11 Classic Bipolar vs Rail-to-Rail Output Stage for CMOS and Bipolar

Figure 11-1 compares a classic bipolar output stage to a CMOS and bipolar rail-to-rail output stage. The classic bipolar output swing is limited by a base-to-emitter drop of Q1 and the saturation of Q3. This configuration is sometimes called a common emitter push-pull output. The rail-to-rail bipolar output stage uses a common collector configuration. The output swing limitation for the rail-to-rail bipolar common collector circuit is the saturation voltage of the output transistor. For bipolar devices the saturation voltage is between 0.2 V and 0.3 V. Thus, for bipolar rail to rail the best output swing is approximately 0.2 V from the supply rail.

The rail-to-rail CMOS common-drain configuration is only limited by the voltage drop when the MOS transistor is driven into triode or ohmic region (V_DS ≤ V_SAT)⁽¹⁾. This minimum voltage is set by the physical size (channel width / length) of the transistors. The size sets the minimum resistance for the transistor that occurs when the transistor is fully turned on. The example shows the output swing to the positive rail is limited by the minimum drain-to-source voltage of the PMOS output transistor. The voltage depends on the current flowing through the transistor and the resistance of the transistor. A large (W / L) output transistor has a small resistance and can have a very low voltage drop if the output current is small. It is not unusual for CMOS output stages to have output swing within millivolts from the supply rail when the output current is low. However, increasing the load current causes a larger voltage drop on the output transistor so the output swing is degraded.

Figure 11-1 Classic Bipolar Output Stage vs CMOS and Bipolar Rail-to-Rail Output Stage

One key difference between the bipolar device and CMOS device is that the CMOS device acts like a resistor when fully-on, but the bipolar device has a 0.2-V saturation voltage that is relatively constant for different currents. For low output currents CMOS rail-to-rail devices can swing very close to the rail, whereas bipolar devices are limited by the 0.2-V saturation voltage. However, at higher currents the output swing on the CMOS degrades but the bipolar output swing remains relatively constant at 0.2 V. Figure 11-2 shows that for very low output currents the CMOS device swings within millivolts from the rail, but at 20 mA of output current the swing is degraded by approximately 1 V. Conversely, the output swing bipolar device remains relative constant from 0 mA to 30 mA of output current.

Figure 11-2 Bipolar Rail-to-Rail vs CMOS Rail-to-Rail

One common request is to design an amplifier that can swing all the way to the supply rail (0-V swing limitation). Unfortunately, rail-to-rail output swing is not practical. Even if the output transistors are sized to be quite large it is always necessary to have an output-stage bias current in the output stage for linear operation. Thus, a current flows in the output transistor even when the load current is zero. This bias current creates a voltage drop (V_DS – R_ON × I_Q(OUT)) on the output transistor and limits the output swing. Increasing the size of the output transistor helps minimize this limitation but cannot eliminate the output swing limitation. Furthermore, there is a practical limit to the size of the output transistors from a cost and performance perspective.