Say no to mediocrity and see how ADI composite amplifiers can “play to their strengths and avoid their weaknesses”!

A composite amplifier consists of two separate amplifiers configured in such a way that one can realize the advantages of each amplifier while attenuating the disadvantages of each amplifier.

Figure 1. Simple composite amplifier configuration.

Referring to Figure 1, AMP1 has the excellent DC precision and noise and distortion performance required by the application. AMP2 meets the output drive requirements. In this configuration, an amplifier (AMP2) with the desired output specifications is placed in the feedback loop of an amplifier (AMP1) with the desired input specifications. Some of the techniques involved in this configuration and their benefits are discussed below.

When you first encounter a composite amplifier, the first question you might ask is how to set the gain. To help with this, it helps to think of the composite amplifier as a single non-inverting op amp contained within a large triangle, as shown in Figure 2. Imagine the large triangle is black so we can't see what's inside, then the gain of the non-inverting op amp is 1 + R1/R2. Uncovering the composite configuration inside the large triangle doesn't change anything; the gain of the entire circuit is still controlled by the ratio of R1 and R2.

In this configuration, it is easy to think that changing the gain of AMP2 via R3 and R4 will affect the output level of AMP2, indicating that the composite gain will change, but this is not the case. Increasing the gain around AMP2 via R3 and R4 will only reduce the effective gain and output level of AMP1, while the composite output (AMP2 output) remains unchanged. Alternatively, reducing the gain around AMP2 will increase the effective gain of AMP1. Therefore, the gain of the composite amplifier is generally determined only by R1 and R2.

Figure 2. Composite amplifier viewed as a single amplifier.

This article will discuss the key benefits and design considerations for implementing a composite amplifier configuration. The article will focus on the impact on bandwidth, dc accuracy, noise, and distortion.

One of the main advantages of implementing a composite amplifier is the wider bandwidth compared to a single amplifier configured for the same gain.

Referring to Figures 3 and 4, let's assume we have two separate amplifiers, each with a gain-bandwidth product (GBWP) of 100 MHz. By combining them into a composite configuration, the effective GBWP of the entire combination will increase. At unity gain, the -3 dB bandwidth of the composite amplifier is about 27% higher, albeit with a small amount of peaking. This advantage becomes more pronounced at higher gains.

Figure 3. Unity-gain composite amplifier.

Figure 4. -3 dB bandwidth improvement at unity gain.

Figure 5 shows a composite amplifier with a gain of 10. Note that the composite gain is set to 10 by R1 and R2. The gain around AMP2 is set to about 3.16, forcing the effective gain of AMP1 to be the same. Splitting the gain equally between the two amplifiers produces the maximum possible bandwidth.

Figure 5. Composite amplifier configured for a gain of 10.

Figure 6 compares the frequency response of a single amplifier at a gain of 10 with the frequency response of a composite amplifier configured for the same gain. In this case, the -3 dB bandwidth of the composite amplifier is approximately 300% higher. How is this possible?

Figure 6. -3 dB bandwidth improvement for a gain of 10.

For a specific example, see Figure 7 and Figure 8. We require a system gain of 40 dB, using two identical amplifiers, each with an open-loop gain of 80 dB and a GBWP of 100 MHz.

Figure 7. Distributing gain to achieve maximum bandwidth.

Figure 8. Expected response of a single amplifier.

To achieve the highest possible bandwidth for the combination, we will divide the required system gain equally between the two amplifiers, each amplifier adding 20 dB of gain. Therefore, setting the closed-loop gain of AMP2 to 20 dB forces the effective closed-loop gain of AMP1 to also be 20 dB. With this gain configuration, both amplifiers operate at a lower point on the open-loop curve than either would at 40 dB gain. Therefore, the composite amplifier will have a higher bandwidth at 40 dB gain than a single amplifier solution of the same gain.

Although it seems relatively simple and easy to implement, proper care should be taken when designing the composite amplifier to obtain the highest possible bandwidth without sacrificing the stability of the combination. In real applications, the amplifiers have non-ideal characteristics and may not be identical, which requires the use of appropriate gain configurations to maintain stability. Also, it should be noted that the composite gain will roll off at a rate of -40 dB/decade, so care must be taken when allocating the gain between the two stages.

In some cases, evenly dividing the gain may not be possible. In this regard, to divide the gain equally between the two amplifiers, the GBWP of AMP2 must always be greater than or equal to the GBWP of AMP1, otherwise peaking will result and possible circuit instability. In cases where the AMP1 GBWP must be greater than the AMP2 GBWP, redistributing the gain between the two amplifiers can often correct the instability. In this case, reducing the gain of AMP2 results in an increase in the effective gain of AMP1. The result is a decrease in the closed-loop bandwidth of AMP1 because it operates at a higher point on the open-loop curve, and an increase in the closed-loop bandwidth of AMP2 because it operates at a lower point on the open-loop curve. If the slowing down of AMP1 and the speeding up of AMP2 are fully applied, the stability of the composite amplifier is restored.

In this article, AD8397 is selected as the output stage (AMP2) and connected with various precision amplifiers AMP1 to demonstrate the advantages of composite amplifiers. AD8397 is a high output current amplifier that can provide 310 mA current.

Table 1. Bandwidth Extension for Various Amplifier Combinations, Gain of 10, V _OUT = 10 Vp-p V _OUT = 10 Vp-p

Figure 9. Op amp feedback loop.

In a typical op amp circuit, a portion of the output is fed back to the inverting input. The error present at the output (created in the loop) is multiplied by the feedback factor (β) and then subtracted. This helps maintain the fidelity of the output relative to the input multiplied by the closed-loop gain (A).

Figure 10. Composite amplifier feedback loop.

For the composite amplifier, amplifier A2 has its own feedback loop, but A2 and its feedback loop are inside the larger feedback loop of A1. The output now contains larger errors caused by A2, which are fed back to A1 and corrected. The larger correction signal causes the accuracy of A1 to be preserved.

The effect of this composite feedback loop can be clearly seen in the circuit of Figure 11 and the results shown in Figure 12. Figure 11 shows a composite amplifier consisting of two ideal op amps. The composite gain is 100, and AMP2 gain is set to 5. V _OS 1 represents the 50 μV offset voltage of AMP1, while V _OS 2 represents the variable offset voltage of AMP2. Figure 12 shows that when V _OS 2 is swept from 0 mV to 100 mV, the output offset is not affected by the magnitude of the error (offset) contributed by AMP2. Instead, the output offset is proportional only to the error of AMP1 (50 μV times the composite gain of 100), and it remains at 5 mV regardless of the value of V _OS 2. Without the composite loop, we would expect the output error to be as high as 500 mV.

Figure 11. Offset error contribution

Figure 12. Composite Output Offset vs. V _OS 2

Table 2. Output Offset Voltage at Gain of 100

The output noise and harmonic distortion of the composite amplifier are corrected in a similar manner to the DC errors, but for the AC parameters, the bandwidth of the two stages also comes into play. We will use an example to illustrate this using output noise; it should be understood that the distortion cancellation is much the same.

Referring to the circuit in Figure 13, as long as the first stage (AMP1) has enough bandwidth, it will correct for the larger noise of the second stage (AMP2). As the bandwidth of AMP1 begins to be exhausted, the noise from AMP2 will begin to dominate. However, if AMP1 has too much bandwidth and there is peaking in the frequency response, then a noise peak will be generated at the same frequency.

Figure 13. Composite amplifier noise sources.

For this example, resistors R5 and R6 in Figure 13 represent the intrinsic noise sources of AMP1 and AMP2, respectively. The upper curve in Figure 14 shows the frequency response for various AMP1 bandwidths and the frequency response for AMP2 for a single fixed bandwidth. Recall from the Gain Division section that if the composite gain is 100 (40 dB) and the gain of AMP2 is 5 (14 dB), the effective gain of AMP1 will be 20 (26 dB), as shown here.

The lower curve shows the broadband output noise density for each case. At low frequencies, the output noise density is dominated by AMP1 (1 nV/√Hz times a composite gain of 100 equals 100 nV/√Hz). This situation persists as long as AMP1 has enough bandwidth to compensate AMP2.

If AMP1 bandwidth is smaller than AMP2 bandwidth, the noise density will begin to be dominated by AMP2 as the AMP1 bandwidth begins to roll off. This can be seen in the two traces in Figure 14, where the noise rises to 200 nV/√HZ (40 nV/√HZ times the gain of AMP2, which is 5). Finally, if AMP1 has a much larger bandwidth than AMP2, resulting in peaking in the frequency response, the composite amplifier will exhibit a noise peak at the same frequency, as shown in Figure 14. The magnitude of the noise peak will also be higher due to the excess gain caused by the peaking of the frequency response.

Figure 14. Noise performance vs. first-stage bandwidth

Table 3 and Table 4 show the effective noise reduction and THD+n improvement, respectively, when using different precision amplifiers as the first stage to form a composite amplifier with the AD8397.

Table 3. Noise reduction using different front-end amplifiers, effective gain = 100, f = 1 kHz

Table 4. THD+n comparison using different front-end amplifiers, effective gain = 10, f = 1 kHz, I _LOAD = 200 mA

Figure 15. DAC output driver application circuit

In this example, the goal of the DAC output buffer application is to provide a 10 V pp output to a low impedance probe with a current of 500 mA pp, requiring low noise, low distortion, excellent DC precision, and the highest possible bandwidth. The 4 mA to 20 mA current output from the DAC will be converted to a voltage by the TIA and then to the input of the composite amplifier for further amplification. The AD8397 at the output meets the output requirements. The AD8397 is a rail-to-rail, high output current amplifier that is capable of providing the required output current.

AMP1 can be any precision amplifier with the required dc precision for the configuration. In this application, a variety of front-end precision amplifiers can be used with the AD8397 (and other high output current amplifiers ) to achieve the excellent dc precision and high output drive capability required by the application.

_{Figure 16. V OUT} and I _OUT of the AD8599 and AD8397 Composite Amplifier

Table 5. AD8599+AD8397 Composite Amplifier Specifications

This configuration is not limited to the AD8397 and AD8599. Other amplifier combinations are possible as long as the output drive requirements are met and excellent dc precision is provided. The amplifiers in Table 6 and Table 7 are also suitable for this application.

Table 6. Amplifiers with High Output Current Drive Capability

Table 7. Precision Front-End Amplifiers

The two amplifiers are combined into a composite amplifier that achieves the best specifications of each amplifier while compensating for the limitations of each. An amplifier with high output drive capability combined with a precision front-end amplifier can provide a solution for very challenging applications. When designing, it is important to consider stability, noise peaking, bandwidth, and slew rate to achieve the best performance. There are many possible solutions to meet various application requirements. The right implementation and combination can achieve the right balance for the application.