Class-D Amplifier Design: High-Quality Audio Technology Comes of Age
By Peter Baasch
The high efficiency of Class-D devices means
they run cool and can be made very compact. It
is no surprise, then, that demand has boomed
as consumers’ desire for smaller and lighter
mobile phones, portable computers, Personal
Digital Assistants (PDAs) and MP3 players with
high quality audio output has grown.
But designing Class-D amplifiers into a portable
application is not completely straightforward.
These are complex, fast switching digital devices
and it is all too easy to end up with an amplifier
that exhibits high Total Harmonic Distortion (THD),
audible noise and worse, that radiates Electro Magnetic
Interference (EMI) affecting both internal
and external circuits.
By adhering to some guidelines, a designer
can avoid the chief causes of noisy designs and
produce a cost effective, compact, powerful
device with good quality audio output.
Figure 1: STMicroelectronics’ TS4962M integrated amplifier
Understanding Amplifier Specifications
THD is perhaps the most significant measure of
the quality of a Class-D design. It is important
to note how the THD is typically specified when
comparing Class-D reference designs. For
example, characterizing an amplifier from 0Hz
(DC) up into the megahertz range results in a high
(and hence poor) THD since the switching frequency
is part of the measured range. For this reason,
THD is usually measured across the audible range
of 20Hz to 22kHz using a second order low pass
filter. The filter is only applied in tests, because in
normal use, the speaker itself (and the human ear)
attenuate higher frequencies.
Output noise is often specified with, and then
without, an A-weighting filter. The A-filter
resembles the human ear’s response and is a
way to eliminate the high frequency content of the
output signal generated by the amplifier’s
switching frequency. The value is specified in mV
or decibels referenced to maximum output power.
The output noise is often combined with THD (i.e.
THD+N), since both affect the audio quality in
almost the same way. As an example, output noise
from a 1-2W amplifier
is typically 50µV. These
figures are quoted from
the specifications, for
the TS4962M (Figure 1)
from STMicroelectronics
and the NCP2820 from
ON Semiconductor.
In addition to the THD+N
of a Class-D amplifier,
other undesirable effects
such as ‘click’ and ‘pop’
can be generated at the
speaker when power is
applied or removed from
the amplifier. To avoid this, most contemporary
amplifiers are equipped with a shut down or
sleep mode that allows the amplifier to switch
on or off in a controlled manner.
The penalty is a slight delay from switching on until
sound comes from the speaker, typically between
5ms and 10ms. From sleep mode, the start-up
is quicker at the expense of a constant standby
current of a few mA.
Eliminating the pop when switching on is easier
for certain types of amplifier without an output
filtering capacitor. A configuration with a center
ground is preferred. This can be done in two ways:
creating an artificial ground at half the supply
voltage, or introducing a switched capacitor
voltage inverter to generate a negative supply and
referencing the load to the true ground. The latter
is preferable, since the half supply voltage type is
susceptible to Electro-Static Discharge (ESD) and
can be difficult to stabilize.
Power Supply Rejection Ratio (PSRR) is also an
important amplifier parameter. In fact, it is more
important for a Class-D design than for Class-A
and Class-B types. In a Class-D amplifier, the
speaker is directly connected to the supply when
the switching element is closed. This means that
the amplifier is very sensitive to variations in
supply voltage.
PSRR is often specified at 217Hz, which is the
burst frequency of a GSM phone. A PSRR of 60dB
is usually enough to make sure that the 217Hz
disappears in the background noise from the
environment. For comparison, it is reasonable to
expect approximately, a 45dB signal-to-noise ratio
from an analog telephone line.
Today’s integrated amplifiers usually include a
range of safety features. For example, most
include a thermal shutdown capability built into
the amplifier. Some
parts also monitor
the current in each
output transistor
cycle-by-cycle and
shut down in case
of over current.
The NCP2820 from
ON Semiconductor,
for example, offers
this feature.
Most amplifiers are
equipped with differential
input stages.
This is mainly to
improve the immunity to common mode noise.
This can be hard to control, particularly in
compact mobile phone designs. In addition,
most codecs or digital-to-analog converters
designed for portable use, have a differential
output. This has the added advantage of
eliminating coupling capacitors. The differential
layout also shortens start up time from power on
or when disabling the power down mode.
The only minor drawback is that the gain setting
usually requires two resistors that have to be
matched. In some amplifiers, the gain setting is
internally fixed so these resistors can be omitted.
If the amplifier
is used in single
ended input mode,
each input should
be subject to
the same impedance,
and coupling
capacitors should
be added on both
input pins.
Figure 2: Class-D Devices’ high efficiency makes them ideal for portable music players
Boosting Output Power
To keep costs down and the power supply
compact, most portable equipment operates from
a single supply voltage. This is typically the 3.6V
output from a Li-ion battery, linearly regulated. If
the amplifier is single ended, the speaker has to be
AC-coupled directly to the amplifier. Consequently,
a bulky output capacitor must be used to maintain
satisfactory low frequency response.
An alternative solution is to employ two amplifiers
working in anti-phase. This removes the need
for the bulky capacitor and doubles the output
voltage swing, thus quadrupling output power and
providing more volume. This technique is known
as Bridge Tied Load (BTL) and has been used in
Class-A and Class-B amplifier powered car audio
systems for many years.
Applying BTL to Class-D is actually simpler than
with a Class-A or Class-B amplifier because most
of the input circuit can be shared. It only requires
two extra output transistors plus two inverters to
convert a single ended amplifier to a full H-bridge.
The fact that the output transistors act as switches
(either on or off) eliminates the need for biasing.
The demand for higher sound levels – for example
to clearly hear a call made in a noisy environment
on a mobile phone – has made the power
boosting BTL topology a popular alternative. Over
the past few years, the average power demand has
increased from around 250mW to 2W, and there
is no indication this trend will change as speakers
continue to shrink.
A drawback with the BTL topology is that each
half of the amplifier is subject to only half the
impedance of the speaker, demanding higher output
currents and lower on-resistance in the output
transistors.
Another simple way to increase output power
is to lower the speaker impedance. A few years
ago, the standard speaker in portable equipment
had a 25mm diameter and 16Ω impedance. But
the demand for smaller portable equipment has
shrunk the speaker to a mere 11mm diameter and
4Ω impedance. More power is needed to boost the
lower frequencies from such a small speaker and
to generate more sound.
Figure 3: Full second order filter (above) and half filter
Combating EMI
While a fast switching amplifier minimizes losses,
it can increase the potential for EMI. Often a
designer applies a low pass filter between the
amplifier and speaker to restrict EMI. This could be
a full second order filter comprising two inductors
and three capacitors or a half filter comprising one
inductor and one capacitor (see Figure 3).
A full filter combines differential and common
mode filters and provides the best attenuation of
both Electric and Magnetic (E and H) fields. The
half filter acts as a differential mode filter only
attenuating the H field and is recommended when
there are cost and space restrictions.
Some manufacturers are promoting filterless
Class-D amplifiers in order to reduce cost and
eliminate the space the filter takes up on the PCB.
The theory is that the inductance of the speaker
acts as a smoothing element. The amplifier and
speakers need to be placed close together to limit
the length of radiating lines (which is relatively
easy in portable equipment).
If the speaker is driven without a filter, the only
path for the high frequency switching currents
is through the speaker - and since the speaker
is partly inductive and partly resistive, power is
dissipated in the speaker. Consequently, there
are higher quiescent currents in the amplifier and
greater losses in the speaker. In comparison, a
filter is purely reactive, with minimal losses and
power dissipation. In addition, using the voice coil
of the speaker as a filter can lead to EMI radiating
from the coil.
Slew-rate limitation and active overshoot clamping
is an alternative method to reduce EMI and
eliminate the need for an output filter.
In summary, the designer faces two choices: if
quiescent current and losses in the speaker are
important, a half or full filter should be considered.
If cost and size are the overriding considerations,
then, provided the amplifier is placed close to the
speaker, it is possible to eliminate the filter; but
remember to consider the power dissipated in the
speaker and the EMI radiation from the speaker
coil.
The potential for EMI can be also be mitigated
through careful PCB layout – for example,
minimizing loop area to keep radiation low – and
careful consideration of the circuits that are placed
next to each other.
All amplifiers should be designed so that current
flow in the output stage is properly decoupled from
the sensitive parts of the circuit. The majority of
the current in the output stage is delivered by the
decoupling capacitors, so good design principles
dictate that tracks should be kept short and the
capacitors should be the types with low effective
series resistance.
Finally, it is also possible to use a spread
spectrum technique to limit EMI. The spread
spectrum technique works by modulating the
switching frequency slightly in order to lower
peaks in the noise spectrum.
Figure 4: Applying BTL to Class-D amplifiers boost sound output from the small
speakers typically used in today’s mobile phones
The Future for Class-D
While it is difficult to predict the future, the
next stage for amplifier integration for portable
applications is likely to be the addition of tone
and volume control. It also seems likely that the
switching operation to change between various
transducers will be integrated into the amplifier.
Whatever the technical trends though, the
future for Class-D amplifiers is assured because
of their efficiency and suitability for the booming
portable products market. This opens up lucrative
opportunities for designers able to produce
efficient Class-D designs with low THD+N and
powerful outputs.
And help is at hand: Future Electronics’
applications engineers are able to offer advice
on Class-D amplifier design, helping to
balance efficiency, cost, size and the quality of
the audio output. Future also supplies a range of
integrated amplifiers and support circuitry from
ON Semiconductor, STMicroelectronics, Zetex and
passives suppliers that offer the best combination
of price and quality on the market.
Figure 5: NCP2820 monitors current and shuts down when it detects over currents
Previous Page
