Capacitors 573

Ceramic and mica dielectric capacitors are also considered. Ceramic capacitors are
used extensively in high power, high frequency switched mode power supplies where

16
they offer small size, low cost, and good performance. The voltage and capacitance
ranges for the four main types of dielectric capacitors are shown in figure 16.1.

Capacitors
Selection of the correct type of capacitor is important in all applications. Just satisfying
capacitances and voltage requirements is usually insufficient. In previous chapters,
capacitors have been used to perform the following functions:
• turn-off snubbering (8.3.1)
• dv/dt snubbering (8.1)
• (RFI filtering (10.2.4, 14.7)
• transient voltage sharing of series connected devices (10.1.1)
• switched-mode power supply output filtering and dc blocking (15)
• dc rail splitting for multilevel converters (14.4)
• power L-C filters
as well as Figure 16.1. Voltage/capacitance boundaries for the principal types of capacitors.
• ac power factor correction and compensation
• dc rail decoupling
• voltage multipliers 16.1 Capacitor general properties
• motors for single phase supplies
The following general principles, properties, and features are common to all capacitor
• cascaded multilevel inverters for VAr compensation
dielectric types.
which is just to name a few uses of capacitors in electrical power applications. In each
application, the capacitor is subjected to stresses, such as high temperature, dv/dt or
16.1.1 Capacitance
high ripple current, which must be taken into account in the design and selection
process. To make the correct capacitor selection it is necessary to consider various
The primary function of a capacitor is to store electrical energy in the form of a charge.
capacitor types, their construction, features, and uses.
The amount of electrical charge, Q, is given by
Two broad capacitor types are found extensively in power electronic circuits, namely: Q = CV (C) (16.1)
• liquid and solid (wet and dry) electrolyte, oxide dielectric capacitors, for while the stored energy is given by
example an aluminium electrolytic capacitor E = ½QV = ½CV 2 (J) (16.2)
• plastic film dielectric capacitors, for example a polyester capacitor. The value of capacitance, C, is directly proportional to surface area, A, and inversely
The first capacitor group has a metal oxide dielectric which offers large capacitance for proportional to the thickness of the dielectric layer, W; that is
a small volume. The second capacitor group, which uses a thin plastic film as a A
dielectric, offers high ac electrical stress properties. C = ε rε o (F) (16.3)
W
574 Power Electronics Capacitors 575

The dielectric constants ε r or alternatively K for materials in common usage, are j

summarised in table 16.1. Z = Rs − (Ω) (16.6)
ω CR
Table 16.1. Dielectric constants for common dielectric materials.
and
Rs 1 real power
Dielectric material Relative dielectric constant tan δ = ω CR Rs = = = (16.7)
X c Q reactive power
εr
where δ is the loss angle and tan δ is termed the dissipation factor, which is the inverse
Vacuum 1 of the circuit quality factor, Q. The angle δ is that necessary to make the capacitor
Air (1 atmosphere) 1.00059 current lead the terminal voltage by 90°, figure 15.2, as for the ideal capacitor.
Polystyrene 2.5 If the insulating or dielectric dc resistance, Ri (= ρi ℓ/A), is included, then
Polypropylene 2.5
1
Polycarbonate 2.8 tan δ = + ω CR Rs (16.8)
Polyethylene-terephthalate 3 ω CR Ri
Impregnated paper 2-6 and at low frequency
Mica 6.5 - 8.7 1
Al203 7 tan δ A = (16.9)
ω CR Ri
Glass 4 - 9.5
Ta203 10 - 25 while at high frequency
Ceramic 20 -12,000 tan δ u ≈ ω CR Rs (16.10)
Both Rs and Xc are dependent on temperature and frequency as shown in figure 16.3.
16.1.2 Equivalent circuit Figure 16.3a shows that the rated capacitance illustrated has a positive temperature
coefficient, the value of which also depends on capacitance and rated voltage. Also
The impedance of a capacitor can be modelled by the capacitor equivalent circuit shown is the negative temperature dependence of equivalent series resistance ESR.
shown in figure 16.2. In series with the ideal capacitor, CR, termed rated capacitance, Figure 16.3b shows that CR and ESR both decrease with frequency.
is an equivalent series resistor Rs (ESR) and equivalent series inductor Ls (ESL). Rs is Since CR and ESR are temperature and frequency dependent, and are related to tan δ
determined by lead and junction resistances, while Ls is the inductance of the electrodes and Z, then tan δ and Z are frequency and temperature dependent as illustrated in
due to the construction and the supply lines. The value of Ls is usually given for a figures 16.3c and 16.3d. Figure 16.3c shows the typical characteristics of the
specific package and capacitor type, and is generally neglected at lower frequencies, impedance of an oxide dielectric capacitor versus frequency, at different temperatures.
below the self-resonant frequency, which is given by At low frequencies the negative slope of Z is due to the dominance of the capacitive
1 reactance, Z ≈ Xc = 1/ωCR, whereas the horizontal region, termed the resonance region,
ωr = (rad/s) (16.4) is where Z is represented by the ohmic resistance Rs, that is Z ≈ Rs. At higher
Ls CR Q frequencies the inductive reactance begins to dominate, whence Z ≈ ωLs and tan δ=Rs
/ωsL.
I I Figure 16.3d shows how the dissipation factor, tan δ, increases approximately
δ proportionally with frequency to a high value at resonance, as would be expected from
equation (16.9). At lower frequencies tan δ may be considered as having a linear
V P frequency dependence, according to tan δ = tan δo + k f.
Figure 16.2. Capacitor equivalent circuit. The service life of a capacitor occurs when its parameters fall outside the specification
limit, termed degradation. Such parameters are usually the capacitance, dissipation
The electrical impedance Z of a capacitor, neglecting Ri the insulation resistance which factor, impedance, and leakage current. The service life is specified under specific
is usually large, is operating conditions such as voltage, ambient temperature, and current, and will
Z = Rs + jX (Ω) (16.5) increase
Since the ESL is neglected, at lower frequencies, since ωL is small
576 Power Electronics Capacitors 577

• the lower the ambient temperature, Ta 16.1.3 Lifetime and failure rate
• the lower the ripple current or voltage, Ir
• the lower the operating voltage in proportion to the rated voltage, Vop /VR Other factors may be relevant to specific dielectrics.
• the higher the ac load frequency, f. Lifetime is the period until a given failure rate is reached. The failure rate, λ, is the ratio
of the number of failures to the service life expected. It is usually indicated in failures
per 109 component hours and is an indicator of equipment reliability.
If, in a large number N of identical components, percentage ∆N fail in time ∆t, then the
failure rated λ, averaged over ∆ t is expressed as
∆N
λ= (/h) (16.11)
N × ∆t
If the sample N is large, then the failure rate in time can be represented by a continuous
‘bathtub’-shaped curve as shown in figure 16.4, such that
1 dN
λ= (/h) (16.12)
N dt
This figure shows the three distinct failure periods, and the usual service life is
specified according to the failure λo, which is constant.

Figure 16.4. The bathtub curve showing variation of failure rate with operating hours.

In the case of voltage, current, and other stresses including temperature, which differ
from those under which λo is specified, conversion or acceleration factors are used to
calculate the new failure rate.
Typical conversion factors are given in table 16.2 for ambient temperature Ta, and
Figure 16.3. Variation of capacitor equivalent circuit parameters with frequency and operating voltage Vop, in relation to rated voltage VR. Alternatively conversion graphs
temperature for a high voltage (47 µF, 350 V) metal oxide liquid dielectric: are also used or the Arrhenius’ law
(a) Rs and CR as a function of temperature; (b) Rs and CR as a function of frequency; n  1 1 
(c) impedance Z as a function of frequency and temperature; and (d) tan δ as a  Vop  − E KT − KTo 
λ = λo   e (16.13)
function of frequency and temperature.
 VR 
578 Power Electronics Capacitors 579

λ 55,½V = 2 × 0.2 × λo
Table 16.2. Stress conversion factors for an aluminium electrolytic capacitor
= 0.8 fit
The conversion factors are cumulative and the failure rate decreases from 2 fit to 0.8
Vop Temperature failures in time.
% Conversion factor Conversion factor
VR Ta (°C) ♣
100 1 ≤40 1
75 0.4 55 2 If the number of units surviving decreases exponential with time, then the probability
50 0.2 70 5 of failure after a service time t is given by
25 0.06 Tjmax 10 F ( t ) = 1 − e − λt (16.14)
10 0.04 Equipment failure rate can be calculated by summing the failure rates of the individual
(a) (b) components, that is
λtotal = λ1 + λ2 ..... + λn (16.15)
If the failure rate is to be constant, then the instantaneous failure rate of the number of
Example 16.1: Failure rate faults per unit time divided by the number of non-failure components must yield a
constant
A component has a failure rate λo = 2 x 109/h, commonly termed 2 fit (failures in time)
1 dF ( t )
using 109/h as reference. =λ (16.16)
With reference to table 16.2, what is the failure rate if 1 − F ( t ) dt
i. the ambient temperature, Ta, is increased to 55°C For n components in a system the probability of system survival is
ii. the operating voltage is halved 1 − F ( t ) = (1 − F1 ( t ) ) × (1 − F2 ( t ) ) × .... (1 − Fn ( t ) ) = e − λ t × e− λ t × .....e − λ t
1 2 n

iii. i. and ii. occur simultaneously? (16.17)

= nλ
Solution if, since the units are identical, λ1=λ2=…= λn.
The meantime between failure (mtbf) is given by
Assume λo applies to conditions at Ta ≤ 40°C and VR. 1 ∞ ∞ 1
mtbf = = ∫ 1 − F ( t ) dt = ∫ e − λt dt = (16.18)
λtotal 0 0 λ
i. If the ambient temperature is increased from 40°C to 55°C, then using a
The service operating life τ for a specified probability of failure is therefore given by
conversion factor of 2 from table 16.2b
1 1
λ55 = 2 × λo τ = An (16.19)
λ 1− F
= 4 fit
that is, the failure rate has doubled, from 2 fit to 4 fit. Example 16.2: Capacitor reliability

ii. Similarly, by halving the operating voltage, a conversion factor of 0.2 is A capacitor has a failure rate λ of 200 x 10-9 failure/hour, 200 fit. Calculate
employed from table 16.2a. The new failure rate is i. the probability of the component being serviceable after one year
λ½V = 0.2 × λo ii. the service life if the probability of failure is chosen to be 1% or 0.1%
= 0.4 fit iii. the mean time between failure
iv. the mean time between failure for 10 parallel connected capacitors
that is, the failure rate has decreased by a factor of 5, from 2 fit to 0.4 fit.
v. the probability of survival for 1 year and of failure for units, if 1000 units each
have 10 parallel connected capacitors.
iii. If simultaneously both the ambient temperature is increased to 55°C and the
operating voltage is halved, then
580 Power Electronics Capacitors 581

Solution In the case of a defect in the dielectric oxide layer of an electrolytic capacitor, the
maximum field strength is reached first in the defective region. This is effectively the
i. The probability of the capacitor being serviceable after 8760 h (1 yr) is given by process which occurs during the formation of the oxide layer, which results in the
1 − F (1 yr ) = e − λt formation of new oxide, thereby repairing the defect. The reforming process is
9 relatively slow compared with the healing time for non-polarised capacitors.
= e −200×10 ×8760 = 0.998 (99.8%) By contrast, the high electric field at the defect in a plastic film capacitor causes an arc
which evaporates the metallisation in the breakdown region, thereby isolating the faulty
ii. Component lifetime is given by dielectric within a few microseconds.
1 1
τ= An
λ 1− F 16.1.5 Temperature range
109 1
τ (1% ) = An = 50,000 h = 5.7 years The operating temperature upper and lower limits are either dictated by expected
200 1 − 0.01 service life or the allowable variation limits on the nominal capacitance. Most
9
10 1 capacitors can be used outside their nominal temperature limits, but at reduced lifetime,
τ ( 0.1% ) = An = 5,000 h = 0.57 years
200 1 − 0.001 hence with reduced reliability. The extremes -55°C to 125°C are common, but
obviously electrolytic capacitors must be restricted to a smaller range if the electrolyte
iii. The mean time between failure, given by equation (16.18) is is not either to freeze or to boil.
109
mtbf = 1/ λ = = 5 x 106 h = 570 years
200 16.2 Liquid and solid, metal oxide dielectric capacitors

iv. The failure rate for 10 capacitors is 10λ = 2000 fit and the mtbf is The oxides of metals such as aluminium and tantalum are capable of blocking current
1 109 flow in one direction and conducting in the other. Operation of metal oxide dielectric
= = 57 years capacitors is based on the so-called valve effect of these two metals.
10λ 2000

v. For 1000 units, each with a failure rate of 10λ, the probability of one unit surviving 1 16.2.1 Construction
year is
−9 The capacitor dielectric layer consists of aluminium oxide Al203 or tantalum oxide
1 - F (1 yr) = e-10 x 200 x 10 x 8760 = 98.2 per cent Tn203 which is formed by an electrochemical oxidising process of aluminium foil or
The probable number of first year failures with 1000 units is sintered tantalum powder. These starting metals form the capacitor anode. The oxide
F (1 yr) = 1 - e-200 x 10
−9 x 8760
= 0.002 pu = 2 units layer withstands high electric field strengths, typically 8 x 108 V/m for Al203 which
represents 1.25 nm per volt, and are excellent insulators (hence result in a high
♣
capacitor loss factor). This field strength is maintained during the oxidising process, so
that the oxide thickness is dependent and practically proportional to the forming
The reliability concepts considered are applicable to all electronic components and
voltage VF. To avoid changing the oxide thickness during normal use, the component
have been used to illustrate capacitor reliability.
operated rated voltage VR should always be lower than the forming voltage, as shown
in figure 16.5. The difference VF - VR is the over-oxidisation voltage and substantially
16.1.4 Self-healing
determines the capacitor operational reliability. For general-purpose electrolytic
capacitors, the value of VR / VF is about 0.8, while solid capacitors are rated at 0.25.
One failure mode of a capacitor is voltage breakdown in a defective area of the
dielectric. As a result of the applied voltage, the defective area experiences an
abnormally high electric field which may cause failure by arcing. Oxide capacitors
using an electrolyte and plastic film dielectric capacitors exhibit self-healing properties,
which in the case of plastic film dielectrics allow the capacitor to remain functional
after voltage breakdown.
582 Power Electronics Capacitors 583

Figure 16.5. Current dependence on voltage of Al electrolytic capacitors.

The oxide dielectric constant εr is approximately 10 for Al203 and 25 for Ta203, while in
comparison paper-based dielectrics have a value of approximately 5. An oxide
thickness of W = 0.7 µm is sufficient for high voltage capacitors (≥ 160 V) as compared
with minimum practical paper dielectric thickness of about 6 µm. The metal oxide type
capacitors potentially offer high capacitance per unit volume. To further improve the
capacitance per unit volume, before oxidation, the aluminium anode surface area is
enlarged 10-300 times by electrochemical deep etching processes. In the case of
tantalum capacitors, the sintered tantalum structure results in the same increase of area
effect.
The capacitor is formed by the placement of the cathode on to the oxide layer. In the
case of the electrolytic capacitor, a highly conductive organic acid electrolytic (based
on dimethylacetamide) which is impregnated into porous paper forms the capacitor
cathode. The electrolyte largely determines the ESR hence it must have a low Figure 16.6. Construction of metal oxide capacitors.
resistivity over a wide temperature range. It must also have a breakdown voltage well
above the capacitor rated voltage at maximum operating temperature. For long life,
electrolytes with a water content must be avoided. Teflon spacers are sometimes used 16.2.2 Voltage ratings
rather than paper. In the case of solid capacitors, a high conductive cathode is formed
by a solid semiconductor metal oxide, such as manganese dioxide. The electrical Basic electrolyte (electrolytic) capacitors are suitable only for unipolar voltages, where
contact to the cathode is a layer of etched aluminium, which has a thin oxide layer. In the anode is positive with respect to the cathode. In the case of the aluminium
solid oxide capacitors, the manganese dioxide is dipped into graphite which is coated electrolytic capacitor, the cathode connection metal does have a thin air-oxide layer
with silver epoxy for soldering. which corresponds to an anodically generated layer with a blocking voltage capability
The four possibilities are shown in figure 16.6. A porous paper or glass fibre is used as of about 2 V. Above this voltage level, an electrolytic generated dielectric oxide film
a space keeping agent in order to avoid short circuits and direct mechanical contact. would be formed on the cathode foil. The effect is to decrease the capacitance and
Long strips of the cross-sections are wound into cylindrical bodies and encased as cause high internal heating and gas formation, which can lead to failure.
shown in figure 16.6. Operation at high voltages causes oxide growth and the Solid, oxide capacitors are in principle capable of supporting bipolar voltage since the
production of hydrogen. Any gas pressure relief valve should be orientated upwards. cathode is a semiconductor, manganese oxide. In practice, impurities such as moisture
584 Power Electronics Capacitors 585

restrict the reverse voltage limits to 5-15 per cent of VF. The usable reverse voltage
decreases with increased ambient temperature.
The rated voltage VR may be exceeded under specified intermittent conditions, resulting
in a maximum or peak voltage limit VP, as shown in figure 16.5, where
for VR ≤ 315V VP = 1.15 VR
for VR > 315V VP = 1.1 VR
Both VR and VP, are derated with increasing temperature.

16.2.3 Leakage current

When a dc voltage is applied to capacitors, a low current, Iℓ k called the leakage current,
flows through every capacitor, as implied by the presence of Ri in the equivalent circuit
model in figure 16.2. With oxide dielectric capacitors, this current is high at first and
decreases with working time to a final value, as shown in figure 16.7.
A low final leakage current is the criterion of a well designed dielectric, thus leakage
current can be considered as a measure for the quality of the capacitor. The current is a
result of the oxidising activity within the capacitor. The leakage current depends on
both dc voltage and ambient temperature, as shown in figure 16.8. The purity of the
anode metal, hence oxide dielectric determines the leakage current.
Figure 16.8. Typical leakage current of oxide capacitors versus:
(a) voltage and (b) temperature.

Liquid, oxide capacitors have the lower leakage currents at rated voltage since when a
voltage is applied; anions in the electrolyte maintain the dielectric electrochemical
forming process. The Mn02 in solid oxide capacitors has lower reforming capabilities.
0.9VR From figure 16.8 it will be seen that leakage increases with both temperature and
voltage. The increase in leakage current with temperature is lower in liquid capacitors
than in the solid because, once again, the electrolyte can provide anions for the
dielectric reforming process.
For an aluminium electrolyte capacitor at 85°C, an expected lifetime of 2000 hours is
achieved by selecting VR / VF = 0.8. However, VF is inversely proportional to absolute
temperature so for the same leakage current at 125°C, the ratio of VR / VF must be
decreased to
VR 273 + 85
= 0.8 × = 0.7
VF 273 + 125
For higher temperature operation, a higher forming voltage is required. But since VF ×
CR is constant for any dielectric/electrode combination, CR is decreased.
When connecting electrolytic capacitors in series, parallel sharing resistors are
necessary to compensate for leakage current variation between the capacitors. The
Figure 16.7. Leakage current variation with working time for a liquid aluminium oxide design of the sharing network is as for the steady-state voltage sharing for
capacitor.
semiconductors presented in 10.1.1. Additionally, the resistors provide a discharge path
for the stored energy at power-off. When parallel connecting capacitors, highest
reliability is gained if identically rated capacitors (voltage and capacitance) are used.
586 Power Electronics Capacitors 587

16.2.4 Ripple current The maximum power dissipation P l is dependent on the thermal dissipation properties
d

of the capacitor, and from equation (5.4)

The maximum superimposed alternating current, or ripple current Ir is the maximum l = h A ∆T
P (W) (16.21)
d
rms value of the alternating current with which a capacitor is loaded, which produces a
where h = heat transfer coefficient (W/m2K)
temperature difference of 10 K between the core and ambient. Ripple current results in
A = capacitor outer surface area (m2)
power being dissipated in the ESR, according to ∆T = temperature difference between capacitor surface, Ts, and ambient, Ta (K)
P = I R
2

d r s
(W) (16.20) Thus the maximum ripple current is given by
which results in an internal temperature rise until equilibrium with the ambient occurs. l
P hA∆Tl
I r = d
= (A) (16.22)
Rs Rs
The ESR, Rs, is both temperature and frequency dependent, hence rated ripple current
Iro, is specified at a given temperature and frequency, and at rated voltage VR. Due to
the square root in equation (16.22), conversion to other operating conditions is
performed with the frequency multiplier √r and temperature multiplier √k, such that
I r = k r I ro = k r I ro (A) (16.23)
Typical multiplier characteristics for aluminium oxide capacitors are shown in figure
16.9. It will be seen from figure 16.9a that electrolytic capacitors are rated at 85°C,
while as seen in figure 16.9b solid types are characterised at 125°C. For each type, a
reference frequency of 100 Hz is used. Electrolytic capacitors usually have a thermal
time constant of minutes, which can be exploited to allow intermittent overloads.

Example 16.3: Capacitor ripple current rating

A 1000 µF, 385 V liquid, aluminium oxide capacitor has an rms ripple current rating Iro
of 3.7 A at 100 Hz and 85°C.
Use figure 16.9a to calculate the allowable ripple current at
i. 60°C and 1 kHz
ii. lowest stress conditions.

Solution

i. Using equation (16.23)

I r = k r I ro = k r I ro (A)
where from figure 16.9a at 60°C, √k = 1.85
at 1 kHz, √r = 1.33
whence I r = 1.33 × 1.85 × 3.7A
= 9.1 A

ii. This capacitor experiences lowest stressing at temperatures below

Figure 16.9. Frequency and temperature ripple current conversion multipliers for: 40°C, where √k =2.25 and at frequencies in excess of 2 kHz when √r = 1.37.
(a) liquid and (b) solid A1203 capacitors. Under these conditions the ripple current rating is
588 Power Electronics Capacitors 589

I r = 2.25 ×1.37 × 3.7A Ripple current ratings may not be specifically given for some capacitor types, for
example solid tantalum capacitors. In this case an indirect approach is used. In
= 11.4A satisfying ac voltage limitations as illustrated in figure 16.10, and any series resistance
♣ requirement, allowable ripple currents can be specified for a given temperature.
Non-sinusoidal ripple currents have to be analysed such that the individual frequency 16.2.5 Service lifetime and reliability
components satisfy
I2
I r ≥ ∑ n 16.2.5i - Liquid, oxide capacitors
2
(16.24)
n rn
As considered in 16.1.3, the reliability and lifetime of a capacitor can be significantly

where I r is for the appropriate rated ambient and reference frequency as indicated in improved by decreasing the thermal and electrical stresses it experiences. Stress
figure 16.9. reduction is of extreme importance in the case of liquid aluminium oxide capacitors
Liquid tantalum capacitors have a ripple current rating which is determined by the since it is probably the least reliable commonly used component.
physical dimensions, independent of temperature over a wide range, and independent The reliability and service lifetime of an aluminium oxide electrolytic capacitor are
of frequency above 50 Hz. dominated by its ripple current, operating temperature, and operating voltage. Figure
16.11 in conjunction with figure 16.9a can be used to determine service life.

Figure 16.11. Service life for an aluminium oxide liquid capacitor. Temperature
Figure 16.10. Rms voltage limits of solid tantalum capacitors for different physical dependence of lifetime variation with: (a) ripple current and (b) operating voltage.
dimensions, temperature, voltage rating, and frequency.
590 Power Electronics Capacitors 591

Example 16.4: A1203 capacitor service life The failure rate of all capacitors can be improved by decreasing the stress factors such
as temperature and operating voltage. But reliability of solid tantalum capacitors can be
A 1000 µF, 385 V dc aluminium oxide liquid capacitor with a ripple current rating Iro increased by placing a series resistor (low inductance) in the circuit. The improvement
of 2.9 A at 100 Hz and 85°C ambient is used at 5 A, 4 kHz, in a 65°C ambient and on a is illustrated by the following design example, which compares the lifetime of both
240 V dc rail. What is the expected service lifetime of the capacitor? liquid and solid tantalum capacitors based on the conversion curves in figure 16.12.

Solution
From figure 16.9a at 4 kHz, √r =1.35, whence
Io 1 5A 1
× = × = 1.28
I ro r 2.9A 1.35
From figure 16.11a, the coordinates 1.28 and 65°C correspond to a 24,000 hour
lifetime with less than 1 per cent failures. Since a 385 V dc rated capacitor is used on a
240 V dc rail, that is, a ratio 0.64, an increase in service lifetime of 17½ per cent can be
expected, according to figure 16.11b. That is, a service lifetime of 28,000 hours or
greater than 3½ years is expected with a relative failure rate of less than 1 per cent.
Generally, between 40 and 85°C aluminium electrolytic capacitor lifetime doubles for
every 10°C decrease in ambient temperature. A service lifetime of 7 years could be
obtained by decreasing the ambient temperature from 65°C to 55°C.
♣

With aluminium electrolytic capacitors, degradation failures are mostly due to factors
such as
• aggressiveness of the acidic electrolyte
• diffusion of the electrolyte
• material impurities.

16.2.5ii - Solid, oxide capacitors

The failure rate of solid aluminium and tantalum capacitors is determined by the
occurrence of open and short circuits as a result of dielectric oxide layer breakdown or
field crystallisation. In general, for a given oxide operating at rated conditions, liquid
capacitors have a shorter lifetime than the corresponding solid type. Solid aluminium
capacitors are more reliable than solid tantalum types and failure is usually the
degradation of leakage current rather than a short circuit.
In comparison with liquid, electrolytic capacitors, solid types, and, in particular,
tantalum type capacitors, have a number of desirable characteristics:
• higher capacitance per unit/volume due to the higher permittivity of Ta203 and
the intrinsically high effective area per unit volume due to the sintered
construction
(b)
• changes in parameters (C, tan δ) are less because the specific resistance of
Mn02 and hence temperature coefficient, is lower than that of liquid Figure 16.12. Stress conversion factors for:
electrolytes (a) solid tantalum capacitors and (b) liquid tantalum capacitors.
• electrolyte is stable, does not evaporate or corrode.
592 Power Electronics Capacitors 593

Example 16.5: Lifetime of tantalum capacitors poly p ropylene P

poly s tyrene S
A 22 µF tantalum capacitor is required to operate under the following conditions: polyphenylene sulph i de I
ambient temperature Ta, 70°C The letter shown after each type is the symbol generally used to designate the film
operating voltage Vop, 15 V type. The symbol K is used to designate plastic, which is Kunststoff in German.
circuit resistance i. 1 Ω Two basic types of plastic film dielectric capacitors are common. The first type
ii. 100 Ω involves metallisation deposited on to the plastic and the metal forms the electrodes.
Typically such a capacitor would be termed MKP, that is metallised - M, plastic - K,
Calculate the expected lifetime for solid and liquid tantalum capacitors. polypropylene - P. A foil capacitor, the second type, results when metal foil is used for
the electrode. Typically such a capacitor would be termed KS, that is plastic - K,
Solution polystyrene - S. The plastic type is generally designated by the fifth letter of the plastic
name, that is the letter after poly, with two exceptions.
Capacitor used CR = 22 µF
VR = 25 V 16.3.1 Construction
For each capacitor type (solid or liquid) the voltage stress factor is
Vop /VR = 0.60 16.3.1i - Metallised plastic film dielectric capacitors
For the solid tantalum, the circuit resistance factor is given by
The dielectric of these capacitors consists of plastic film on to which metal layers of
i. Ri' = 1 Ω /15 V = 0.07 Ω/V which is < 0.1 Ω/V
approximately 0.02-0.1 µm are vacuum deposited. A margin of non-coated film is left
ii. i
'
R = 100 Ω /15 V = 6.6 Ω/V which is > 3 Ω/V as shown in figure 16.13a. The metallised films are either wound in a rolled cylinder or
Based on figure 16.12, the capacitor lifetime calculation is summarised below. flattened to form a stacked block construction. In this construction, the metallised films
are displaced so that one extends out at one end of the roll and the next layer extends
Liquid tantalum Solid tantalum out the other end as shown in figure 16.13a. This displaced layer construction is termed
R Ω 1 and 100 1 100 extended metallisation and facilitates electrical contact with the electrodes. A hot metal
Ri Ω n/a 0.1 3 spray technique, called schooping, is used for making electrical contact to the extended
Σ Ri (1) 12 1 edges of the metallised plastic winding. This large disk area contact method ensures
Χ good ohmic contact, hence low loss and low impedance capacitor characteristics result.
2.2 0.10 0.10 The most common metallised plastic film capacitors are those employing polyester,
at Vop/VR=0.6 and 70°C
λo /h 4×10
-8
1×10
-8
1×10
-8
MKT and polypropylene, MKP.
λ 2.2×4×10
-8
12×0.1×10
-8
1×0.1×10
-8
Polyester has a higher dielectric constant than polypropylene, and because of its
/h -8 -8 -8
= λo×Σ 8.8×10 1.2×10 0.1×10 stronger physical characteristics it is available in thinner gauges than is polypropylene.
fit 88 12 1 Very high capacitance values result in the smallest possible space. But polypropylene
τ
h
45,000 83,000 100,000 has a higher dielectric strength and lower dielectric losses, hence is favoured at higher
(% failures) within λ∆t (0.4%) (0.1%) (0.1%) ac voltages.
♣
16.3.1ii - Foil and plastic film capacitors
16.3 Plastic film dielectric capacitors
Foil capacitors normally use a plastic film dielectric which is a flexible bi-axially
aligned electro-insulator, such as polyester. Aluminium foils and/or tin foils are used as
Plastic (polymer) dielectric type capacitors are non-polarised capacitors and in general
the electrodes. The thin strips are wound to form the capacitor as shown in figure
offer high dv/dt and pulse rating capability compared with oxide type capacitors.
16.13b. An extended foil technique similar to the extended metallisation method is used
The most common dielectric plastics used are:
to enable contact to be made to the extended foil electrodes.
polye t hylene-terephthalate (polyester or PEPT) T
poly c arbonate C
594 Power Electronics Capacitors 595

metallised paper electrodes. The coarse porous nature of the paper allows for improved
fluid impregnation of the dielectric material, which counters the occurrence of gas air
bubbles in the dielectric. This construction has the electrical advantages of high
dielectric strength, low losses, and a self-healing mechanism, all at high voltages.
Two plastic dielectrics can be combined, as shown in figure 16.13d, to form a mixed
layer capacitor. It involves a double metallised polyethyleneterephthalate film and
polypropylene films. These dielectric combinations give low inductance, high
dielectric strength, and low losses with high ac voltage capability.

16.3.2 Insulation

The insulation characteristics of a capacitor are indicated either as a resistance value Ri

as shown in Figure 16.2 or as a time constant, τ = Ri CR. The resistance comprises the
insulation resistance of the dielectric (layer to layer) and the insulation resistance
between layer and case. This later resistance is determined by the quality of the case
insulating material and by the length of the surface leakage paths.
Both the time constant and resistance are dependent on voltage and temperature, as is
shown in figure 16.14. These characteristics illustrate that extremely high insulation
schooping resistance values can be obtained.
contact
16.3.3 Electrical characteristics

16.3.3i - Temperature dependence

The capacitance of plastic film capacitors changes with both temperature and
frequency, as shown in figure 16.15. The dependence is strongly dependent on the
dielectric film although some foil types are virtually independent of frequency. Table
16.3 summarises capacitance temperature dependence for a wide range of dielectrics.
schooping The temperature coefficient is measured in parts per million per degree Kelvin, ppm/K.
contact The temperature dependence of dissipation factor is shown in figure 16.21a.
16.3.3ii - Dissipation factor and impedance
Figure 16.13. Plastic capacitor constructions: (a) extended single metallisation; (b) Figure 16.16a shows the typical frequency dependent characteristics of the dissipation
extended foil; (c) mixed dielectric; and (d) mixed dielectric, double metallisation. factor for a range of plastic dielectric capacitor types. It is important to note that
polyester types have 50-100 times the losses of polypropylene capacitors. A low loss
characteristic is important in power pulse applications where capacitor package heat
16.3.1iii - Mixed dielectric capacitors dissipation may be a limiting factor.
Generally, tan δ rises with increased frequency and increased capacitance. Tan δ is
To further improve the electrical stress capabilities of a capacitor, combinations of
dominated by dielectric losses and the contact resistance of the leads. The extended
different dielectrics are commonly used. Such capacitors use combinations of
foil/metallisation and schooping contact methods provide not only a low and constant
metallised plastics, metallised paper, discrete foils and dielectrics, and oil
ohmic contact, but because of the large contact area, result in a low self-inductance.
impregnation.
The resonant frequency of such capacitors, because of their self-inductance and their
Figure 16.13c shows the layers of a mixed dielectric paper and polypropylene
capacitance, is high as shown by the minimum impedance in figure 16.16b. Minimum
capacitor. A thin gauge of polypropylene dielectric is combined with textured
596 Power Electronics Capacitors 597

impedance decreases with increased capacitance and each capacitor in the range, here
1.5 nF to 4.7 µF, has its own Y-shaped impedance curve. The self-resonant frequency f = 1kHz
decreases with increased capacitance. In figure 16.16b, the full impedance curves for
maximum and minimum capacitance only have been shown.

Figure 16.15. Plastic film dielectric capacitance variation with: (a) ambient
temperature and (b) frequency.
Figure 16.14. Plastic dielectric insulation resistance temperature dependence
characteristics:(a) resistance Ri and (b) time constant τ.
598 Power Electronics Capacitors 599

Table 16.3. Capacitor temperature coefficient for various dielectric materials

Dielectric Temperature coefficient (ppm/K)

type metallised other film/foil
Polypropylene -170 -120
Polyester 400 400 (non-linear)
Polycarbonate 150 -50 to -150
Polystyrene -125
Paper 300 300
Mica 100
Ceramic + 1000 to -1000 (non-linear)
Aluminium 1500
Tantalum
(solid and liquid) +200 to +1000

16.3.3iii - Voltage derating

The ac and dc voltages which may be applied continuously to a capacitor vary with
ambient temperature and also frequency in the case of ac voltage rating. Typical
characteristics showing frequency and temperature dependence are shown in figure
16.17 for plastic dielectric capacitor types. It will be seen that the ac voltage rating is
significantly less than the dc voltage rating, while both voltage ratings are derated
above 85°C and at higher frequencies. In all situations, the sum of the dc voltage and
peak value of superimposed ac voltage must not exceed the rated dc voltage.
An alternative approach for calculating the maximum ac voltage, allowable Vac, for a
capacitor is based on the power dissipation limits, P, of the package.
If we neglect Ri and ESL in the capacitor equivalent circuit shown in figure 16.2, then
VR2
P= = I 2 Rs
s
(W) (16.25)
Rs
and
Rs2
VR2 = Vac2 (16.26)
s
1
Rs2 + 2 2
ω CR
Since from equation (16.10) for plastic dielectric capacitors
tan δ = ω CR Rs
then equation (16.25) can be written as
P = ( Rs CR ) ω 2 CRVac2 (W) (16.27)
or alternatively Figure 16.16. Frequency characteristics for plastic dielectric capacitors:
(a) maximum dissipation factor, tan δ and (b) typical impedance characteristics, Z, for
P = tan δ ω CRVac2 (W) (16.28) metallised plastic dielectric capacitors.
600 Power Electronics Capacitors 601

The value of tan δ for equation (16.28) is available from figure 16.16a or, alternatively,
the value of RsCR for equation (16.27) is available from figure 16.18.
The maximum permissible power dissipation, P l which depends on the package
dimensions and ambient temperature, is given in figure 16.19. Thus when the power
dissipation, for a given ac voltage, has been calculated, figure 16.19 can be used to
specify the minimum size (dimensions) capacitor capable of dissipating that power.
The following example illustrates the design approach outlined.

Figure 16.18. Maximum product of series resistance, Rs, and rated capacitance, CR,
as a function of frequency.

Example 16.6: Power dissipation limits - ac voltage

A 0.1 µF plastic capacitor is used in a 100 V ac, 10 kHz and 50°C ambient application.
Select suitable metallised polypropylene and polyester capacitors for this application.

Solution

i. Metallised polyester capacitor (MKT)

From equation (16.27)
P = ( Rs CR ) ω 2 CRVac2 (W)
From figure 16.18, RsCR = 2 × 10-7 at 10 kHz. Thus
P = (2 × 10-7 ) × (2π × 104 ) 2 × (0.1× 10-6 ) × (100) 2
= 780 mW
From figure 16.19, at 50°C a MKT capacitor of dimensions 11×20×31 (mm) can
Figure 16.17. Plastic dielectric capacitor, voltage derating characteristics: dissipate 930 mW. The applicable capacitor must have an ac voltage rating in excess of
(a) dc voltage derating with ambient temperature; (b) ac voltage derating with 100 V ac. From figure 16.17b, it can be seen that a 0.1 µF, 400 V dc MKT capacitor is
temperature; and (c) ac voltage derating with frequency. necessary, given that the dimension constraints are met.
602 Power Electronics Capacitors 603

16.3.3iv - Pulse dVR /dt rating

Related to the ac voltage rating and power handling capabilities of a capacitor is the
rated pulse slope dVR /dt, which from i = CR dv / dt is specified by
VR V
R= = R (16.29)
CR dV/dtmax I
where R is the minimum series resistance including the ESR. Generally for a given CR,
dv/dt capability increases with rated voltage VR, and decreases as the distance between
the metallised electrode contacts increases. If the capacitor operating voltage Vop is
decreased below VR, at which voltage, dv/dt capability is specified, dv/dt capability
increases according to
d Vop d VR VR
= × (V/s) (16.30)
dt dt Vop
The dv/dt capability depends on both the dielectric type and layer construction.
Generally polystyrene (KS) and polyester (KT) foil type capacitors are not applicable
to high dv/dt applications. Metallised polycarbonate capacitors offer slightly better
dv/dt properties than those of metallised polyester. Metallised paper capacitors can
withstand very high levels of dv/dt, 30-50 times higher than those for metallised
polyester. Capacitors using polypropylene, or even better a mixed dielectric involving
polypropylene, offer extremely high dv/dt capability. With the construction shown in
figure 16.13d, a 1 µF metallised polypropylene capacitor with VR of 2000 V dc and
1000 V ac, a 2500 V/µs capability is attainable. Practically the dv/dt limit may be
Figure 16.19. Maximum power dissipation for metallised plastic capacitors
restricted by the external connections. Such ratings are obtainable with polypropylene
as a function of ambient temperature and capacitor dimensions.
because of its extremely low losses, tan δ, as indicated in figure 16.16a. Under such
high dv/dt stresses, it is important to ensure that the power dissipated does not exceed
ii. Metallised polypropylene capacitor (MKP)
the package limit.
From equation (16.28)
P = tan δ ωCRVac2 (W) 16.3.4 Non-sinusoidal repetitive voltages
From figure 16.16a, tan δ = 4.0 × 10-4 at 10 kHz, for a 600 V dc type. Thus
P = (4.0 × 10-4 ) × (2π × 104 ) × (0.1× 10-6 ) × 100 2 Capacitors used for repetitive transient suppression, and for turn-off snubbers on GTO
= 25.6 mW thyristors and diodes, experience high-magnitude short-duration voltage and current
pulses which are not sinusoidal. High dv/dt capacitors based on metallised
From figure 16.19, at 50°C, the smallest volume MKP capacitor, of dimensions
polypropylene are used, which are limited by their internal power losses, hence
6.5×15×26 mm, can dissipate 300 mW. From figure 16.17c it can be seen that a 0.1 µF,
temperature rise and package power dissipation limit.
630 V dc (250 V ac) MKP capacitor is necessary.
A restrictive graphical design approach for capacitor selection with sinusoidal,
From figure 16.17c it can be seen that a 250 V dc 0.1 µF polypropylene foil capacitor
sawtooth, and trapezoidal pulse trains is shown in figure 16.20. The design approach is
(KS) is capable of 160 V ac at 10 kHz. Figure 16.16a shows the dissipation factor of
illustrated by the following example.
KP type capacitors to be under half that of the metallised equivalent. That is, the
expected losses are only
P = (1.4 × 10-4 ) × (2π × 104 ) × (0.1× 10-6 ) × 100 2
= 9 mW
♣
604 Power Electronics Capacitors 605

The peak to peak requirement is 100 V, hence only a 630 V dc 0.1 µF MKT
capacitor can fulfil the specification.
♣
An alternative approach to specify the voltage limits for non-sinusoidal repetitive
voltages is to sum the power contribution due to each voltage harmonic. The total
power due to all harmonics must not exceed the capacitor package power limits.
The non-sinusoidal voltage v can be expressed in the form
v = ∑ Vi sin ( iω t + φi ) (16.31)
∀i

where Vi is the magnitude of the ith voltage harmonic, which has an rms value of
V
vi = i
2
From equations (16.10) and (16.27), assuming capacitance is frequency independent
Pi = ( Rs CR )i ωi2 CR vi2
i
(16.32)
or
Pi = tan δ i ωi CRi vi2 (16.33)
The total power dissipated is the sum of the powers associated with each frequency.
The near-linear frequency dependence of tan δ and RS CR, as shown in figures 16.16a
and 16.18, may be utilised to simplify the calculation procedure. Assuming the rated
Figure 16.20. Metallised polyester capacitor selection graph capacitance is independent of frequency may be a valid and helpful simplification,
for sinusoidal and non-sinusoidal voltages. while the temperature dependence of CR initially could be accounted for by using a
value at 10 K above ambient.
Example 16.7: Capacitor non-sinusoidal voltage rating
Example 16.8: Capacitor power rating for non-sinusoidal voltages
A 0.15 µF MKT capacitor is used to generate a 10 kHz maximum and 25 µs risetime
minimum, sawtooth ac voltage waveform. What voltage rated capacitor is applicable if The applied voltage across a 1 µF MKP capacitor, at 40°C ambient is √2 100 sin(2π ×
the output voltage maximum is 100 V p-p? 104t) + √2 Y sin(2π× 3 x 104t)
What is the maximum allowable voltage Y?
Solution
Solution
Worst-case conditions are at maximum frequency, 10 kHz, and minimum
risetime, 25 µs. From equation (16.33), the total power is given by
With reference to figure 16.20, use Pi = tan δ1 ω1CR v12 + tan δ 3 ω3CR v32
f = 10 kHz (repetition frequency) 1 3

τ = 25 µs (risetime) From figure 16.15b we may assume that capacitance is independent of frequency
C = 0.15 µF (capacitance) for polypropylene types. From figure 16.15a, at 50°C, rated capacitance has
According to the dashed line in figure 16.20, starting from f = 10 kHz, yields reduced by only 1 per cent - thus temperature effects on CR may be neglected.
VR = 100 V dc gives maximum peak voltage of 27 V From figure 16.16, for a 600 V MKT capacitor
VR = 250 V dc gives maximum peak voltage of 38 V tan δ1 at 10 kHz (ω1) = 2.5 × 10-4
VR = 400 V dc gives maximum peak voltage of 47 V tan δ3 at 30 kHz (ω3) = 4.2 × 10-4
VR = 630 V dc gives maximum peak voltage of 59 V From figure 16.19b it can be seen that 880 mW can be dissipated in the largest
606 Power Electronics Capacitors 607

package at 50°C. Total power is given by shown in figure 16.16a. These films also offer good insulation properties as shown in
0.88W = 2.5 × 10-4 × 2π × 104 × 1× 10-6 × 1002 figure 16.14. Variation of capacitance with frequency and temperature is shown in
figure 16.15, while percentage variation of losses, tan δ, with frequency and
+ 4.2 × 10-4 × 6π × 104 × 1× 10-6 × Y 2 (W)
temperature is shown in figure 16.21. The typical capacitance range of X2 capacitors is
Solving for Y, Y = 30.2 V rms. from 10 nF to l µF, rated for 250 V ac application.
♣
The key properties of plastic type non-polarised capacitors are summarised in table
16.4. The excellent dielectric properties of the polypropylene lead to metallised
polypropylene capacitors being extensively used in power applications.

Table 16.4. Properties of non-polarised plastic type capacitors

dielectric type εr tanδ λo dv/dt self-healing

polypropylene low low good high good
polyester medium high poor medium good
polystyrene low low good high poor
polycarbonate low medium good medium good
mixed dielectric medium medium good medium good
paper high high very good high very good

16.4 Emi suppression capacitors

Non-polarised capacitors are used in rfi filters for electrical appliances and equipment,
as was introduced in 10.2.4. The capacitors used between line and neutral are termed
class X while those used to earth are termed class Y.

16.4.1 Class X capacitors

X capacitors are suitable for use in situations where failure of the capacitor would not
lead to danger of electric shock. X capacitors are divided into two subclasses according
to the ac power line voltage applied.
• The Xl subclass must support a peak voltage in excess of 1.2 kV in service, while
• X2 capacitors have peak service voltage capabilities of less than 1.2 kV.
In order to obtain the peak voltage requirement of Xl capacitors, a construction
comprising impregnated paper dielectric and metal foil electrodes is essential. The
common capacitance range is 10 nF to 0.2 µF.
The lower peak voltage requirement of X2 capacitors allows the use of a metallised
plastic dielectric, of which polyester and polypropylene are common. Impregnated
paper dielectrics may also be employed. Advantageously, metallised plastic film
suppression capacitors yield high dv/dt capability with low associated losses, tan δ, as Figure 16.21. RFI capacitance variation with:
(a) ambient free-air temperature and (b) frequency.
608 Power Electronics Capacitors 609

16.4.2 Class Y capacitors

Class Y capacitors are suitable for use in situations where failure of the capacitor could
lead to danger of electric shock. These capacitors have high electrical and mechanical
safety margins so as to increase reliability and prevent short circuit. They are limited in
capacitance so as to restrict any ac current flowing through the capacitor, hence
decreasing the stored energy to a non-dangerous level.
An impregnated paper dielectric with metal foil electrodes is a common construction
and values between 2.5 nF and 35 nF are extensively used. Capacitance as low as 0.5
nF is not uncommon.
A Y-class capacitor for 250 V ac application can typically withstand over 2500 V dc
for 2s, layer to layer. On an ac supply, 425 V ac (√3 VR) for 1000 hours is a common
continuous ac voltage test.
If dv/dt capability is required, polypropylene film dielectric Y-class capacitors are
available, but offer lower withstand voltage capability than paper types. Generally
paper dielectric capacitors offer superior insulation resistance properties, as shown in
figure 16.14a.
Metallised paper capacitors are also preferred to metallised plastic types because they
have better self-healing characteristics. Breakdown in metallised plastic film dielectrics
causes a reduction of the insulation resistance because of a higher carbon deposit in the
breakdown channel than results with paper dielectrics.

16.4.3 Feed-through capacitors

Feed-through or four-terminal capacitors are capacitors in which the operating current

flows through or across the electrodes. High frequency rfi is attenuated by the
capacitors and the main power is transmitted unaffected. That is they suppress emi
penetration into or from shielded equipment via the signal or power path.
Figures 16.22a and b show three terminal feed-through capacitors while figure 16.22c Figure 16.22. Feed-through capacitors for RFI attenuation: (a), (b) three user
is a four-terminal capacitor. A three-terminal coaxial feed-through, wound capacitor terminals; (c) four terminals; and (d) coaxial feed-through capacitor construction.
cross-section is shown in figure 16.22d. The feed-through rod is the central current-
carrying conductor: the outer case performs the function of an electrode plate and
connector to produce an RF seal between the capacitor case and shielding wall. 16.5 Ceramic dielectric capacitors
These capacitors are effective from audio frequencies up to and above the SW and
VHF band (>300 MHz). Current ratings from signal levels to 1600 A dc, 1200 A ac are Ceramic capacitors as a group have in common an oxide ceramic dielectric. The
available, in classes Xl and X2, rated at 240 V ac, 440 V ac and 600 V dc. Class Y dielectric is an inorganic, non-metal polycrystalline structure formed into a solid body
feed-through capacitors rated at 25 A and 440 V ac, 600 V dc are available. by high temperature sintering at 1000 to 1300°C. The resultant crystals are usually
between 1 and 100 µm in diameter.
The basic oxide material for ceramic capacitors is titanium dioxide (Ti02) which has a
Important note: This section on emi-suppression capacitors does not imply those relative permittivity of about 100. This oxide together with barium oxide (Ba02) forms
requirements necessary to conform with governmental safety and design standards. barium titanate (BaTi03) which is a ferro-electric material with a high permittivity,
typically 104. Alternatively strontium titanate may be utilised. These same materials are
used to make positive temperature coefficient resistors - thermistors, where dopants are
added to allow conduction.
610 Power Electronics Capacitors 611

Table 16.5. Ceramic dielectric capacitor characteristics

I II
Dielectric class (εr < 500) (εr > 500)
Low K Moderately high K High K
EIA designation* COG X7R Z5U
IEC/CECC designation CG 2C1 2F4
Temperature range °C -55 to 125 -55 to 125 + 10 to 85
Dielectric constant εr 13 - 470 700 to 50,000
Temperature coefficient of (N150) (X7R) (Z5U)
CR (typical) -150 ± 60 ppm ±15% +22% / -56%
Dissipation factor tan δ 0.15% @ 1 MHz 2.5% 3%
C nF < 0.2 < 4.7 < 40
VR V 500-1k 100 to >2k
* In EIA designation, first letter and number indicate temperature range while last letter indicates capacitance change.

Metal plates of silver or nickel (with minimal palladium and platinum) are used to form
the capacitor. Single plate, or a disc construction, is common as is a multi-layer
monolithic type construction.
The ceramic dielectric is split into two classes, as shown in table 16.5.

16.5.1 Class I dielectrics

This class of dielectric consists mainly of Ti02 and additions of Ba0, La203 or Nd205,
which provides a virtually linear, approximately constant and low temperature
coefficient as shown in figure 16.23a.
COG capacitors belong to the class 1 dielectrics and have a low temperature coefficient
over a wide temperature range, as seen in table 16.5. They provide stability and
minimum dissipation properties. In attaining these properties, a low dielectric constant
results and these capacitors are termed low K. Because of the low dielectric constant,
capacitance is limited.

16.5.2 Class II dielectrics

Ceramic capacitors in this class are usually based on a high permittivity ferroelectric
dielectric, BaTi03, hence termed hi K. Large capacitance in a small volume can be
attained, but only by sacrificing the temperature, frequency, and voltage properties, all
of which are non-linear. Typical characteristics are shown in figure 16.23. Their
characteristics are less stable, non-linear, and have higher losses than class I ceramic,
as seen in table 16.5. Also see table 16.6.
612 Power Electronics Capacitors 613

performed by a saturable reactor, as considered in 8.3.4. Exploitation of voltage

dependence capacitance is generally outside the capacitor specification.
Advantageously, the disc ceramic capacitor has low inductance, but the high
dissipation factor may limit the frequency of operation. Multi-layer ceramic capacitors
can be used in switched mode power supply input and output filters.

16.6 Mica dielectric capacitors

The dielectric mica is a mineral which has a plane of easy cleavage enabling large
sheets of single crystal to be split into thin plates, typically 50 µm thick. Stacks of mica
plates are interleaved with silver metal foils as shown in figure 16.24.

Figure 16.23. Typical properties of commercial ceramic capacitors: (a) capacitance

change with temperature; (b) dissipation variation with temperature; (c) capacitance
change with dc voltage; (d) ESR change with frequency; (e) capacitance change with
ac voltage; (f) dissipation factor variation with ac voltage; (g) capacitance change
with frequency; and (h) dissipation factor variation with frequency.

Table 16.6. Characteristics of class I and II type dielectrics

Class I Class II
Almost linear capacitance/temperature
Non-linear capacitance/temperature function
function
No voltage dependency of capacitance
and loss angle
No ageing Slight ageing of capacitance
High insulation resistance High insulation resistance
Extremely high capacitance value per unit
volume
Very small dielectric loss
High dielectric strength
Normal capacitance tolerance Normal capacitance tolerance
±1% to ±10% ±5% to -20+80%

16.5.3 Applications
Flat circular disc ceramic (Z5U dielectric, high K) capacitors have a 2000 V dc, 550 V
ac rating with capacitances of up to 47 nF. An exploitable drawback of such a ceramic
capacitor is that its permittivity decreases with increased voltage. That is, the
capacitance decreases with increased voltage as shown in figure 16.23c. Such a
capacitor can be used in the turn-off snubber for the GTO thyristor and diodes which
Figure 16.24. Silver mica capacitor:
are considered in 8.1.3 and 8.1. High snubbering action is required at the (a) exploded construction view and (b) electrode pattern of a silvered mica plate.
commencement of turn-off, and can subsequently diminish without adversely affecting
losses or the switching area trajectory tailoring. The capacitor action is a dual to that
614 Power Electronics Capacitors 615

The metal foils, to which the leads are spot-welded, are made of silver, copper, brass,
tin or lead. The stack is held together either by the encapsulation or a metal crimp. The
assembled unit is encapsulated by dipping it into high melting temperature
microcrystalline wax or by coating it with epoxy resin.

16.6.1 Properties and applications

Mica capacitors are non-polar, low loss, and stable up to about 30 MHz, where the lead 0V/Vs or
ground
length and electrodes dominate as inductance. Because of their relatively high cost of
manufacture, as a result of the high labour content and diminishing number of mines,
the ceramic capacitor, particularly the monolithic multi-layer type, is favoured.
Maximum ratings are a few nanofarads at 5000 V, with dissipation factors of 0.1 per
cent at 1 kHz. For capacitance less than 1 nF, a 0.1 per cent dissipation factor is
obtainable at 1 MHz. An insulation resistance of 105 MΩ at 20°C down to 104 MΩ at
125°C is common for capacitance to 10 nF, after which resistance falls off. Typical
operating temperature range is from -55°C to 125°C, with a capacitance temperature
coefficient of 0 to +70 ppm/K.
The maximum current depends on the edge connections and electrodes, so for each
physical design the factor is different and is expressed in mA/pF. This rating may range
between 1.6 mA/pF for smaller packages (9×8×8 mm) down to 0.12 mA/pF for larger
packages (44×32×33 mm). A maximum VA limit must also be observed, typically 50
VA for smaller sizes up to 820 VA for the larger sizes.
Mica dielectric capacitors are sensitive to pressure. Figure 16.25 Component stray capacitance C:
(a) when isolating power devices; (b) between transformer windings; (c) in opto-
16.7 Appendix: Minimisation of stray capacitance couplers; and (d) between terminals of metal oxide semiconductor devices.

Unexpected component stray capacitance, and inductance, can have disastrous circuit
consequences. Figure 16.25 shows four examples of electronic components which have Interwinding capacitance, shown in figure 16.25b, is important in switch mode power
stray capacitance between two parts of the component used at different potentials. supplies and other applications using transformers. By winding the primary and
When the isolated part rapidly changes its relative potential, a charging current flows secondary in different bobbin sections, the interwinding capacitance is decreased since
according to i = C dv/dt. With just 1 pF of capacitance, and at 10,000 V/µs, which is their physical separation is increased. Alternatively, an overlapped copper foil ground
possible with MOSFETs and IGBTs, 10 mA of current flows. This current coupled from shield layer is wrapped between the two windings. The copper strip is a connected to a
the power level to the signal level would affect cmos or ttl circuitry, leading to supply rail or earthed so that charging currents bypass sensitive circuitry.
malfunction and possible failure, if precautions are not taken. Experimentation will reveal the best connection potential and location position. The
Figure 16.25a shows a power package electrically isolated from its heatsink, which is copper foil overlapped turn ends must not make electrical contact, otherwise a short
grounded (to 0V or Vs) in order to minimise rfi radiation. Large power blocks have circuit turn results. Minimise winding start to finish turns capacitance by using the
over 100 pF of isolation capacitance. Other than injecting noise, the level may be winding method shown in figure 17.19b.
sufficient to activate earthing leakage circuitry, if connected to ground. Increasing the A similar solution is used in opto-coupler packages. A grounded Faraday’s grid is
ceramic substrate or mica thickness decreases capacitance according to equation (16.3), placed between the emitter and receiver in order to divert charging current. High dv/dt
but at the expense of increasing thermal resistance. Al2O3 reduces the thermal opto-couplers, with less than 1 pF capacitance input to output, are guaranteed to 15000
impedance compared to aluminium nitride, but at the expense of increased cost. V/µs at 200 V dc levels. This dv/dt limit decreases to 1000 V/µs on a 600 V dc rail.
The effects of capacitive charging current can be minimised by driving the emitting
diode from a low impedance source, both when on and off. Speed and current transfer
ratio can be traded for higher dv/dt capability by increasing separation. For high
616 Power Electronics

voltages and high dv/dt a fibre optic is an expensive alternative, but unlike the pulse
transformer, has now lower cut-off frequency.

Figure 16.25d shows the Miller capacitance associated with the MOSFET and IGBT.
During switching, the Miller capacitance charging and discharging currents slow the
switching transition as power level current is injected into and from the gate level
circuitry. A low impedance gate drives reduces the Miller capacitance effects.
A commonly overlooked capacitively injected current is that associated with the use of
oscilloscope probes, when measuring power level signals. The scope probe ground
should be physically connected to an appropriate power ground point, rather than
signal ground. Always use the highest voltage step-down ratio probes as possible, since
capacitance tends to decrease with increased step down ratio.

Reading list

Siemens, Components, 1986.

Mullard, Book 3 Parts 1b and 1e, 1984/85.

Rifa, Capacitors, 1986.