Operational Amplifiers (opamps) (part 2)

<< cont. from part 1

4. Different Types of Operational Amplifiers and Application Considerations

In a modern approach, practical opamps available from the component manufacturers could be grouped into several categories, such as

1. Voltage feedback opamps (the most common form of opamp, similar to the 741 type).

2. Current feedback opamps.

3. Micro-power opamps.

4. Single supply opamps.

5. Chopper stabilized opamps.

6. Wideband, high-speed, high-slew-rate opamps.

As many excellent references describe the design techniques for voltage feedback opamps, the discussion here is limited to special types that have appeared on the market during the last ten years. For details on voltage feedback opamps, see Horowitz and Hill (1996) and Analog Devices (1987, 1990).

4.1 Voltage Feedback Opamps

The voltage feedback operational amplifier (VOA) is used extensively throughout the electronics industry, as it’s an easy-to-use, versatile analog building block. The architecture of the VOA has several attractive features, such as the differential long-tail pair, high-impedance input stage, which is very good at rejecting common mode signals. Unfortunately, the architecture of the VOA provides inherent limitations in both the gain-bandwidth trade-off and the slew rate. Typically, the gain-bandwidth product (f r) is a constant and the slew rate is limited to a maximum value determined by the ratio of the input stage bias current to the dominant-pole compensation capacitor. As a comprehensive discussion on the characteristics and applications of VOAs are well documented, only few examples of applications that highlight the importance of some parameters are described here.

FIG. 11 Some application examples showing the importance of VOA parameters" (a) Sensitive photodiode amplifier, (b) High-impedance amplifier (1014 f2), (C) Low-noise/low-distortion RIAA preamplifier, (d)Low-droop positive peak detector. (of Burr-Brown Inc.)

For example, in many applications, bias current can cause offset errors and errors in the transfer function of the circuit. For example, FIG. 11 (a) shows a photodiode amplifier application. Photodiodes generate a current output proportional to the light intensity falling on the photodiode. These currents, however, usually are in the sub-nano-amp range. The circuit of FIG. 11(a) is simply a current-to-voltage converter. An ideal opamp would cause all the current from the photodiode to flow through the feedback resistor, causing the output voltage of the circuit to be linearly proportional to the current from the photodiode. A real opamp would lose some of the photodiode current into its input terminal. If the photodiode current is 10 nA, then an opamp with a 1 nA bias current would cause only 9 nA of signal to pass, causing a 10% error. If the amplifier has a 75 fA bias current, then the error caused by the amplifier is only 0.00075%. In a pH transducer application such as FIG. 11(b), a low bias current amplifier is used for two reasons: The OPA111 is an FET input device, thereby giving it an extremely high input impedance. Since the pH probe is a high-impedance element, the input impedance must be even higher. Low bias current also is necessary, since with a source impedance of 500 Mr2, a 1 nA bias current would cause an offset error of 500 mV. The signal output of the probe is only 50 mV; offset error is ten times the signal.

In critical applications where noise can contribute significant errors, such as data acquisition system front ends or audio applications, opamps with a low voltage noise density are crucial. FIG. 11(c) shows an audio application.

Preamplifiers for audio deal with small signal levels; noise that is negligible in large signal applications suddenly becomes a concern when the signal is only an order of magnitude above the noise (an S/N ratio of 20 dB).

FIG. 12 Simplified VOA and noninverting amplifier: (a) Architecture, (b) Noninverting amplifier

4.1.1 Closed- Loop Gain and Bandwidth of VOA

Equivalent open-loop voltage gain-related components of VOAs are shown in FIG. 12(a). With A = gmZ z, where gm is the transconductance of the input stage and Zz is the parallel impedance formed by Rz and Cz. The transfer function of the closed-loop gain of the amplifier, con figured for the noninverting case ( FIG. 12(b)) is given by ...

It’s important to stress at this stage that, since the VOA is a high differential input impedance device, when negative feedback is applied in this way, voltage is sampled at the output and fed back as a voltage signal to the input. The VOA generally is internally compensated to give a low-frequency dominant pole, fo, and so the open-loop gain is A = A0[1 -4-j(f/fo)],

where fo ~ 1/(2rrRzCz)

... and Ao is the low-frequency open-loop voltage gain. Substituting this into equation (14) gives ...

...where (Ao + G)fo ~ Aofo = fT, since A0 >> G, and this is the gain-bandwidth product of the VOA, which is constant. The closed-loop bandwidth (fp) of the amplifier is fp = f r/G; therefore increasing G results in a decrease in the closed-loop bandwidth, while a decrease in G leads to an increase in f p. This is the "classical" gain-bandwidth trade-off exhibited by a voltage amplifier with a single dominant-pole frequency response. The maximum closed-loop bandwidth of f r will be obtained when 100% feedback is applied; that is, when G = 1.

Generally, VOAs are internally compensated for resistive feedback to guarantee stable operation for any value of G, including G = 1.

The equivalent circuit for a voltage feedback opamp is shown in FIG. 13(a). FIG. 13(b) shows the classic gain versus frequency response obtained with dominant pole compensation. The 6 dB/octave roll-off is the optimum choice for best phase margin and fastest settling time. The product of the closed-loop gain and the closed-loop bandwidth (gain-bandwidth product) is constant for fixed compensation. The closed-loop bandwidth therefore varies inversely with the closed-loop gain.

FIG. 13 Equivalent circuit and frequency response of VOA: (a) Equivalent circuit, (b) Frequency response

FIG. 14 A simplified schematic of a high-speed VOA. (of Analog Devices, Inc.)

FIG. 15 A simplified schematic of the AD847. (Analog Devices, Inc.)

4.1.2 High-Speed Voltage Feedback Opamps

Within the last decade, many opamps with high-speed capability have entered the market to supply the demand for high-speed A/D converters, video signal processing, and other industrial needs. Many of these opamps combine high speed and precision. For example, the devices introduced by Analog Devices in the late 1980s, such as the AD840, 841, and 842 devices, are designed to be stable at gains of 10, 1, and 2 with typical gain-bandwidths of 400, 40, and 80 MHz, respectively.

A simplified schematic of high-speed voltage feedback is shown in FIG. 14. For purposes of discussion, the amplifier is shown in the inverting mode. The amplifier consists of a differential input stage (common emitter), a voltage amplification stage, and a class AB (push-pull) output driver.

An example of such a device is the AD847 from Analog Devices. The AD847 is fabricated on Analog Devices' proprietary complementary bipolar (CB) process, which enables the construction of PNP and NPN transistors with similar values of f r in the 600-800 MHz region. The AD847 circuit (FIG. 15)

includes an NPN input stage followed by fast PNPs in the folded cascade intermediate gain stage. The CB PNPs also are used in the current-amplifying output stage. The internal compensation capacitance that makes the AD847 unity gain stable is provided by the junction capacitance of transistors in the gain stage.

The capacitor, C F, in the output stage mitigates against the effect of capacitive loads. At low frequencies and with low capacitive loads, the gain from the compensation node to the output is very close to unity. In this case, C F is bootstrapped and does not contribute to the compensation capacitance of the part. As the capacitive load is increased, a pole is formed with the output impedance of the output stage. This reduces the gain, and therefore, C F is completely bootstrapped. Some fraction of C F contributes to the compensation capacitance, and the unity gain bandwidth falls. As the load capacitance is increased, the bandwidth continues to fall and the amplifier remains stable. For more details related to high-speed VOAs, see Analog Devices (1990).

4.1.3 Grounding and Bypassing

In designing practical circuits with devices such as AD847, remember that, whenever high frequencies are involved, some special precautions are in order.

Circuits must be built with short interconnection leads. A large ground plane should be used whenever possible to provide a low-resistance, low-inductance circuit path as well as minimizing the effects of high-frequency coupling. Sockets should be avoided because the increased interlead capacitance can degrade bandwidth.

Feedback resistors should be of low enough value to assure that the time constant formed with the capacitance at the amplifier summing junction won’t limit the amplifier performance. Resistor values of less than 5 k~ are recommended. If a larger resistor must be used, a small (< 10 pF) feedback capacitor in parallel with the feedback resistor, RF, may be used to compensate for the input capacitance and optimize the dynamic performance of the amplifier. Power supply leads should be bypassed to ground as close as possible to the amplifier pins. Ceramic disc capacitors of 0.1 uF are recommended.

4.2 Current Feedback Opamps

The current feedback operational amplifier (CFOA) is a relatively new arrival to the analog designer's tool kit. The first monolithic device was produced by Elantec Inc. in 1987.

Current feedback operational amplifiers were introduced primarily to overcome the bandwidth variation, inversely proportional to closed-loop gain, exhibited by voltage feedback amplifiers. In practice, current feedback opamps have a relatively constant closed-loop bandwidth at low gains and behave like voltage feedback amplifiers at high gains, when a constant gain bandwidth product eventually results. Another feature of the current feedback amplifier is the theoretical elimination of slew-rate limiting. In practice, component limitations do result in a maximum slew rate, but this usually is much higher (for a given bandwidth) than with voltage feedback amplifiers.

The current feedback concept is illustrated in FIG. 16. The input stage now is a unity-gain buffer, forcing the inverting input to follow the noninverting input. Thus, unlike a conventional opamp, the latter input is at an inherently low (ideally zero) impedance. Feedback always is treated as a current and, because of the low impedance inverting terminal output, R2 always is present, even at unity gain. Voltage imbalances at the inputs cause current to flow into or out of the inverting input buffer. These currents are sensed internally and transformed into an output voltage. The transfer function of this transimpedance amplifier is A(s); the units are in ohms.

It can be shown that, if A(s) is high enough (like the open-loop gain of a conventional opamp), very little current flows in the inverting input at balance.

The overall closed-loop transfer function becomes

Vout = 1 + Vin

…which is the same as a conventional opamp.

FIG. 16 Current feedback opamp model (a) Simplified version, (b) Analytical model

However, if the dominant pole is created by feeding the current imbalances into the compensation capacitor, the time constant will be set by the product of this capacitor (Ct) and the feedback resistor Rz. The closed-loop bandwidth now is given by ....

...and is independent of closed-loop gain. A more complete mathematical analysis can be obtained using the representative model shown in FIG. 16(b). Notice that the input buffer has been given a finite output impedance (Rinv) to model practically realizable buffers. The error current from the buffer is mirrored and fed into a transimpedance stage consisting of Rt and Ct, where the current-to-voltage conversion takes place. The voltage generated here is buffered by another unity-gain stage and fed to the main amplifier output. Because the value of the small-signal trans-resistance, Rt, is very high (often in the megaohm range), only minute error currents are needed to change the voltage at node 2 by several volts.

Consequently, the amount of current that must flow into or out of the inverting input terminal under steady-state conditions is extremely small. The feedback network, even though it may be formed from quite low-value resistors, therefore presents a very light effective load on the output of the input buffer.

Applying Kirchhoff's law to nodes of circuit in FIG. 16(b), the overall transfer function and the like can be derived. It can be shown that, if the product of Rt and Abuf is large, the closed-loop gain approaches [1 + (R2/R1)] as the transconductance approaches infinity, which is the same result as a VOA with large open-loop gain. The AC behavior would appear to be somewhat less intuitive; however, if the product of Rt and Abuf is large enough, the closed-loop pole frequency can be closely approximated by ...

This result indeed is very different from the constant gain-bandwidth product of a voltage feedback amplifier. At low gains, when R2 7> R1, and assuming Rbuf << R2 (which always is true in practice), the closed-loop pole frequency is dictated predominantly by the compensation capacitor and the feedback resistor, R2, and is substantially independent of the exact gain setting. Therefore, the choice of feedback resistor is of somewhat more importance than in the case of voltage feedback amplifiers, and most manufacturers will state a suggested minimum to avoid oscillation problems.

At high gains (in practice 50 or more could be considered high), the Rinv term becomes dominant and the amplifier asymptotically assumes a constant gain-bandwidth product given by ...

The gain of the output buffer, Abuf, also plays its part in determining the closed-loop pole frequency. As the main amplifier output is loaded, the gain drops below unity and causes a reduction in closed-loop bandwidth predicted by equation (19). This also tends to make practical amplifiers more stable when heavily loaded. With a constant load, the bandwidth reduction can be compensated for by reducing the value of the feedback resistor.

This theory, although useful, in practice (as always) is compounded by second-order poles and zeroes and stray capacitive and parasitic inductive effects.

Such limitations manifest themselves principally as peaking in the small signal response, overshooting and ringing in the transient response, and a sensitivity of overall AC behavior to loading conditions.

The most significant advantages of the CFOAs are:

• The slew-rate performance is high, where typical values could be in the range of 500-2500 V/laS (compared to 1-100 V/us for VOAs).

• The input referred noise of CFOA has a comparatively lower figure.

• The bandwidth is less dependent on the closed-loop gain than with VOAs, where the gain-bandwidth is relatively constant.

FIG. 17 Equivalent circuit of current feedback opamp

4.2.1 Practical High-Speed Current Feedback Amplifiers

The equivalent circuit for a current feedback opamp is shown in FIG. 17.

An ideal current feedback amplifier has no input impedance at its inverting input (Rin = 0), infinite input impedance at its noninverting input, and the voltage at the inverting input is held at that of the noninverting input by a unity gain buffer.

The transfer function of this amplifier (Vout/lin) is a dimensional quantity with the dimension of a resistance, not a ratio, as in the case of voltage feedback opamps. Because of the very low (ideally zero) inverting input impedance, the current feedback opamp has a bandwidth more or less independent of closed-loop gain for a fixed feedback resistor R FS. This implies that the product of the closed-loop gain and the closed-loop bandwidth is not a constant; hence, it’s inappropriate to apply the term gain-bandwidth product to current feedback amplifiers. Examples of practical devices are AD844 and AD846 from Analog Devices and HA5004 from Harris Semiconductors. FIG. 18 compares the frequency response of CFOA with VOA.

FIG. 18 A comparison of frequency response: (a) VOA, (b) CFOA

FIG. 19 Simplified current feedback opamp. (Analog Devices, Inc.)

The practical current feedback amplifier is based on a common base input stage, as shown in FIG. 19. This configuration is characterized by a high impedance noninverting input, which drives the inverting input via a unity gain buffer. The open-loop inverting input impedance is approximately equal to any resistance in series with the emitter (RE) plus the dynamic input impedance of the grounded base transistor (rE). The output voltage is controlled by the input current and is related to it by the transimpedance gain expressed in ohms. Because of the low-impedance inverting input common base stage, a stepped input voltage A Vin will produce a stepped current in the emitter and collector of Q1. This will yield a slew rate of Iin/Cc. Therefore, the rise and fall times of the output essentially are constant regardless of the output voltage swing. This attribute provides for exceptional full-power bandwidth. Bandwidth characteristics for several current feedback opamps are shown in Table 2-2. For further details on applications and design techniques on CFOAs, see Analog Devices (1990), Lidgey and Hayateleh (1997), Toumazou et al. (1990, Section 16), and Mancini and Lies (1995).

TABLE 2 Bandwidth Comparison of Few CFOA. (Analog Devices, Inc.)

4.3 Single-Rail Opamps

Single-supply operation is becoming an increasingly important requirement as systems get smaller, cheaper, and more portable. Portable systems rely on a battery as the primary power source. Consequently, power consumption and operating time per battery charge are high on the designers' priority list. This makes low-voltage operation and low power consumption critical.

Most general purpose opamps are designed to operate from 4-15 V supplies.

Trying to make them work at low voltages may require special attention because few are specified to operate at lower voltages. Many even won’t function at 5 V or less. In modern products designed to operate from batteries (such as two 1.5 V cells), designers must consider some aspects that are less serious in high rail voltage-based systems. Some important ones follow:

1. A low signal swing compresses signal-to-noise performance.

2. The noise floor tends to rise at low currents.

3. Choosing the ground reference becomes important.

4. The bandwidth suffers as the supply current drops.

4.3.1 Effect of Reduced Signal-to-Noise Performance

The most immediate effect on an amplifier circuit running on single supply at a reduced voltage is the reduction in signal swing. Even if the noise floor remains constant (highly unlikely), the signal-to-noise performance will drop as the signal amplitude decreases.

Most opamps that are designed to operate from 4-15 V usually have no problem with output headroom: 3-4 V is sufficient for most applications.

However, if the supply voltages drop to, say + 12 V, a 3 V headroom requirement (at each rail) would limit the signal swing to 6 V P-P (from +3 V to -t-9 V). Dropping the voltage further to +5 V will render these opamps inoperable, as no output swing capability is left.

For this reason and others, low-voltage single-supply opamps are designed so that their output stage swings as closely to both rails as possible. This helps to restore some of the lost signal-to-noise performance. It’s useful if the output can swing to the negative rail. Many opamps have this capability. Single-rail opamps are designed to require 1 or 2 V at most as headroom. Some special designs could have even less headroom, allowing them to operate at -t-5 V or less. The drawback is that most of these opamps are not designed to source much output current, usually 5 mA or less.

In addition to being compressed at the high-level end by signal reduction, the signal-to-noise performance generally is squeezed from the bottom end as well, as noise floor tends to rise. This is because the single supply usually accompanies an inevitable drive toward lower supply current consumption, which tends to increase noise. The designer must decide how these issues affect the final performance target of the system and make the necessary trade-offs.

4.3.2 Ground Reference

Most dual-rail opamps use the signals referenced to 0 V, which is the midpoint of 4-Vcc rail. This is the most convenient, as there is ample supply headroom to work with. Such is not the case for single-supply circuits. Ground reference can be chosen anywhere within the supply range of the circuit. There is no standard to follow. Indeed, the choice of ground reference depends on the types of signals used. To illustrate this point, choosing the negative rail as the ground reference may optimize the signal dynamic range of an opamp designed to swing to 0 V. On the other hand, the signal may have to be level shifted to be compatible with another device that is not designed to operate at 0 V input because the signal no longer can be inverted. These limitations can force an inefficient design, increasing the cost of the system.

Choosing the negative rail as the ground reference is the simplest and most natural choice for setting the ground reference. Since the negative rail is the return of the supply, its impedance is very low and it makes an ideal ground reference. All signals can be returned to this point without concern about its ability to sink the current, assuming the current is within the rated capability of the supply. Nevertheless, the voltage drop in negative rail ground returns can be a problem, and their resistance must be kept low.

Setting the ground reference at the negative rail usually allows the maximum signal dynamic range, as it establishes the limit of one end of the signal range.

However, this may present a problem interacting with other devices, as their inputs or outputs are not designed to operate near the negative rail. Therefore, it’s important to know what types of devices are needed for the application, knowing that the signal will' swing to the negative rail.

Not all single-supply circuits work well with the 0 V ground system. For example, audio or video signals may be best handled using a false-ground system by biasing the amplifiers to the midpoint of the supply. Then the signal can be AC coupled through the amplifier chain. This removes the necessity of level shifting at each amplifier stage, as otherwise would be required if the 0 V ground system is used. It also saves the cost of additional components to do the level shifting.

It often is assumed that a false-ground circuit need be only a simple buffer amplifier without the bypass capacitors and compensation. In some cases, one can get by without them as long as the false-ground node sees no dynamic or transient load changes. These can occur when driving the ground pin of a D/A or an A/D converter. In these applications, the false-ground node must hold its voltage constant with minimum perturbation. In the presence of reference "bounce," conversion error may result or noise may be injected into the circuit.

The choice of the quality of the false-ground rests on whether the circuit is sensitive to false-ground perturbation, both in DC and AC terms. Not all applications necessarily require a high-quality, low-impedance false-ground. In fact, in some cases, it may be sensible to use both false-ground and negative-rail ground references in different parts of the circuit. One needs to observe the level shifting requirement at interfaces of the two; it depends on what works best.

A solid false-ground can be implemented easily using a voltage divider or reference voltage buffered by an amplifier. The choice of the opamp and the implementation are critical to a good reference node with no reference "bounce." An example is shown in FIG. 20, and further details may be found in Analog Devices (1992).

FIG. 20 A supply splitter reference generator for OP-90. (of Analog Devices Inc.)

4.3.3 Handling 0V Input Signals

Dual-rail opamps, which generally are designed with NPN input transistor pairs, usually cannot handle signals that are at or near the negative rail, due to the requirement of 2-3 V headroom at either rail. Operating at or near the negative rail would cause the amplifier to go nonlinear as saturation is approached. Even worse, it may cause the output to reverse its phase.

Opamps designed for single-supply operation typically use PNP input pairs to allow the input to operate linearly at or near the negative rail. Another input architecture that can operate at 0 V uses MOSFET transistors in a CMOS amplifier.

An example is OP-80 from Analog Devices, which has a P-channel input pair.

4.3.4 Handling 0 V Output Signals

Ideally, a single-supply opamp's output should be capable of swinging to the negative rail, especially if it’s feeding into the input of the next stage that can operate at 0 V. Beware that, as the output swings very near the negative rail, the output accuracy can drop off rapidly, as a function of the (output sink)

load current. This is because a single-supply amplifier's output stage usually has either an active transistor pull-down or it may require a resistor to pull-down to the negative rail. Whatever the means, the pull-down's finite resistance develops a voltage drop that prevents the output from swinging fully to the negative rail.

If accuracy is required at or near 0 V, the load current must be kept as small as practical. Always keep in mind that the total load current also includes the current flow in the feedback resistor. As mentioned previously, some opamps have an active output drive to the negative rail. Others require a resistor pull-down to achieve a complete swing to 0 V. FIG. 21 shows both topologies.

FIG. 21(a)'s OP-80 output stage relies on the internal on-resistance of QouT2 to pull the output down to the negative rail. Including a source resistance, the total pull-down resistance typically is 400 f2. As long as the load current is less than I~A, its output will swing to within 1 mV of the negative rail.

The OP-90 utilizes a conventional push-pull emitter-follower output stage, as shown in FIG. 21(b). This provides a low-impedance drive except for output voltages less than 0.5 V to the negative rail. Below this voltage, the bottom output transistor Qour2 saturates, clamping the output at a base-emitter voltage above the negative rail. If no external pull-down resistor is provided, the top output transistor Qourl turns off completely, as its base voltage tends to pull less than its base-emitter voltage, cutting off the base drive. However, a pull-down resistor would keep the QouT1 active, and therefore linear operation would continue as the output swings to the negative rail. For applications and other design data about these examples, see the device data sheets.

FIG. 21 Zero-volt swing output stage designs: (a) OP-80 output stage, (b) OF-90 output stage. (Analog Devices Inc.) Opamps whose output is designed to swing from one rail to the other are ideally suited for single, reduced supply operation. Theoretically, the maximum signal-to-noise performance for a given supply voltage can be achieved. One such rail-to-rail device is the OP-295 dual operational amplifier (Analog Devices, 1992). Most industry standard precision opamps are designed to operate with dual, higher voltage (4-15 V) supplies. Their precision performance is optimized for these supply voltages. As such, trying to operate them at lower voltages may be inadvisable, as their specifications may no longer apply. Indeed, many don’t function even if their input is forced to the negative rail or if they are powered by a single +5 V supply.

Very few single-supply opamps exhibit sufficiently low input offset and offset drift to qualify them as precision opamps. The OP-90, OP-290, and OP-490 come close to precision performance yet can operate off a single supply.

For a comprehensive practical discussion on single-supply operation, devices, applications, and restrictions, see Analog Devices (1992).

4.4 Micro-Power Opamps

Numerous opamps on the market perform well at supply currents in the 500 gA-1 mA range, but certain applications require devices that operate at even lower currents. For example, applications that rely on batteries or solar cells need to keep current drain to a minimum. Low-current operation also is essential for minimizing power dissipation in equipment containing large quantities of tightly packed active components.

Micro-power opamps can meet these needs. Although definitions of the term vary, all micro-power devices perform at currents lower than the 500 uA minimum of "low-power" devices. Opamps that operate below about 250 ~A supply currents generally can be categorized as micro-power devices. Table 2-3 indicates a representative group of micro-power opamps and their characteristics.

4.5 Chopper-Stabilized Opamps

The best bipolar opamps may have offsets as low as 10 ~V. When a design requires an extremely low input offset voltage (Vos) with virtually no offset-voltage drift over time or temperature, chopper-stabilized devices are available from many manufacturers.

Low offset-voltage drift probably is the most attractive characteristic of chopper-stabilized opamps. Applications that take advantage of this feature include strain-gauge amplifiers, thermocouple amplifiers, and precision data collection in environments where periodic adjustments are either impractical or impossible. Most often these applications involve low-frequency signals (less than 10 Hz).

TABLE 3 Representative Micro-Power Opamps

FIG. 22 Basic operation of a chopper-stabilized opamp (a) Storing of offset of nulling amplifier (b) Storing of offset of main amplifier

FIG. 22 shows the basic operation of a chopper-stabilized opamp. The device consists of a clock generator, a main and a nulling amplifier, and two holding capacitors, which can be either internal or external. The main amplifier has fixed connections to the input and output pins.

The hulling amplifier alternately performs two tasks. First, it uses a holding capacitor (CI) to store a correction for its own offset, which is measured while the amp's input pins are shorted. The hulling amp then measures the offset of the main amplifier and stores a correction voltage in another capacitor (C2). The amplifier alternates between these tasks at the rate of the clock. Because the opamp is constantly correcting its offset, offset voltages are low and drift is virtually nonexistent.

Practical examples of such opamps are ICL7650S from Intersil or Harris Semiconductor and MAX430 from Maxim Integrated Circuits. Typical specifications of the ICL76505 type devices are 5 uV of Vos and an input bias current of 10 pA at 25°C. Offset drift averages 0.02 uV/°C within the range of -25 to 80°C.

TABLE 4 Comparison of Precision Opamps vs. Choppers. (of Analog Devices Inc.)

The ICL76505 requires external holding capacitors to operate, but devices such as MAX430 have 0.1 uF chip capacitors bonded to the lead frame within the device. Such devices that operate from +15 V power supplies allow you to directly replace conventional LM741 or LM 108 devices.

Early chopper-stabilized amplifiers actually used relays to chop the signal, but today such amplifiers are monolithic and use MOS switches to do the job.

Early monolithic amplifiers of this type had high noise at the chopping frequency (which may be between a few hundred Hz and a few tens of kHz). This high-frequency noise is less of a problem in the latest devices, which may contain quite effective filters at the chopper frequency (although careful layout and supply decoupling still are important with these parts), but switching noise remains the major problem with chopper-stabilized opamps. Because of the technology used to manufacture them (frequently CMOS), many chopper-stabilized opamps have a voltage noise of several uV in the band 0.1-l0 Hz, and it therefore is necessary to integrate their output for several seconds or tens of seconds to obtain the low offset of which they are potentially capable. Not all systems allow such long integration times, and so a compromise becomes necessary between offset and speed. Table 4 compares the precision opamps versus chopper stabilized amplifiers.

For amplifiers that have no built-in capacitors, such as the Intersil 7650S and the 420 series from Intersil and Maxim, you must select the hold capacitors carefully. Noise really is not the critical issue; any low-leakage capacitor is adequate in this regard. But settling time is an issue; capacitors with a high dielectric absorption, such as ceramic capacitors, can take several seconds to settle after power first is applied. Therefore, you should use Mylar or polypropylene capacitors if you need fast initial settling.

When a chopper-stabilized amplifier is overloaded due to saturation, capacitors can acquire excessive charges and it could take a long time to recover when the overload is removed. To avoid the recovery time problem, many chopper-stabilized opamps use clamp circuits that prevent the amplifier from reaching saturation during an overload. FIG. 23 shows a circuit that prevents the opamps from reaching saturation. When V OUT approaches either rail, the appropriate clamp transistor begins conducting. Connecting the clamp pin to the amplifier's inverting input pin puts that transistor in parallel with the gain resistor. As the transistor conducts, it reduces the gain of the amplifier, thus preventing saturation.

But using the clamp has two drawbacks. First, the available output range of the amplifier is reduced by as much as 1 V from each rail. Second, the leakage of the clamp transistors shows up as additional bias current at the input and reduces accuracy. Typical leakage currents range from 1 to 10 pA.

FIG. 23 A technique for preventing saturation of chopper-stabilized opamps

The sampling techniques that chopper-stabilized opamps use to eliminate drift can result in intermodulation as well as clock noise. Interaction between the input signals and the clock can generate intermodulation products in the form of sum and difference signals. If the input signal's frequency is close to the rate of the clock, the difference product shows up as additional offset error. You can eliminate this error by filtering the input signal to keep its frequency range well below the sampling clock frequency. To avoid intermodulation problems, many chopper amps allow the designer, using an external signal, to set the clock frequency of the amplifier. For further details, see Quinnell (1989) and Harris Semiconductor (1993-94).

4.6 High-Voltage Power Opamps

FIG. 24 Power operational amplifiers. (Apex Microtechnology Corporation.)

For power amplifier applications, special components are available from a limited number of manufacturers. Today, high-voltage amplifiers with total (rail-to-rail) voltages of 1200 are capable of driving around 75 mA to a load. The PA 89 hybrid IC from Apex Microtechnology Corporation is an example. Similarly, high-current amplifiers, which can handle as high as 30 A, with 150 V rail to rail, are available from the some manufacturers. The PA03 power operational amplifier from Apex is an example. Some applications of such special power opamps are sonar transducer drivers, piezoelectric transducer drivers, high-voltage instrumentation, programmable power supplies, and linear and rotary motor drivers.

FIG. 24 shows some of these hybrid devices.

In these types of opamps, the designer has to consider many special situations, such as power supply performance, thermal management, safe operating area, and stability. A discussion of these topics is beyond the limits of this section.

Application notes in Apex Microtechnology Corp. (1996, pp. D 1 -D 105) provide an excellent discussion.