feature

Low Drift: 1.5 ppm/c; Low Initial Error: 1 mV; Pin Programmable Outputs: +10V, +5V, +65V Tracking, –5V, –10V; Flexible Output Force and Sensing Terminations; High-impedance ground sensing; machine dipping package; MIL-STD-883 compliant version available.

General Instructions

The AD588 represents a significant advance in state-of-the-art monolithic voltage references. The low initial error and low temperature drift give the AD588 absolute accuracy not previously available in monolithic form. The AD588 uses a proprietary ion implanted buried Zener diode, as well as high stability thin film resistors for laser wafer dicing low cost and excellent performance.

The AD588 consists of a basic reference unit and three additional units providing amplifiers with pin-programmable output ranges. This amplifier is laser trimmed for low offset and low drift to maintain reference accuracy. The amplifier has been configured to allow Kelvin connections to the load and/or booster to drive long lines or high current loads, providing the full accuracy of the AD588 required in application circuits. The low initial error allows the AD588 to be absolutely accurate as a reference in precision measurement applications where the system requires 12 bits. In such a system, the AD588 can provide known voltages in software for system calibration, as well as drift allowing other components in the system to compensate for drift. Manual system calibration and periodic costs can therefore eliminate recalibration. Also, the mechanical instability of the trimmer potentiometer was used in conjunction with the AD588 in conjunction with the automatic calibration software.

There are four versions of the AD588. The AD588JQ and AD588KQ and grade packages are specified for 0°C to 70°C operation in 16 lead cermet. The AD588AQ and BQ grades are packaged in 16 lead cermet and specified for the industrial temperature range of -25°C to +85°C.

Product Highlights

1. AD588 provides 12-bit absolute precision without any user adjustment. An optional trim connection is available for applications requiring higher precision. Trimming can be done without changing the zener or buffer operating conditions of the amplifier and therefore will not increase temperature drift.

2. The output noise of AD588 is very low, usually 6μV pp. A pin is provided for additional noise filtering using an external capacitor.

3. No external components are required, and the tracking mode can be connected using the Kelvin output with an accuracy of ±5 V. Tracking error is less than 1 mV, and fine-tuning requires precise symmetry between the +5 V and -5 V outputs.

4. Pin strapping capability allows multiple output configurations: ±5 V, +5 V, +10 V, –5 V and –10 V dual output or +5 V, -5 V, +10 V and -10 V single output.

theory of operation

The AD588 consists of a buried Zener diode reference, amplifiers for pin-programmable output ranges, and associated thin-film resistors as shown in Figure 1. A temperature compensation circuit provides the device with a temperature coefficient of 1.5ppm/°C or less.

Amplifier a1 performs multiple functions. The main function of a1 is to amplify the zener voltage from 6.5v to the desired 10v output. In addition, A1 provides external adjustment of the 10 V output via pin 5 gain adjustment. Using a bias compensation resistor between the Zener output and the non-vertical input of A1, a capacitor can be added at the noise reduction pin (Pin 7) to form a low pass filter and reduce the Zener's noise contribution to the circuit. Two matched 10 kΩ nominal thin film resistors (R4 and R5) split the 10 V output in half. Pin VCT (pin 11) provides access to the center of the voltage range and pin 12 (bal adj) can be used for fine tuning of this partition.

The ground sense of the circuit is provided by amplifier A2. The non-vertical input (pin 9) senses the system ground, and the ground is transmitted to the point on the circuit where the inverting input (pin 10) is connected. This could be pin 6, 8 or 11. The output of A2 drives pin 8 to the appropriate voltage. So if pin 10 is connected to pin 8, the voltage on the vlow pin will be the same as the system ground voltage. Alternatively, if pin 10 is connected to the VCT pin, it will be ground, and pins 6 and 8 will be +5 V and -5 V, respectively.

Amplifiers a3 and a4 are internally compensated to buffer the voltages at pins 6, 8 and 11 and provide full Kelvin outputs. Therefore, the AD588 is fully Kelvin capable by providing a means of sensing the system ground and providing forcing and sensing outputs referenced to that ground.

Application of AD588

The AD588 can be configured to provide +10V and -10V reference outputs, as shown in Figure 2A and 2C, respectively. It can also be used to provide a +5 V, -5 V, or ±5 V tracking reference, as shown in Figure 2B. Table I details the appropriate pin connections for each output range. In each case, pin 9 is connected to system ground and power is supplied to pins 2 and 16.

The AD588's architecture provides ground sensing and an uncommitted output buffer amplifier, giving the user a great deal of functional flexibility. The AD588 is specified and tested in the configuration shown in Figure 2A. The user can choose to take advantage of the many other configuration options provided by the AD588. However, the performance of these configurations is not guaranteed to meet the extremely stringent data sheet specifications.

As shown in Table I, amplifier A4 in a +10V configuration can be used to provide a +5V buffered output (Figure 2A). -5V buffered output available with -10V amplifier A3

configuration (Fig. 2c). Specifications are not guaranteed for the +5 V or -5 V output in these configurations. The performance will be similar to that specified for the +10 V or -10 V output. Unbuffered outputs are available at pins 6, 8, and 11. Loading these unbuffered outputs will hurt circuit performance.

Amplifiers a3 and a4 can be used interchangeably. However, the AD588 was tested (and guaranteed to specification) with an amplifier connected as shown in Figure 2A and Table I. When A3 or A4 is not used, its output force and sense pins should be connected, and the input should be grounded.

Two outputs of the same voltage can be obtained by connecting A3 and A4 to the appropriate unbuffered outputs on pins 6, 8 or 11. The performance of these dual output configurations is generally within data sheet specifications.

calibration

In general, the AD588 will meet the requirements of a precision system without additional tuning. The initial output voltage error is 1 mV, and the output noise specification is 10 μV pp, allowing an accuracy of 12 to 16 bits. However, in applications requiring higher accuracy, additional calibration may be required. Trimming has been done by using gain adjustment pins and balance adjustment pins (pins 5 and 12 respectively).

The AD588 provides a precision 10 V span with a center tap (VCT) used with a buffer and a ground sense amplifier to achieve the voltage output configurations in Table I. Gain adjustment and balance adjustment can be used in any of these configurations to fine-tune the magnitude of the span voltage and the position of the center tap within the span. Gain adjustment should be done first. Although trimming is not interactive within the device, gain trimming moves the balance trimming point when changing the range size.

Figure 2b shows gain and balance trimming in +5v and -5v tracking configurations. Each trim uses a 100 kΩ 20-turn potentiometer. The gain trim pot is connected between pin 6 (vHigh) and pin 8 (vLow) and the wiper is connected to pin 5 (gain adjustment). Adjust the potentiometer so that it produces 10 V between pin 1 and pin 15 of the amplifier output. Then adjust the balance potentiometer (also connected between pins 6 and 8 and the wiper to pin 12 (BAL ADJ)) to adjust the range from +5 V to –5 V.

Trim works exactly the same way in other configurations. When generating +10 V and +5 V, the gain adjustment is used to fine-tune +10 V, and the balance adjustment is used to fine-tune +5 V. In the -10 V and -5 V configurations, the gain adjustment is again used to fine-tune the magnitude of the range -10 V, and the balance adjustment is used to fine-tune the center tap, -5V.

In a single output configuration, the gain adjustment is used to trim the output using full scale (+10 V or -10 V), while the BAL ADJ is used to trim the output using half scale (+5 V or -5 V). The input impedance on the gain trim and balance trim pins is approximately 150 kΩ. The gain adjustment trim network effectively attenuates 10 volts by a factor of 1500 through the trim pot to provide -3.5 mV to +7.5 mV with a resolution of approximately 550 µV/rev (20-turn pot). The bal adj trim network attenuates the trim voltage by a factor of approximately 1400, providing a trim range of ±4.5 mV and a resolution of 450 µV/rev.

Trimming the AD588 produces no additional temperature error, so precision potentiometers are not required.

For a single output voltage range, or in a balanced case, no adjustment is required, pin 12 should be connected to pin 11. If gain adjustment is not required, pin 5 should be left floating.

Noise performance and reduction

The noise produced by the ad588 is typically less than 6µv pp over the 0.1hz to 10hz frequency band. Noise in a 1MHz bandwidth is about 600µV pp. The main source of this noise is the buried Zener, which contributes about 100 nV/√. In contrast, the contribution of the op amp is negligible. Figure 3 shows the 0.1Hz to 10Hz noise of a typical AD588.

If further noise reduction is required, an optional capacitor, cn, can be added between the noise reduction pin and ground, as shown in Figure 2b. This will form a low pass filter with 4kΩrb on the output of the Zener unit. A 1µf capacitor has a 3 dB point at 40 Hz and reduces high frequency (to 1 MHz) noise to about 200µV pp. Figure 4 shows the 1 MHz noise of a typical AD588 with and without a 1µf capacitor.

NOTE: When using the AD588 in –10 V mode, a second capacitor is required for noise reduction (Figure 2c.). In this mode, the noise reduction capacitor is limited to a maximum of 0.1µf.

opening time

When power is applied (cold cranking), the time it takes for the output voltage to reach its final value within the specified tolerance is the turn-on settling time. The two components usually associated with this are: the time the active circuit settles and the time the thermal gradient on the chip settles. Figures 5a and 5b show the turn-on characteristics of the AD588. It shows settling in about 600 microseconds. Note that when the horizontal scale is extended to 2 ms/cm in Fig. 5b, there is not any thermal tail.

The output on-time is modified when an external noise reduction capacitor is used. When present, this capacitor provides additional load to the current source of the internal Zener diode, resulting in a slightly longer turn-on time. For a 1µf capacitor, the initial turn-on time is about 60 ms (see Figure 6).

NOTE: If noise reduction is used in a ±5 V configuration, a 39 kΩ resistor is required between pins 6 and 2 for proper startup.

temperature performance

The AD588 is designed for precision reference applications where temperature performance is critical. Extensive temperature testing ensures that the device's high performance levels remain unchanged over the operating temperature range.

Figure 7 shows the typical output voltage drift of the AD588BD and illustrates the test method. The sides of the box in Figure 7 are defined by the operating temperature limits, and the top and bottom are defined by the maximum and minimum output voltages measured over the operating temperature range. The slope of the diagonal line drawn from the lower left corner of the box determines the performance level of the device.

Each AD588A and B grade unit is tested at -25°C, 0°C, +25°C, +50°C, +70°C, and +85°C. This method ensures that the output voltage change that occurs when the temperature changes within the specified range will be contained within a box with a diagonal slope equal to the maximum specified drift. The position of the box on the vertical scale changes from device to device as the initial error and the shape of the curve change. Within the appropriate temperature range, the maximum height of the box is shown in Figure 8. Reproducing these results requires a combination of high accuracy and stable temperature control in the test system. Evaluation of the AD588 will produce a curve similar to Figure 7, but the output readings may vary depending on the test method and equipment used.

Kelvin connection

Force and sense connections, also known as Kelvin connections, provide a convenient way to eliminate the effects of voltage drops in circuit lines. As shown in Figure 9, the load current and wire resistance generate an error (verror=r×il) when loaded. The Kelvin connection in Figure 9 overcomes this problem by including the wire resistance in the forced loop of the amplifier and sensing the load voltage. The amplifier corrects for any errors in the load voltage. In the circuit shown, the output of the amplifier is actually 10V+ and the load voltage is the desired 10V.

The AD588 has three amplifiers that can be used to implement Kelvin connections. Amplifier a2 is dedicated to the ground force sensing function, while amplifiers a3 and a4 are not committed to do other force sensing work for free.

In some single output applications, one amplifier may not be used.

In this case, the unused amplifier should be connected as a unity gain follower (force + sense pins connected together) and the input should be grounded.

The unused portion of the amplifier can also be used for other circuit functions. Figures 10 to 14 show the typical performance of a3 and a4.

Dynamic performance

The output buffer amplifiers (A3 and A4) are designed to provide static and dynamic load regulation to the AD588 over a less complete reference.

Many a/d and d/a converters impose transient current loads on the reference, and poor reference response can degrade the converter's performance.

Figures 15a and 15b show the characteristics of the AD588 output amplifier driving a 0 mA to 10 mA load.

Figures 16a and 16b show the output amplifier characteristics driving a 5 mA to 10 mA load, a common situation found when a reference is shared among multiple converters or used to provide bipolar bias currents.

In some applications, the variable load can be resistive and capacitive in nature, or connected to the AD588 via long capacitive cables.

Figures 17a and 17b show the output amplifier characteristics driving a 1000 pf, 0 mA to 10 mA load.

Figures 18a and 18b show the crosstalk between the output amplifiers. The top trace shows the output of A4, DC coupled and offset by 10 volts, while the output of A3 is subjected to a load current step of 0mA to 10mA. The transient settlement at a4 is about 1 μs, and the load-induced offset is about 100 μv.

Attempting to drive large capacitive loads (over 1000 pF) may result in ringing or oscillation, as shown in the step response photo (Figure 19a). This is due to the additional pole formed by the load capacitance and amplifier output impedance, which consumes phase margin. Figure 19b shows the recommended method for driving capacitive loads of this magnitude. The 150Ω resistor isolates the capacitive load from the output stage, while the 10 kΩ resistor provides a DC feedback path and maintains output accuracy. A 1µf capacitor provides a high frequency feedback loop. The performance of this circuit is shown in Figure 19c.

Using AD588 with converter

The AD588 is an ideal reference for various A/D and D/A converters. Below are a few representative examples.

14-bit digital-to-analog converter AD7535

High-resolution CMOS D/A converters require a high-accuracy reference voltage to maintain rated accuracy. The combination of the AD588 and AD7535 leverages the initial accuracy, drift, and full Kelvin output capability of the AD588 and the resolution, monotonicity, and accuracy of the AD7535 to produce a subsystem with outstanding characteristics. See Figure 20.

16-bit digital-to-analog converter AD569

Another makes full use of the AD588 to provide a reference for the AD569, as shown in Figure 21. Amplifier a2 senses the system common and forces vct to assume that value, producing +5v and -5v at pin 6 and pin 8 respectively. Amplifiers A3 and A4 buffer these voltages to the corresponding reference force sense pins of the AD569. The full Kelvin scheme eliminates the effects of circuit traces or wires and wire connections on the AD588 and AD569 themselves that would otherwise degrade system performance.

replace internal references

Many converters include built-in references. Unfortunately, these references are the main source of drift for these converters. Drift performance can be greatly improved by using a more stable external reference, such as the AD588.

The AD574AAD574A 12-bit analog-to-digital converter uses its on-chip reference and specifies gain drift from 10 ppm/°C to 50 ppm/°C (depending on the grade). The reference value typically accounts for 75% of this drift. So the total drift can be improved by a factor of 3 to 4 using the ad588 to provide the reference.

Using this combination may result in a significant increase in full-scale error due to differences between the on-board reference of the laser trimming device and the external reference of the actual application device. The on-board reference voltage is specified as 10 V ± 100 mV, while the external reference voltage is specified as 10 V ± 1 mV. This can result in an apparent full-scale error of up to 101 mV that exceeds the AD574 gain error specified by ±25 mV. External resistors r2 and r3 allow this error to be zero. Their contribution to full-scale drift is negligible.

The high output drive capability allows the AD588 to drive up to six converters in a multi-converter system. The gain error of all converters will track to better than ±5 ppm/°C.

RTD excitation

A resistance temperature detector (rtd) is a circuit element that characterizes resistance with a positive temperature coefficient. Resistance measurements represent the measured temperature. Unfortunately, the resistance of the wires that cause RTDs often adds to measurement errors. 4-wire ohm measurement overcomes this problem. This method uses two wires to bring an excitation current to the RTD and two other wires to pump down the resulting RTD voltage. If these two extra wires are connected to a high input impedance measurement circuit, the effect of their resistance is negligible. Therefore, they transmit the true rtd voltage.

A practical consideration when using 4-wire ohm technology with a RTD is the excitation current over the temperature of the RTD. The designer must choose the smallest practical field current that still provides the required resolution. RTD manufacturers typically specify the self-heating effects for each model or type of RTD they make.

Figure 24 shows the AD588 providing accurate excitation current for a 100Ω RTD. In a resistance temperature detector, a small excitation current of 1 mA consumes only 0.1 mW of power.

In RTD current source applications, the output drive capability of amplifier A3 limits the load current to ±10mA. If more drive current is required, a series transistor can be inserted in the feedback loop to provide higher current. Pass transistors do not affect accuracy and drift performance.

Bridge driver circuit

A Wheatstone bridge is a common sensor. In its simplest form, a bridge consists of four terminal elements, two connected in a quadrilateral, the excitation source is connected along one of the diagonals, and the detector contains the other diagonal. Figure 26a shows a simple bridge driven by a unipolar excitation supply. The differential voltage eo is proportional to the deviation of the element from the initial bridge value. Unfortunately, this bridge output voltage is riding at a common mode voltage approximately equal to vin/2. Further processing of this signal must be limited to high common-mode rejection techniques such as instrumentation or isolation amplifiers.

Figure 26b shows the same bridge sensor, this time driven by a pair of bipolar power supplies. This configuration ideally eliminates common-mode voltages and relaxes constraints on any subsequent processing elements.

As shown in Figure 27, the AD588 is the best choice for the control components in a bipolar bridge driver scheme. Transistors Q1 and Q2 act as series pass elements, increasing the current drive capability to the 28mA required for a typical 350Ω bridge. If the bridge balance is not ideal, a differential gain stage may still be required. Such gain stages can be expensive.

Additional common-mode voltage reduction is achieved by using the circuit shown in Figure 28. a1 is the ground sense amplifier that serves the power on the bridge, maintaining a virtual ground at a center tap. Instead the voltage appearing on the center tap is now single-ended (referenced to ground) and can be amplified by a less expensive circuit.