Use Careful Component Selection, Topology, and Layout to Achieve 7.5-Digit Signal-Chain Accuracy

While many resolution-readout requirements can be satisfied with four or even five digits, applications such as lab-rated digital multimeters (DMMs), field-meter calibration, weighing scales/lab balances, and seismic instrumentation require higher accuracy with a meaningful display of 7.5 (7½) digits. This performance is needed to accurately measure small, low-frequency signal changes in the presence of unavoidable large DC signals and offsets.

Achieving this level of accuracy requires a multifaceted effort with careful attention to factors related to component selection and physical implementation. Designers must understand the multiple potential error sources, the implications of short- and long-term behavior, and circuit stability.

This article briefly reviews the design issues associated with developing a meaningful and accurate readout of an analog signal that is precise to 7.5 digits. It then introduces suitable components from Analog Devices that designers can use to achieve this performance.

Component selection for high-precision readouts

A high-precision system begins with its active and passive components. While high integration can simplify design and layout to provide an assured level of performance, designers can frequently achieve higher performance using optimized, single-function ICs in a suitable arrangement and physical layout. These components’ processes, fabrication, thermal issues and gradients, and packaging and associated stresses can be optimally matched to the application requirements.

At the heart of a high-accuracy 7.5-digit system (Figure 1) are a pre-amplifier, matched gain-setting resistors, an analog-to-digital converter (ADC), and a voltage reference.

Figure 1: At the center of a 7.5-digit system’s signal chain are a pre-amplifier, gain-setting resistors, an ADC, and a voltage reference. (Image source: Analog Devices, modified by Bill Schweber)

The low-level analog signal is sent to a low-noise pre-amplifier, where matched precision resistors set the gain. There may also be an electromagnetic interference (EMI) filter. The amplified signal then passes through a high-resolution ADC, which provides digitized values using a precision voltage reference for accuracy. The converted output is transferred to the system process via one of several input/output (I/O) formats.

The pre-amplifier: Here, two critical parameters are noise and drift, which impact consistency and accuracy. One suitable pre-amplifier candidate is the ADA4523-1BCPZ-RL7 (Figure 2, left), an 8-lead, 36 V, low-noise, zero-drift operational amplifier (op amp). This op amp offers precision DC performance over a wide supply range of 4.5 V to 36 V. Offset voltage and 1/f noise are suppressed, allowing it to achieve a maximum offset voltage of ±4 μV and a typical input noise voltage of 88 nV peak-to-peak (p-p) at frequencies ranging from 0.1 Hz to 10 Hz. The device comes in an 8-lead surface-mount package, and shown (Figure 2, right) is its referred input voltage noise density from near DC to 10 MHz.

Figure 2: The ADA4523-1BCPZ-RL7 (left) is housed in an 8-lead surface-mount package; shown (right) is its input-referred voltage noise density from near DC to 10 MHz. (Image source: Analog Devices)

The self-calibrating circuitry of the chopper-stabilized ADA4523-1BCPZ-RL7 yields low offset-voltage drift with temperature (0.01 μV/°C maximum) and zero drift over time. Additionally, the ADA4523-1BCPZ-RL7 uses on-chip filtering to achieve high immunity to EMI.

Gain-setting resistors: Programmable gain is often needed to match the circuit to different input signal amplitudes and formats. For precision performance, it is more important that the gain-setting resistor pair be matched and track each other closely with temperature changes, rather than having the correct absolute value. A standalone device incorporating these pairs can generally provide better performance than resistors that are integrated into the amplifier die.

For example, the LT5401AHMSE#PBF (Figure 3, left) is an ultra-precision matched resistor network optimized for use with fully differential amplifiers, with excellent matching specifications over the entire temperature range. It contains two strings of matched resistors, each providing three tap points. Using a pair of ADA4523-1BCPZ-RL7 op amps and these paired gain-setting resistors makes the desired amplifier configuration possible (Figure 3, right). The resulting matched ratios are well-suited for precisely setting a differential amplifier’s gain or attenuation.

Figure 3: The LT5401AHMSE#PBF (left) contains three pairs of matched resistors; it is key to a high-precision programmable gain stage comprising two ADA4523-1BCPZ-RL7 amplifiers (right). (Image source: Analog Devices, modified by Bill Schweber)

Among the LT5401AHMSE#PBF’s key precision and stability attributes are:

• 0.003% resistor ratio matching (maximum)
• 96.5 dB common-mode rejection ratio (CMRR) (minimum)
• ±25 ppm gain error (maximum)
• ±0.5 ppm/°C matching temperature drift (maximum)
• 8 ppm/°C absolute resistor value temperature drift
• long-term stability: <8 ppm at 6,500 hours

The ADC: Once the signal has been amplified and conditioned, it is ready to be digitized. While there are many ADCs with different architectures and attributes, the sigma-delta approach is well-suited for precision applications because it can provide a balance between conversion time and resolution.

A suitable ADC example is the AD7177-2BRUZ-RL7 (Figure 4), a 32-bit, 10 kilosample/second (ksps) low-noise multiplexed converter with 100 microsecond (µs) settling time and rail-to-rail input buffers that ease interfacing to the pre-amplifier output. Its multiple input channels can be arranged as two fully differential channels or four single-ended channels via its crosspoint multiplexer.

Figure 4: The multichannel AD7177-2BRUZ-RL7 sigma-delta ADC features high conversion resolution and input-channel configuration flexibility. (Image source: Analog Devices)

Note that while it is a highly integrated device, this integration is not detrimental to the precision analog performance, as much of it is on the digital and I/O side. The multiple input channels are useful because many high-precision applications require a comparison between side-by-side channels or use one channel for baseline readings in real-world data acquisition scenarios.

This converter also provides 85 dB filter rejection of 50 Hz and 60 Hz interference to maintain signal integrity and does so with a settling time of 50 ms. It also includes an on-chip 2.5 V reference (±2 ppm/°C drift) and can use its internal clock for conversion timing or be supplied with an external clock. While the on-chip voltage reference is more than adequate for many applications, it is not acceptable in applications requiring higher precision. Therefore, the AD7177-2BRUZ-RL7 allows the user to supply an external reference if needed.

The voltage reference: The performance of the voltage reference is the defining element in the signal chain. An internal voltage reference for the ADC is beneficial in most cases, as it reduces component count, saves on board real estate, and ensures a defined level of converter performance.

However, an on-chip reference cannot match the performance of a dedicated standalone device, which is designed, fabricated, trimmed, and tested to do one thing and do it very well: provide a highly accurate, stable, and low-noise voltage. With few exceptions, a system’s precision, accuracy, and stability cannot surpass that of the reference. However, second- and third-order error effects, such as die and package stresses from self-heating and thermal gradients, can affect the reference performance.

For this reason, Analog Devices offers the ADR1399 precision voltage references with their design, process, and packaging optimized for this single function. To further improve their performance, their highest-precision voltage references include an on-board heater to maintain a constant temperature, as temperature variations have a significant impact on stability.

The ADR1399s are precision buried Zener shunt voltage-reference ICs with a fixed 7.05 V output, featuring excellent temperature stability over a wide range of voltage, temperature, and quiescent current conditions. A temperature-stabilizing loop is incorporated with the active Zener on a monolithic substrate, which nearly eliminates changes in voltage with temperature.

The subsurface Zener circuit is fully specified at a quiescent current of 3 milliamperes (mA) and offers ultra-low noise of 1.44 μV p-p from 0.1 to 10 Hz, and 1.84 µV_RMS from 10 Hz to 1 kHz. It also has an extremely low temperature coefficient of 0.2 ppm/℃ along with excellent long-term stability of 7 ppm/root kilohours (ppm/√kHrs).

Two versions of this device are available. The ADR1399KHZ (Figure 5, top, left) is in a simple 4-pin, hermetically sealed TO-46 package placed inside a plastic thermal insulator. The insulator helps keep ambient fluctuations to a minimum, reducing the required heater power.

In contrast, the ADR1399KEZ (Figure 5, bottom, left) is in a non-insulated, 8-pin, surface-mount leadless chip carrier (LCC). Two of the additional four pins are not internally connected, while the other two split the active reference into Kelvin-connection force-and-sense action to enable higher accuracy. The effect of package type on reference voltage versus temperature shows the negligible difference between the ADR1399KHZ in a TO-46 can (Figure 5, top, right) and the ADR1399KEZ in an LCC (Figure 5, bottom, right).

Figure 5: The effect of package type (left) on reference voltage versus temperature shows the negligible difference between the ADR1399KHZ in a TO-46 can (top, right) and the ADR1399KEZ in an LCC (bottom, right). (Image source: Analog Devices)

Circuit topology

To achieve precision, designers must also use topologies and architectures that inherently reduce or even cancel error sources. Some signals may need differential configurations to balance and cancel induced noise. Matched and tracking resistors, as noted, can improve differential performance in an amplifier, especially over temperature. Furthermore, the four-arm Wheatstone bridge is often employed to create a ratiometric measurement arrangement, where unwanted drifts in the bridge arms cancel each other out, leaving only the signal of interest.

Physical implementation

The physical construction of a precision design plays a significant role in performance. There are many factors to consider, including EMI shielding and thermocouple effects. Any connection of dissimilar metals forms a thermoelectric junction, generating a small temperature-dependent voltage (the Seebeck effect). These can be the dominant error source in low-drift circuits. Connectors, switches, relay contacts, sockets, resistors, and solders are all candidates for significant thermal electromotive force (EMF) generation.

Even junctions of copper wire from different manufacturers can generate thermal EMFs of 200 nV/°C, which is over 10 times the maximum drift specification of the ADA4523-1BCPZ-RL7. Figure 6 illustrates the potential magnitude of EMF voltages and their temperature sensitivity.

Figure 6: Shown are the thermal EMFs generated by the junction of two copper wires from different manufacturers (left) and solder-to-copper (right). (Image source: Analog Devices)

Of course, the multiple circuit grounds are a major consideration. Substantial, low-impedance, separate analog and digital grounds are essential. The current flows need to be mapped and directed away from sensitive areas, with only one connection point between the two ground regimes. The usual considerations about the use of carefully placed bypass capacitors between power rails and grounds adjacent to their loads also apply.

Calibration

It may seem that calibration of the final unit is the most direct way to resolve accuracy and stability problems, but that is usually not the case. Calibration to this level of resolution accuracy requires an extremely costly standard with a carefully arranged setup and is a time-consuming process. The unit would also need periodic recalibration at a defined cycle.

Calibration results are used in several ways to correct or compensate for the reading error. Calibration is most effective for verifying a design’s performance, rather than as a tactic to achieve desired goals.

Conclusion

Achieving meaningful precision and accuracy to 7.5 digits is a significant challenge in analog and mixed-signal design. The design solution must combine the right components, circuit topology, physical layout, and the appropriate calibration. Top-tier components, expertise, and application support from Analog Devices, combined with careful attention to design subtleties, make meeting this challenge possible.