Design a High Precision Industrial Sensing System Front-End

By Bonnie Baker

Contributed By Digi-Key's North American Editors

Industrial and process control applications collect extensive precision temperature, pressure, and strain data for upstream decision making. The challenge for designers is that these applications require multiple, high precision channels that can maintain high accuracy in the frequency domain.

This article discusses the key components and parametric requirements of an accurate, high performance industrial sensing and signal conversion front-end. As noise is a determining factor with respect to accuracy, the final suitable solution resolves noise issues.

System overview

A high precision 18-bit industrial sensing front-end system should comprise a cost-effective, isolated, multiple channel data acquisition (DAQ) structure that can manage industrial signal levels. From input to output, the multi-channel, high precision circuit to be described starts with an eight-input multiplexer, configurable to single-ended or differential input channels (Figure 1). These multiplexer inputs receive various sensor inputs for process control, such as those from temperature, pressure, and optical sensors.

Diagram of eight-input, multi-channel, high precision circuit

Figure 1: An eight-input, multi-channel, high precision circuit for multiple sensor inputs starts with an input multiplexer configurable to single-ended or differential input channels. (Image source: Bonnie Baker)

In Figure 1, a programmable gain instrumentation amplifier (PGIA), denoted as “PGA,”, follows the input multiplexer with similar input and output swing voltage capability. Both the multiplexer and PGIA stages are capable of managing high voltage inputs up to ±10 volts.

The common-mode voltage and the wide voltage output swing of the PGIA is not consistent with the 18-bit analog-to-digital converter’s (ADC’s) single supply input range. To prepare the signal voltage range for the ADC, the system requires a funnel amplifier. The funnel amplifier executes three functions: a signal level shift, conversion from single-ended to differential, and attenuation to meet the single supply 18-bit ADC’s input requirements.

After the 18-bit ADC, a digital isolator provides galvanic isolation. This style of isolation allows differing common-mode voltages between each side without interference to the signal fidelity.

Circuit details

As so far described, the isolated multi-channel DAQ system has a multiplexer, PGIA stage, ADC amplifier driver, and a precision, fully differential, successive approximation register (SAR) ADC. The system monitors eight channels using a single ADC. However, the ADC drivers and the ADC are the primary noise contributors (Figure 2).

Schematic for an isolated multi-channel DAQ system with an 18-bit ADC (click to enlarge)

Figure 2: Shown is the schematic for an isolated multi-channel DAQ system with an 18-bit ADC. The ADC and ADC drivers are the primary noise contributors.  (Image source: Analog Devices)

The noise level is one specification that dictates the type of components that will fit into this application circuit.

Selecting the right components

In Figure 2, the input multiplexer is Analog Devices’ ADG5207BCPZ-RL7, a high voltage, latch-up proof, 8-channel differential multiplexer with an ultra-low capacitance of 3.5 picofarads (pF) and charge injection of 0.35 picocoulombs (pC). This low charge injection makes these switches ideal for sample-and-hold DAQ circuits which require low glitch rates and fast settling times. The ADG5207 can be configured to receive both single-ended and differential input signals. The complex programmable logic device (CPLD) shown in the circuit selects the ADG5207’s active channel by using its address pins.

The PGIA is Analog Devices’ AD8251ARMZ-R7. This device provides selectable gains of 1, 2, 4, and 8. Following that, Analog Devices’ AD8475ACPZ-R7 selectable gain, fully differential funnel amplifier provides a level shift for a ground common-mode voltage to 2.048 volts, and gain settings of 0.4 and 0.8. The AD8475 has a low output noise spectral density of 10 nanovolts per root square hertz (nV/√Hz). The PGIA and funnel amplifiers’ gains combine to provide appropriate full-scale input signals to the Analog Devices AD4003BCPZ-RL7 18-bit SAR ADC (Table 1).

AD8251 gain AD8475 gain Cumulative gain Full-scale input range Full-scale output range
1 0.4 0.4 ±10.24 V 0 V to 4.096 V
2 0.4 0.8 ±5.12 V 0 V to 4.096 V
4 0.4 1.6 ±2.56 V 0 V to 4.096 V
8 0.4 3.2 ±1.28 V 0 V to 4.096 V

Table 1: The input and output voltage range corresponding to four gain configurations for the AD8251 PGIA. The PGIA and the AD8475 funnel amplifiers’ gains combine to provide appropriate full-scale input signals to the AD4003BCPZ-RL7 18-bit SAR ADC. (Table source: Bonnie Baker)

The AD4003BCPZ-RL7 is a fully differential, 2 megasample/second (MSPS), 18-bit precision SAR ADC that has a typical signal-to-noise ratio (SNR) of 98 decibels (dB) for a 4.096 volt reference.

System noise analysis

Due to its impact on accuracy, noise must be given serious consideration when designing higher speed precision DAQs. Noise is a phenomenon in the frequency domain that impacts both the AC and DC accuracy of the ADC’s digital output. Noise is a random event: it is possible that a noisy circuit will give the absolute correct result for a single conversion, and with the next conversion create a profoundly inaccurate result. The challenge for designers is to determine the acceptable noise contributions of all the devices in the circuit.

The total system root-mean-square (rms) noise equals the root sum square of all devices in the circuit referred to the input of the AD4003 ADC and is calculated using Equation 1:

Equation 1 Equation 1

Where:

VnADG5207 = ADG5207 multiplexer rms noise contribution

VnAD8251 = AD8251 PGIA rms noise contribution

VnAD8475 = AD8475 funnel amplifier rms noise contribution

VnAD4003 = AD4003 18-bit ADC rms noise contribution

The calculated system rms SNR uses the AD4003’s full-scale input range, or VREF, and is calculated using Equation 2:

Equation 2 Equation 2

AD4003 ADC noise: The AD4003 ADC noise is a function of the converter’s quantization error and internal thermal noise. The calculation for the AD4003’s rms input voltage noise uses the full-scale input voltage (VREF) and the operating SNR, per Equation 3:

Equation 3 Equation 3

The data sheet specification for AD4003’s SNR with a VREF equal to 4.096 volts is approximately 98 dB.

AD8475 funnel amplifier noise: The AD8475 rms output noise is a combination of the amplifier’s spectral noise density (𝜖AD8475) at 1 kilohertz (kHz) and the bandwidth limit of the amplifier circuit. The AD8475 bandwidth with a gain of 0.4 V/V equals 150 megahertz (MHz). The 3 dB corner frequency of the following resistor-capacitor (RC) filter is 6.63 MHz. The combination of the AD8475 and the output RC filter creates a bandwidth limit of 6.63 MHz, per Equation 4:  

Equation 4 Equation 4

Where:

𝜖AD8475 = 10 nV/√Hz.

R = 200 ohms (Ω)

C = 120 pF

BWRC = 1 / (2xp x R x C) ~ 6.63 MHz

AD8251 PGIA noise:  The rms noise contribution of the AD8251 is a function of its referred-to-input AD8251, 1 kHz spot noise (𝜖AD8251) with units of nV/√Hz, its gain setting (GAD8251), the gain of the AD8475 (GAD8475), and the noise filter bandwidth at the input of the AD4003 (BWRC). It is calculated using Equation 5:

Equation 5 Equation 5

The value of 𝜖AD8251 equals 40 nV/√Hz for a gain of 1 V/V and 18 nV/√Hz for a gain of 8 V/V.

ADG5207 multiplexer noise: The Johnson-Nyquist noise equation provides the multiplexer’s noise spectral density and the resulting rms noise, Equation 6:

Equation 6 Equation 6

Where:

kB = Boltzmann’s constant = 1.38 x 10-23

T = temperature in Kelvin

RON = multiplexer “on” resistance (per ADG5207 datasheet)

The use of this formula (Equation 6) is appropriate because the multiplexer acts like a series resistance.

The multiplexer’s spectral density value (ϵnADG5207) yields the ADG5207 rms noise contribution using Equation 7:

Equation 7 Equation 7

Noise analysis summary

The total calculated noise contributions for each component in Figure 2 and resulting SNR for a cumulative gain of 3.2 is 84.7 dB. The most significant contributors to the total noise are the AD8251 PGIA and the AD4003 ADC (Table 2).

Cumulative gain ADG5207 AD8251 AD8475 AD4003 Total
en
(nV/√Hz)
Vn
(μVRMS)
en
(nV/√Hz)
Vn
(μVRMS)
en
(nV/√Hz)
Vn
(μVRMS)
Vn
(μVRMS)
Vn
(μVRMS)
SNR
(dB)
0.4 2.04 2.29 40 44.7 10 28 35.4 63.6 93.2
0.8 2.04 4.57 27 60.4 10 28 35.4 75.5 91.7
1.6 2.04 9.15 22 98.4 10 28 35.4 108.6 88.5
3.2 2.04 18.3 18 161 10 28 35.4 168.2 84.7

Table 2: The calculated SNR performance of the multi-channel DAQ system for a cumulative gain of 3.2 is 84.7 dB. (Data source: Analog Devices)

Circuit evaluation and test

To evaluate and test this circuit, designers can use the EVAL-CN0385-FMCZ circuit evaluation kit, which contains the Figure 2 circuit (Figure 3).

Image of Analog Devices EVAL-CN0385-FMCZ evaluation board

Figure 3: The EVAL-CN0385-FMCZ evaluation board can be used to experiment with the DAQ front-end design described in this article. (Image source: Analog Devices)

The CN-0385 Design Support Package contains the complete circuit schematic and layout support material. The evaluation kit also contains the EVAL-SDP-CH1Z controller board to facilitate data capture (Figure 4).

Diagram of test setup functional layout to evaluate the DAQ front-end

Figure 4: Test setup functional layout to evaluate the DAQ front-end. (Image source: Analog Devices)

The performance results of the EVAL-CN0385-FMCZ board show values that closely match the noise calculations (Table 3).

Cumulative gain SNR (dB) Noise (μVRMS) THD (dB)
0.4 93.9 55.2 -99.2
0.8 92.8 62.6 -98.5
1.6 90.6 80.7 -97.0
3.2 88.0 108.9 -94.6

Table 3: SNR, noise, and total harmonic distortion (THD) performance of the EVAL-CN0385-FMCZ board for a 10 kHz full-scale sine wave input for cumulative gains of 0.4, 0.8, 1.6, and 3.2. (Data source: Analog Devices)

An Audio Precision SYS-2700 generated the signal into a differential input mode. The 10 kHz input signal fast Fourier transform (FFT) plots are shown (Figures 5, 6, 7, and 8).

Graph of FFT for 10 kHz, 20 volts p-p input

Figure 5: FFT for 10 kHz, 20 volts p-p input for gain = 0.4 on single, static channel. (Image source: Analog Devices)

Graph of FFT for 10 kHz, 10 volts p-p input

Figure 6: FFT for 10 kHz, 10 volts p-p input for gain = 0.8 on single, static channel. (Image source: Analog Devices)

Graph of FFT for 10 kHz, 5 volts p-p input

Figure 7: FFT for 10 kHz, 5 volts p-p input for gain = 1.6 on single, static channel. (Image source: Analog Devices)

Graph of FFT for 10 kHz, 2.5 volts p-p input

Figure 8: FFT for 10 kHz, 2.5 volts p-p input for gain = 3.2 on single, static channel. (Image source: Analog Devices)

As shown by the plots, the performance of the ADG5207, AD8251, AD8475, and AD4003 signal chain inside the EVAL-CN0385-FMCZ evaluation board maps very close to earlier calculations.

Conclusion

In industrial and process control environments there are extensive data collecting activities, including the collection of precision temperature, pressure, and strain data. These applications require multiplexed high precision channels while maintaining high accuracy with low noise in the frequency domain. The ideal analog measurement front-end has a multiplexer, PGIA, and an 18-bit, 2.0 MSPS precision ADC. The ADC samples the signal from the active multiplexer channel. This article provides accurate calculations and complementary test data for a suitable circuit. Test results show that the actual performance of the ADG5207, AD8251, AD8475, and AD4003 signal chain inside the EVAL-CN0385-FMCZ evaluation board maps very close to calculated values.

Disclaimer: The opinions, beliefs, and viewpoints expressed by the various authors and/or forum participants on this website do not necessarily reflect the opinions, beliefs, and viewpoints of Digi-Key Electronics or official policies of Digi-Key Electronics.

About this author

Bonnie Baker

Bonnie Baker is a contributing author at Digi-Key Electronics. Burr-Brown, Microchip and Texas Instruments facilitated her involvement in analog design and analog systems for the last 30+ years. Bonnie holds a Masters of Science in Electrical Engineering from the University of Arizona (Tucson, AZ) and a bachelor’s degree in music education from Northern Arizona University (Flagstaff, AZ). In addition to her analog design fascination, Bonnie has a drive to share her knowledge and experience through the authorship of over 450 articles, design notes, and application notes.

About this publisher

Digi-Key's North American Editors