Power Lines: The Next “Digital” Revolution?
Inside the future of switching power amplifier technology
By
Pat Quilter
|
|
 |
As everyone knows, digital technology has revolutionized the handling
and processing of audio information, bringing a new level of clarity,
error correction, and programmability to audio systems. |
It’s important to clarify the difference between “digital” and “switching”
technology, and explain the advantages and drawbacks of forthcoming “switching”
or “Class D” amplifiers.
EARLY METHOD
Analog signal processing represents the audio waveform as a continuously
variable voltage level inside a signal processing device. This was the
first method of processing audio.
The microphone converts sound pressure waves into electrical waves, which
travel through connectors to mixers, equalizers, and limiters, and finally,
to power amplifiers, where the signals are then magnified until they are
large enough to drive the speaker cones to the desired sound level. All
the way through this chain, the electrical signal bears a close relationship
to the sound it represents. It is an electrical analog of the audio waveform.
(Figure 1)
|
|
|
The whole process has a compelling directness, it was easy to get started with, and the signal quality has improved steadily over the years. So what’s the problem? The problem is that analog signals are subject to many forms of degradation, especially when recorded and stored for long periods. |
These degradations accumulate until the signal becomes noisy, distorted,
or muffled. Each form of degradation blends in with the original signal
and cannot, then, be separated. This limits how much can be done with
the signals, and the various processors are subject to tolerances and
aging that produce erratic results.
The earliest computers used analog signals and their designers quickly
discovered that random noise and small changes in the signals degraded
accuracy and prevented getting exact answers to their problems. The solution
was to encode data as exact numbers rather than voltage levels. To ensure
the greatest possible reliability, the simplest possible encoding system
was used the binary system, in which we distinguish only two values,
“one” and “zero”. These values were conveniently being represented as
“on” or “off” states of relays, tubes, and eventually, solid-state logic
gates.
By using enough on-off “bits” and certain combining rules, any desired
level of accuracy can be achieved, and minor disturbances of the signal
have absolutely no effect, as long as “on” and “off” remain distinguishable.
As data is processed and re-copied, it is actually “cleaned up” so that
minor errors are eliminated.
Therefore, digital processing, by definition, is the art of representing
audio signals as a stream of numbers (digits), generally within a computer
data processor. The process starts with one initial approximation, in
which the exact analog signal is encoded to the nearest available digital
value.
PRECISION & REPEATABILITY
By using enough bits, and enough samples per second, the analog signal
can be represented as accurately as required. After making the initial
analog-to-digital (A-D) conversion, the data is then handled with no further
losses, using the amazing power of modern computers. Only at the end do
we return to the analog world with a digital-to-analog (D-A) conversion.
(Figure 2) Thus, we have clean, clear audio files that can be downloaded,
copied, and edited with no further degradation.
|
|
|
Digital signal processing (DSP) allows the audio signal to be manipulated in many ways, with better precision and repeatability than similar analog techniques, and performing operations such as signal delay, pitch correction, or echo cancellation that are almost impossible with analog methods. And, digital networks permit many audio devices to be linked together, with dynamic “addressing” that allows the system to be reconfigured at will without unplugging cables. |
All of these processes are getting less expensive. Computer designers
are constantly upgrading the amount of data that their systems can handle.
The circuitry required to process each bit keeps shrinking. But power
amplifiers have to produce a definite amount of power to drive loudspeakers
to a desired sound level, and this requires large transistors, power transformers,
heat sinks, high voltages, strong enclosures, and many protection circuits
to prevent runaway conditions and breakdowns.
Thanks to many decades of development, modern linear amplifiers (the usual
form of analog) handle thousands of watts of power with extremely high
resolution and low noise floors. However, the power handling circuitry
is bulky and generates a lot of waste heat, requiring heat sinks and cooling
fans.
For many decades, there has been another method of producing high power
signals, called “switching” or “Class D” technology. To appreciate the
difference, we need to understand the two concepts.
MAGNIFIED REPLICA
The classic linear technique starts with the power supply. This converts
AC power into a steady voltage reservoir, ready to supply the needs of
the amplifier. High power transistors, called output devices, valve a
certain amount of this power to the speakers, in a continuously variable
manner, controlled by the input signal so that the loudspeaker is driven
by a magnified replica of the input signal. (Figure 3)
The technique is relatively straightforward, and the two main problems
are error correction (making the output track the input exactly), and
the waste energy lost by the valving process. This waste energy appears
as heat, and requires multiple power transistors, large cooling systems,
and careful protection to prevent burnout during overloads. The waste
heat is inherent in the process, and although it can be reduced with some
effort, it will always limit the amount of power that can be handled.
|
|
|
The best way to reduce the waste heat is to eliminate the “valving” process. An output transistor uses no power when it is turned off, and passes almost all the power supply voltage into the load when it is fully turned on, with minimal internal losses. It is only the in-between, partially activated state where most of the power is lost. Unfortunately, this is the normal operating region for a linear amplifier. |
To correct this, we need a method to operate the output transistor only
as an “on-off” device, while still delivering all the in-between levels
into the speaker.
The solution is to switch the output transistor fully on or fully off
at an extremely rapid rate, and control the average output power by varying
the percentage of “on” and “off” time. This technique is called Pulse
Width Modulation (PWM), and basically “maps” a continuously varying input
voltage level into a continuously varying pulse width.
If the switching occurs rapidly enough, the switching frequency can be
filtered away at the output, leaving only a magnified version of the input
signal. It is still an analog process; instead of a continuously variable
voltage level, we have a continuously varying on/off ratio. (Figure
4)
This technique is theoretically capable of almost perfect efficiency,
thus promising much higher power with much smaller heat sinks. However,
it is a more indirect method of producing high-power audio, so a number
of new problems are introduced. The output transistors must be able to
switch almost instantly, at rates about 15 times higher than the desired
audio limit, thus requiring at least a 300 kHz switching frequency to
handle full range audio up to 20 kHz.
Output accuracy is affected by power supply fluctuations, small delays
in switching, thermal changes in the power devices, the modulator used
to produce the PWM signal, the quality of the output filter, and other
variables. These problems are being solved, and we can expect more use
of switching technology in the future.
DON’T BE MISLED
One unfortunate confusion has entered our vocabulary, however. Because
switching amplifiers use “on-off” operation, they superficially resemble
binary logic circuitry, and are frequently called “digital” amplifiers.
This is quite misleading, as the switching amplifier is still very much
an analog system, with all the usual problems of analog error correction.
Although it may seem like a minor point, by misusing “digital” in this
way, we create the danger that users will assume that all “digital amps”
are inherently accurate, like other digital technologies. Actual switching
amplifiers are considerably less accurate than well-designed linear amplifiers,
and this has been one of the barriers to wider use.
Users need to understand that high-fidelity switching amps are even harder
to design than high-fidelity linear amps. Further, the complexities of
switching circuitry have so far cost more than conventional solutions
with similar power.
QSC Audio is actively researching switching amplifier technology, actively
seeking to solve the problems of cost, fidelity and reliability. When
perfected, it will dramatically reduce waste heat and may even eliminate
the need for cooling fans.
However, it will be years before switching amplifiers can meet or beat
the very high standard set by linear amplifiers, especially those that
use switching power supplies to reduce size and weight. We expect to use
switching amplifiers initially in those applications that absolutely require
more power in confined spaces with lower heat loss.
As cost is reduced and fidelity improved, this technology will spread
more widely into general-purpose power amplifiers.
Patrick Quilter is the “Q” in QSC Audio, founding the company in 1968 and remaining senior amplifier designer for the past 30-plus years. QSC has shipped about 2 billion watts worth of his product designs.
October 2003 Live Sound International