How to avoid instability in rail-to-rail CMOS amplifiers

MOS transistors have been shrunk in size since their invention decades ago. Reductions in gate oxide thickness, channel length, and width have driven significant reductions in overall circuit size and power consumption. The reduction in gate oxide thickness reduces the maximum allowable supply voltage, while the reduction in channel length and width reduces the form factor and speeds up the product. These improvements have driven the performance of high-frequency CMOS rail-to-rail input/output amplifiers to meet the increasing demands of today's system designers for a new type of analog circuit that must operate at the same low supply voltages as digital circuits.

This application note answers some of the unique questions about the latest generation of CMOS rail-to-rail amplifiers. The article begins with an overview and description of the traditional voltage feedback and current feedback amplifier circuit topologies, as well as the most common causes of feedback amplifier oscillations. For ease of analysis and discussion, the CMOS rail-to-rail amplifier circuit is divided into four major blocks: input, intermediate gain, output, and feedback network stages. The frequency-dependent gain and phase shift of each stage are shown, followed by a complete system simulation that includes all four basic circuit blocks. The second part shows and discusses the mechanisms, tradeoffs, and advantages of three practical solutions for solving amplifier oscillation problems.

Voltage Feedback Amplifier

Figure 1 shows a simplified implementation of the EL5157 - a very popular high bandwidth voltage feedback amplifier. This implementation uses a classic differential input stage to drive a folded cascode second stage, which converts the differential voltage at the input stage into a current at a high impedance gain node that is implemented with the amplifier's high voltage gain. Essentially, the second stage current source output impedance becomes an output signal at a high impedance node, increasing any current differences created in the signal path transistors. The output stage is a push-pull class AB buffer that buffers the high voltage gain to the single-ended output of the amplifier.

Figure 1: Voltage Feedback Amplifier

Output Sensing

An inductor is an electronic component whose impedance is frequency dependent: it is low at low frequencies and increases at high frequencies. An "ideal" op amp output impedance is zero, but in reality the output impedance of an amplifier is inductive and increases with frequency, just like an inductor. Figure 2 shows the output impedance of the EL5157. A common challenge in applications that utilize op amps is driving a capacitive load. This is challenging because the inductive output of the op amp combines with the capacitive load to create an LC tank topology where the capacitive load, along with the inductive driving impedance, creates additional phase lag as the feedback closes around the loop. This reduction in phase margin can cause the amplifier to oscillate. When oscillating, the amplifier can get very hot and may even self-destruct. There are several well-known solutions to this problem.