7 Rail-to-Rail CMOS Amplifiers

CMOS amplifiers also can have input common mode from the negative to positive op-amp power supply (rail-to-rail common-mode range). In many cases, the input common mode actually extends beyond the supply rails. For example, a rail-to-rail CMOS device with a ±2.5-V supply can have common-mode range 200 mV beyond the supply rails (–2.7 V < V_CM < +2.7 V ). From an internal operation perspective, one common way that this feature is implemented is to use both a PMOS and NMOS differential input stage. Observe in the previous CMOS example that a PMOS input stage has common-mode range to the negative supply and NMOS input stage has common-mode range to the positive supply. The rail-to-rail input structure uses both types of input transistors with a circuit to enable and disable the transistors according to their valid common-mode range. There is a small common-mode range, called the crossover region, where both sets of transistors are turned on.

One disadvantage to this approach is that the offset of the PMOS transistor pair is different from the offset of the NMOS pair. This difference causes an abrupt transition of offset voltage when the common-mode voltage transitions through the crossover region. This offset transition introduces crossover distortion that is inherent with this type of rail-to-rail amplifiers. Note that for precision devices, the offset of the two input pairs is often trimmed so that the mismatch is minimized. Even in trimmed devices some crossover distortion is introduced as the offset changes in the crossover region because both input pairs are turned on simultaneously. Figure 7-1 illustrates a simplified rail-to-rail input stage and the associated offset response versus common-mode input. Note that the common-mode range is broken into the PMOS region, crossover region, and NMOS region.

Figure 7-2 illustrates how a sinusoidal signal is impacted by crossover distortion from the rail-to-rail input stage. In this example, the input signal is applied to a unity gain buffer that has an input stage with crossover distortion. The input signal is an ideal sinusoidal waveform, and the output signal tracks the input until the common mode transitions through the crossover region. When passing through the crossover region, the amplifier offset changes so the output signal shifts up or down according to the offset shift. This offset change can range from microvolts to millivolts depending on the amplifier offset. Generally, this shift is too small to notice on an oscilloscope, so this image is exaggerated for illustrative purposes. This kind of distortion is noticeable in the frequency domain or when THD is calculated. The amplifier from Figure 7-3 has a crossover region at 3.75 V and is configured in a buffer configuration. When the input signal avoids the crossover region the distortion is low (THD = 108.5 dB). When the signal passes through the crossover region, the distortion increases (THD = 83.8 dB). Furthermore, inspecting the FFT graph shows significant increase in the harmonic components.

Amplifiers with crossover distortion often provide a specification for common-mode rejection ratio (CMRR), that is defined according to the crossover region. Table 7-1 shows a typical CMRR specification for a rail-to-rail amplifier with crossover distortion. Notice that the CMRR is significantly better when the input range is limited below the crossover region (CMRR_MIN = 76 dB, V_IN< (V+) – 1.4 V) as compared to the entire input range (CMRR_MIN = 65 dB, V_IN < 5.7 V).

A different approach to rail-to-rail common-mode range is to use a single PMOS input stage with an internal charge pump to boost the supply. For this approach, the charge pump generally boosts the internal supply of the input stage by approximately 1.8 V. An amplifier with a 5-V supply is internally boosted to 6.8 V. The internal PMOS stage operates linearly to approximately 1.0 V from the positive supply, so amplifier common mode extends to 5.8 V (6.8 V – 1.0 V = 5.8 V ). Note that although the common-mode range can extend to 5.8 V, the ESD input structure clamps the input at about 5.3 V. This example is illustrated in Figure 7-5. Although the values can differ for different devices, the principle is the same for all zero-crossover devices. Since this approach only uses one input transistor pair, it does not have crossover distortion and is called a zero-crossover amplifier. Figure 7-4 shows the FFT of a zero-crossover op amp (OPA320). This measurement was made under the same test conditions and hardware as was used for Figure 7-3, but the amplifier was changed to a zero-crossover type. Observe by comparing the two figures that when the signal is below 3.75 V, the performance for the two circuits is similar. This is because neither op amp is in a crossover region for this common-mode range. However, when the signal passes through 3.75 V, the OPA316 shows crossover distortion but the OPA320 does not have crossover distortion.

For the zero-crossover device, the input stage does not consume much power, so this charge pump does not need external components to function. Nevertheless, this zero-crossover type input stage has some additional noise from the charge pump circuit. This noise can show up in the time domain as a ripple on the noise signal, or in the noise spectral density curve as a tone at the charge pump switching frequency. However, this noise is quite low for most modern zero-crossover devices, and can be insignificant for many applications. Figure 7-6 illustrates two examples of how the noise transient can show up in the time and frequency domain measurements of a device with the internal charge pump. Notice that the OPA320 has noise tones at 6 MHz and harmonics of this frequency. These tones are related to the switching of the internal charge pump. Comparing the OPA320 to the OPA365 shows larger noise tones for the OPA365. The OPA320 is a modern version of this zero-crossover technology and significant effort has been placed on minimizing the charge pump noise. Also observe that the time domain noise for the OPA365 shows a periodic charge pump signal whereas the OPA320 charge pump signal is buried in the broadband noise. In general, for zero-crossover devices, the charge pump noise is mainly a concern for low-gain applications. In higher-gain applications the amplifier bandwidth is normally below the charge pump switching frequency, so the noise tones are attenuated. Furthermore, an external filter can be added to further minimize any charge pump switching noise.