The common-mode voltage range of an amplifier is the range of usable input voltages allowing for linear operation. Depending on the input stage topology, amplifiers may have common mode input range (V_CM) which may be limited relative to one or both supply rails; best case scenario is achieved when input voltage range extends slightly beyond both supply rails (rail-to-rail operation). Limitations in the V_CM range come from operating voltages needed to bias the transistors in the input stage, and to ensure operation within a linear range (saturation range for MOSFET or active range for bipolar transistors). We’ll illustrate these limitations for MOSFET amplifier input stages. Figure 1-1 shows a simplified representation of an N-channel MOSFET input stage. The stage consists of a current source (single NMOS, Q3, shown for simplicity), a differential pair with input voltage applied to the gate of each transistor, and an active load PMOS current mirror, Q4 and Q5. The NMOS differential pair has an input common mode voltage limitation with respect to the negative rail, –V_s. Performing Kirchhoff’s voltage walk from –V_s to V_in+, we have:

Therefore, the N-channel MOSFET V_CM range is limited to -V_s by a certain voltage as detailed in Equation 1. Note that we make some important assumptions, namely that V_sat of both NMOS and PMOS transistors are perfectly matched and equal to 0.1V (a typical value). Similarly, we assume that V_gs for both NMOS and PMOS transistors are equal and have a value of 0.9V, a typical value to turn on the transistors. With these assumptions in mind, we can say that this simplified NMOS input stage allows input common mode voltage range operation about 1 V from V_s. Performing Kirchhoff’s walk from the opposite end, +V_s to V_in±, we get:

Given that V_gs typically exceeds V_ds for a MOSFET at the edge saturation (V_sat at a minimum, V_gs at a maximum), the maximum V_CM comes within V_sat, or 100 mV, of the positive rail, +V_s.

Conversely, the P-channel MOSFET input stage is limited on the positive side, usually on the order of 1 V from positive rail, +V_s. On the negative side, the common mode voltage range of a P-MOSFET can come within V_sata, or 100 mV from the negative rail,-V_s.

To avoid the limitations of the single differential pair input stage, a complimentary N-channel and P-channel MOSFET (CMOS) input stage design can be used. This design uses two input differential pairs (an N-channel MOSFET pair and a P-channel MOSFET pair), a current steering scheme, and a double-folded cacose summing the two input signals (Figure 1-2). V_set is a voltage source used to control the functionality of the diverting transistor Q₈. For common mode voltage below +V_s – V_set, Q₈ is off, and the drain current (I_d) from Q₅ (current source) flows straight through the P-channel differential pair (Q₁ and Q₂). The double-folded cascode allows the drains of Q1 and Q2 to be biased down to V_sat above – V_s, resulting in V_CM swing below the negative rail. This allows the V_CM to extend a certain voltage, ΔV_P, below the negative rail. Similarly, for common mode voltage above +V_s – V_set Q₈ is on, and I_d is steered from the P-channel pair to the N-channel pair via the current mirror (Q₆ and Q₇). The V_CM range can therefore exceed the positive rail, +V_s, by a certain voltage, ΔV_N. To sum it up, this gives an op-amp with this input stage topology a rail-to-rail V_CM range as detailed in Equation 3.

Now that we understand the rail-to-rail operation of a complementary input stage, we more elaborately address ΔV_N and ΔV_P. Looking deeper into the complementary input stage amplifier in Figure 1-2, we can see that the rail-to-rail input performance is dependent on the second stage. Using Kirchhoff’s voltage law from the positive rail down to the input, similar to the one performed previously on Figure 1-1.

We can see from Equation 4 above that the complimentary input stage amplifier has input common mode range 0.7 V above the positive rail. We can find the common mode input voltage range to the negative rail by performing the same procedure.

From Equation 4 and Equation 5 we find that the common mode input voltage range extends beyond the positive and negative rail typically by about 0.7 V, which is represented by the aforementioned terms ΔV_N and ΔV_P. In data sheets, you will notice that most rail-to-rail amplifiers are specified up to 0.1 V (not 0.7 V) beyond the supplies. This is due to the protection diodes between the input and each rail.

As an example, let’s assume that we are working with OPA391 operational amplifier, which we have in a gain of 100 V/V, powered with a 5.0-V single supply. Let's say we want to measure current between 0 and 50 A, and we choose a 1-mΩ shunt resistor. This means we will see differential input in the range of 0 to 50 A × 1 mΩ = 50 mV. The minimum V_CM value for OPA391 is 0.1V below ground, so our input conditions are in agreement with the data sheet requirements. However, OPA391 A_OL output conditions are specified in the range of –V_s + 0.1 V < V_out < +V_s – 0.1 V. Therefore, between 0 and 0.1 V and 4.9 V to 5 V of output the op-amp may encounter some non-linearity, which is undesirable. We can solve the issue quite easily:

• We can level-shift the input common mode up by 1 mV (which is amplified by the closed loop gain), or level shift -V_s down by 100 mV. Level shifting the input can be achieved through a simple voltage divider circuit, or providing headroom for the output can be accomplished by providing a small negative rail. If no negative rail is present in the system, a negative charge pump like the LM7705 solves that problem in a single IC. In either case, it will put V_out at least 100 mV above –V_s under 0 current condition, which resolves the output swing issue on the low end. Take note that either solution also exceeds the maximum linear operating range on the high end, which is in violation of the output voltage swing specifications. We can solve that issue by decreasing the gain of the circuit or the value of the shunt resistor slightly to bring V_{out (max)} at or below +V_s – 100 mV.

While the complementary input stage described offers an excellent solution to the input common mode problem, it is important to keep in mind that the transition between each pair will generate a change in the input offset voltage of the amplifier, also known as input crossover distortion. The transition can be eliminated by keeping both pairs conducting at all times, but this is often avoided because of the excessive power dissipation required. The input crossover distortion can be circumvented more elegantly by using a zero-crossover amplifier (Zero-crossover Amplifiers: Features and Benefits tech note). These amplifiers use a single transistor pair and an integrated charge pump to push the internal voltage supply enough beyond the nominal value to remain in linear operation.

The output swing range of an amplifier is the range of output voltages allowing for linear amplifier operation. As with V_CM, the output swing (V_out) limitations are related to operating voltages of transistors in the output stage. Depending on the application and topology, V_out may be more or less limited relative to the rails, regardless whether they are single, dual, or asymmetric. Perfect rail-to-rail performance does not exist in practice, although some of the complimentary MOSFET designs come fairly close.

Many applications require V_out swing to only one rail, typically the negative rail. The earliest op-amp output stages accomplished this by having an NPN emitter-follower configuration with a resistive pull-down (Figure 3-1 A). A pull-down resistor to the negative rail allows the output to approach the negative rail, but this greatly limits the sinking current and results in slow output response. A similar design (bipolar or MOSFET) utilized NPN/NMOS current sources in place of the pull-down resistor, offering higher gain and near to-negative-rail output swing (Figure 3-1 B).

With the advent of modern complimentary bipolar processes, better matched, high speed PNP and NPN transistors became available. As a result, the complimentary emitter-follower output stage (Figure 3-1 C) was developed with its most significant advantage being low output impedance. The major drawback of this topology is its limited output swing, typically on the order of 1V or more to the rail due to transistor operating voltages in the stage. Specifically, the minimal forward-bias voltage across the PNP current source (V_FB-P), and the base-emitter voltage (V_BE-N) limit swing to positive rail, whereas the V_FB-N of the NPN current source and the V_BE-P limit swing to negative rail. Full output voltage swing of the complimentary bipolar output stage is thus:

More recent complimentary common-emitter or common-source output stages (Figure 3-1 D and Figure 3-1 E), allow the op-amp output swing much closer to rail, but both of these stages have fairly high output impedance. For the bipolar version of this stage, the output swing limitation to each rail comes from V_{ce, (sat)}, or the minimal collector-emitter voltage needed to keep each transistor operating in the linear region. The typical V_sat for a bipolar transistor is 300mV at 25ºC and changes by roughly -2mV per each ºC increase in temperature.

We can perform a similar analysis to the MOSFET version of this stage, illustrated in Figure 3-1 (E). The output swing limitation comes from the MOSFET on-resistance (R_on) while in the triode region, which causes an output voltage range limitation relative to the rail equal to I_d × R_on. In essence, in the non-linear operating region, the MOSFET acts as a small resistor and produces a voltage drop. Under unloaded conditions, I_d = I_q, the limitation caused by the voltage drop is on the order of 5 mV to 50mV, which is considered almost truly rail-to rail performance.

Keep in mind, that under normal operation I_d is equal to the quiescent current of the output transistors, I_q, plus the load current. In other words, the output swing will decrease as the load current increases. A plot illustrating this effect is included in data sheets – look for the output voltage swing vs output current plot (often known as a claw curve). Operate within the range of the curves to remain in linear operation