AN542
P
E R F OR MA N C E
I
MP RO VE ME NTS I N
C
LASS
D A
U D I O
A
M P L I F I E R S
U
S IN G
S i 8 2 4
X
A
U D I O
D
R IV E R S
1. Introduction
The Si8241/44 Audio Gate Drivers facilitate high-fidelity, Class D amplification while offering many other
advantages over competing gate drivers. Based on Silicon Labs' proprietary CMOS isolation technology, these
drivers offer the benefits of:
Precise dead time adjustment for low total harmonic distortion (THD) and high system efficiency
Individually-isolated drivers to facilitate simpler system topologies and external component flexibility
High functional integration for small installed size and competitive cost
The gate driver IC is a critical system component because it influences both system architecture and system
performance. The Si8241 Audio Gate Driver is a high-voltage driver capable of switching at frequencies up to
8 MHz to realize pre- or post-filter control. These devices have integrated precision dead time generators for low
THD and high efficiency, as well as isolated output drivers that eliminate the need for input level shifting and
facilitate the straightforward implementation of a two-state, half-bridge Class D amplifier.
This application note discusses the Silicon Labs Class D reference design, a stereo, two-state, half-bridge Class D
amplifier that delivers 120 W per channel into 8
or 150 W into 4
.
The reference design and demo board is
available at
www.silabs.com/isolation
and uses the Si8241 Audio Gate Driver driven by a self-oscillating PWM
modulator, as shown in Figure 1.
VBOOT
Bootstrap
5V
Audio
In
Si8241
ISOdriver
VDDI
VDDA
VOA
PWM
GNDA
DT
VDDB
VOB
GNDB
+50V
Error Amplifier
-
+
Comparator
-
+
R
DT
Speaker
GND1
Filter
-50V
Feedback Amplifier
-
+
Figure 1. Si8241-Based Class D Amplifier Block Diagram
Rev. 0.1 10/10
Copyright © 2010 by Silicon Laboratories
AN542
AN542
2. The Si8241 Audio Gate Driver
Every so often, a new IC is introduced that challenges the current technological hegemony. With features that
make these products the perfect drivers for Class D amplification, the Silicon Labs Si8241/44 Audio Gate Drivers
represent a new standard for the Class D amplifier industry. Key features are outlined in the following sections.
2.1. Programmable Dead Time
It is well documented that a precise dead time setting is critical in Class D amplifiers. During dead time, both the
high-side and low-side MOSFETs are off. However, the low-side MOSFET body diode continues to conduct
current, which manifests itself as output distortion. Too short a dead time causes shoot-through current that
reduces system efficiency; too long a dead time increases THD, negatively impacting audio quality. While
competing audio drivers typically have coarse, digital dead-time settings (i.e. 1 of
n
delay values), the Si8241/44
Audio Gate Drivers have a precise linear dead time setting that programs with a single external resistor. This
feature provides the resolution necessary to precisely set dead time for optimal system performance. The Audio
Gate Driver dead time equation is shown in Equation 1.
DT
=
10R
DT
where Dead Time (DT) is in ns and R
DT
is in k
Equation 1. Audio Gate Driver Dead Time
Per Equation 1, the Silicon Labs Class D amplifier uses a 2 k resistor to generate 20 ns of dead time. Changing
this dead time value to 18 ns only requires changing the RDT to 1.8 k (connected from the dead-time pin (DT)
input to ground). This setting mechanism allows dead time to be incrementally increased or decreased in
nanosecond increments, instead of tens of nanoseconds like competitive products.
2.2. Input/Output Isolation
Implementing a two-state, Class D amplifier can be difficult due to input level shifting requirements, and most
available Class D drivers lack the capability to eliminate level shifting. Drivers that do eliminate level shifting have
other peculiarities making them less-than-ideal for Class D operation (example: driver output ground terminal
referenced to the –VBUS rail, requiring the input drive signal to be level-shifted). This is not the case with the
Si8241 Audio Gate Driver where the isolation (i.e. level shift function) is implemented internally and is transparent
to the user. The Si8241 Audio Gate Driver controlled by TTL input signal levels drives the outputs to ±VBUS, and
only a single TTL PWM input signal is required to drive a two-state Class D amplifier.
2.3. High-Voltage Outputs
The Si8241 is capable of switching very high voltages (up to a 1,500 Vdc peak driver-to-driver differential voltage is
possible) allowing a ±750 VBUS. For practical Class D amplifier designs, a voltage of ±100 Vdc can deliver an
astounding 600 W of audio power into 8
.
2.4. Output Current Drive
Class D amplifier switching MOSFETs should not be "slammed" on and off by excessively high current gate drivers.
With its 0.5 A peak current outputs, the Si8241 Audio Gate Driver hits the sweet spot for Class D operation up to
400 W. Power levels beyond 400 W typically require larger MOSFETs and, consequently, more gate drive. For
applications of this type, the Si8244 (4A) Audio Gate Driver provides the required added gate drive, where rise and
fall times can be adjusted with a series gate resistor.
2.5. High-Frequency Operation
One of the best attributes of the Si8241 Audio Gate Driver is its 8 MHz maximum switching frequency, making it the
fastest driver on the market for Class D operation. The Silicon Labs Class D reference design operates at
approximately 500 kHz, and operating the amplifier between 500 kHz and 1 MHz dramatically reduces the high-
frequency artifacts, resulting in a remarkably clean audio waveform.
2
Rev. 0.1
AN542
3. Gate Drive Structure
Figure 2 illustrates the simplicity of the Si8241 driving a two-state, half-bridge Class D amplifier. The boot supply
tied to D1 must be 12 V higher than the –50 V reference (–38 V) so that the MOSFETs each have a 12 V drive
signal.
-38V
+50V
D1
C1
R1
Si8241
ISOdriver
PWM
PWM
NC
VDDA
VOA
GNDA
NC
NC
VDDB
VOB
GNDB
Speaker
+5V
C2
SHDN_HI
VDDI
GNDI
DISABLE
DT
R2
R3
NC
VDDI
C3
-38V
C4
-50V
Figure 2. Si8241 Audio Gate Driver Gate Drive Circuit
4. Filter Inductor Selection
In early Class D amplifiers, the most common type of inductor used was a toroid of Type 2 iron material
manufactured by Micrometals (Figure 3). This was because Type 2 material has a very low magnetic permeability
(µ in Equation 2), slightly higher than an air core inductor, which has a linear BH curve and will not saturate.
B
-
= ---
H
Equation 2. Magnetic Permeability
However, when an iron core inductor operates further up on the BH curve, its permeability decreases with the
falling slope of the BH curve and approaches zero as the core saturates. When this happens, there are no more
iron particles to align within the core, and the inductance approaches zero.
Recently, a number of manufacturers developed inductors specifically for Class D amplification, such as Sagami
Elec. Co., whose inductors are used in the Silicon Labs Class D reference design. The inductance value of these
devices does not decrease with current as happens in most inductors. Instead, it remains relatively flat over the
range of interest and is a good choice for Class D amplifier applications.
Figure 3. Micrometal Inductor BH Plot
Rev. 0.1
3
AN542
5. Reference Design Board Architecture
The Silicon Labs Class D reference design architecture uses a phase-shift, self-oscillating modulation approach
capable of achieving a far greater signal-to-noise ratio than that of a clock-driven amplifier. This self-oscillating
implementation eliminates the circuitry necessary to generate the triangle waveform (see "8. Self Oscillation" on
page 6 for details). To keep the circuit as simple as possible, a two-state, half-bridge is implemented and
demonstrates the benefits of using the Si8241 Audio Gate Driver.
5.1. Control Theory Basics
Looking at the equation derivation in the frequency domain, Figure 4 shows input signal R(s), which is the Laplace
transform of the time domain regulation input signal R(t). C(s) is the Laplace transform of the time-domain-
controlled output signal C(t), where s = j.
R(s)
+
-
E(s)
G(s)
C(s)
H(s)
Figure 4. First Order Control Loop Model
E
s
=
R
s
–
C(s)H(s)
Equation 3.
C
s
=
E(s)G(s)
Equation 4.
Substituting C(s) from Equation 4 into Equation 1, and solving for E(s)/R(s) yields:
E(s)
1
---------- = -------------------------------
-
-
R(s)
1
+
G(s)H(s)
Equation 5.
Similarly, substituting E(s) from Equation 4 into Equation 3 and solving for C(s)/R(s) yields:
C(s)
G(s)
-
---------- = -------------------------------
-
R(s)
1
+
G(s)H(s)
Equation 6.
Examining these equations within the context of a Class D control loop, G(s) is the forward transfer function; H(s) is
the feedback transfer function, and E(s) is the error signal (i.e., the difference between the input and output).
4
Rev. 0.1
AN542
6. Closed Loop Transfer Function
The closed loop transfer function is defined as the ratio of the controlled variable to the input variable. The
controlled variables are the speaker terminals, and the input variable is the MP3 player input, the CD input, or
some other input source connected to the amplifier. Therefore, the equation for the closed-loop transfer function is
given by Equation 7.
G(s)
-
Closed_Loop_Transfer_Function
= -------------------------------
1
+
G(s)H(s)
Equation 7. Closed Loop Transfer Function
The closed loop gain describes how the output responds over the audio bandwidth to the input regulation signal. It
is understood that the output should have a specific closed loop gain with respect to the input regulation signal, and
that gain should be as flat as possible over the audio bandwidth. The inductor between the controlled variable and
the speaker terminals plays a crucial role in the performance of the amplifier as previously discussed. The closed
loop gain of the Silicon Labs Class D reference design is implemented such that approximately 1 Vpp input will
yield full output power into an 8
load.
7. Open Loop Transfer Function
The open-loop transfer function is obtained by breaking the loop at some arbitrary point and traversing the entire
loop back to the same point. When H(s) = 1, the open loop and the forward transfer functions are identical.
Therefore, the open loop transfer function is given by Equation 8.
Open_Loop_Transfer_Function
=
G(s)H(s)
Equation 8.
The open loop transfer function determines whether the loop is stable, as well as determining what the overall open
loop gain of the amplifier will be over the audio bandwidth. The higher the open loop gain, the lower the error signal
and, therefore, the more easily the control loop can keep the output following the input command. Some early
Class D amplifier designs used an integrator for the error amplifier. This produced high gain at low frequencies but
low gain at high frequencies due to the pole produced by the integrator. This caused the THD to increase
dramatically above 5 kHz, destroying the high-frequency response of the amplifier. A better solution is to keep the
open loop gain constant and as high as possible throughout the audio bandwidth. This should yield a constant THD
response, and, indeed, it does, as will be shown in the performance curves in "13. Performance" on page 8.
Care should be taken in designing the open loop response of the amplifier. The three key elements are the
bandwidth, phase margin, and gain margin. In designing a Class D amplifier, the target is to have 45° of phase
margin with a bandwidth of approximately half the switching frequency. The control loop cannot compensate for the
LC filter response since the filter is outside of the loop. The entire LC filter is designed as a Bessel function with a
load resistance of 6
.
Therefore, the filter is slightly underdamped at 8
and slightly overdamped at 4
.
This can
be seen by placing a 100 mVpp square wave into the input and looking at the output response with an 8
load
and a 4
load.
Rev. 0.1
5
淘e淘
京公网安备 11010802033920号
Copyright © 2005-2026 EEWORLD.com.cn, Inc. All rights reserved
电子工程世界
