The electronics industry continues to see a significant growth in the number of portable devices being produced. Portable units rely on batteries for their power and battery power is most efficient when the circuitry is restricted to a single supply voltage. To further these designs, analog circuit manufacturers provide many operational amplifiers designated for "single supply" and "rail-to-rail" operation.
The data sheets that describe these devices contain specifications and typical performance curves that never appeared in data sheets for classical operational amplifiers. The inclusion of this new data can give rise to questions concerning the impact on performance. For best performance, recognize the limitations and performance tradeoffs and be aware of those application circuits that minimize the effects on overall system operation.
Operational amplifier designs now push the signal operating range as close as possible to the supplies on both input and output to achieve the greatest dynamic range possible for the given supply voltage. Classical operational amplifiers were designed to operate from 15V supplies with performance specified for 10V signals. Designing an operational amplifier with output voltage swing to both supply rails requires some deviations from classical circuit structures.
A classical output stage would be structured with complementary common emitter stages as shown in Figure 1. Transistor Q2 acts as a current source feeding the base of Q1. The minimum voltage drop from the supply rail to the output is the VCE-SAT of Q2 added to the VBE of Q1. This total voltage drop would be excessive for the operational amplifier to be considered rail-to-rail output in most cases.
Figure 1. Classical operational amplifier output stage.
This stage does have the advantage of a low output resistance. As the output stage is a unity gain circuit, there is a minimal gain change with change in load resistance.
A rail-to-rail output stage is generally composed of complementary common-emitter or common-source stages as shown in Figure 2. The voltage drop from the supply rail is just the VCE-SAT of a single transistor. While this reduces the voltage drop from the output to the supply rail, it makes the output a gain stage. The gain of this stage is now dependent on the load resistance. This can affect two primary parameters, the open loop gain under load and the output impedance.
Figure 2. Output stages that swing very close to the supply rails are known as rail-to-rail outputs although the output voltage does not actually reach the rail voltage.
The significance of this dependency is seen as characteristic curves include gain vs. load resistance. As the load impedance decreases the gain decreases.
Figure 3. Open loop gain is dependent upon the load resistance as is shown in this typical curve for the TLV2731.
Changing open loop gain with load impedance may be significant in high accuracy designs. Gain relationships are given for both inverting and non-inverting configuration operational amplifiers below. Notice the error term that contains the open loop gain variable is the same for both circuits and the gain error is a function of the equivalent closed-loop, non-inverting gain.
To evaluate the impact of this variable gain on a design, consider the error term values shown in Table I. For an amplifier with an open-loop gain of 120dB (gain ratio of 1,000,000V/V) and a closed-loop gain of unity the response is almost perfect, however, that same amplifier configured for a gain of 1000 will show a 0.1% error. If a change of load resistance causes a 10dB drop in open loop gain, the gain of 1000 circuit drops from 999 to 996.8. As the open-loop gain of the amplifier decreases the affect of the changing open-loop gain becomes more pronounced.
Table I. Error term values dependent on the open loop gain and non-inverting closed loop gain of the operational amplifier.
It is important to note that the gain specification is taken at a very low frequency and that gain drops off with increasing frequency. Therefore, an amplifier stage may have acceptable performance at DC but may degrade significantly at higher frequency. The Gain vs. Frequency plot will predict the gain at the higher frequency.
Another variable introduced with the rail-to-rail output stage performance is the connection point of the load impedance. A stringent condition is with the load returned to a voltage that is half way between the supply rails. Under this condition, both the positive side and negative side transistors will each be conducting during part of the cycle. If the load is returned to ground (negative rail) then the negative side transistor would only be turned on during slewing of the output signal. Specifying an operational amplifier with the load resistor connected to ground, actually the negative supply rail, can allow the device to appear better in the data sheet than in the real application.
Understanding the internal topology of a rail-to-rail operational amplifier can help in designing for optimum performance in the application. Maximum signal swing will occur when the load current is kept to a minimum. In those cases, where the load is returned to one of the supply rails the operational amplifier will usually be able to output a voltage closer to that rail than the other.
One area of misunderstanding might be termed "Output Offset". The basic inverting gain stage shown in Figure 4 illustrates the concept. In this circuit, the voltage at the non-inverting input is ground. With the negative supply below ground, the output can reach that level, however, if the negative supply is ground the output cannot swing to the actual value. It appears that the amplifier has an input voltage offset. The situation is further confused when the user attempts to measure the input offset voltage of the operational amplifier. As the gain of the amplifier is changed, the error voltage at the output is expected to change. In this circuit configuration, it may not. If the input offset attempts to drive the output negative the output will remain at the lowest possible positive voltage which has no relationship to the actual input offset. This issue can be avoided by generating a pseudo-ground or signal-ground voltage that is above the actual power ground.
Figure 4. Basic inverting gain stage illustrates the "Output Offset" error in rail-to-rail operational amplifiers.
As the rail-to-rail output specification impacts the circuit design so does the single supply qualifier on the input stage. The term "single supply", when used to describe an operational amplifier, is not strictly defined within the industry. Several manufacturers interpret it to include any operational amplifier that has at least one of the supply rails within the common mode voltage (CMV) range. Perhaps the most common circuit topology to accomplish the full CMV range beyond both supplies (rail-to-rail input) is the parallel input stage as shown in Figure 5. While this circuit shown is realized in MOS FET devices a similar topology could be realized with bipolar transistors.
Figure 5. Parallel input structure that allows CMV to range outside both power-supply rails.
This input stage has three modes of operation. For a CMV from less than the negative supply to some voltage less than the positive rail, the P-channel pair of FETs (Q3 and Q4) is active. For a CMV from less than the P-channel upper limit to a voltage greater than the positive rail, the N-channel pair of FETs (Q1 and Q2) is active. Since one stage turns off after the other stage turns on there is a range of inputs where both stages are active. Each stage may have a different voltage offset, CMRR, power supply rejection ratio (PSRR), and voltage offset drift. The amplifier may show a third set of parameters when operated within the transition region.
The circuit shown in Figure 6 demonstrates the three distinct operating regions. The resistor values set the differential gain at 100. The triangle wave signal generator sweeps the CMV over its entire range.
Figure 6. A circuit with three distinct operating regions.
The scope trace presented in Figure 7 shows the voltage waveforms associated with the circuit in Figure 6 for two samples of the same model operational amplifiers. The top trace is the input CMV that goes from --2.5V to +2.5V. This is near the full rated CMV. The center trace is the output from one amplifier and the bottom trace is the output from a second operational amplifier. All three operating modes are visible in both output traces. On the left side is the response from the P-channel input pair. The right side of the trace is the N-channel pair response. The transition zone shows very plainly.
Figure 7. Top trace is the CMV applied to operational amplifier. Center trace is output from one amplifier and bottom trace is output from second amplifier.
To view the output traces in terms of the operational amplifier performance specifications the voltage at any point on the trace equates to the input voltage offset at that CMV. Normal practice is to specify the input voltage offset at zero CMV (the vertical mid-line in the trace). It should be noted that the voltage offset for any input pair might be close to that of the associated input pair but this is not necessarily the case. It is also seen that the offset of either stage may be positive or negative. Some devices are laser-trimmed at the wafer level to adjust the voltage offset to near zero. This can result in very good performance. The center trace on Figure 7 shows a shift of approximately 330mv and the bottom trace shows approximately 50mV shift. Since the operational amplifier is in a gain of 100, these levels translate to 3.3mV and 0.5mV respectively at the amplifier input.
The slope of the trace relates to the CMRR of that input pair. It is observed that the CMRR in the transition region can be very poor. Almost all data sheets that specify a CMRR value over the full CMV range report the end-point average of the trace shown in Figure 7.
While this characteristic of single-supply operational amplifiers appears large, its affect on the signal can be minimized or totally avoided with careful design of the application circuit.
The CMV experienced by an operational amplifier is determined by the application circuit and input signal. The circuits of Figure 8 show various configurations for single ended input signals. To avoid the "Output Offset" as discussed above, the operating zero point is established with the bias voltage marked VB.
Figure 8. Basic signal treatments affect how the operational amplifier input stage experiences the signal.
The extent to which this input transition step causes a problem is determined by whether the input common-mode signal crosses the transition point and the gain setting of the amplifier. A full-scale signal into the simple voltage follower shown in Figure 8A will cross the transition point and the effect on the output signal may be significant. The operational amplifier response shown in Figure 7 center trace will cause a 3.3mV step in the output signal. For a full five-volt output signal, this is a --64dB response or almost 3 bits in a 12-bit system. In a 16-bit system, this covers over 43 bits.
Circuit B, in Figure 8 can cause the most problem. If the circuit is set for gain and the value of VB is set at the transition voltage, an error will appear in the output waveform. This error signal can be very difficult to isolate because the transition point differs over a range of devices and the magnitude of the step can vary by more than ten to one from one unit to the next. The transition voltage for a great number of amplifiers is 1.5 to 1.6V below the positive supply rail. This creates a zone of greater error in those designs where the supply voltage is three volts and the signal ground or zero reference is created at one-half of the supply voltage.
The circuit in Figure 8C is very forgiving. Even if the value of VB is set at the transition voltage, there will be no output distortion because the common mode voltage does not change with changing input signal.
Perhaps the circuit with the greatest potential for trouble is the differential amplifier of Figure 6. This circuit, with the gain of 100, will add a step error signal in the output when the common-mode voltage crosses the transition region. Significant confusion will exist because this error signal would not be associated with the signal of interest but the signal that is being eliminated, the common-mode signal.
The single most significant specification that will indicate a minimum step across the transition region is the CMRR over the full-scale range of CMV. The OPA350, with a minimum CMRR of 76dB for an input change from --0.1V to 5.6V will show an error step of less than 1mV referred to the input.
The performance of rail-to-rail output and single supply operational amplifiers has improved greatly. When used with care these devices can prove useful in designing low power and battery-supplied systems. The secret to success is to know where the possible pitfalls are located and then design to avoid them.
Editor's note: Many years in the shadow of the great Jerry Graeme, Bill Klein - with more than 30 years of applications experience - is a superstar in his own right. We look forward to many more of his articles and commentaries.