Low voltage amplifiers are often used in single-supply voltage circuits and make use of the rail-to-rail output feature these amplifiers have. While the output of these amplifiers can swing close to zero volts, they do not reach zero volts at the output. This can cause problems when driving analog/digital converters (ADC) due to the loss of some portion of the low side of the ADC's input range. In another case, when amplifiers are used in a series of gain stages, the saturated output of an amplifier is amplified by the gain of the succeeding stage, and the minimum output is the output saturation of the first stage multiplied by the gain of the succeeding stage.
This problem with low-voltage amplifiers in a single-supply voltage application can be fixed by supplying a small, negative-supply voltage to the amplifier. This voltage enables the amplifier's output to swing to zero volts while keeping the total supply voltage at less then the maximum operating-supply voltage specification of the amplifier. This article discusses several methods of creating small, negative-supply voltages for use with amplifiers.
The Problem
In many cases, sensor signal conditioning requires that a sensor's output be scaled to a range that includes zero volts. For example, in Figure 1, an ADC is using a 2.5-volt reference voltage so the ADC's input range is 0 to 2.5 volts. To use the full dynamic range of the ADCs input span, the sensor's output is required to be scaled to range from 0 to 2.5 V.
When amplifiers are operated from a single-supply voltage, the outputs will not swing to zero volts, and rail-to-rail outputs will have some output-saturation voltage relative to ground. In the sensor's signal-conditioning path, the result is the loss of the lower range of the ADC's input.
For example, if an amplifier's output swing low is 50 mV and the output is driving a 12-bit ADC with a 2.5 volt reference, the lowest 82 values of the ADC's input are lost. [(0.050 V)/(2.50 V/4096) = 81.9] The 50 mV minimum output swing represents 2% of the ADC's input range under these conditions. Additionally, as output saturation is approached, the amplifier's gain decreases, which adds nonlinearities and distortion to the output signal, near the saturation level.

Figure 1: Loss of ADC dynamic-range example
(Click on image to enlarge)
In other situations, the signal chain uses several amplifiers in series and the output saturation voltage of an amplifier is amplified by following amplifier stages, thus increasing the minimum output voltage that can be output. Figure 2 is a schematic that details this problem.
Figure 2 shows a two-amplifier circuit, where each amplifier has the same characteristics, with the gain of A1 equal to 1+RF1/RG1. The gain of second amplifier, A2, is equal to 1+RF2/RG2. If the signal input is zero, ideally, it would be desirable to have the VOUT2 voltage to equal zero.
In the case of amplifiers operating from a single supply, the output saturation voltage of A1, VOUT1, is amplified by A2. For example, if the signal input is 0 V, the gain of A2 is 10 and the output saturation voltage of A1 is 50 mV, then the minimum output voltage of A2 is 250 mV (50 mV x 5 = 250 mV). The result is that the output swing of A2 will not go below 250 mV, even though the output of A2 could swing to 50 mV.

Figure 2: Amplified saturation-voltage example
(Click on image to enlarge)
Figure 3 is a schematic for testing an amplifier's output swing and is used for the measurements shown in this article. This circuit will source or sink about 1 mA (peak) when the output is near the supply voltage rails. The jumper J1 provides a choice of ground or a -0.23 voltage source for the negative supply voltage.

Figure 3: Amplifier output-swing test circuit
(Click on image to enlarge)
Figure 4 shows the output signals of LMP7731, LMC7101 and LMV851 amplifiers with a single 5-volt supply and a noninverting gain of two using the test circuit of Figure 3. The input signal is a triangle wave which ranges from 0 to 2.5 volts. Figure 4A, 4C, and 4E shows the input and output voltage with respect to time, while Figures 4B, 4D, and 4F show the X-Y plot of the input and output signal.
The amplifiers display the different levels of output saturation and different levels of presaturation distortion. The white crosshairs in Figures 4B, 4D, and 4F is the X=0 and Y=0 point, while the horizontal segment of the green trace is the output saturation voltage and the sloped portion is the active region of the amplifier's output. The curvature between the sloped segment and the horizontal segment is a transition zone where the amplifier is loosing gain and output signal distortion is occurring.
The LMP7731 and LMV851 plots show a relatively sharp transition between the active and saturation regions of the output stage. These transitions allow the output to swing closer to the output-saturation level before significant distortion is added to the output signal, when compared to amplifiers with less output drive such as the LMC7101. The LMC7101 plots show a higher output-saturation voltage and gradual transition between the active and saturated regions of the output stage.
This data shows the two effects of driving the output close to the voltage-supply rails. The output saturation will limit the voltage swing, and distortion will be added if the output is driven close to the output saturation level. Recent amplifier designs, such as the LMP7731 and LMV851, have increased the output drive to reduce the distortion near the saturation level and lower the output saturation voltage.

Figure 4: Output vs. input signal showing saturation and distortion
(Click on image to enlarge)
A Proposed Solution
Most low-voltage amplifiers have a maximum operating voltage of 5.5 V, while many of the 5-V voltage regulators have an output of 5V ±5% specification, or a 4.75 to 5.25 V range. The maximum output of 5.25 V from the regulator is 0.25 volts lower then the maximum amplifier operating voltage. If a small negative voltage of -0.25V is used as the negative supply for the amplifier, the amplifier's output would be able to swing to 0 V while maintaining a total supply voltage of less then 5.5 V. A negative supply voltage of -0.23 V ±5%, or a range of -0.217 to -0.242 V, would meet this requirement.
A Negative Bias Generator
A device such as the LM7705 is designed to generate a small negative voltage from a positive supply voltage, through the use of a charge pump inverter and a low dropout regulator. It will sink up to 20 mA from the negative supply pin of an amplifier. Figure 5 is a schematic showing the LM7705 being used with the output swing test circuit shown in Figure 3.

Figure 5: Amplifier with a -0.23 volt negative-supply generator
(Click on image to enlarge)
Figure 6 shows the measurement results for the same LMP7731, LMC7101, and LMV851 amplifiers measured for the data in Figure 4 and in the three cases the amplifiers' output swings to zero volts. Also observable is the offset voltage difference between the LMP7731, LMC7101 and the LMV851. The X-Y plot of the LMC7101 and LMV851 shows a shift in the negative Y direction and represents the output referred offset voltage of the amplifier.
The noise gain of the test circuit is two, so the input offset voltage is half of the output referred offset voltage. As can be seen by comparing Figures 4 and 6, the LM7705 and its -0.23 V output is effective at providing the amplifier with an output range that includes zero volts.

Figure 6: Output vs. input signal with negative bias on negative-supply pin
(Click on image to enlarge)
The LMP7705 is capable of sinking up to 20 mA, which is adequate for a majority of sensor interfacing and signal-path circuits being used. In those cases where greater then 20 mA of current is required, the circuit in Figure 7 can be used to sink up to 40 to 50 mA.

Figure 7: Higher-current negative-bias generator
(Click on image to enlarge)
The circuit in Figure 7 uses the LM2662, a charge pump inverter, to generate a -5 volt supply voltage and is used to supply a negative voltage to the LM8261 amplifier. The LM8261 is a high output-current drive with unlimited capacitive-load capability and is configured as an inverting amplifier with a gain less then one.
In this example, a 4.096 V voltage reference is scaled down to a -0.23 V value. The voltage reference is typically available for use with the ADC. If a different value voltage reference is used, for example 2.048 or 2.500 volts, the ratio of the gain resistors R1 and R2 will have to be recalculated to scale that reference voltage to -0.23 V. The unlimited capacitive-load drive of the LM8261 allows the use of bypass capacitors on the signal chain amplifiers being used.
In conclusion, this article discussed two problems with signal fidelity which may occur when operating amplifiers using a single supply voltage, and how a small negative voltage can be used to remove those problems and achieve a real, true-zero output. The small negative-supply voltage is particularly useful with low-voltage amplifiers so that their maximum operating voltage is not exceeded. The LM7705 was shown to be an effective solution for generating the negative bias voltage for most situations, while the LM2661 and LM8261 combination could be used if the negative supply-current requirement exceeds 20 mA.
About the Author
Walter Bacharowski is an amplifier-applications manager at National Semiconductor Corporation, where he has worked for 15 years. He has a bachelor's degree in electrical engineering from Cleveland State University and has had continuing education in engineering, management, marketing, and technology. His personal interests include electronics, model rocketry, and alternative-energy technology. He can be reached at Walter.Bacharowski@nsc.com.