Class-D audio amplification circuits have been available for decades,offering high efficiency and small size compared to the more prevalentlinear Class-AB topology.
The voltage across the output device in a Class-AB amplifier equals thedifference between the speaker and rail voltages and changes accordingto the audio signal. The power loss is, therefore, this voltage timesthe output current.
Typical efficiencies are 30 percent. As a result, Class-AB outputstages usually require heat sinks and possibly a fan, particularly atpower levels exceeding 50W.
By contrast, the output devices' switching and conduction lossesdominate in a Class-D power amplifier. These transistors – typicallypower MOSFETs operating in saturation – routinely deliver efficienciesin excess of 90 percent. A passive LC filter smoothes the power stage'srailto- rail output to recover the audio signal.
Class-D headphone amplifiers help preserve battery life and smallsize in many portable systems. High-power applications use Class-Dtopologies to reduce size, weight and cost—notably by reducing oreliminating heat sink mass.
Several circuit behaviors affect Class-D audio performance. Notableamong these are the switching waveform's rise and fall times, thedifference in delays through the upper and lower bridge switches, thedeadtime between upper and lower conduction intervals, andswitch-timing jitter.
The rise and fall times depend on the FET driver's output currentcapability and the FET's total gate charge. Dead time produces a signaldependent gain error, resulting in output distortion. Noise affectingthe FET's switching times causes jitter, which significantly affectstotal harmonic distortion (THD) and sonic performance.
To reduce THD and noise, semiconductor manufacturers such asInternational Rectifier provide solid-state drivers that deliver 1A ormore gate-drive current with ±100V capability for drivinghalf-bridge topologies to power levels of 500W into 8 ohm .
These drivers may include laser-trimmed throughput-delay matching of20ns maximum between the upper and lower switches.
More recently, manufacturers have developed drivers that haveprogrammable fixed dead times to solve simultaneously three sources ofTHD: dead time accuracy, delay matching, and switching jitter. Forthese devices, laser trimming not only fixes dead time to severaldiscrete durations but it inherently eliminates the need for delaymatching.
After the driver switches one FET off, it minimizes switching jitterby forcing a programmable dead time before it turns on the oppositeswitch. In the interim, the system rejects noise signals that couldaffect the switch timing.
Dead-time programmability lets the user set the driver's timing tomatch the needs of the circuit's bridge FETs. As amplifier designersincrease the maximum output power for a given topology, they must uselarger bridge FETs, which have larger total gate charge. For a givengate-drive current, a larger amplifier will need a longer dead time.
|Figure1: IC gate drivers help simplify Class-D-amplifier design and provideTHD+N performance on par with the best Class-AB amplifiers.|
The most common objective assessment of amplifier performance isthe THD-plus-noise (THD+N) measurement. A halfbridge Class-D driver ICwith the features mentioned above and a careful layout result inexcellent THD+N (Figure 1 above ).
The driver is a 400kHz self-oscillating design with a feedback pathfrom the switch node. The resulting THD+N equals or improves on some ofthe best results obtainable by high-end Class-AB amplifiers.
One of the challenges of audio amplifier design is fault protection,especially against the over-current (OC) conditions that shortedoutputs commonly cause during system installation or reconfiguration.Discrete robust protection systems use scores of parts, which take upboard space and adversely affect reliability. They also requiresignificant engineering effort if they are to remain sonicallytransparent during normal operation.
|Figure2: The IC driver provides configurable over-current shutdown,under-voltage lockout, and a floating front-end interface.|
By including programmable OC protection for both upper and lowerswitches, a half-bridgedriver can enhance the protection function, ensure sonictransparency, and reduce both amplifier-design risk and cycle time (Figure 2 above ).
The protection circuit uses each switch's RDS(on) as thecurrent-sense element. As a result, only a few external parts arenecessary to configure complete bulletproof OC protection. Figure 3 below shows OC protectionat work. The outputs shut off following an OC condition. The outputinductor safely releases its stored energy and the output rests at zerovolts.
UVLO (under-voltage lock out) is another important protectionfeature. If the power supply falls below the IC's minimum operatingvoltage, it isn't able to drive the gates properly, which can result inshoot-through currents in the bridge FETs.
Under these conditions, the driver's UVLO turns off both gates. Forthe purpose of under-voltage detection, the driver IC senses gate-drivecontrol power between VB and VS for the upper rail and between VCC andCOM for the lower rail.
|Figure3: The over-current shutdown – a critical feature – gracefullydischarges the filter inductor as it brings the output voltage to zero.|
Amplifier designers can use the CSD pin to program the driver'sfault response. The pin provides five functions: self-resetting timer,shutdown, latched protection, shutdown status output and power-up delaytimer.
The self-resetting timer uses an external timing capacitor to setthe shutdown interval, after which the driver restarts. This functionalso provides power-up delay so that power-supply levels can stabilizebefore the amplifier begins to drive the speakers.
The CSD pin can serve as a remote shutdown input. If an externaldevice, such as an open drain microcontroller I/O pin pulls the CSD pinlow, the driver shuts down. Once the external device releases the pin,an internal current source ramps the external timing capacitor as inthe previous case.
A resistor and FET can implement a fault-protection latch thatrequires an explicit reset signal before the driver will restart. Justa few more parts extend the latching-mode circuit to include afault-status output.
As if sonic improvements and integrated fault protection weren'tenough, today's Class-D drivers also offer floating control inputs.This structure greatly simplifies the interaction between the amplifierand the system's front end.
The IC features three isolated voltage wells; the IC substrateserves as COM on the low-side rail, the high-side rail references VS,and the floating input references VSS. Control circuitry can referencesystem ground or the lower rail. The IC includes five level shifters:Three communicate protection-circuit signals; two are for gate-drivesignals. Figure 2 shows thethree isolated voltage wells and the five level shifters.
Being direct with MOSFETs
Low layout and component parasitics are essential for achievinghigh-performance results in any high-frequency system. DirectFETMOSFETs have no bond wires and exhibit small lead inductances.
This typeof DirectFETdevice reduces both the gate and power-train circuitinductances. Lower gate-circuit inductance reduces switching delays.Lower package inductance in the drain and source leads enables greaterdV/dt with less EMI than a TO-220 package. The DirectFET package offersthe additional benefits of low thermal impedances and twosided cooling,which can simplify the thermal design.
Optimizing the switch characteristics obtains the highestefficiency. Power-supply size and cost drop with increasing switchingfrequency. At 400kHz, a single-stage two-element LC output filter givesthe results in Figure 1 .
The best MOSFET for the application does not necessarily have lowestRDS(on). Switching losses play an increasingly dominant role asfrequency rises and, therefore, designers must weigh total gate chargeagainst RDS(on) to find the best combination.
Figure 1's results also depend on low MOSFETpackage inductance. This120W, 4OHM design requires neither a heat sink nor forced air. Today,MOSFETs for specific Class-D applications are compatible with highvoltages and switching frequencies.