This is only a preview of the November 2021 issue of Practical Electronics. You can view 0 of the 72 pages in the full issue. Articles in this series:
|
Circuit Surgery
Regular clinic by Ian Bell
Capacitor Dielectric Absorption
–
+
+
–
–
–
+
–
+
–
+
–
+
+
–
+
–
–
–
+
C harging
vo ltage
+
+ + + + + + + + + + + + + + + + + + + + + + + +
R ecove ry
vo ltage
–
–
+
+
+
–
+
–
+
+
–
–
–
+
+
–
–
+
–
+
+
+
–
–
–
+
+
–
+
–
52
+
+
Most, but not every month, LTSpice
is used to support descriptions and
analysis in Circuit Surgery.
The examples and files are available
for download from the PE website.
+
Simulation files
–
As the name suggests, dielectric absorption
is due to the properties of the capacitor’s
dielectric – the insulating material which is
placed between the two conducting plates
to form the capacitor. The dielectric is an
insulator, so unless the breakdown voltage
is exceeded, very little electric charge
flows through it when a voltage is applied
across the capacitor. However, significant
internal charge redistribution does occur
in the form of polarisation of dipoles.
Dipoles are parts of a material which can
have a distribution of charge (in simple
terms, positive at one end and negative at
the other). This can occur at the atomic,
–
+
–
Dielectrics
C apacitor
vo ltage
U npolarised
+
about capacitors on the EEWeb forum:
‘I have a question, but have not found
an answer on the Net. An electrolytic capacitor, if it is in short circuit for a long
time, discharges and its voltage goes to
zero. But if you remove the short circuit,
the voltage on the terminals slowly rises
again. Why? Thank you very much.’ As
was quickly pointed out by forum expert
Peter Traneus Anderson, this effect is called
‘dielectric absorption’. Other names are
also used, such as ‘voltage recovery’ and
‘dielectric soakage’.
As John indicates, dielectric absorption
is a non-deal phenomenon which occurs
in capacitors – it is not exclusive to
electrolytic capacitors but tends to be
larger in capacitors of this type. The effect
is most obvious if a capacitor is charged
for a long time, then discharged to zero
volts very quickly and immediately left
open circuit. After this, both real and ideal
capacitors will start at zero volts. An ideal
capacitor will stay at zero volts, but a real
capacitor will develop a non-zero ‘recovery’
voltage across it after a time delay due to
dielectric absorption.
–
J
ohn Curtin asked a question
– – – – – – – – – – – – – – – – – – – – – – – –
P olarised by an applied electric f ield
Fig.1. Polarisation of dipoles in a
capacitor dielectric.
ionic, molecular and higher structural
levels (such as cells in biological tissues).
An electric field applied to a dielectric
(in our case, a voltage applied across a
capacitor) causes the dipoles to rotate to be
more aligned with the applied field (with
no field applied they are randomly aligned)
– see Fig.1. The field created by the aligned
dipoles interacts with the applied field and
threby allows the capacitor to store more
energy (hold more charge) than if there
was just a vacuum between the plates. The
more polarisation, the more additional
storage (so more capacitance for a given
physical size of device). When the field
is removed (capacitor discharged) the
dipoles return to their random orientation
– referred to as ‘relaxation’.
The polarisability of the material is
characterised by its permittivity. The
permittivity of a material can be stated
as an absolute value (symbol – Greek
epsilon), or as relative permittivity ( r),
which is relative to the permittivity of
free space (vacuum) ( 0); that is, = r 0.
The symbol (Greek kappa) is also used
for relative permittivity. Materials used for
capacitor dielectrics tend to have a large
relative permittivity (sometimes referred
to as ‘high K’) as this allows for smaller
capacitors for a given capacitance value
and voltage rating.
The polarisation of dielectrics is a
complex process because there are typically
multiple types of dipole within materials
and the polarisation and depolarisation
C harging D ischarging
O pen circuit
T ime
Fig.2. Behaviour of a capacitor exhibiting
dielectric absorption.
(relaxation) is not instantaneous. In
simple terms, dielectric absorption is
due to the fact that not all of the dipoles
depolarise in the time taken to rapidly
discharge a capacitor to zero voltage. They
continue to hold charge (‘absorbed’ by
the dielectric). Initially, if the capacitor
is open circuited after discharge, this
does not result in a voltage at the plates.
However, the charge will get transferred
to the capacitor plates over time as the
dipoles depolarise. This will cause a
charge build up on the plates leading to
the recovery voltage that John asks about.
Capacitor behaviour
and measurement
The behaviour of a capacitor exhibiting
dielectric absorption is illustrated in Fig.2,
which is a graph of the voltage across a
capacitor against time, undergoing the
charge/discharge/open-circuit sequence just
mentioned. Subjecting capacitors to this
process can be used to make measurements
to observe or characterise their dielectric
absorption. A circuit concept for doing
this is shown in Fig.3. The specifics of
implementation may vary; for example, in
R 1
1
S1
2
3
V c
R 2
C
H igh- impedance
vo ltmeter
Fig.3. Circuit concept for observing or
measuring capacitor dielectric absorption.
Practical Electronics | November | 2021
the simplest cases manual switching could
be employed, but automated switching
using relays and various improvements to
the basic idea can be incorporated.
In the first part of the graph the capacitor
has been charged for a long time, so the
voltage across it is effectively equal to the
charging voltage. With the circuit in Fig.3
this is achieved by applying VC through
R1 with S1 in position 1. A ‘long time’
could be from minutes to hours. Typically,
real capacitors would be charged to a
suitable voltage (maybe their rated working
voltage) when making dielectric absorption
measurements, taking care to safely limit
the initial current surge when charging is
started. Applying the maximum working
voltage to a discharged capacitor through
too small a resistance (R1 in Fig.3) may
result in a damaging current surge (eg,
by connecting the capacitor directly to a
power supply at that voltage).
Once fully charged, the capacitor is
discharged quickly relative to the charge
time, this is achieved by connecting a
suitable resistor across the capacitor (R2 in
Fig.3, with S1 in position 2). The switchover
from charge to discharge mode should
occur quickly with the capacitor briefly
open-circuit (S1 break-before-make). The
capacitor will hold its charge during this
time. Capacitors tend to discharge slowly
if left charged and open circuit, so if the
aim is to observe dielectric absorption,
the time between charge and discharge
should be brief.
To allow the dielectric absorption
phenomenon to be observed, the discharge
time should be just sufficient to get very close
to zero volts as quickly as possible. This
means discharge through a low resistance
(R2) – but, for real tests, the current must be
limited to a value that will not cause any
problems. Standard measurements may
use fixed times and resistance values for
discharge to ensure the same conditions
are used when comparing devices.
After the quick discharge the capacitor
is disconnected from the discharge resistor
so that it is open circuit (S1 in position 3 in
Fig.3). The capacitor voltage will rise, as
shown in the final section of the graph in
Fig.2. Initially, the increase will be rapid,
but the rate of change will slow over time.
The final voltage reached is the recovery
voltage. The voltage across it over time
can be measured using a voltmeter, which
should have a very high input impedance
if accurate results are required.
Dielectric absorption (DA) figures for
capacitors are defined as the percentage
of the recovery voltage (VR) relative to
the charging voltage (VC) under specified
conditions (charge and discharge time,
discharge resistance, open-circuit time after
which voltage is measured, temperature
and so on):
DA(%)= (VR/VC) × 100
Practical Electronics | November | 2021
As well as describing the behaviour of a
capacitor exhibiting dielectric absorption,
the preceding discussion has sketched a
basic approach to measuring the DA value.
You will find references to MIL-C-19978D
and EN 60384-1 as standard test produces,
but their relevance in the context of modern
electronic systems may be debatable. For
further discussion on this, and a more
comprehensive exploration of measuring
dielectric absorption, readers might be
interested in a paper published online
by Leslie Green: Practical Exploration of
Dielectric Absorption in Capacitors for
the 21st Century (see: http://lesliegreen.
byethost3.com/articles/Dielectric.pdf).
Effects of dielectric absorption
The recovery voltage due to dielectric
absorption varies significantly for different
types of capacitors. Capacitors with polymer
dielectrics, such as polyphenylene sulphide,
polypropylene and polystyrene tend to have
the lowest DA values, which can be well
below 0.1%. Dielectric absorption is higher
for ceramic capacitors (around 0.5 to 2.5%)
but depends on the type of ceramic, with
Class 1 C0G (NP0) having lower DA values
than class 2 X7R. The highest DA values
are for electrolytic capacitors, which can
reach 10% or 15%.
Dielectric absorption makes itself felt
(literally in some cases) in different ways
depending on the type of capacitor and the
application. For high-value, high-voltage
capacitors there is a possibility for the
recovery voltage to deliver a significant,
even lethal electric shock to someone
handling the capacitor. Such capacitors
are usually shipped and stored shorted
out, but can still recover after the short is
removed, so in situations where this is a
potential problem people need to follow
appropriate safety procedures.
Dielectric absorption can also cause
problems in some signal processing
applications, with one of the most important
examples being sample and hold circuits
(see Fig.4) – these are commonly used to
provide steady input voltage samples to
analogue-to-digital converters (ADCs). The
input signal is sampled by rapidly charging
a capacitor to the value of the input voltage
at the instant that a conversion is required
(by closing the sample switch briefly). The
sample switch is then opened, and the
capacitor holds the value steady while
the conversion is performed, buffered by
a high-impedance unity-gain amplifier to
prevent the capacitor from discharging.
When used with a single input signal,
in many cases the voltage will not change
much between successive samples (with a
smoothly changing signal) and dielectric
absorption will not be a problem. However,
some signals will have abrupt changes,
and it is common for ADCs to be used
with multiplexers so that a single ADC
Multiplexe r
I n1
I n2
I n3
–
Sample
switch
+
O utput
( to AD C )
C
I nN
Fig.4. Sample and hold with input
multiplexer. The switches are electronic
(typically MOS transistors).
can convert signals from multiple input
channels (see Fig.4). Here it is quite likely
that the capacitor will be charged to near
the maximum input voltage for one sample
and close to zero on the next – if different
channels are at opposite ends of the input
voltage range. This creates a situation
very similar to that shown in Fig.2. Even
if the conversion takes place relatively
quickly compared with the time taken to
reach the full recovery voltage, the effect
of dielectric absorption will result in an
unwanted change in voltage on the sample
capacitor – remember, the rate of change is
relatively fast initially. This can result in
errors in the converted value. For example,
for a 12-bit ADC a change of only 0.025%
of the full-scale voltage is sufficient for a
change of the converted value (by 1 leastsignificant bit).
Another circuit where dielectric
absorption may have a significant effect
is the op amp integrator (see Fig.5), one of
the ‘standard’ op amp applications. The
output of this circuit is proportional to
the integral of the input voltage over time.
In some situations, it is necessary to reset
the integrator (or more precisely, set the
initial conditions of the integration). This
is shown in Fig.5, implemented as a switch
discharging the capacitor, although setting
a specific voltage other than zero may be
required. The situation for the capacitor in
Fig.5 is again very similar to the scenario in
Fig.2. The capacitor may be at a relatively
large value when the reset (discharge)
occurs, and the resulting recovery voltage
will be added as an error to the output
from the circuit after the switch is opened.
Model circuit
It is possible to model the dielectric
absorption of a capacitor using an equivalent
circuit of the form shown in Fig.6, which
R eset switch
C
V in
R
–
V out
+
Fig.5. Basic op amp integrator circuit.
53
R A1
R A2
R A3
R AN
C 1
C A1
C A2
C A3
C AN
Fig.6. Equivalent circuit model for capacitor
dielectric absorption.
Fig.7. Equivalent
circuit model
with five RC pairs
used for dielectric
absorption
and other nonideal capacitor
characteristics.
R S
L S
R L
was published by Paul Dow in 1958 in a
paper in the IRE Transactions on Electronic
Computers. The model helps us understand
Ω
the behaviour of capacitors with respect
to dielectric absorption and simulate these
I nitially
effects at circuit level without needing a
uncharged
1 V
0 .1 µ F
(0 V )
deep knowledge of the physics of dielectrics
at the atomic and molecular level. The
equivalent circuit comprises a set of N
Fig.8. Example circuit for RC time
RC circuits (RA1 and CA1 to RAN and CAN)
constant – charging.
in parallel with an ideal capacitor, C1. C1
has the rated value of the capacitor. The
capacitors modelling the absorption (CA)
will have a smaller value than C1 (the
I nitially
charged
Ω
0 .1 µ F
total absorption capacitance relative to
to 1 V
C1 determines the final recovery voltage).
The resistor values will typically be large
(megohms to giga-ohms) to account for
Fig.9. Example circuit for RC time
the slow change of the recovery voltage.
constant – discharging.
Intuitively, the
model in Fig.6
explains the
1 .0 V
behaviour shown
0 .9 V
in Fig.2. The
0 . 8V
long charge time
0 .7 V
ensures that all the
6 3% charged
0 .6 V
af ter R C = 0 . 1 ms
capacitors in the
0 . 5V
circuit are charged
0 .4 V
to the applied
D ischarged to 37 %
af ter R C = 0 . 1 ms
voltage, despite the
0 . 3V
large resistor values
0 .2 V
in the absorption
0 .1 V
part. The quick
0 V
discharge through
0 . 0 ms 0 . 2 ms 0 . 4 ms 0 . 6 ms 0 . 8ms 1 . 0 ms 1 . 2 ms 1 . 4 ms 1 . 6 ms 1 . 8ms 2 . 0 ms
a relatively small
resistor removes
Fig.10. Example RC charge and discharge waveforms based on
the charge from C1,
Fig.6 (0 to 1.0ms) and Fig.7 (1.0 to 2.0ms).
R A1
R A2
R A3
R A4
R A5
C A1
C A2
C A3
C A4
C A5
C 1
but the large resistors in series with the
other capacitors mean that little charge is
removed from them. During the open circuit
period the capacitors in the equivalent
circuit will share charge – specifically
charge will move off CA1 to CAN, and onto C1
until the voltages on the capacitors equalise
(if the capacitors are at different voltages this
difference will appear across the resistors,
causing current to flow, moving the charge
between the capacitors). We observe the
voltage on the terminals (across C1) rise –
this is the recovery voltage, or the capacitor
appearing to recharge itself. The large
resistor values mean that this will take a
relatively long time compared with the
discharge of C1 just performed.
The circuit in Fig.6 only models the
effect of dielectric absorption, but other
components can be added to cover other
non-deal characteristics of capacitors
if required. An example is shown in
Fig.7, which uses five RC pairs to model
absorption and also includes equivalent
series resistance (ESR) (RS), equivalent
series inductance (ESL) (LS) and leakage
resistance (RL).
Curves and time constants
The concept of RC time constants is
important in the dielectric absorption
equivalent circuit. This can be illustrated
by the basic charge and discharge circuits
shown in Fig.8 and 9. The well-known
exponential charge and discharge curves
for these ideal RC circuits are shown in
Fig.10. The time constant for an RC circuit
is the value obtained by multiplying the
resistor and capacitor values together (the
result has units of time). For example, in
the circuits in Fig.8 and 9 we get RC =
1.0kΩ × 1.0µF = 0.1ms. When a capacitor is
charged via a resistor from a fixed voltage,
the voltage across the capacitor will be
63% of the applied voltage after time RC
has elapsed (as seen in Fig.10). Similarly,
when a capacitor is discharged through a
resistor the voltage will reach 37% of the
initial value after the time constant time
has elapsed.
The time constants of each of the set of
RC circuits in the equivalent circuit in Fig.6
typically increase by a factor of ten along
the sequence from 1 to N. Time constants
for dielectric absorption can be very long –
implying many hours or more to approach
the final recovery voltage. It is also worth
Fig.11. Simulation example.
54
Practical Electronics | November | 2021
noting that a dielectric absorption means
that for real capacitors the charge and
discharge curves do not follow the perfect
exponential characteristic of ideal devices.
Simulation example
The circuit in Fig.11 is an LTspice schematic
which can be used as a starting point to
explore the equivalent circuit in Fig.6.
The switches (S1 and S2) are controlled
by the pulse waveforms from V2 and V3 to
operate the circuit in a similar way to that
described earlier for Fig.3. The simulation
is configured with very idealised switches
(very high off and very low on resistance,
as defined by the SW1 model), so this is not
a realistic circuit in terms of what is done
to the capacitor. The switch model can be
changed, or resistors added (as in Fig.3)
to make this more like a real test circuit.
Just three absorption RC circuits are
used to keep the simulation simple, with
their time constants increasing by a factor
of five, so the simulation does not have to
be too long. The longest of these is R3C4
which is just under one second. Thus, it is
reasonable in this context that charging the
capacitor for a ‘long time’ has a duration
of five seconds – an RC circuit charges to
more than 99% of the applied voltage in
5RC. V2 is configured to close switch S2
from simulation time 1s to 6s to charge
all the capacitors in the equivalent circuit
(see the simulation waveforms in Fig.12).
Immediately after this, C1 is discharged very
quickly by closing S1 for 200µs, after which
both switches are open. The short discharge
duration and relatively large values of R1
to R3 compared to S1’s resistances mean
that C2 to C4 will lose very little charge
while S1 is closed – this can be confirmed
by zooming in on the waveform.
It is difficult to see the details of the
recovery voltage in Fig.12, so Fig.13
provides a zoom-in. The final voltage
is 100mV, which is about 10% of the
applied voltage. We can also obtain this
figure by analysing the charge sharing
– the total absorption capacitance is
3 × 37nF = 111nF. The overall total
capacitance with C1 is 1111nF. With
Fig.13. Zoom in to show recovery voltage.
Practical Electronics | November | 2021
Fig.12. Simulation results from the circuit in Fig.11.
C1 starting at 0V and C2 to C4 at 1V,
the four capacitors sharing charge will
result in a final voltage of (111/1111)
× 1V = 100mV. The recovery time is
dominated by the R3C4 time constant,
and we simulate the recovery for 5 ×
R3C4, so we are close to the final value.
The simulation can be modified
to study the effect of different
dielectric absorption characteristics or
measurement scenarios. For example,
Fig.14 shows the effect of adding a 10MΩ
resistor across C1 – which could be the
resistance of a voltmeter. This effect is
to discharge C1 at the same time as it
is being charged from the absorption
capacitors, so the recovery voltage peaks
and then decreases.
Fig.14. Recovery voltage with 10MΩ resistor across C1.
55
|