Key considerations when replacing an op amp
Article By : Peter Semig
This article discusses, at a high level, three considerations when replacing a general-purpose or precision voltage-feedback op amp.
A few months ago, I wrote a Signal Chain Basics article titled “3 considerations when selecting an op amp.” While that article is very useful for selecting operational amplifiers (op amps) for new designs, it does not cover what you need to consider when replacing an op amp. Many specifications that look evident, like common-mode voltage range, turn out to have nuances that you may overlook.
The purpose of this article is to discuss, at a high level, three considerations when replacing a general-purpose or precision voltage-feedback op amp. The three considerations include the op-amp input stage topology, output stage topology, and process technology. Each of these have potentially unintended consequences that may affect an op amp’s performance or functionality or both in a given design.
For example, the output stage topology determines the output swing range but can also affect the stability of the circuit, depending on the open-loop output impedance, Zo. Other important considerations when replacing an op amp such as back-to-back input diodes, offset voltage drift and input bias current drift are outside the scope of this article, but I do recommend investigating them.
Consideration No. 1: The input stage topology (Vos vs. Vcm)
Two of the most common op-amp input stage topologies are the traditional single-input pair and complementary-input pair, as shown in Figure 1. Figure 1a depicts an input stage comprising a single pair of N-channel P-channel N-channel (NPN) transistors. Such a topology typically has an input common-mode voltage range that includes the negative supply voltage but may only extend to within 1 V or 2 V of the positive supply voltage. As a trade-off for this shortcoming, op amps with only one input stage transistor pair have a relatively constant offset voltage because the transistors are inherently well matched.
Figure 1 Single PNP transistor pair (a) and complementary-input pair (b) are common op-amp input stage topologies. Source: Texas Instruments
Besides the reduced common-mode voltage range of devices with such an input stage, these devices may suffer from phase inversion, which occurs in some op amps when the input common-mode voltage exceeds the linear input common-mode range. During phase inversion, the output voltage swings to the opposite rail. Although there are board-level techniques to prevent this from occurring, a simpler solution exists: the complementary-pair input stage (Figure 1b).
This topology has one pair of N-channel metal-oxide semiconductor (NMOS) transistors (active when the common-mode voltage is near the positive supply voltage) and one pair of P-channel metal-oxide semiconductor (PMOS) transistors (active when the common-mode voltage is near the negative supply voltage). This topology protects against phase inversion and extends the common-mode voltage range across the entire input supply voltage range.
While this topology extends the input common-mode range, switching between the PMOS and NMOS transistor pairs creates an offset voltage “transition region,” as depicted in Figure 2. The common-mode voltage where this transition occurs and the magnitude of the offset voltage change will depend on the op amp’s design and process technology. Devices that have large offset voltage changes are typically not considered to have rail-to-rail input (RRI), but for those with well-matched input stage transistor pairs, digital correction techniques or offset trimming typically have RRI. Specifications other than the offset voltage—such as common-mode rejection ratio, bandwidth, noise, slew rate and open-loop gain—typically have degraded performance in the NMOS operating region.
Figure 2 Input stage with complementary-input pairs comprise non-RRI (a), RRI (b) and RRI-trimmed (c). Source: Texas Instruments
One topology worth mentioning is the zero-crossover amplifier. Zero-crossover amplifiers use an internal charge pump and have only one pair of input stage transistors. The charge pump internally boosts the supply voltage of the device, for example 1.8 V, which ensures linear operation of the input stage transistor pair across the entire supply voltage range (RRI) and with no input offset voltage transition region (Figure 3).
Figure 3 Zero-crossover op amp uses an internal charge pump and has one pair of input stage transistors. Source: Texas Instruments
In summary, when replacing an op amp, make sure that both the common-mode voltage range and input stage topology are compatible with the original device.
Consideration No. 2: The output stage (Zo)
The second major consideration to take into account when replacing an op amp is Zo. This becomes especially important when driving a capacitive load because Zo and the load create a pole in the op amp’s loop gain curve. This pole can cause stability issues by adding delay in the feedback path, thereby reducing the phase margin of the circuit.
One of the most common solutions for stabilizing a capacitive load is to place an isolation resistor, Riso, between the load and the op-amp circuit. Riso compensates for the pole by creating a zero in the transfer function. The placement of the zero (and therefore the value of Riso) depends on Zo, however. Therefore, it is important to understand not only the magnitude of Zo, but also how it changes over frequency. Figure 4 depicts a variety of Zo curves to illustrate this concept.
Figure 4 The op-amps show a variety of Zo curves. Source: Texas Instruments
If the replacement op amp has a different output stage—and therefore a different Zo curve—you may need to adjust the compensation component(s). Running PSpice® for TI simulations is a relatively quick and easy way to check the phase margin of the design and adjust component values, if needed. Be sure to measure the small-signal overshoot on the bench to verify the simulation results and real-world reliability of the design.
Consideration No. 3: Process technology (bipolar vs. CMOS)
Finally, process technology affects many op-amp specifications, including the offset voltage, drift, common-mode and output swing ranges, the output voltage vs. output current (claw curves), noise, and input bias current. Digging into all of these specifications is beyond the scope of this article, but a couple specifications to highlight include input bias current and noise.
Bipolar amplifiers, or at least op amps with bipolar transistor input stages, have a relatively large input bias current compared to complementary metal-oxide semiconductor (CMOS) amplifiers. That is because the input bias current of bipolar input stage op amps depends on the magnitude of the transistor base current, which is typically in the range of nanoamperes.
While there are techniques for reducing the input bias current in bipolar amplifiers—for example, input bias current cancellation—CMOS amplifiers have considerably smaller input bias currents that are usually in the picoampere or even femtoampere range because their input bias current is caused by the leakage current of the electrostatic discharge diodes that protect the input pins of the device. The input bias current specification is particularly important in applications that have large resistors in the feedback network and when interfacing with high-impedance signal sources. So be careful if you are replacing a CMOS amplifier with a bipolar amplifier in one of these applications.
In addition to input bias current, consider an op amp’s input voltage noise spectral-density curve when replacing an op amp. The curve plots the noise in nanovolts per square root hertz versus frequency. There are two primary regions of this curve: the 1/f and broadband regions. The 1/f region represents the low-frequency noise component, which decreases as the frequency increases. The broadband region is higher-frequency noise, which is typically constant over frequency. The 1/f “noise corner” is where the 1/f region transitions to the broadband region. It is often considered a figure of merit when comparing op-amp noise performance. In general, bipolar amplifiers have lower-frequency 1/f noise corners than CMOS amplifiers. Figure 5 depicts the input voltage noise spectral-density curves for a bipolar and CMOS amplifier.
Figure 5 A comparison of input voltage noise spectral density is shown for bipolar and CMOS amplifiers. Source: Texas Instruments
You should also consider the effects of broadband noise and bandwidth when replacing an op amp. For example, it seems logical that you could easily replace a 6-nV/√Hz, 10-MHz op amp with a 3-nV/√Hz, 50-MHz op amp. However, if the design has no external filtering—for example, an RC filter on the output—the lower-noise, wider-bandwidth op amp actually contributes more noise at the output than the higher-noise, lower-bandwidth op amp.
Other important considerations
The next time you need to replace an op amp, be sure to consider more than supply voltage, package and pinout. For instance, even though two devices may have the same common-mode voltage range, they may have different input stage designs. Depending on the original and replacement op amps, this may introduce a transition region that causes reduced performance when the input signal approaches the positive supply.
Similarly, two op amps may have the same output swing range but very different open-loop output impedance plots. In such cases, you should simulate the design to ensure sufficient phase margin.
Next, replacing CMOS op amps with bipolar op amps (or vice versa) has a plethora of implications. Two of note include input bias current and noise. If the original design has large resistors in the feedback network, interfaces with a high-impedance signal source, or both, compare the input bias current plots.
Finally, take care when comparing op-amp noise performance specifications in conjunction with bandwidth. Just because an op amp has a lower-frequency 1/f noise corner or lower broadband noise does not necessarily mean that it contributes less noise to the signal path.
This article was originally published on Planet Analog.
Peter Semig is the applications manager for general-purpose amplifiers at Texas Instruments.