Electronic Design

# Composite Amplifier Boosts Precision Of Programmable Voltage Source

Engineers often need dc voltage sources with very high precision and good resistive and/or capacitive load-driving capability for industrial applications. However, precision benchtop power supplies can be very expensive if high accuracy also is required. IC-based dc voltage sources, such as references and regulators, cost less but only partially satisfy the requirements. References have good dc accuracy yet lack resistive or capacitive load-driving capability. On the other hand, regulators have the load-driving power but not the necessary accuracy and software programmability.

A power output stage that can drive capacitive loads is shown in the figure. It has a precision input circuit and voltage programmability from ­2.5 to +2.5 V, with 16-bit accuracy. The circuit employs a dual auto-zero amplifier, A1a-A1b (AD8552); a high-output current amplifier, A2 (AD8531); a 16-bit serial-input digital-to-analog converter (DAC), A3 (AD5542); and a precision voltage reference, A4 (AD780). If programmability isn't needed, the circuit can use a fixed reference instead of the DAC.

Circuit operation goes as follows: The DAC (A3) is configured for bipolar operation when combined with A1a. Its output can be programmed to ±VREF, which is set to 2.5 V by the AD780. In combination, the DAC and amplifier produce a negligible error. Worst-case, bipolar output-voltage error, VOUT-BIP, can be calculated as shown in the Equation Listing, where:

D = code loaded to the DAC
VREF = reference voltage = +2.5 V
VGE = gain error = ±5 LSB
ZE = zero-scale error = ±1 LSB
INL = integral linearity = ±1 LSB
VOS = amplifier offset voltage = 5 µV
RD = resistor matching = ±0.00076%
A = amplifier open-loop gain
= 1,000,000

For example, when the code loaded to the DAC is 1 LSB (code 1000 0000 0000 0001), the worst-case error will be approximately 350 µV. These small errors can be stored as CAL factors in a calibration table if even higher accuracy is required.

The programmed voltage is fed into the unity-gain composite amplifier composed of A1b and A2. This composite amplifier performs close to the ideal, as the input stage of the AD8552 (A1b) provides a very high common-mode rejection ratio (CMRR), power-supply rejection ratio (PSRR), open-loop gain, and very low offset voltage.

The composite amplifier also has a very high output current and good capacitive and resistive load-driving capability due to the AD8531. Additionally, both the AD8552 and AD8531 have rail-to-rail inputs and outputs. Using the equation below, users can calculate the worst-case error of this composite amplifier:

VOUT = (VIN ­ VIN_DIFF)
VIN = VOUT_BIP
VIN_DIFF = VOS + en + VOUT/A +VICM/CMRR +ΔVS/PSRR

where:

en = amplifier input noise = 20 nV/√Hz
CMRR = 120 dB
PSRR = 120 dB

When light loads are driven (RL > 500 Ω), the composite amplifier circuit adds less than 10 µV of error to the programmed signal coming out of the DAC. For larger than ±120-mA output currents, A2 limits the output performance.

The above circuit can function as a precision, programmable reference with sink and source output current, or as a low-dropout regulator with excellent load regulation and efficiency. Proper bypassing of the am-plifiers and additional pins on the DAC and DSP are omitted for clarity. 