Design tradeoffs in data acquisition
The analog signal path is full of tradeoffs and decisions that affect overall performance.
By Martin Rowe, Senior Technical Editor -- Test & Measurement World, 6/1/2008
The analog acquisition section is the heart of any data-acquisition system. Microprocessors, digital-signal processors, memory, firmware, software drivers, operating systems, and software applications may form the brains of a system, but they’re only as good as the analog circuits. To build a system that has the necessary speed, resolution, and accuracy for a given application, you need to find the right combination of analog data converters, op amps, multiplexers, and voltage references.
Figure 1 shows the basic analog signal path through the analog-to-digital converter (ADC). Every data-acquisition system uses some form of this basic configuration. The choice you make for each component will affect the options available for the other components.
 Figure 1. Analog channels consist of components such as a programmable-gain amplifier (PGA), an offset amplifier, and a low-pass filter (LPF). Courtesy of National Instruments. |
The analog signal path begins at the input connectors. Most data-acquisition systems have some form of circuit protection before the analog circuits. Components such as fuses or clamping diodes limit the voltage or current that enters a system, thus protecting its components from harm.
Data-acquisition systems seldom have a single measurement channel. Digital multimeters (DMMs) typically have one channel, but you can combine a DMM with relays to increase channel count. Data-acquisition systems, whether plug-in boards, USB modules, or stand-alone systems, may have a dedicated ADC per channel, or they may have a single ADC with a multiplexer (mux) connected to multiple channels. The use of a dedicated ADC for each channel lets the system simultaneously sample on all channels.
| Analog IC manufacturers offer extensive technical information on designing with ADC. You can find application notes, data sheets, online seminars, technical papers, and simulation software available at no cost. See resource information. |
Following the mux (if the system has one), a programmable-gain amplifier (PGA) amplifies or attenuates the input voltage from a sensor or other signal source to best match the ADC’s input voltage range. Some systems may include a second op amp that adds a DC offset voltage to the input signal. The offset voltage shifts the signal so it’s centered in the input range of the ADC. The ADC’s input voltage range, therefore, has a major influence on your choice of PGA.
You may also use additional clamp circuits to protect the ADC. Just prior to the ADC, most system designers add a low-pass anti-aliasing filter. This filter limits the signal path’s bandwidth and is the last chance to minimize aliasing before the ADC digitizes the signal.
To successfully digitize analog signals, ADCs need a reference voltage, V ref. Some ADCs have an internal reference while others use an external reference source.
“We prefer an external voltage reference,” said Kevin Cawley, senior principal engineer at Keithley Instruments. “We believe that external voltage references are more stable than internal ones.”
Alex Ivchenko, engineering manager at United Electronic Industries (UEI), went a step further. “If you have an external reference, you can adjust the gain of the input path by controlling the ADC reference voltage,” he said. “If your input voltage is too high, you need to provide a higher V ref.”
An ADC’s digital output can be in either serial or parallel form. The serial bus offers improved analog performance because fewer lines will change state at a given time, which minimizes bouncing on the power and ground lines and reduces overall system noise. But because serial interfaces run at higher clock speeds than parallel buses for the same number of bits, you must carefully route signals to keep the noise down.
Choosing an ADCThe choice of ADC leads to numerous design tradeoffs that you must consider. Most ADCs used in data-acquisition systems use successive-approximation register (SAR) or sigma-delta (ΣΔ) architectures. In general, SAR devices yield higher speeds than ΣΔ ADCs, but ΣΔ architectures produce finer resolution. If you need better than 18-bit resolution, you’ll need a ΣΔ converter.
The sample rate and power-supply voltages of the ADC will determine the type of support circuits you can use. Consider, for example, the supply voltage. Most of today’s ADCs are made from a CMOS process rather than from a bipolar process. CMOS devices operate with considerably less power than bipolar devices. They can also operate with lower power-supply voltage rails. Where bipolar devices may need 12-V or 15-V rails, CMOS devices run from unipolar power supplies of 5 V, 4 V, 3.3 V, 2.5 V, and even 1.8 V.
Although the low voltages reduce power consumption, they also compress an ADC’s dynamic range. An ADC that operates at 12 V has six times the dynamic range of a 0–4 V device. Thus, an equal amount of noise will impact the 12-V system far less than on a 4-V system. You must, therefore, keep the noise entering an ADC to less than 1 least-significant bit (LSB). You need an op amp with a noise level consistent with the dynamic range of an LSB in front of the ADC. That means you’ll need lower noise for a 24-bit ADC than you will for a 16-bit ADC.
To get the best dynamic range, you should push the high-level signals as far into the analog channel as possible, according to Cawley. He noted that Keithley’s DMMs provide the best accuracy at the 10-V range where they need neither amplification nor attenuation of the incoming signal.
Designers push backBecause of the better dynamic range that higher-voltage rails offer, many designers of industrial data-acquisition systems demand such rails for their op amps and data converters. As a result, ADC manufacturers have developed CMOS data converters that operate at 16-V power rails. These devices can handle sensor inputs up to 15 V, noted Chris Hyde, senior field applications engineer at Analog Devices.
Another compensation for the low dynamic range is to digitize your sensor signal as early as possible. “High-speed ADCs have come down in price to the point where oversampling makes sense,” said UEI's Ivchenko.
With oversampling, you can use digital filtering to reduce noise. The more you oversample and filter, the better the noise immunity, but the slower the system. Ivchenko pointed out that over-sampling by 22n and using a digital averaging filter will improve noise performance. The table lists how much oversampling is needed to improve noise performance by a given number of bits.
| The amount of oversampling needed to improve noise performance by a given number of bits |
|
| Noise improvement (bits, n) |
Amount of oversampling |
|
1 |
4 |
|
2 |
16 |
|
3 |
64 |
|
4 |
256 |
Following the ADC, Ivchenko applies a “brick wall” (120 dB/octave) digital finite impulse response (FIR) filter to reduce noise and extract the spectrum of interest. Then, he decimates a portion of the data or he applies a moving average to make the sample rate acceptable for the application.
Low-voltage ADCs and op amps need sufficient current to supply and keep signals stable during data conversion. “Designers often pick op amps and voltage references that don’t have enough drive,” said Hyde. “A voltage reference may need to both source and sink current.” An ADC may have a dynamic input impedance and may need a low-impedance signal source with sufficient coupling to maintain the reference voltage level.
“SAR converters need a very low output impedance source to keep the input signal from varying during conversion,” said Luis Orozco, analog design engineer at National Instruments. “Because SAR ADCs usually present a highly dynamic load to their power supply, we carefully bypass all devices.” He noted that matching the correct op amp to an ADC is critical.
“An op amp with the required performance to achieve the advertised ADC specs may consume several times the current that an ADC uses,” Orozco said. The reference input on ADCs behaves similarly to the signal input. Low-power devices such as voltage references may need capacitors or buffers to maintain their output at a stable level as the ADC samples its reference.
“Not only that,” added Ivchenko, “but you should use low equivalent-series resistance [ESR] bypass capacitors. Use X7R ceramics rather than tantalum capacitors whenever possible. A capacitor must charge or discharge quickly enough to feed sufficient peak current to an ADC during a conversion cycle.” A high ESR will increase a capacitor’s charge and discharge time.
Figure 2. Voltage references often need (a) a bypass capacitor or (b) a capacitor with a buffer amplifier.
Figure 2 shows both ways to provide ample current. In Figure 2a, a capacitor stores energy and supplies it when the ADC needs additional current to keep the reference voltage stable. A 22-µF capacitor is usually sufficient, but check your ADC’s data sheet to be sure. In Figure 2b, an op amp buffers the voltage reference from the ADC. The op amp provides the voltage reference with a high-impedance input while its low-impedance output provides enough current for the ADC. While the op-amp solution is more elegant, it adds an offset voltage to V ref, it adds noise to the system, it increases power consumption, and it is more expensive.
To improve dynamic range and noise rejection, you should use differential inputs in your data-acquisition system. With differential (as opposed to single-ended) inputs, any signals common to both signal lines will be negated to the best of the ability of the common-mode rejection (CMR) amplifier or the ADC. If your sensor’s output is single ended, you can use a single-ended-to-differential converter driver circuit (Figure 3). You can design your data-acquisition system to use single-ended or differential inputs.
 Figure 3. A single-ended-to-differential converter circuit lets you digitize differential signals. Courtesy of Analog Devices. |
Many data-acquisition systems have a mux that adds channels. Resistance and capacitance in a mux can affect signal integrity. For example, charge injection from a mux can turn DC signals into AC signals. The on-resistance (R on) combines with parasitic capacitance to form a low-pass filter, which has an RC time constant. Figure 4 shows what happens if the time constant is too long relative to sampling time.
 Figure 4. Charge injection, on-resistance, and parasitic capacitance can cause leakage across adjacent channels in multiplexed systems. |
If the time constant is fast compared to the sample rate, then you should see a square wave at one-half the sample rate. But if the time constant is too long, you’ll get what resembles a triangle wave because of charge injection between the channels.
“R on should be no more than a few ohms,” said Hyde of Analog Devices. “On-resistances of a few hundred ohms are too much for many of today’s data-acquisition applications.” National Instruments’ Orozco contends that a few hundred ohms isn’t too much because of the high input resistance of the upstream op amp.
Hyde also pointed out that a mux’s on-resistance can change based on the amplitude of the system’s input signal. If you change channels from one voltage rail to the other, you need to know the RC time constant of the channel. While R on changes with voltage, channel capacitance causes an impedance change with frequency. These impedances work against the capacitance to form a variable low-pass filter and cause distortion.
“The channel must settle within the accuracy limits of the ADC to prevent charge-induced errors,” said Hyde, and he added that newer muxes have lower capacitance than older models.
Technical data
 Figure 5. Reference design boards provide support circuits and communications that let you test an ADC. Courtesy of Analog Devices. |
Data sheets also provide design and layout information, but as Keithley’s Cawley found out, information on data sheets and reference design boards may differ. When designing a 500-ksamples/s, 18-bit data-acquisition system, Cawley relied on the design information in a data sheet, only to find that the ADC produced between 3 and 7 LSBs of noise (5 µV/LSB). “When I switched to the layout recommended in the reference design, the noise dropped to within 1 LSB” he said. “The reference design used four layers of ground under the quad-flat-pack (QFP) device. Nine vias connected the ground planes, but the data sheet used a trace from the ADC to a bypass cap instead of using a ground plane.”
Analog IC manufacturers offer extensive technical information on designing with ADC. You can find application notes, data sheets, online seminars, technical papers, and simulation software available at no cost.
RESOURCES
"ABCs of ADCs" by Nicholas Grey, National Semiconductor, www.national.com/appinfo/adc/files/ABCs_of_ADCs.pdf
"The ABCs of ADCs: Understanding How ADC Errors Affect System Performance," Maxim Integrated Products, www.maxim-ic.com/appnotes.cfm/an_pk/748
"Accurately measuring ADC driving-circuit settling time," by Rajiv Mantri and Bhaskar Goswami, Texas Instruments, www.ti.com/litv/pdf/slyt262
A/D Converters - Application Notes, Analog Devices
www.analog.com/en/cList/0,2880,760%255F788%255F43,00.html
ADCPro Analog-to-Digital Converter Evaluation Software, Texas Instruments, focus.ti.com/docs/toolsw/folders/print/adcpro.html
"ADC Quantization: Error and Noise" online course by Nicholas Grey, National Semiconductor, www.national.com/AU/design/0,4706,259_0_,00.html
"ADC Requirements for Temperature Measurement Systems," by Mary McCarthy and Eamonn Dillon, Application Note AN-880, Analog Devices,
www.analog.com/UploadedFiles/Application_Notes/42082002AN_880.pdf
ADC Simulation, Software, and Behavioral Models, Analog Devices
www.analog.com/en/content/0,2886,760%255F788%255F68056%255F15,00.html
"A Glossary of Analog-to-Digital Specifications and Performance Characteristics" by Bonnie Baker, http://focus.ti.com/lit/an/sbaa147a/sbaa147a.pdf
"ADS121x Analog-Digital Converter Applications Primer," by William P. Kline, Texas Instruments, www.ti.com/litv/pdf/sbaa022
"Amplifiers and Bits: An Introduction to Selecting Amplifiers for Data Converters (Rev. B)," by Bruce Carter, Patrick Rowland, Jim Karki, and Perry Miller, Texas Instruments, www.ti.com/litv/pdf/sloa035b.
"Analog ICs for Low Voltage Systems," Application Note 3741, Maxim Integrated Products,
www.maxim-ic.com/appnotes.cfm/an_pk/3741
"Demystifying Sigma-Delta ADCs," Application Note 1870, Maxim Integrated Products,
www.maxim-ic.com/appnotes.cfm/an_pk/1870
"The design and performance of a precision voltage reference circuit for 14-bit and 16-bit A-to-D and D-to-A converters," by Perry Miller, Application Specialist—Data Converters, Texas Instruments, and Doug Moore, Thaler Corp., Tucson, Arizona, www.ti.com/litv/pdf/slyt168
"Interfacing op amps and analog-to digital converters," by Bruce Carter, Texas Instruments, www.ti.com/litv/pdf/slyt104
"Matching the noise performance of the operational amplifier to the ADC," by Bonnie Baker, Texas Instruments, www.ti.com/litv/pdf/slyt237
"Measuring Single-Ended 0V to 5V Signals with Differential Delta-Sigma ADCs," by Michael Ashton, Texas Instruments, www.ti.com/litv/pdf/sbaa133
"RLC Filter Design for ADC Interface Applications (Rev. A)," by Michael Steffes, Texas Instruments, www.ti.com/litv/pdf/sbaa108a
"Shielding and Guarding: How to Exclude Interference-Type Noise" by Alan Rich, Application note AN-347, Analog Devices www.analog.com/UploadedFiles/Application_Notes/41727248AN_347.pdf
"Using the ADS7800 12 BIT ADC with Unipolar Input Signals," by R. Mark Stitt and Dave Thomas, Texas instruments, www.ti.com/litv/pdf/sbaa044
"Understanding High Speed ADC Testing and Evaluation" by Brad Brannon and Rob Reeder, Application Note AN-835, Analog Devices,
www.analog.com/UploadedFiles/Application_Notes/
4115557206466474685142207552745732150239440755569051663372515871138132239AN_835_0.pdf
"Understanding the Effects of Clock Tolerances on 50/60Hz Noise Rejection in High Performance Sigma Delta ADCs" Application note 3418, Maxim Integrated Products,
www.maxim-ic.com/appnotes.cfm/an_pk/3418.
"Understanding SAR ADCs," Application note 1080, Maxim Integrated Products,
www.maxim-ic.com/appnotes.cfm/an_pk/1080
Talkback
Related Content
Related Content
