CN0318

The circuit in Figure 1 is based on the AD5760, a true 16-bit, un-buffered voltage output DAC that operates from a bipolar supply of up to 33 V. The AD5760 accepts a positive reference input range of 5 V to V_DD − 2.5 V and a negative reference input range of V_SS + 2.5 V to 0 V. The AD5760 offers a relative accuracy specification of ±0.5 LSB maximum, and operation is guaranteed monotonic with a ±0.5 LSB DNL maximum specification. Output noise is 8 nV/√Hz, and the AD5760 also exhibits an extremely long term linearity error stability of 0.00625 LSB.

Figure 1 shows the AD5760 configured in the unity gain mode with amplifier input bias current compensation in order to generate a symmetrical bipolar output voltage range. This mode of operation uses an external output operational amplifier, as well as on-chip resistors (see AD5760 data sheet), to provide the input bias current compensation. These internal resistors are thermally matched to each other and to the DAC ladder resistance, resulting in ratiometric thermal tracking.

The AD8675 precision op amp has low offset voltage (75 μV maximum) and low noise (typical values are 2.8 nV/√Hz; and 0.1 μV p-p, 0.1 Hz to 10 Hz) and is an optimum output buffer for the AD5760. The AD5760 has two internal matched 6.8 kΩ feedforward and feedback resistors, which can either be connected to the AD8675 op amp to provide a 10 V offset voltage for a ±10 V output swing, or connected in parallel to provide bias current cancellation. In this example, a bipolar ±10 V output is shown, and the resistors are used for bias current cancellation. The internal resistor connection is controlled by setting a bit in the AD5760 control register (see the AD5760 data sheet).

The ADR4550 is a high precision voltage reference that offers excellent temperature stability (2 ppm/°C maximum , B-grade) and ultra-low output voltage noise (2.8 μV p-p, 0.1 Hz to 10 Hz). These features make it an ideal reference for the AD5760.

In order to obtain a ±10 V output voltage range, the +5 V reference voltage from the ADR4550 is amplified to ±10 V (as shown in Figure 1) by using the AD8675 and AD8676 (dual AD8675).

The output buffer is again the AD8675, used for its low noise and low drift. This amplifier in conjunction with the AD8676 (AD8675 dual) are used to amplify the +5 V reference voltage from the low noise ADR4550 to +10 V and ---10 V respectively. R1, R2, R3 and R4 in this gain circuit are precision metal foil resistors with 0.01% tolerance and a temperature coefficient resistance of 0.6 ppm/°C. R6 and C4 form a low-pass filter with a cutoff frequency of approximately 10 Hz. The purpose of this filter is to attenuate voltage reference noise.

The two AD8675 op amps in the circuit can be replaced with a single AD8676 dual amplifier if desired. However the EVAL-AD5760SDZ board was designed for flexibility in the output stage, therefore two AD8675 op amps were chosen.

The digital input to the circuit is serial and is compatible with standard SPI, QSPI, MICROWIRE^®, and DSP interface standards.

Linearity Measurements

The precision performance of the circuit shown in Figure 1 is demonstrated on the EVAL-AD5760SDZ evaluation board using an Agilent 3458A multimeter. Figure 2 shows that the integral nonlinearity as a function of DAC code is well within the specification of ± 0.5 LSB.

Figure 3 shows that the differential nonlinearity as a function of DAC code is within the ±0.5 LSB specification.

Noise Drift Measurements

To realize high precision, the peak-to-peak noise at the circuit output must be maintained well below 1 LSB, which is 152 μV for 16-bit resolution and a +10 V unipolar voltage range, and 305 μV for a 20 V peak-to-peak voltage range.

A real world application will not have a high-pass cutoff at 0.1 Hz to attenuate the 1/f noise, but will include frequencies down to dc in its pass band. With this in mind, the measured peak-to-peak noise is shown in Figure 4 for a +10 V unipolar voltage range and in Figure 5 for a ±10 V bipolar voltage range. In both cases, the noise at the output of the circuit was measured over a period of 100 seconds, effectively including frequencies as low as 0.01 Hz in the measurement.

Figure 4 shows the noise performance of the signal chain for a 10 V output span (1 LSB = 152 μV). The 10 V range is obtained by grounding the V_REFN input of the AD5760.

The peak-to-peak output noise for the 10 V range in Figure 4 is summarized below:

• Zero scale = 0.96 μV p-p = 0.006 LSB p-p
• Half scale = 7.46 μV p-p = 0.05 LSB p-p
• Full scale = 12.88 μV p-p = 0.08 LSB p-p

The zero-scale output voltage exhibits the lowest noise because it represents the noise from the DAC core only because the V_REFN input is connected to ground. The noise contribution from each voltage reference path is attenuated by the DAC when the zero-scale code is selected. At low frequencies, temperature drift and thermocouple effects become contributors to noise. These effects can be minimized by choosing components with low thermal coefficients. In this circuit, the main contributor to low frequency 1/f noise is the voltage reference. It also exhibits the greatest temperature coefficient value in the circuit of 2 ppm/°C.

Figure 5 shows the noise performance of the signal chain for a 20 V output span (1 LSB = 305 μV).

The peak-to-peak noise for the 20 V range in Figure 5 is summarized below:

• Zero scale = 18 μV p-p = 0.06 LSB p-p
• Half scale = 2.47 μV p-p = 0.008 LSB p-p
• Full scale = 9.22 μV p-p = 0.03 LSB p-p

The noise is lowest at half scale because the DAC core provided the most attenuation of the references at this point. The noise at zero scale is larger than full scale because the negative reference passes through an additional buffer stage.

Complete schematics and layout of the printed circuit board can be found in the CN-0318 Design Support package: www.analog.com/CN0318-DesignSupport.