In the midst of the flurry of multimillion-transistor digital ICs assailing the marketplace, it's easy to forget that a wide variety of loads demands a lot of volts, amps, and watts to drive them. Some examples are motors, actuators, vibration-cancellation systems, and, of course, loudspeakers. The representative sampling of monolithic power amplifiers in Table 1can provide that needed drive. However, you must choose among them prudently to ensure that both the input and output characteristics satisfy the needs of your application.

Picture One

Of the two sets of characteristics, those of the output stage are probably the more crucial to the success of your design. The worst that can happen if you choose an amplifier with inappropriate input parameters are inaccuracies accruing from offset and drift and, perhaps, undue loading of the source because of too-low input impedance.

Far graver woes can result from the choice of an amplifier whose output structure doesn't match the needs of the load. These woes include premature thermal shutdown and outright destruction of the amplifier IC. A discussion of the various output structures in Fig 1 may offer some insight into the appropriateness of the various amplifier types for your application.

Figs 1a, b, and c are simplified output structures that use bipolar power transistors. You're not likely to find the full-complementary topology in Fig 1a in monolithic amplifiers, because high-power pnp transistors are not very amenable to monolithic processing. More's the pity, because, as Ref 1 explains, the full-complementary stage is easy to use. It offers symmetrical output impedance and low crossover distortion (assuming proper class-AB biasing).

The quasi-complementary stage in Fig 1b is amenable to IC processing and provides performance similar to that of the full-complementary topology. The two lower transistors in the circuit form a "composite-pnp" transistor. The drawback to this structure is the heavy local feedback in the lower transistor loop, which can lead to instability in the presence of inductive loads.

The all-npn output stage in Fig 1c is an early approach to fabricating monolithic power amplifiers. Its main disadvantage is the large difference in output impedances when sourcing and sinking current. The upper (sourcing) transistor is an emitter follower with inherently low output impedance. The lower (sinking) transistor operates in common-emitter mode with inherently high output impedance. Another disadvantage is the large difference in open-loop gain in sourcing and sinking modes.

Bipolar power transistors are cursed with a negative temperature coefficient for V_BE, the base-emitter voltage required to yield a given collector current. So, unless a circuit designer incorporates compensation techniques into the output-stage drive circuitry, the power transistors are subject to thermal runaway. Power causes heating, heating causes V_BE reductions, and, with a given V_BE drive voltage, the transistor turns on even more fully. Without compensation precautions, this positive thermal feedback causes the transistor to self-destruct.

It's a tricky proposition to provide such V_BE compensation. A designer must strike a fine balance between two almost-conflicting criteria: maintaining adequate quiescent current to prevent crossover distortion and preventing thermal-runaway conditions.

An important consideration in matching bipolar output stages to the needs of a load is to respect the "safe operating area" (SOA) carefully. Ref 2 gives a detailed explanation of how to use SOA curves for bipolar power transistors. Fig 2 shows an SOA plot for a typical bipolar device. The maximum safe current is a function of V_CE.

At low values of V_CE, the transistor can deliver maximum current to the load. In this area of the SOA, the only worry is to avoid excessive currents that could overstress wire bonds or die metallization. As V_CE increases, the power dissipation increases to the point that self-heating raises the junction temperature to the maximum rated value (usually 200øC). Note that in this broken-line region, V_CE x I_C (power) is a constant.

As V_CE increases further, the safe output current decreases more rapidly. This area of the SOA plot is the second breakdown (also called secondary breakdown) region. Here, the transistor develops "hot spots" in regions of its geometry, points at which current flow concentrates. This condition leads to localized thermal runaway and, finally, device destruction. The final region of the SOA is the transistor's breakdown voltage.

Note the solid lines on the right side of the SOA plot in Fig 2. They indicate that you can exceed the second-breakdown limitations for short power pulses. Thanks to the thermal time constant of the silicon structure, such short-duration bursts of power don't give the transistor enough time to develop suicidal tendencies.

As Ref 2 explains, you might be tempted to check for safe operation only at the point of maximum output voltage and current. But, this point may not be the most stressful. Fig 3 is a plot of power dissipation as a function of the output voltage delivered to a resistive load. You can see that the worst-case dissipation for the amplifier occurs at 50% output voltage. You should use this condition to check for SOA compliance.

A short circuit to the negative rail is also a very stressful condition in the SOA. It forces the full power-supply voltage across the conducting transistor, and the current far exceeds the SOA limit. For this reason, amplifier designers incorporate thermal-shutdown circuitry in the ICs. For Burr-Brown's OPA544, for example, the short-circuit current is about 4A, and the safe SOA current is 1.5A. Thermal shutdown prevents device destruction.

Make note that the voltage in the horizontal axis of the SOA is not the amplifier's output voltage, but rather the voltage across the conducting transistor in the output stage. The vertical axis is the current delivered to the load. Also, be aware that capacitive or inductive loads can produce exceedingly stressful conditions in the output stage.

If a capacitive load is charged to the negative supply voltage and the output tries to go positive, the output instantaneously delivers maximum rated current at the full supply-voltage span. And, because an inductor acts as a current source, an analogous condition arises when the output tries to go negative.

National Semiconductor's bipolar-output parts (Table 1) incorporate a dynamic SOA-protection mechanism, called SPiKe, which stands for self-peak instantaneous Kelvin temperature-protection circuitry. National claims the circuitry makes the ICs nearly impervious to damage from instantaneous temperature peaks and overvoltage and overcurrent conditions.

Advances in IC processing make it possible to fabricate monolithic high-performance MOSFET amplifiers, a product line that was once the domain of multichip-hybrid manufacturers. Two advantages MOSFET amplifiers enjoy are low quiescent current and the absence of second-breakdown thermal runaway. The surface-mountable Apex PA43, for example, works with a total supply span of 350V and consumes only 2-mA quiescent current.

SOA considerations are not as crucial for MOSFET structures as they are for bipolar outputs. MOSFET output stages have two limitations: the current-handling capability of the die metallization and the temperature of the MOSFETs. For user convenience, Apex does publish SOA curves for its devices. Fig 4 shows the SOA curves for the PA43, which has an output structure similar to that of Fig 1d. Note, again, that you can exceed SOA limits for short power pulses.

In Table 1, you can see that Apex's MOSFET amplifiers are high-voltage devices. High-voltage MOSFETs are much more amenable to IC processing than are bipolar transistors. These Apex amplifiers are general-purpose units, unlike the National and SGS-Thomson devices, which are characterized for audio applications. However, their application specificity doesn't prevent you from using them for other applications needing high-power drive.

The TDA729x Series from SGS-Thomson uses a mixed bipolar-CMOS-DMOS process to provide output-power levels from 60 to 100W to a loudspeaker. Fig 1e illustrates their output structure. SGS developed this novel architecture to ensure accurate control of quiescent current and to prevent crossover distortion. The quiescent current is a function of I_REF and the area ratio of Q₁ and Q₂, two adjacent DMOS transistors.

The local feedback afforded by the unity-gain block, A, in conjunction with the I_REF, Q₁ bias circuitry, prevents either transistor from cutting off completely. Complete cutoff would engender crossover distortion because of the relatively long recovery time of a cut-off MOSFET.

Like National, SGS employs a type of dynamic overload protection in the output stages of the TDA729x Series. Although the MOSFET output transistors are not subject to second breakdown, they are subject to destruction by overload. The protection circuitry senses the voltage and current--hence, power--in the output transistors and keeps the devices within SOA limits. It also uses a local temperature-sensing technique that dynamically controls the maximum dissipation.

In selecting a power-amplifier IC, you'll want to take a careful look at the devices' input characteristics. Table 2 gives data-sheet parameters for the amplifiers listed in Table 1. Factors influencing your choice include source impedance, accuracy requirements, common-mode conditions, and noise performance.

The amplifiers from National Semiconductor and SGS-Thomson use bipolar input stages. This fact is evident from the relatively high, 0.5-µA maximum input-bias currents. Therefore, you'd probably not select these types for applications having extremely high driving-source impedances. In Table 2, note that these amplifiers offer no offset-drift specs. As stated, they're designed principally as audio amplifiers, so drift is of relatively limited importance.

In judging whether the noise performance of an amplifier suits your application, you'll have to do some translation between the different style specs of the various amplifiers. For example, Burr-Brown's OPA544 specs 36 nV/ÖHz typical, and the other amplifiers spec either wideband or limited-band rms or p-p voltage figures.

Maximum offset voltages range from the OPA544's 5 mV to the PA43's 30 mV. If the 30 mV seems high, remember that the PA43 is a 350V amplifier intended for applications such as telephone-ring generators, piezoelectric positioning, and electrostatic transducers and deflection circuits--applications in which offset is of little importance.

Just when we thought the amplifier-IC world was safe from the "fully on, fully off" mentality of digital technology, along comes Apex Microtechnology's SA01, a high-power device that occupies only 2 in.2 and delivers 2 kW to its load (Ref 3). This hybrid amplifier uses pulse-width modulation (PWM), or Class-D operation, to provide power amplification.

This device is a breakthrough in that such power levels were formerly available only from large "bricks" and chassis-mounted PWM amplifiers. The SA01 provides a full H-bridge (push-pull) output, effectively doubling the power available from a given single power supply.

Although this device's operating principle is not strictly "Boolean," you can consider it a digital circuit in the sense that its output stage only switches fully on and off and spends no time (except during transitions) in the linear region. The result is >90% efficiency, as opposed to the wasteful 50% inherent in linear amplifiers. The device is useful in applications such as brush-type motor control, magnetic coils for MRI systems, and vibration cancellation.

If the idea of rolling your own switching amplifiers appeals to you, read Jeff Sherman's EDN article (Ref 4), which shows how to use Harris Semiconductor's MOSFET-driver ICs to configure highly efficient, low-distortion amplifiers.

You now have available a variety of monolithic power amplifiers to fit almost any conceivable application. In selecting one, take care to match the input and output characteristics and limitations to the needs of your system.

1. Apex Microtechnology, Improve Your Analog Power, Seminar III.

2. Trump, Bruce, Power Amplifier Stress and Power Handling Limitations, Burr-Brown Application Note AB-039.

3. Schweber, Bill, "First hybrid PWM amplifier requires just 2 in.2," EDN, March 30, 1995, pg 9.

4. Sherman, Jeffrey, "Class D amplifiers provide high efficiency for audio systems," EDN, May 25, 1995, pg 103.