
20
dB
/d
20
ec
.
Characteristics of
ferrite beads
(Sample:
BLM21PG221)
30
40
1
10
100
1000
frequency (MHz)
Figure 5-9 An example of insertion loss characteristics for a ferrite bead
5.3.3 Improving high frequency characteristics of ferrite beads
As shown in Figure 5-7, ferrite beads also have stray capacitance EPC, becoming the
main cause for a decrease in impedance at high frequencies above 100MHz. Products with
improved high-frequency impedance by making this EPC smaller have been
commercialized, and used for high frequency noise measurement for frequencies above
several hundred MHz. Figure 5-10 shows an example of impedance characteristics for a
component with reduced EPC.
37
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
2000
impedance (ohm)
1500
Improved winding wire
configuration + material
BLM18GG471
1000
Improved winding wire
configuration BLM18EG471
500
Conventional Type
BLM18PG471
0
1
10
100
1000
10000
frequency (MHz)
Figure 5-10 Impedance characteristics of ferrite beads with improved
high frequency characteristics by reducing EPC
5.4 Characteristics for a combination of capacitors and inductors
When inductors are used for power supplies, often capacitors are used in combination.
Therefore, we will introduce characteristics of LC filters where capacitors and inductors
are used in combination.
Ideal forms of an LC filter’s insertion loss characteristics are shown in Figure 5-11.
When the measurement system impedance is fixed and the ratio between L and C are
established appropriately, we can get frequency characteristics with 20dB/dec. slope per
element. 9)
Construction and frequency characteristics and LC filters
With One Element
With Two Elements
With Three Elements
Inductor
Capacitor
120
Insertion Loss (dB)
Insertion Loss (dB)
100
60
80
100
0.1
1
10
Frequency
100
120
40
60
.
80
40
.
ec
60
20
c
/de
ec.
0
20
/d
40
0
T-Type
dB
60
20d
B/d
20
π-Section
L-Type
dB
40
Insertion Loss (dB)
0
L-Type
80
100
0.1
1
10
Frequency
100
120
0.1
1
10
100
Frequency
Figure 5-11 Insertion loss characteristics of LC filters
Generally, impedance is not constant for power supply circuits, so that it is difficult to
match the ratio between L and C for all frequencies. Also, a capacitor is always placed
between an inductor and an IC as described in Section 5-1; the filter becomes L type or
π-section. Therefore, examples of the cases where the ratio between L and C is off from the
measurement system impedance are shown in Figure 5-12.
38
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
1µH
1µH
1µF
1µF
10nF
π-Section
L-Type
0
0
Inductor 1µH only
Inductor 1µH only
20
insertion loss (dB)
20
insertion loss (dB)
Jul.20,2010
40
60
Capacitor 1µF only
80
L-Type Filter
1µF+1µH
100
40
Capacitor
10nF only
Capacitor
1µF only
60
80
π-Section Filter
1µF+1µH+10nF
100
120
120
0.1
1
10
frequency (MHz)
100
1000
0.1
1
10
frequency (MHz)
100
1000
Figure 5-12 Examples of insertion loss characteristics of L-type and
π-section filters (calculated values)
In addition, characteristics of a capacitor and an inductor comprising an LC filter are
also indicated in Figure 5-12. When the ratio of L and C does not match like this, the slope
of the curve representing the attenuation region is not constant, having a point of
inflection. Meanwhile, as we have described previously, capacitors and inductors do not
operate ideally under high frequency. Therefore, in order to predict actual frequency
characteristics, this influence must be taken into consideration as well.
Results of calculating insertion loss, when an L-type filter is formed in combination of an
MLCC and a ferrite bead, are shown in Figure 5-13. As shown in this graph, the frequency
characteristics of an actual LC filter differ from Figure 5-11. As an overall tendency, total
insertion loss can be made larger by combining a ferrite bead.
0
Ferrite bead only
insertion loss (dB)
20
(BLM21PG221)
Capacitor only
(Assuming 1µF ESL=2nH)
40
60
L-Type Filter
(Ferrite bead + capacitor)
80
100
1
10
100
frequency (MHz)
1000
Figure 5-13 Characteristics of L-type filter using a ferrite bead (calculated values)
Figure 5-14 shows an example of confirming the fluctuation in noise control effect when
combining a capacitor and a ferrite bead through an experiment. In this example, the
source of the noise is an IC operating at 4MHz, and the noise passing through the
decoupling circuit is observed with voltage fluctuation and spectrum. The first row shows a
measurement without a ferrite bead (using a 1µF decoupling capacitor), the second row is a
measurement of an L-type filter by adding a ferrite bead to the first row, and the third row
39
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
is a measurement of a π-section filter by adding a 10µF capacitor to the second row. In the
second and third row examples, where a ferrite bead is used, it is confirmed that both
voltage fluctuation and spectrum are greatly improved.
Noise Generation
Observing IC power supply
noise switched at 4MHz
(Observed 3V power supply
with AC coupling
Decoupling
Circuit
Measurement
Connecting to power supply
Waveform
Spectrum
60
100
MLCC 1µF
Capacitor is not enough for noise
suppression, and noise passes
through under high frequency
50
level (mV)
noise level (dBuV)
50
1µF
0
-50
40
30
20
-100
-300
-200
-100
0
100
200
10
300
0
50
100
150
200
frequency (MHz)
250
0
50
100
150
200
frequency (MHz)
250
0
50
100
250
time (ns)
Ferrite Bead
MLCC 1µF+Ferrite Bead
(BLM18PG221 Rated Current 1.4A) 1µF
Noise level improved by about 10dB
overall
60
100
50
noise level (dBuV)
level (mV)
50
BLM18PG221
(220Ω@100MHz)
0
-50
40
30
20
-100
-300
-200
-100
0
100
200
10
300
time (ns)
Ferrite Bead
300
60
100
10µF
noise level (dBuV)
50
1µF
50
level (mV)
MLCC 1µF + Ferrite Bead + 10µF
(BLM18PG221 Rated Current 1.4A)
Noise level is improved further by
forming a π-section filter
300
0
-50
BLM18PG221
(220Ω@100MHz)
30
20
-100
-300
40
10
-200
-100
0
100
200
300
time (ns)
150
200
300
frequency (MHz)
Figure 5-14 An example of noise suppression and improvement by ferrite beads
5.5 LC filter
As described in Chapter 2, the insertion loss of a capacitor tends to become smaller for
frequencies above 10MHz. In order to make an improvement, overall insertion loss is made
larger by combining a capacitor with small capacitance and taking advantage of
self-resonance. However, in this case, antiresonance occurs between two capacitors,
diminishing the improvement effect. In order to avoid such a problem, it is effective to use
an LC filter where a ferrite bead is used in combination in place of a capacitor with small
capacitance. 11)
Figure 5-15 shows an example of this LC filter, and Figure 5-16 shows calculation results
of frequency characteristics when combining this filter with a large-capacitance capacitor.
Ferrite Bead
Feed-Through
Capacitor
Equivalent Circuit
Ferrite Bead
Input/Output
Electrode Ground
Input/Output
Electrode Electrode
Figure 5-15 LC filter NFE31 for power supplies
40
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Capacitor
1µF+2200pF
Capacitor
1µF only
(1608-size MLCC)
Jul.20,2010
Capacitor
1µF only
(1608-sizeMLCC)
LC filter only
Capacitor
2200pF only
(1608-size MLCC)
NFE31PT222(2200pF)
Capacitor 1µF+LC filter
(a) When combining two capacitors
(b) When combining a capacitor and a LC filter
Figure 5-16 Frequency characteristics of a power supply filter (calculated values)
Figure 5-15 shows a component with a combination between a feed-through capacitor
and a ferrite bead with excellent high-frequency characteristics assuming its use with a
large capacitance capacitor. Input and output electrodes on both ends are connected with a
metal molded terminal, making the serial resistance extremely small while realizing 6A
rated current.
As with the calculation results shown in figure 5-16 (a), when two capacitors are simply
combined, their insertion loss may be smaller than one capacitor at the frequency where
antiresonance occurs. Figure 5-16 (b) shows that when using an LC filter combined with a
ferrite bead in place of a small-capacitance capacitor, this kind of problem can be prevented,
making it possible to form a filter circuit effective at high frequency near 1GHz. We can
assume that using an LC filter has the effect of suppressing antiresonance with the
component resistance of the built-in ferrite bead.
Figure 5-17 is an example of confirming fluctuation in the noise suppression effect when
combining an LC filter with a capacitor through an experiment. As with Figure 5-14, an IC
operating at 4MHz is used as the noise source, and the noise passing through the
decoupling circuit is observed with voltage fluctuation and spectrum. (The spectrum
analyzer is with higher sensitivity to allow for observation of more minute noise.)
The first row shows a case where a 1µF MLCC is combined with a 2200pF capacitor (a
6mm wiring is inserted in the middle to generate strong antiresonance). In this case,
antiresonance between capacitors cause a strong ringing for the power supply voltage, and
a strong spectrum is observed at this frequency.
The second row shows a case where an LC filter with 2200pF capacitance replaces the
2200pF MLCC in the same location, and a 10µF MLCC is added to the second row for the
third row. In either case, significant improvement is made for both voltage fluctuation and
spectrum. Spectrum, especially, is kept at an extremely low level, confirming the fact that
an LC filter with a feed-through capacitor has an excellent suppression effect for high
frequency noise.
41
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Noise Generation
Observing IC power supply
noise switched at 4MHz
(Observed 3V power supply
with AC coupling)
Decoupling
Circuit
Connecting to
power supply
Measurement
Waveform
60
100
6mm apart
50
1µF
2200pF
noise level (dBuV)
50
level (mV)
MLCC 1µF+MLCC2200pF
If capacitors are simply
combined, there are frequencies
where noise may not attenuate
due to antiresonance
(Around 60MHz in this case)
Spectrum
(Since insertion loss of the filter is large,
RBW is set to be 10kHz to lower the
noise floor more than other data)
0
-50
40
30
20
10
-100
-300
-200
-100
0
100
200
0
300
0
50
100
150
200
frequency (MHz)
250
300
50
100
150
200
frequency (MHz)
250
300
time (ns)
100
60
50
NFE31P222
(2200pF)
noise level (dBuV)
50
level (mV)
MLCC 1µF+LC filter (2200pF)
(NFE31PT222Z Rated Current 6A)
1µF
Antiresonance is eliminated and good
noise suppression characteristics over
a wide frequency range
0
-50
40
30
20
10
-100
-300
-200
-100
0
100
200
0
300
0
time (ns)
100
70
60
1µF
Since capacitors are used in several locations in
actual circuits, so this is a common construction
10µF
noise level (dBuV)
50
level (mV)
MLCC 1µF+LC filter +10µF
(NFE31PT222Z Rated Current 6A)
Noise suppression effect improves
further by mounting between two
capacitors
NFE31P222
(2200pF)
0
-50
40
30
20
10
-100
-300
50
-200
-100
0
100
200
300
time (ns)
0
0
50
100
150
200
frequency (MHz)
250
300
Figure 5-17 An example of improving noise suppression using an LC filter
On the other hand, the insertion loss characteristics described in Figure 5-16, etc. are for
the case where the measurement impedance is 50Ω, and impedance of power supplies are
often smaller than that. Therefore, we will show the result of fluctuation in characteristics
as the measurement impedance is varied in Figure 5-18. This is converted from the
calculation results in Figure 5-16.
While with simply combining two capacitors as in Figure 5-18 (a), insertion loss tends to
decrease greatly in low-impedance circuit, there is a tendency for a certain level of effect to
be maintained even for a low-impedance circuit in Figure 5-17 (b) with an LC filter
combined. This is considered to be the case because while the capacitor’s effect decreases in
the low-impedance circuit, the effect of the inductor increases, therefore two effects are
cancelled together when combined with an LC filter. Therefore, the decoupling circuit for
power supplies combined with an inductor is expected to show a noise suppression effect
relatively stable against fluctuation in power supply impedance.
42
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
1µF
2200pF
1µF
NFE31P222
(2200pF)
0.5Ω
2Ω
10Ω
50Ω
(Normally measurement
is taken at 50Ω)
insertion loss (dB)
0.5Ω
insertion loss (dB)
Jul.20,2010
2Ω
50Ω
(Normally measurement is
taken at 50Ω)
frequency (MHz)
frequency (MHz)
(a) When combining two capacitors
10Ω
(b) When combining a capacitor and an LC filter
Figure 5-18 Characteristics when measurement impedance is varied (calculated values)
5.6 Considerations when using an inductor for a power supply
Although the noise suppression effect of an inductor basically increases as impedance
increases (inductance increases), as we have described thus far, impedance stops
increasing under high frequency due to the influence of EPC. Also for the sake of noise
suppression or suppression of resonance, generally, the larger the component resistance in
impedance, the better the effect will become. We need to select a component for noise
suppression from this kind of viewpoint.
On the other hand, from the viewpoint of providing low-impedance power supplies, since
the use of an inductor works to increase impedance, we will need to compensate with a
capacitor having enough capacitance between the inductor and the IC (guidelines for
determining the necessary capacitance is indicated in equation 2-2). Therefore, it is not
recommended to use an inductor with unnecessarily large impedance.
When using an inductor for a power supply, we need to select a component considering
saturation due to direct current resistance and current in addition.
5.6.1 Influence of direct current resistance
Since direct current resistance of an inductor can generally cause energy loss and heat
generation, it is desirable for it to be small when using an inductor for a power supply.
Aside from this, voltage fluctuation due to voltage drop can be problematic.
Since direct current resistance can decrease voltage for circuits where constant current
flows through, power supply voltage must be maintained high to compensate for that.
When there is large current fluctuation, voltage ripple may occur as the following:
∆V ripple = Rdc ⋅ ∆I ripple
(5-1)
Here, ΔVripple, Rdc and ΔIripple represent voltage ripple, direct current resistance of a
component, and current fluctuation respectively. For example, when current fluctuation of
1A is applied to a component with about 100mΩ Rdc, 100mV voltage ripple occurs.
Therefore, for a low-voltage power supply with small allowable voltage ripple, we need to
choose a component with small direct current resistance.
43
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
5.6.2 Influence of current saturation
In general, when magnetic flux density of a ferromagnetic material, such as ferrites,
reaches saturation, their magnetic permeability tends to become smaller. Therefore, we
need to focus on the fact that when large current flows, inductance and impedance
decrease for inductors with magnetic cores. In order to confirm this influence, we need to
observe the noise level while current is large.
5.7 Lineup of inductors suitable for power supplies
Main examples of ferrite beads and choke coils commercialized at Murata are listed
below.
Please refer to our catalog for their details.
5.7.1 Ferrite beads
Models to keep direct current resistance down are in the lineup for ferrite beads for
power supplies.
●0603 size BLM03P series
0.5±0.05
0.25±0.1
0.3±0.03
0.3±0.03
0.6-±0.03
●1005 size BLM15P series
0.15±0.05
1.0±0.05
0.5±0.05
: Electrode
(in mm)
●1005 high frequency improvement
(by structure) BLM15E
: Electrode
(in mm)
●1005 high frequency improvement
(by structure and material) BLM15G
0.25±0.1
1.0±0.05
0.5±0.05
0.5±0.05
0.5±0.05
0.25±0.1
1.0±0.05
0.5±0.05
: Electrode
: Electrode
(in mm)
(in mm)
●1608 size BLM18P series
●1608 for 6A 600Ω BLM18K series
0.4±0.2
1.6±0.15
0.8±0.15
T
0.8±0.15
1.6±0.15
0.4±0.2
0.8±0.15
: Electrode
(in mm)
44
: Electrode
(in mm)
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
●1608 6A high performance BLM18S series
●1608 high frequency improvement
(by structure) BLM18E
1.6±0.15
0.8±0.15
0.8±0.15
T
0.5±0.15
1.6±0.15
Jul.20,2010
0.4±0.2
0.4±0.2
: Electrode
: Electrode
(in mm)
●1608 high frequency improvement
(by structure and material) BLM18G
(in mm)
●2012 size BLM21P series
0.5±0.2
0.8±0.1
0.85±0.2
0.35±0.15
1.6±0.1
0.8±0.1
2.0±0.2
1.25±0.2
EIA CODE : 0805
: Electrode
: Electrode
(in mm)
●3216 size BLM31P series
(in mm)
●4516 size BLM41P series
0.7±0.3
3.2±0.2
1.6±0.2
1.1±0.2
0.7±0.3
1.6±0.2
4.5±0.2
: Electrode
(in mm)
1.6±0.2
: Electrode
(in mm)
45
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
5.7.2 Choke coils
This product group reduces direct current resistance.
●3216 size LQH31C series
●3225 size LQ32C series
2.3±0.2
2.5±0.2
1.8±0.2
1.6±0.2
3.2±0.3
(in mm)
0.7min. 0.7min. 0.7min.
2.5±0.2
2.0±0.2
1.6±0.2
2.5±0.2
3.2±0.3
0.9±0.3 1.3±0.2 0.9±0.3
(in mm)
●3225 size low-direct current resistant type LQH32C_33, 53 series
2.5±0.2
1.55±0.15
2.5±0.2
A
A
3.2±0.3
2.5±0.2
A : 2.8 max.
0.9±0.3 1.3±0.2 0.9±0.3
(in mm)
●4532 size LQH43C series
●5750 size LQH55D series
3.2±0.2
5.0±0.3
3.2±0.2
4.5±0.3
(in mm)
1.0min. 1.0min. 1.0min.
●6363 size LQH66S series
6.3±0.3
4.7±0.3
6.3±0.3
6.3±0.3
1.3 1.7 1.3
min. min. min.
46
5.0±0.3
4.7±0.3
2.6±0.2
3.6±0.2
(in mm)
5.7±0.3
1.3 1.7 1.3
min. min. min.
(in mm)
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
5.8 Lineup of LC filters suitable for power supplies
The NFE series in combination of ferrite beads and feed-through capacitor with excellent
high frequency characteristics, and the BNX series combined with grounds are available.
Generally, the BNX series is used mainly for a power supply input part such as power
supply connectors, not for decoupling circuits described in this manual.
●6816 size NFE61P series
1.0±0.2
0.7±0.2
(1)
(2)
(3)
0.7±0.2
2.6±0.3
0.7±0.2
(1)
(2)
(3)
1.6±0.15
0.7±0.2
1.6±0.3
●3216 size NFE31P series
6.8±0.5
3.2±0.35
1.6±0.3
1.6±0.15
: Electrode
: Electrode
(in mm)
(in mm)
●BNX series
12.1±0.2
9.1±0.2
1.0±0.3
(4)
(4)
7.0±0.2 1.55±0.2
2.5±0.2
(4)
(4) (2)
L3
B
(3) PSG
1.6±0.3
1.05±0.2
BNX022
1.3±0.2
(1)
(1)
L2
0.3±0.1
(3)
4.2±0.3 (2.45)
(1) (3)
3.1±0.2
(3)
L1
(2)
0.3±0.1
(1)
C2
CB (2)
C1
CG (4)
(1)-(4): Terminal Number
PSG: Power Supply Ground
CG: Circuit Ground
CB: Circuit+B
(4)
(2)
: Electrode
1.55±0.2
(in mm)
47
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
6. Suppressing power supply voltage fluctuation
Thus far in previous chapters, we have mainly described noise suppression
performance/insertion loss characteristics of decoupling circuits for power supplies.
Starting in this chapter, we will mainly describe power supply impedance and voltage
fluctuation from the viewpoint of current supply performance of power supplies.
Power supply voltage must be constant in order for a digital IC to operate correctly. As
shown in Figure 6-1, when power supply current fluctuates from the IC operation, power
supply voltage may fluctuate due to the influence by the wiring between the power supply
module and IC, as well as the influence by the response characteristics of the power supply
module itself, causing malfunction in the IC or affecting peripheral circuits with noise. It
can also reduce operating speed and signal integrity in some cases.
In order to prevent this voltage fluctuation, a capacitor called a decoupling capacitor,
which temporarily supplies current in the vicinity of the IC, is used. This ability of a
decoupling capacitor to supply current to a capacitor is sometimes called charge supply
capability.
In this chapter, we will describe the mechanism of suppressing voltage fluctuation by a
decoupling capacitor using a circuit to go through simple current fluctuation as a model.
The performance necessary for a capacitor to keep down voltage fluctuation is also
introduced.
Current
Fluctuation
Voltage
Fluctuation
V
PCB
Power Supply Module
Power Supply Terminal
Power Supply Impedance
I
t
t
PDN
IC
Temporary supply
of current (charge
supply ability)
Decoupling
Capacitor
Response delay by
the power supply
Cable Inductance
Pattern Inductance
Figure 6-1 Current and voltage fluctuations
6.1 Relationship between power supply impedance and voltage fluctuation
The size of the voltage fluctuation occurring at the power supply, when a power supply
terminal for the IC is connected to the power distribution network (PDN), and power
supply current of the IC fluctuate as in Figure 6-2, can be expressed by the equation below:
∆V = ∆I ⋅ Z P
(6-1)
In this equation, ∆V, ∆I and ZP correspond with voltage fluctuation (V), current
fluctuation (A) and power supply impedance of PDN (Ω) respectively. This ∆V is voltage
observed at the power supply terminal of the IC and ZP is likewise impedance observed at
the power supply terminal.
48
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Power Supply Voltage
(Voltage Fluctuation ∆V)
Ground
Power Supply
Wiring (PDN)
Power Supply Current
(Current Fluctuation ∆I)
IC
PDN
Digital IC
Power
Supply
Power
Supply
Voltage
Measurement
Power Supply Impedance
ZP
Power Supply Terminal
Figure 6-2 Measurement position of voltage fluctuation
An example of voltage waveform when power supply voltage fluctuates is shown in
Figure 6-3. In this case it is understood that the power supply impedance is resistivity
without frequency characteristics. Figure 6-3 (a) shows a case where the current simply
varies stepwise, and Figure 6-3 (b) fluctuates in a slightly more complex pattern. In either
case, voltage fluctuation appears in correspondence with current fluctuation (when current
increases voltage decreases).
a (A)
Current
Waveform
Voltage
Waveform
∆I
0 (A)
v (V)
∆V
v-Zp・a (V)
t
(a) When current fluctuates step-wise
a (A)
Current
Waveform
0 (A)
Voltage
Waveform
v (V)
t
(b) When current fluctuate in a complex manner
Figure 6-3 An example of voltage fluctuation in relation with current variation
6.2 Voltage fluctuation when a capacitor is present
In an actual PDN, impedance would have reactance as well as frequency characteristics
instead of resistivity due to the inductance from the decoupling capacitor and wiring, and
49
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
consequently the voltage waveform would appear differently from the current waveform
showing complex changes. An example of the measurement results is shown in Figure 6-4.
3.1
volt (V)
3.1
3.0
3.0
2.9
0
20
40
60
80
100
time (ns)
Figure 6-4 Measurement example of digital IC power supply voltage (3V line)
Therefore, we will consider a voltage waveform where current fluctuates stepwise, in
order to understand the basic voltage fluctuation when using a decoupling capacitor.
Let us consider the voltage fluctuation when current flows through a power supply using
a simple model illustrated in Figure 6-5. For the purpose of simplicity, let us assume there
are no peripheral elements including wiring inductance, and PDN consists of only one
capacitor. Voltage fluctuation is calculated with a circuit simulator then, assuming that the
power supply current of the IC fluctuated stepwise as shown in 6-3 (a). We set power
supply voltage to be 3V, power supply impedance to be 0.5Ω, current amplitude to be 1A,
launching time to be 10ns and pulse width to be 1µs.
PDN
1A
0.5Ω
Cap
Power Supply
3V
ESL
ESR
0A
Semiconductor
Voltage
Figure 6-5 Voltage fluctuation calculation model
Calculation results are shown in Figure 6-6. In order to check the trend, three types of
capacitors, 2µF, 5µF, and 10µF were designated. Each capacitor is understood to have
10nH ESL (includes mounting inductance ESLPCB assuming the use of an electrolytic
capacitor with lead) and 50mΩ ESR.
The graph on the left of Figure 6-6 is voltage waveform, and current waveform is
indicated above for a comparison purpose. The graph on the right of Figure 6-6 indicates
frequency characteristics of the capacitor impedance.
For the voltage waveform, a spike is observed at the launching stage, and the voltage
decreases gradually due to the discharge characteristics of the capacitor. After that, we can
observe a gradual increase in voltage as the current pulse ends and the capacitor begins to
recharge. Since charge and discharge time of a capacitor is relative to the capacitance, the
slope of the curve changes according to the capacitor’s capacitance, causing the range of
50
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
voltage fluctuation to change as well.
As we can see from the result of Figure 6-6, when capacitance is larger, voltage
fluctuation from charge and discharge will become smaller. Also, the scale of fluctuation in
this section corresponds with the size of impedance for the capacitive zone of the low
frequency side for the impedance characteristics on the right. When impedance is smaller
(the lower curve in the graph), the range of voltage fluctuation becomes smaller. In order to
reduce voltage fluctuation by this capacitor charge and discharge, it is necessary to use a
capacitor with a large enough capacitance.
2
Current
-1
2.8
-4
2.6
-7
2µF
5µF
2.4
0
impedance (mohm)
3
10000
2µF
10µF
current (A)
voltage (V)
3.2
1000
100
5µF
10
-10
2
4
6
8
time
time (µs)
(us)
10µF
1
0.001
10
0.01
0.1
1
frequency (MHz)
10
100
1000
Amplitude of voltage fluctuation changes
according to capacitance
(Shows in low-frequency impedance)
Figure 6-6 An example of the voltage waveform when using a capacitor (calculated values)
In order to observe the spike at the launching time for the calculation results of Figure
6-6 in more detail, Figure 6-7 shows an expanded time axis for a 10µF capacitor.
Calculation results for an ideal capacitor without ESR or ESL is also graphed for
comparison purposes. While no spike occurs for an ideal capacitor, a spike occurs for the
capacitor including ESR and ESL at the launching time. The size of the spike tends to
correspond mainly with ESL. In the impedance characteristics on the right side, the
capacitor corresponds with the high frequency impedance where the capacitor becomes
inducible (small impedance: the curve toward the bottom has a smaller spike).
When this spike is deep, voltage fluctuation may not stay within the rated power supply
voltage even if a capacitor with a sufficient capacitance is used. If this spike causes a
problem, ELS must be reduced.
Ideal Capacitor
Current
2
-1
Level changes mainly
relative to ESR
2.8
2.6
2.4
-4
Near-Real Capacitor
-7
Peak changes mainly relative to ESL
0.9
1
1.1
time
time (µs)
(us)
-10
1.2
1.3
current (A)
voltage (V)
3
10000
impedance (mohm)
3.2
Near-Real Capacitor
1000
100
Ideal Capacitor
10
1
0.001
0.01
0.1
1
10
frequency (MHz)
100
1000
Spikes occur at the launching part of the current from
increase in impedance in high-frequency region
Figure 6-7 Launching part of the voltage waveform when using a capacitor
Since the size of this spike also corresponds with the size of current fluctuation (dI/dt) in
51
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
addition to capacitor’s ESL, the spike may not be very evident when voltage fluctuation is
relatively small. This calculation is based on the launching time of 10ns (dI/dt = 1 x 108
A/s).
6.3 Suppressing spikes with parallel capacitors
In order to suppress the spike from the current fluctuation at launching time, relatively
small-capacitance capacitors with excellent frequency characteristics can be used in
parallel. Figure 6-8 shows an example of a waveform when capacitors are used in parallel.
In this case, in addition to a 10µF capacitor, a 1µF MLCC (ESL is 2nH taking ESLPCB into
consideration and ESR is 10mΩ) is supposed to be used. The calculation results of the
figure show that the depth of the spike is halved by adding those capacitors. This effect
corresponds with the fact that the high frequency impedance is reduced in terms of the
impedance characteristics shown in the right side of the figure.
We can also add capacitors with smaller capacitance to further suppress this spike.
0.5Ω
Current
-1
2.8
-4
Spikes decrease when a
1µF capacitor is added
2.6
-7
Near-Real Capacitor
2.4
0.9
1
1.1
time (µs)
time
(us)
1.3
2nH
50mΩ
10mΩ
Pw=1µs
Tr=10ns
Near-Real Capacitor
1000
100
When adding a
1µF capacitor
10
Ideal Capacitor
-10
1.2
1µF
10nH
10000
current (A)
voltage (V)
3
2
impedance (mohm)
Ideal Capacitor
3.2
10µF
1
0.001
0.01
0.1
1
frequency (MHz)
10
100
1000
As impedance decreases in high-frequency
region, spike level becomes smaller
Figure 6-8 Change in waveform when capacitors connected in parallel are added
As seen above, connecting capacitors in parallel has a certain effect for suppressing the
spike. However, antiresonance between neighboring capacitors may cause a problem, as
described in Chapter 3. Figure 6-9 shows an example of intentionally creating such a
condition.
In this case, the calculation results for using a 0.1µF capacitor in place of the 1µF
capacitor mentioned above are shown. Although the size of the spike is unchanged from the
1µF capacitor, a strong surge in the waveform passed the spike (ringing with an
approximately 0.2µs frequency). The frequency of this surge corresponds with the
antiresonance frequency over the impedance curve. When a surge is large like this, we
must take caution in avoiding going over the acceptable limit for power supply voltage
fluctuation.
52
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
0.5Ω
10µF
0.1µF
10nH
2nH
50mΩ
10mΩ
Jul.20,2010
Pw=1µs
Tr=10ns
Ideal Capacitor
Current
3.2
Near-Real Capacitor
2
10000
Antiresonance
-4
2.6
Spikes are reduced when a 0.1µF
capacitor is added, but surge occurs -7
Near-Real Capacitor
2.4
0.9
1
current (A)
voltage (V)
2.8
impedance (mohm)
-1
3
1000
100
Ideal Capacitor
1
0.001
-10
1.1
time (µs)
(us)
1.2
When a 0.1µF
capacitor is added
10
0.01
0.1
1.3
1
frequency (MHz)
10
100
1000
Surge in relation to antiresonance frequency
appears in the waveform
Figure 6-9 When parallel capacitors are resonating
6.4 Suppressing spikes with low-ESL capacitors
Using a capacitor with small ESL is another way to keep the spike small. Figure 6-10
shows the results of the calculation. We will assume that the low-ESL capacitor introduced
in Chapter 4 is used, and capacitance is set to be 10µF, ESL is set to be 0.2nH, and ESR is
set to be 50mΩ. As we can see from this figure, under these conditions, the spike is almost
totally eliminated and ringing in the case of combined capacitors is not observed either.
Therefore, we can see that low-ESL capacitors are advantageous for suppressing voltage
fluctuation.
0.5Ω
10µF
Pw=1µs
Tr=10ns
0.2nH
50mΩ
Ideal Capacitor
3.2
Current
2
10000
2.8
-4
When exchanging with a low-ESL
capacitor (Spikes are almost eliminated)
2.6
-7
current (A)
voltage (V)
-1
impedance (mohm)
Near-Real Capacitor
3
2.4
-10
1
1.1
time (µs)
(us)
100
10
Low-ESL Capacitor
Ideal Capacitor
Near-Real Capacitor
0.9
1000
1.2
1.3
1
0.001
0.01
0.1
1
10
frequency (MHz)
100
1000
Impedance decreases in high-frequency region
overall and spikes are almost eliminated
Figure 6-10 When using a capacitor with small ESL
6.5 Voltage fluctuation when the pulse width is wide
When the pulse width of the current is wide, we will need to wait for the power supply
response, since a capacitor alone cannot maintain the voltage. In Figure 6-11, delay in
53
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
power supply response is expressed as inductance LPowerDelay, to simply simulate voltage
fluctuation when the pulse width is long. The calculation results in this case are indicated
in Figure 6-12.
LPowerDelay
0.1µH
PDN
0.1Ω
1A
Cap
Power
Supply
3V
0A
ESL
ESR
Semiconductor
Voltage
Figure 6-11 Simulation circuit when pulse width is wide
Current
Power Supply Impedance
2
2.8
-1
-4
10µF
2µF
2.6
-7
impedance (mohm)
50µF
3
10000
current (A)
voltage (V)
3.2
1000
100
10
50µF
Spike from capacitor ESL
2.4
-10
0
2
4
6
time (µs)
time
(us)
8
1
0.001
0.01
10µF 2µF
0.1
10 appears at a frequency
Vibration
where impedance is high
1
frequency (MHz)
10
100
1000
Figure 6-12 Calculation results when pulse width is wide
From Figure 6-12, we see that there is a possibility that more low frequency surge is
observed over the voltage waveform depending on the capacitance of a capacitor. We can
assume that there is a form of resonance occurring between the response time of the power
supply and the capacitance of the capacitor. The frequency of this surge corresponds with
the frequency with high impedance for the impedance characteristics on the right in
Figure 6-12.
As we can see from the result of comparing capacitors in Figure 6-12, this surge is more
evident when the capacitance is relatively small (2µF in the figure). Therefore, we will
consider the necessary capacitance for the capacitor to eliminate the surge.
When we consider the circuit in Figure 6-11 as a type of RLC serial resonance circuit,
damping condition for the resonance circuit is as the following: 13)
C ≥
4LPowerDelay
R2
(6-2)
In this case, C represents the capacitance of the capacitor and R represents output
resistance of the power supply (for the purpose of simplicity, we have neglected the
influence of ESR and ESL). The possibility for the surge to occur becomes small if we use a
capacitance larger than C. For example, if we apply the conditions indicated in Figure 6-12
to equation (6-2), the necessary capacitance for the capacitor becomes approximately 40µF
54
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
(when we look at the case with 50µF capacitor, the surge is completely eliminated).
We have assumed that the main cause of delay in the power supply response is the
inductance of the wiring from the power supply module to the IC, and expressed
characteristics of the power supply as LPowerDelay. In some cases we need to take delay in
response of the power supply module itself. In such a case, we will let response time be
TPowerDelay(s) and we will make a model assuming LPowerDelay to be approximately
LPowerDelay = R ⋅TPowerDelay
(6-3)
from the time constant of the RL serial circuit.
In a real-life situation, since it is not plausible to express the response characteristics of
a power supply as inductance, the value calculated with the method above is purely an
estimate. Also, with a smoothing capacitor used for the power supply output, we cannot
apply the method above since we need to consider totally different elements such as power
supply capacitance and output voltage range.
55
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
7. Location of a capacitor for suppressing power supply impedance
In Chapter 6, we described the relationship between voltage fluctuation and power
supply impedance, as well as voltage fluctuation waveform when using a decoupling
capacitor. However, we did not touch on the influence from capacitor mount wiring. In
order to suppress voltage fluctuation in an IC, impedance relative to the power supply
terminal of an IC must be made small. However, normally there is some kind of printed
wiring between a power supply terminal and a capacitor. Since inductance of this wiring is
significant compared with the capacitor’s ESL, it is necessary to make the wiring
inductance small in order to make impedance relative to the power supply terminal small.
Inductance of this wiring is influenced by a pattern configuration or the distance to the
capacitor. In this chapter, we will describe printed circuit design to reduce inductance for
the wiring connecting an IC and a capacitor, in order to suppress power supply impedance
below a certain value under high frequency.
7.1 Power supply impedance relative to an IC
Although power supply wiring between an IC and a capacitor does not have a set
configuration, making it difficult to make a model, we will assume that it can be expressed
as a micro strip line (MSL) for the purpose of discussion. Let us consider impedance of
PDN with several capacitors attached as illustrated in Figure 7-1.
Let us assume that capacitors are laid out hierarchically in order of 10µF, 2.2µF and
0.47µF (in Figure 7-1, the size of the illustrated capacitors corresponds with their
capacitance). A small-size and small-capacitance capacitor is located right near the IC and
a large-size and large capacitance capacitor is located relatively far from the IC.
Figure 7-2 shows the calculation results of impedance at PDN relative to IC power
supply terminals A, B, and C in such a model. The blue line represents impedance where
all capacitors are connected to PDN in Figure 7-1, and the red line represents impedance
with only the capacitor nearest to the power supply terminal connected.
From the results of Figure 7-2, we can tell that the power supply impedance is mainly
formed from the nearest capacitor when frequency is over 10MHz. This signifies that
impedance in a high-frequency zone is dominated mainly by inductance, and influence
from relatively distant capacitors can be negligible due to the increase in effective ESL
from the wiring inductance. (In this diagram, the PDN as a whole and the closest capacitor
show a relatively large difference for point A. This is considered to be due to the fact that
the nearest capacitor and the rest of the PDN are connected parallel to each other left and
right.)
Since in the high-frequency zone impedance of the nearest capacitor becomes dominant,
in order to reduce impedance of a PDN below a certain value, we only have to consider the
nearest capacitor and the wiring it is connected to. We will focus on wiring design up to the
nearest capacitor with such a premise.
56
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
A
Jul.20,2010
Via to Ground
B
7
IC Ground Terminal
2
5
12
20
15
7
Power
Supply
0.1Ω
Power Supply Terminal
10
3
Capacitor
10µF
Wiring
Width
5mm
C
3mm
2mm
1mm
ESL=1nH
ESR=3mΩ
2.2µF
ESL=1nH
ESR=10mΩ
0.47µF ESL=1nH
ESR=15mΩ
(Dielectric material thickness 0.8mm
Dielectric Constant 4.5)
Figure 7-1 Modeling diagram of power supply impedance calculation
Point B
Only the nearest
capacitor is connected
Entire
PDN is
connected
impedance (ohm)
impedance (ohm)
Point A
Only the nearest
capacitor is connected
Entire
PDN is
connected
frequency (MHz)
frequency (MHz)
(a) Point A
(b) Point B
impedance (ohm)
Point C
Only the nearest
capacitor is connected
Entire
PDN is
connected
frequency (MHz)
(c) Point C
Figure 7-2 Calculation results of power supply impedance
7.2 Simple estimation of power supply impedance relative to an IC
Assuming that the wiring from the IC’s power supply terminal up to the nearest
capacitor may be expressed as an MSL, by modeling it as in Figure 7-3, impedance of this
capacitor relative to the power supply terminal, ZPowerTerminal can be expressed with the
following equation:
Z PowerTerminal = Z cap + Z line
(7-1)
57
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
In this equation, Zcap is impedance of the capacitor and Zline is impedance of the wiring
leading to the capacitor. Zcap includes impedance of the pad for mounting the capacitor and
a via.
Via to Ground
Wiring Impedance Zline
Ground Terminal
Capacitor
Impedance Zcap
IC
Power Supply
Wiring
Power Supply Terminal
Impedance
ZPowerTerminal
Wiring Cross-Section Structure
l
w
h
Ground Layer
Figure 7-3 Wiring model to the nearest capacitor
Zline, which is impedance for the wiring, can be considered short-circuited at the tip
(since a capacitor is connected to the model illustrated in Figure 7-3, it can be considered
short-circuited under high frequency); if the length of wiring is sufficiently short compared
against the wavelength of the corresponding frequency, the inductance can be
approximated by inductance. We will let inductance of the wiring be Lline. Also, in the
high-frequency zone exceeding self-resonant frequency, the impedance of capacitor Zcap is
formed by the ESL of the capacitor, ESLcap. Therefore, impedance relative to the power
supply terminal of an IC, ZPowerTerminal, can be expressed as the following equation:
Z PowerTerminal = Z cap + Z line ≅ j 2πf ( ESLcap + Lline )
(7-2)
We can use inductance per unit length of MSL multiplied by length l as the value for Lline.
Various approximation equations are proposed based on the inductance per unit length of
MSL correlating with characteristic impedance. 14)
However, when handling a case with
wide wiring width like a power supply, its equation may become complex. Therefore, the
following equation is proposed to very roughly approximate Lline:
Lline
h
= 0 .4 l  
w 
0.6
× 10 − 6
(H)
(7-3)
In this equation, h is the thickness of the dielectric material considering the power
supply wiring as an MSL, w is the width of the wiring, and l is the length of the wiring
(units are all millimeters). By substituting Lline with ESLcap of the capacitor used in the
equation (7-2), we can estimate the impedance relative to the power supply terminal of the
IC in the high-frequency region (where the capacitor becomes inducible). Note that we
need to include inductance from the capacitor mounting pad and via (ESLPCB) in ESLcap
used here.
7.3 Possible range for placing the closest capacitor of an IC
We can calculate by inversion the length of wiring necessary to control power supply
58
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
impedance below a target value, approximating inductance of wiring with a simple
equation in (7-3). Let the target impedance relative to the power supply of the IC be ZT and
the target frequency, the largest frequency necessary to satisfy this impedance is fT@PCB,
the maximum allowable length of wiring lmax is as the following.
As mentioned earlier, impedance relative to the power supply terminal shows
inducibility under high frequency; we will only consider the inductance. We can derive the
largest allowable inductance for the wiring, Lline_max, by substituting ZT for the power
supply impedance ZPowerTerminal and fT for the frequency f in the equation (7-2).
Lline_max ≅
ZT
− ESLcap
2πf T
(7-4)
We can derive the maximum allowable length for the wiring, lmax, by substituting Lline
with this Lline_max in the equation (7-3):
l max = 2.5
Lline _ max
h
 
w 
0.6
× 106 ≅ 0.4
Z T − 2πf T @ PCB ESLcap
h
f T @ PCB  
w 
× 106
0.6
(m)
(7-5)
As shown in figure 7-4, when placing the closest capacitor within this lmax from the
power supply terminal of an IC, we can achieve the target impedance in the high-frequency
zone. We will call this lmax the maximum allowable wiring length. When lmax is large, we
have more flexibility in the location of the capacitor.
impedance (ohm)
PDN Impedance
10
High frequency area of the
target impedance is achieved
1
IC
ZT
0.1
0.01
0.1
1
10
100
[email protected]
1000
Placing a capacitor
within lmax
frequency (MHz)
Figure 7-4 Placing a capacitor within lmax
On the other hand, in terms of the capacitor, lmax can be seen as the capacitor’s effective
range for containing the power supply impedance to less than ZT. As shown in Figure 7-5,
when the power supply terminal of an IC is placed less than lmax from the capacitor, one
capacitor can suppress multiple power supply impedances of the IC at less than ZT. As we
can see from the equation (7-5), this capacitor has a wide effective range since lmax of a
capacitor with small ESLcap becomes large.
When one capacitor covers multiple IC power supplies as in Figure 7-5, current may
become large when the timing of IC operations match, necessitating the change in the
target impedance ZT. Also, when the wiring connecting the capacitor and multiple ICs
overwraps (like two ICs on the right), voltage induced in the overwrapped wiring may
cause noise interference between ICs. When these problems occur, a capacitor should be
59
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
used for every IC.
IC
IC
IC
Placing an IC
within lmax
Figure 7-5 Placing an IC within lmax
If 2πfT ESLcap is larger than ZT in the equation (7-5), a capacitor cannot be used since lmax
will be less than zero. This indicates that the ESL of the capacitor itself is too large,
making it unable to achieve the target impedance ZT even if the wiring is connected with
ideal zero inductance. In such a case, we need to either use a capacitor with small ESL or
use multiple capacitors in parallel to make an equivalence of small ESLcap.
7.4 Guideline for the maximum allowable wiring length, lmax
We can assume from the equation (7-5) that when using a capacitor with small ESLcap,
lmax becomes large, increasing the flexibility of the capacitor location. Also, this lmax should
be influenced by the dimension of intersection for the printed wiring (h and w). Therefore,
in order to confirm this tendency, and show a guideline for wiring design, we have
calculated lmax varying wiring width, thickness of the dielectric material, and mounting
conditions for the capacitor as shown in Figure 7-6. The results are shown between Figures
7-7 and 7-10.
We assume low-ESL capacitors such as 3-terminal capacitor and LW reverse capacitor or
MLCC are used in this case. We set the upper limit frequency, fT@PCB of target impedance
tentatively at 100MHz assuming that the measurement is taken at the power supply
terminal external to an IC. (In an actual situation, the value fT@PCB varies greatly
depending on the IC used.)
From calculation results in Figures 7-7 through 7-10, we can tell that in order to make
lmax large, low-ESL capacitor, thin dielectric substrate, and wide wiring are effective. Also,
the smaller the target impedance is, the larger the fluctuation in lmax tends to be due to
capacitor ESL.
As seen in this section, lmax becomes large with a low-ESL capacitor making capacitor
placement more flexible. Also, the effective range of one capacitor becomes larger, enabling
the use of fewer capacitors to cover a wide area of PDN.
When targeting less than 0.2Ω@100MHz for Figures 7-7 through 7-10, using one MLCC
would cause 2πfT ESLcap to become larger than ZT as described earlier, making it
“impossible to wire”.
In such a case, we need to use multiple capacitors in parallel as
described in Chapter 8, to make ESLcap smaller. Also, the use of a low-ESL capacitor,
introduced in Chapter 4, would be effective in making ESLcap smaller.
The descriptions above presume that the power supply wiring can be treated as MSL and
60
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
target frequency is sufficiently low for the wiring length. Therefore, they would not apply if
a single-sided substrate which cannot be treated as MSL is used or if wiring resonance
occurs due to high frequency.
Capacitor
• 3-Terminal Capacitor (Assuming ESLcap=0.05nH)
• LW Reverse Capacitor (Assuming ESLcap=0.2nH)
• MLCC Best Condition (Assuming ESLcap=0.5nH)
• MLCC Normal Condition (Assuming ESLcap=1nH)
Target Impedance
ZT= 0.1, 0.2, 0.5, 1, 2, 5 (Ω)
Maximum Frequency of Target Impedance
[email protected]=100MHz
IC
Power Supply Terminal
Power Supply Wiring
Wiring CrossSection Structure
w
Calculating maximum allowable
wiring length lmax
w=0.1 to 10 (mm)
h=0.1, 0.2, 0.4, 0.8 (mm)
h
Ground Layer
h=0.1h=0.1 3端子コンデンサ
3-Terminal Capacitor
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
h=0.2h=0.2 3端子コンデンサ
3-Terminal Capacitor
100
100
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
最大
許 容 配Wiring
線 長 lm
ax(m/max
m ) (mm)
Maximum
Allowable
Length
Figure 7-6 Calculation conditions
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5Ω
Less than 0.2Ω
Less than 0. 1Ω
10
1
0.1
h=0.4h=0.4 3端子コンデンサ
3-Terminal Capacitor
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5Ω
Less than 0.2Ω
Less than 0.1 Ω
10
1
1
Wiring配線幅w(mm)
Width w (mm)
10
h=0.8h=0.8 3端子コンデンサ
3-Terminal Capacitor
100
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5Ω
Less than 0.2Ω
Less than 0. 1Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Figure 7-7 3-Terminal Capacitor Best Conditions (ESLcap=0.05nH)
61
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
h=0.4 h=0.4 LW逆転コンデンサ
LW Reverse Capacitor
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
h=0.2
LW Reverse Capacitor
h=0.2 LW逆転コンデンサ
100
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
h=0.1 h=0.1 LW逆転コンデンサ
LW Reverse Capacitor
100
Jul.20,2010
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
h=0.8 LW逆転コンデンサ
h=0.8 LW
Reverse Capacitor
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Figure 7-8 LW Reverse Capacitor Best Conditions (ESLcap=0.2nH)
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring配線幅w(mm)
Width w (mm)
10
h=0.4 MLCC
Best Conditions
h=0.4 MLCC最良条件
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring配線幅w(mm)
Width w (mm)
10
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
h=0.1 MLCC最良条件
h=0.1 MLCC
Best Conditions
100
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
Maximum最大許容配線長lmax(mm)
Allowable Wiring Length /max (mm)
(Cannot be wired when under ZT=0.1Ω)
h=0.2 MLCC
Best Conditions
h=0.2 MLCC最良条件
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
h=0.8 MLCC最良条件
h=0.8 MLCC
Best Conditions
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
Figure 7-9 MLCC Best Conditions (ESLcap=0.5nH)
(Cannot be wired when under ZT=0.2Ω)
62
10
10
C39E.pdf
Note · Please read rating and
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5Ω
Less than 0.2Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring配線幅w(mm)
Width w (mm)
10
h=0.4 MLCC
Normal Conditions
h=0.4 MLCC通常条件
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5Ω
Less than 0.2Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
最大許容配線長lmax(mm)
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
Maximum
Allowable Wiring Length /max (mm)
h=0.1 MLCC
Normal Conditions
h=0.1 MLCC通常条件
100
Maximum
Allowable Wiring Length /max (mm)
最大許容配線長lmax(mm)
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
h=0.2 MLCC
Normal Conditions
h=0.2 MLCC通常条件
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
h=0.8 MLCC
Normal Conditions
h=0.8 MLCC通常条件
100
Less than 5Ω
Less than 2Ω
Less than 1Ω
Less than 0.5 Ω
Less than 0.2 Ω
Less than 0.1 Ω
10
1
0.1
1
Wiring
Width w (mm)
配線幅w(mm)
10
Figure 7-10 MLCC Normal Conditions (ESLcap=1nH)
(Cannot be wired when under ZT=0.5Ω)
63
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
8. Configuration of PDN combined with capacitors
Power supply wiring and decoupling capacitors connected to a power supply terminal of
an IC as a whole is called PDN. 1) One of performance indicators for this PDN is impedance
(power supply impedance) at PDN relative to the power supply terminal of an IC. When
the power supply impedance is smaller for the PDN, the current supply performance and
power integrity (PI) will be higher. As described in Chapter 6, when the power supply
impedance is smaller, voltage fluctuation when power supply current for the IC fluctuates
becomes smaller.
In a large-scale and high-speed IC, power supply current strongly fluctuates and its
frequency is high; power supply impedance must be made small over a wide frequency
range. In such a case, since one capacitor cannot achieve the necessary impedance,
multiple capacitors are positioned hierarchically as shown in Figure 8-1 to achieve the
target power supply impedance. We will now describe the hierarchical positioning of
capacitors to fulfill the target impedance.
On-Chip Capacitance
Package Capacitor
Chip (Semiconductor)
Package
Board Capacitor
Bulk Capacitor
PCB
Power Supply Module or
Power Supply Connector
Figure 8-1 An example of decoupling capacitor placement
8.1 Hierarchical positioning of decoupling capacitors
When capacitors are positioned hierarchically as in Figure 8-1, each of the capacitors is
named depending on its position, as in Figure 8-1, and connected as illustrated in Figure
8-2. 15) On-chip capacitance (capacitance formed on silicone) is not a component, but is
added because it shares the same function.
Those capacitors function as a “reservoir of charge” from the viewpoint of the current
supply function of the PDN. In other words, by instantaneously handling a local current
request near a semiconductor, they maintain the time-to-respond and voltage for power
supply modules. Also from the viewpoint of the frequency characteristics of power supply
impedance, as frequency increases impedance increases for the power supply module
without any help, capacitors are placed near ICs to reduce impedance in high-frequency
region.
64
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
PCB
Package
Delayed
response cable,
etc.
Board wiring, etc.
Socket, BGA,
board wiring,
etc.
Wire bonding, internal wiring
for semiconductor, package
wiring, etc.
Current
Fluctuation
Semiconductor
Power
Supply
I
t
Bulk
Capacitor
Board
Capacitor
Package
Capacitor
On-Chip
Capacitance
Figure 8-2 Current supply model from capacitors
As described in Chapter 7, we need to consider wiring inductance in addition to
capacitors for power supply impedance relative to the IC. In Figure 8-2, the influence of the
wiring between the semiconductor and each of the capacitors is expressed in terms of
inductance (for the purpose of simplicity, capacitance and resistance of the wiring is
neglected). Since the wiring inductance of the far capacitors becomes large, impedance
cannot be reduced under high frequency. Conversely, we can expect capacitors near the
semiconductor to stay effective in the high frequency.
In this sense, if we can gain sufficient capacitance from the on-chip capacitance, it would
be ideal for reducing power supply impedance. In reality, this is difficult due to space
constraints. Therefore, we place capacitors hierarchically from near to far from the
semiconductor as shown in Figure 8-2 to achieve target power supply impedance.
8.2 Impedance of PDN
The target value for power supply impedance necessary for IC operation is called target
impedance (ZT), and it is necessary to stay below the target value for the frequency range
necessary as shown in Figure 8-3 (although the target value is a constant in the figure, it
may vary depending on the frequency).
PDN consists of a power supply, a decoupling capacitor, and wiring to connect them, etc.
PDN must be designed to meet the target impedance as a total. (Although a target
impedance must be selected taking operations of the IC and the circuit in consideration, it
may not be explicit in some cases. We will introduce the guideline for its set-up in Section
8.7.)
Ideally, power supply impedance should be expressed in terms of impedance relative to
the transistor on the silicon wafer of the model in Figure 8-2. However, it is not practical to
take a measurement over the wafer. In reality, the power supply impedance needs to be
expressed by establishing a point of measurement such as a BGA terminal on the package
or power supply pad over the PCB (generally, the value varies depending on the
measurement location). According to the description below, it is impedance (a virtual value
65
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
since it is unmeasurable in reality) relative to the semiconductor element unless stated
Impedance
otherwise.
Target Impedance
ZT
Measurement
Power Supply Impedance
Frequency
Maximum frequency
necessary for IC
operations
fT
Figure 8-3 Target impedance
8.3 Hierarchical positioning of capacitors
The frequency characteristics of the impedance from the entire PDN when positioning
capacitors hierarchically as shown in Figure 8-2 will become that of Figure 8-4. Target
impedance is met as a total by combining frequency ranges covered by each of the
ita
t
Lim
De
for laye
po d re
we
r s spon
up
ply se
r
cito
pa
Ca
or
cit
pa
Ca
ce
an
cit
pa
Ca
r
hip
ito
-C
ac
ap
On
eC
ag
ck
Pa
ard
Bo
lk
Bu
logZ
ion
by
PC
Lim
Bw
i
t
irin
BG ati
g
A a on
nd
b
PC y
Lim
s
B
wir ock
bu itati
ing et,
mp on
an by
Lim
s
d
e
p
i
m
t
ac
wir ati
ka icon
ing on
ge
d
in by p
wir ucto
se
o
ing r
w
mi
co er s
nd
uc upply
tor
capacitors. 1)
Target
Impedance
ZT
Bold line: Impedance achieved by
combination of capacitors
logf
fT
Figure 8-4 Simulation diagram for impedance from a combination of capacitors
The impedance of each capacitor shown in Figure 8-4 is not from components alone, but
includes influence by wiring between semiconductor elements and capacitors as shown in
Figure 8-5. The frequency characteristics of impedance at this capacitor relative to the
semiconductor element become roughly V-shaped as shown in Figure 8-6 (wiring
capacitance is neglected for the purpose of simplicity).
The range of this curve where it meets the target impedance ZT is called the effective
frequency range of the capacitor in this case. As shown in Figure 8-6, the lower limit fmin of
66
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
the effective frequency range is limited by the capacitance of the capacitor Ccap and the
upper limit fmax is limited by the inductance of the capacitor ESLtotal. This ESLtotal includes
inductance of the capacitor ESLcap and inductance of wiring Lline. Also, this ESLtotal in turn
includes the ESL of the capacitor itself and the inductance of the capacitor mounting pad
and vias.
Rline Lline
(
)
Ccap
Ccap
Digital IC
(Semiconductor
element)
ESLcap
ESRcap
Capacitor
ESLtotal
ESRtotal
Equivalent circuit of impedance
relative to digital IC
Wiring
Figure 8-5 Equivalent circuit for a single capacitor
|=
2
|Z
to
ta
fE
SL
1
p
ca
fC
2π
log |Z|
l
|=
|Z
Ccap
Effective Frequency Range
π
ZT
ESLtotal
ESRtotl
ESR tota l
Equivalent circuit of impedance
relative to digital IC
fmin =
1
2π ⋅ ZT ⋅ Ccap
fmax =
ZT
2π ⋅ ESLtotal
log f
Figure 8-6 Frequency characteristics for impedance of a single capacitor
As we can see from Figure 8-6, the effective frequency range of a capacitor becomes wider
when ZT is large, and becomes narrow when ZT is small.
The lower limit of the capacitor impedance is limited by ESRtotal. We need to use a
capacitor with its ESR smaller than ZT for the power supply with small ZT..
At the connecting area of the capacitor hierarchy, the capacitor on the low-frequency side
(Capacitor 1) and the capacitor on the high frequency side (Capacitor 2) must be combined
in such a way that the effective frequency range is overwrapped (without any gap) as
shown in Figure 8-7. Therefore, when ESLtotal of a capacitor on the low-frequency side
changes, the capacitance necessary for the capacitor on the high-frequency side changes as
well.
Also, as shown in Figure 8-4, impedance may increase at the connecting area for the
frequency. This is due to the fact that antiresonance may occur between capacitors as
described in Chapter 3. Therefore, connection within the effective frequency range must be
established with sufficient margin.
67
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Effective frequency
range of Capacitor 1
Effective frequency
range of Capacitor 2
log |Z|
ZT
fmin − 2
1
=
2π ⋅ ZT ⋅ CCap −2
fmax −1 =
Impedance of Capacitor 1
ZT
2π ⋅ ESLtotal −1
Impedance of Capacitor 2
log f
Ccap-1
Ccap-2
ESLtotal-1
ESLtotal-2
ESRtotal-1
ESRtotal-2
Equivalent circuit of Capacitor 2
Equivalent circuit of Capacitor 1
Figure 8-7 Hierarchical connection of capacitor impedance
Meanwhile, the effective frequency range of a capacitor varies depending on the level of
target impedance as described previously, allowing hierarchy to be omitted. When current
fluctuation of an IC is small, the effective frequency range expands with relatively high
target impedance. Also, when capacitance of a board capacitor is large and a capacitor with
small ESL is being used, the effective frequency range expands, making it possible to
eliminate a bulk capacitor or a package capacitor front and back and reduce the number of
capacitors used.
An example of simplifying hierarchy is shown in Figure 8-8.
ZT
log |Z|
Power
Supply
Board
Capacitor
On-Chip
Capacitance
log f
(a) When target impedance is high
log |Z|
ZT
Power
Supply Bulk
Capacitor
Board
Capacitor
On-Chip
Package Capacitance
Capacitor
log f
(b) When target impedance is low
Figure 8-8 An example of the hierarchical structure of capacitors
68
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
8.4 Target impedance on PCB
In the capacitor hierarchy shown in Figure 8-2, an on-chip capacitance and a package
capacitor are provided on the IC, so that they cannot be controlled during the PCB design
stage.
Therefore, normally the lower limit of the frequency covered by the on-chip capacitance
and the package capacitor is considered as the upper limit frequency, [email protected], at the PCB
design stage, and designated to be the upper limit frequency for target impedance set for
the power supply terminal external to the IC package. This frequency is generally
considered to be 10MHz to 100MHz.
When designing a decoupling capacitor on PCB, our aim will be to meet target impedance
up to this [email protected] (it is not necessary to target the maximum frequency for IC operation).
The measurement point for this impedance is the power source terminal of the IC package.
In the following section we will describe capacitors used hierarchically on the PCB and
their use.
8.5 Bulk capacitor
A bulk capacitor is a large-capacitance capacitor covering impedance in the
low-frequency region. It is positioned in a section at the power supply area, and in some
cases substitutes as a smoothing capacitor for the power supply module.
As shown in Figure 8-6, the lower limit for impedance of a capacitor is limited by ESR,
and the upper limit of the effective frequency range is limited by ESL and wiring
inductance. Therefore, when using a capacitor with small ESR and ESL, the capacitance of
a board capacitor handling higher frequency can be reduced, or capacitor layout may
become more flexible.
An example comparing the impedance of an electrolytic capacitor and an MLCC is shown
in Figure 8-9. In this case both capacitors are 2.2µF. Even with a low-ESR capacitor using
conductive high polymer, its impedance in the frequency range over 10MHz is larger
compared with MLCC. This indicates that the ESL of the MLCC is small and its upper
limit for the effective frequency range is high.
Aluminum Electrolytic Capacitor
Conductive Polymer Capacitor
MLCC
Figure 8-9 Comparison between the impedance of an electrolytic capacitor and an MLCC
69
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
8.6 Board capacitor
Impedance in a higher frequency region where a bulk capacitor does not function is
handled by a board capacitor located on the PCB near the IC. Normally, an MLCC is used
for this capacitor. One capacitor is sufficient for a relatively small-scale and low-speed IC,
but for high-performance ICs with low target impedance, multiple capacitors in parallel
may be used as shown in Figure 8-10. 2)
Relatively large mounting space
is necessary
Detailed adjustment of
capacitance is necessary to
achieve target impedance
Power
Supply Line
IC (Noise
Source)
Power
Supply Line
IC (Noise
Source)
Multilayer
Ceramic
Capacitor
0.1µF x 10 units
Multilayer Ceramic Capacitor
1µF+10000pF+1000pF+100pF
(a)
(b)
Capacitors with different capacitors
are mounted in parallel, and effective
frequency range is enlarged taking
advantage of their difference in selfresonant frequency
Several capacitors with the same
capacitance are mounted in parallel,
and effective frequency range is
enlarged taking advantage of their
parallel effect
Figure 8-10 Examples of board capacitor parallel layout
Figure 8-10 (a) shows a combination of capacitors with different capacitances. Low
impedance over a wide frequency range is targeted by combining capacitors with different
self-resonant frequencies, taking advantage of the characteristics of capacitors where it
becomes low impedance near the self-resonant frequency.
Caution must be taken in a case where impedance does not become smaller because
antiresonance occurs in the gap between self-resonant frequencies of capacitors as
described in Chapter 3. An example of joint impedance when using four capacitors at 1µF,
10000pF, 1000pF and 100pF in parallel is shown in Figure 8-11. The frequency
characteristics of impedance appear in waves, and in some cases their impedance exceeds
one 1µF capacitor at antiresonance frequencies.
100
pF
100
0pF
100
00p
F
1µF
Joint impedance with
1µF+10000pF+1000pF+100pF
Figure 8-11 Joint impedance when capacitors with differing
capacitance are used in parallel (calculated values)
70
C39E.pdf
Jul.20,2010
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Figure 8-10 (b) shows a case of capacitors with the same capacitance in parallel. In this
case, as the calculation results in Figure 8-12 indicate, problems from antiresonance do not
occur as frequently (calculation assumes that wiring between capacitors may be neglected).
This method has an effect of capacitor impedance becoming parallel, in addition to
impedance of pads and vias becoming parallel (in the case where a via is used for each
capacitor). There is also the advantage that it is relatively easier to increase capacitance
because of the increased number of capacitors.
On the other hand, an increased number of capacitors has the disadvantages of increased
space and cost. Also, as the area increases, mounting positions of capacitors are relatively
far apart, making capacitors less effective from the wiring impedance, gradually
decreasing the effect of the increased number of capacitors.
0.1µ
F
1µF
x
x1
unit
1u
nit
Joint impedance with
0.1µF x 10 units
Figure 8-12 Impedance when capacitors with the same
capacitance are used in parallel (calculated values)
If the method shown in Figure 8-10 gives trouble, using a low-ESL capacitor as shown in
Chapter 4 will give the same effect as an increased number of capacitors with one capacitor.
This is more advantageous for space and cost. Figure 8-13 shows a comparison of
impedance for multiple MLCCs and one low-ESL capacitor. One low-ESL capacitor realizes
impedance equivalent to using 10 MLCCs in parallel.
MLCC
1µF x 1 unit
Low-ESL Capacitor
(LLA) 1µF x 1 unit
Joint impedance when MLCC
0.1µF x 10 units are used
Figure 8-13 Comparison between using MLCCs in parallel
and a low-ESL capacitor (calculated values)
71
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
8.7 Capacitance design of a capacitor
An example of establishing capacitance for a bulk capacitor and a board capacitor from
the target impedance is shown below. As shown in Figure 8-14, we consider the case where
a bulk capacitor and a board capacitor are positioned between the power supply module
and the IC. Wiring is formed with MSL and the mounting positions of capacitors are
roughly predetermined.
Power Supply
Power supply
module or power
supply connector
Bulk Capacitor
Capacitance Cbulk
IC
Ground
Board Capacitor
Total Capacitance Cboad
Residual Inductance ESLcap
Figure 8-14 Model for designing capacitor capacitance
8.7.1 Establishing target impedance
First, target impedance ZT is determined as shown in Figure 8-15. If the target value
and maximum frequency of the power supply impedance necessary for IC operation are
already known, these values are used. If they are unknown, they are established with the
following equation:
ZT =
∆V
∆I
(8-1)
In this case, ΔV is the maximum allowable ripple voltage and ΔI is the maximum
transient current fluctuating from the static state (if unknown, we will let it be about
one-half of the maximum current value of an IC 16) 17)). Maximum frequency [email protected], of ZT
varies depending on the operation speed of an IC. If unknown, set it to be approximately
100MHz.
impedance (ohm)
10
1
ZT
0.1
0.01
0.001
0.001
0.01
0.1
1
10
frequency (MHz)
100
[email protected]
1000
Figure 8-15 Establishing target impedance
8.7.2 Establishing capacitance of a bulk capacitor
Next, we will establish the capacitance of capacitors from the low frequency side. The
first capacitor will be the bulk capacitor. Its model is shown in Figure 8-16.
When we can assume that inductance from cables between the power supply module and
a circuit or printed wiring is the main factor preventing us from achieving target
72
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
impedance at the bulk capacitor mounting position when the power supply module is
operating ideally, we will let this inductance be LPower and establish the bulk capacitor
capacitance Cbulk as shown in Chapter 2.
LPower
2
ZT
C bulk ≥
(8-2)
When the wiring consists only of printed wiring, we can use the following equation below
from Chapter 7 to estimate LPower.
Lline
h
= 0 .4 l  
w 
0.6
(H)
× 10 − 6
(7-3)
In this equation, h is the thickness of dielectric material in MSL, w is the wiring width,
and l is the wiring length.
In a case where the response characteristics of the power supply module itself are not
negligible, this inductance LpowerResponce must be calculated into Lpower in equation (8-3). A
rough estimate can be established as the following from the time constant of inductance as
shown in Chapter 6.
LPowe Re sponce = Z T ⋅t Power Re sponce
(8-3)
In this equation, tpowerResponce is the response speed of the power supply module.
Power supply inductance
relative to mounting
position of bulk capacitor
Wiring
inductance up to
bulk capacitance
h
Lline = 0.4l  
w 
LPower
Power
0. 6
× 10 −6 Supply
IC
Ground
Length l
Establishing bulk
capacitor capacitance
Cbulk ≥
w
Wiring Cross-Section Structure
LPower
ZT
h
Power supply module or
power supply connector
Ground Layer
2
Bulk capacitor
capacitance is Cbulk
impedance (ohm)
10
Response characteristics of power supply
f = ZT/2πLPower
Characteristics of bulk
capacitor capacitance
1
0.1
0.01
f = 1/2πZTCbulk
0.001
0.001
0.01
0.1
1
frequency (MHz)
10
100
1000
Figure 8-16 Establishing capacitance for a bulk capacitor
8.7.3 Establishing a board capacitor
Next, we will establish capacitance for the board capacitor, Cboad, as shown in Figure
8-17. If we let the inductance of the wiring between the bulk capacitor and the board
capacitor be Lbulk, the necessary capacitor at the board capacitor mounting area is
73
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
C boad ≥
Jul.20,2010
Lbulk
2
ZT
(8-4)
as with the equation (8-2). Although, strictly speaking, this Lbulk includes the ESL of the
bulk capacitor and the inductance of all wiring between the IC and the bulk capacitor, in
Figure 8-17, only the wiring between the bulk capacitor and the board capacitor will
represent inductance as a whole.
Wiring inductance up to
board capacitor
h
Lbulk = 0.4l  
w 
0. 6
× 10 −6
Power Supply IC
Ground
Length l
Establishing capacitance for
board capacitor
Cboad ≥
w
Power supply module or
power supply connector
Lbulk
2
ZT
h
Wiring CrossSection Structure
Ground Layer
Board Capacitor
Total capacitance Cboad
Residual inductance ESLcap
Characteristics of board capacitor capacitance
10
impedance (ohm)
Power supply response characteristics
1
0.1
0.01
Characteristics of Lbulk for bulk capacitor wiring
0.001
0.001
0.01
0.1
1
10
frequency (MHz)
100
1000
Figure 8-17 Establishing board capacitor capacitance
8.7.4 Positioning of the board capacitor
Next, the board capacitor will be positioned. By positioning the board capacitor so that
the distance between the IC and the power supply terminal is within the maximum
allowable wiring length lmax as described in Chapter 7, ZT can be met at a frequency up to
[email protected] as shown in Figure 8-18.
l max ≅ 0.4
Z T − 2πf T @ PCB ESLcap
h
f T @ PCB  
w 
0.6
× 106
(m)
(7-5)
In this equation, ESLcap is the ESL of the board capacitor, and it includes inductance
(ESLPCB) from the capacitor mounting pad and via, in addition to ESL of the capacitor
itself.
74
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
Power Supply
Placed by selecting board capacitor
taking maximum allowable wiring length
lmax into consideration
lmax ≅ 0.4
ZT − 2πfT @ PCB ESLcap
h
fT @ PCB  
w 
0 .6
× 10
w
h
Wiring CrossSection Structure
IC
Ground
lmax
Ground Layer
Board Capacitor
Total capacitance Cboad
Residual inductance ESLcap
Placed within lmax
6
10
Response characteristics of power supply
impedance (ohm)
1
0.1
0.01
Characteristics of bulk capacitor
0.001
0.001
0.01
0.1
Characteristics of board
capacitor inductance
1
10
frequency (MHz)
100
[email protected]
1000
Figure 8-18 Placement of the board capacitor
8.7.5 Reducing ESLcap
Depending on the target impedance, one capacitor does not achieve an appropriate
length of lmax being unable to reach target impedance. In such a case, we need to position
multiple capacitors in parallel as indicated in the diagram on the left of Figure 8-19, to
reduce ESLcap equivalently, and enlarge lmax. The use of a low-ESL capacitor as described
in Chapter 4 is effective as well.
Dashed line: When using one low-ESL
capacitor for board capacitor
Solid line: When using multiple
MLCCs for board capacitor
May place a capacitor with capacitance
more than Cboad within lmax
impedance (ohm)
When lmax does not exist for one
capacitor, reducing ESLcap by
combining multiple capacitors,
and enlarging lmax may be
necessary. Total capacitance
must be over Cboad.
Combination of capacitance
should be adjusted to keep
antiresonance down.
When combining MLCCs, make
sure antiresonance part will not
surpass target impedance
10
lmax
1
0.1
0.01
0.001
0.001
0.01
0.1
1
frequency (MHz)
10
100
1000
Figure 8-19 Reduction of ESLcap
A summery of the procedures above is shown in Figure 8-20.
75
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Power supply inductance
relative to Bulk Capacitor
mounting position
STEP 1
Establishing target impedance
∆V
ZT =
∆I
LPower
Wiring inductance
up to capacitor
h
Lline = 0.4l  
w 
∆V: Maximum allowable ripple voltage
∆I: Current Fluctuation (rated current
when unknown)
w
h
Power supply
module or power
supply connector
Establishing maximum frequency fT for IC
power supply terminal
100MHz when unknown
Bulk capacitor
capacitance is Cbulk
STEP 3
STEP 4
STEP 5
Establishing bulk
capacitor capacitance
Establishing board
capacitor capacitance
Board capacitor is selected and placed
taking maximum allowable wiring length
lmax into consideration
LPower
ZT
Cboad ≥
2
Lbulk
2
ZT
lmax ≅ 0.4
ZT − 2πfT @ PCBESLcap
h
fT @ PCB  
w 
0 .6
× 10 −6
Power Supply IC
Ground
× 106
Wiring CrossSection Structure
Ground Layer
lmax
Board Capacitor
Total capacitance Cboad
Residual inductance ESLcap
Located within lmax
Dashed line: When using one low-ESL
capacitor for board capacitor
Placing a capacitor with capacitance
more than Cboad within lmax
Solid line: When using a combination
of MLCCs for board capacitor
10
impedance (ohm)
0 .6
Length l
STEP 2
Cbulk ≥
Jul.20,2010
1
0.1
ZT
lmax
0.01
0.001
0.001
Adjustment of antiresonance is
necessary when combining MLCCs
0.01
0.1
1
10
[email protected]
100
frequency (MHz)
1000
When lmax does not exist for one capacitor,
reducing ESLcap by combining multiple
capacitors, and enlarging lmax may be
necessary. Total capacitance must be
over Cboad.
Combination of capacitance should be
adjusted to keep antiresonance down.
Figure8-20 An example of capacitor design to achieve target impedance
8.8 Making a PDN with ultra-low impedance
When low voltage, large current and high-speed response are required in a power supply
simultaneously such as the core power supply for a large-scale CPU, we may need low
impedance in the magnitude of mΩ. In such a case, it becomes necessary to combine
multiple capacitors for each hierarchy to reach target impedance with the effect of parallel
connections. Impedance design become complex in this case since the number of capacitors
and power supply terminals increases greatly and the power supply wiring configuration
will be complex too. The use of low-ESL capacitors introduced in Chapter 4 may make
power supply design simpler and become advantageous in terms of space and cost from a
reduced number of capacitors. Figure 8-21 shows an example of ultra-low impedance
design with a combination of various capacitors.
76
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
impedance (mohm)
100
PDN Design Target
DC-50MHz 5mΩ
10
1
Power supply internal
impedance 2mΩ
Assuming Lpower to be
10nH
0.1
0.001
0.01
General-Purpose Capacitor
36 units
0.1
1
10
frequency (MHz)
General-Purpose
+ LW Reverse Capacitor
16 units
General-Purpose Capacitor
Organic Polymer
Capacitor 330µF
MLCC 10µF
100
1000
General-Purpose
+ 3-Terminal Capacitor
16 units
Low-ESL Capacitor
LW Reverse Capacitor 100µF
3-Terminal Capacitor 100µF
MLCC 100µF
Figure 8-21 An example of ultra-low impedance design with a combination of capacitors
77
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
9. Summary
In this manual, we have described the configuration and mounting method of decoupling
circuits used for power supplies for ICs. It is assumed that the use of this circuit is for
suppressing noise and supplying enough current for IC operations (achieving PI: power
integrity). Descriptions are made based on insertion loss from the viewpoint of noise
suppression and impedance from the viewpoint of power integration as performance
indicators for decoupling circuits.
Figure 1 shows various decoupling circuits being compared from the viewpoints of noise
suppression and power integrity. For IC power supplies, MLCCs are used as a method of
achieving both of these features simply. When replacing this with 3-terminal capacitors or
low-ESL capacitors, performance improvement can be expected for both noise suppression
and power integrity. Furthermore, when inductors such as ferrite beads are added,
performance can be improved from the viewpoint of noise suppression; however, an
increase in power supply impedance may occur in some cases. In such a case, capacitors
must be reinforced. An increased number of stages for combinations of capacitors and
inductance can further attenuate the noise. These filters should be applied in
correspondence with levels required by circuits.
Filter for Power Supply
IC
Power
Supply
Wiring
Lower power supply impedance (PI)
High PI +
High EMC
MLCC
Low-ESL
Capacitor
3-Terminal
Capacitor
High PI +
High EMC
Ferrite
Bead
High EMC + Ultra-High
Attenuation
High EMC+ High Attenuation
Normal Level
High EMC
Noise suppression level (EMC)
Figure 1 Structure of power supply filters
Characteristics of power supply circuits from the viewpoint of noise countermeasures
include wiring configuration being complex compared with signal circuits, making it
difficult to design characteristic impedance; extremely low impedance in some cases; a
wide range of frequency range for noise countermeasure applications from voice band to
GHz; and having a wide sphere of influence due to being shared by many circuits. In order
for a bypass capacitor to function effectively in such circuits, mounting structure and
78
C39E.pdf
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
Jul.20,2010
wiring design creating small inductance are necessary for low impedance under high
frequency. To this end, this manual describes the wiring configuration for mounting
capacitors as far as possible. We hope you can utilize this information when designing
electronic equipment.
References
1) Hidetoshi Yamamoto, “Pawainteguritei no tame no kondensa no tekiyou (Application of
Decoupling Capacitors for Power Integrity Improvement),” Journal of Japan Institute of
Electronics Packaging, vol. 12 No.3, May 2009
2) Hidetoshi Yamamoto, “Mobairukiki no dengen noizutaisaku (zenpen) (Power Supply Noise
Countermeasure for Mobile Equipment Part 1/2)” pp. 117-129, Electro Magnetic
Compatibility (EMC), September 2007, No. 233
3) Mark I. Montrose, “EMC and the Printed Circuit Board - Design, Theory and Layout Made
Simple,” Wiley-IEEE Press, 1998
4) Brian Young, “Digital Signal Integrity Modeling and Simulation with Interconnects and
Packages,” Prentice Hall PTR, 2001
5) Clayton R. Paul, “Introduction to Electromagnetic Compatibility,” Wiley-Interscience, 1992
6) Tadashi Kubodera, “Kousoku dejitarukairo jissou nouhau (High-Speed Digital Circuit
Mounting Know-How),” CQ Publishing Co., Ltd., 2002
7) “Dejitarukairo ni okeru DCdengenrain no noizutaisaku (Noise Countermeasure for DC
Power Supply Line in Digital Circuits),” Murata Manufacturing Technical Material,
TE13JT, 1996
8) Yukio Sakamoto, “Yokuwakaru dengenrain no EMC/noizutaisaku sekkei (Power Supply
Line EMC/Noise Countermeasure Design Made Simple),” Nikkan Kogyo Shimbun Ltd.,
2006
9) “Noise Suppression by EMIFIL® Basics of EMI Filters,” Murata Manufacturing Technical
Material, 1986
10) Larry D. Smith, “Frequency Domain Target Impedance Method for Bypass Capacitor
Selection for Power Distribution Systems,” pp.119-136, Power Distribution Network
Design Methodologies, IEC, 2008
11) Yukio Sakamoto, “Zukai Noizutaisakubuhin to EMC sekkei (EMI Suppression
Components and EMC Design Illustrated),” Kogyo Chosakai Publishing Co., Ltd., 2005
12) Seiji Sakai, “Chip Ferrite Beads,” pp. 52-57, Electro Magnetic Compatibility (EMC),
November 2007, No. 235
13) Tetsuo Ikeda, “Denki riron dai 2 han (Theory of Electricity 2nd Edition),” Morikita
Publishing Co., Ltd., 2006
14) Stephan H. Hall, Garret W. Hall, James A. McCall, “High-speed Digital System Design; A
Handbook of Interconnect Theory and Design Practices,” Wiley-Inter Science 2000
15) Mikhail Popovich, Andrey V. Mezhiba, Eby G. Friedman, “Power Distribution Networks
with On-chip Decoupling Capacitors,” Springer, 2008
16) Madhavan Swaminathan, A. Ege Engin, “Power Integrity Modeling and Design for
Semiconductor and Systems,” Prentice Hall PTR, 2008
79
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
17) Takahiro Yaguchi, “Purintohaisenban no pawainteguritei sekkei (Power Integrity Design
of PCB),” Journal of Japan Institute of Electronics Packaging, vol. 12 No. 3, May 2009
80
C39E.pdf
Jul.20,2010
Note · Please read rating and
CAUTION (for storage, operating, rating, soldering, mounting and handling) in this catalog to prevent smoking and/or burning, etc.
· This catalog has only typical specifications. Therefore, please approve our product specifications or transact the approval sheet for product specifications before ordering.
C39E.pdf
Jul.20,2010
Fly UP