Making the right choice from the wide array of data acquisition boards and software available can be a daunting task. Ideally, you want to maximize your PC-based system now, and ensure future flexibility as your needs change.
Data acquisition boards translate real-world analog and discrete signals into digital data for processing by a computer. The primary feature of the majority of these boards is the analog-to-digital converter (ADC). Many boards also include a digital-to-analog converter (DAC) for control capabilities, configurable counter/timers, and digital I/O lines for communication with digital devices (Fig. 1). By correctly choosing from the myriad of variations on these features, you can fully optimize your system and avoid paying for features you don't need. Careful review of product specifications is the key to configuring the optimum system.
One of the first things to consider is the voltage range you will be measuring. Begin by choosing the sensor (for example, strain gauge, thermocouple, microphone, or pressure transducer). Determine the range of output voltage the sensor will provide. The data acquisition board should provide a range that matches the maximum output range of the transducer. This results in the ADC utilizing the greatest number of data points in the range to be measured, thereby providing the highest possible resolution.
For example, if the sensor's output varies from 1 to 3 V, choosing a board with a 0-to 5-V input range, rather than a board with a range of ±5 V (0 to 10 V), yields twice as many valid data points and a higher resolution.
Many boards provide multiple input ranges by using software-programmable-gain amplifiers. While this is a flexible solution, a software approach may incur a performance penalty, especially if you require a different gain on every channel. For applications requiring high-speed data acquisition with multiple gains and channels, a channel-gain list is a better alternative. This on-board memory buffer is preloaded with channel numbers and associated gains. It automatically selects channel and gain values, without compromising throughput.
The number of input channels on data acquisition boards typically ranges from four to 64. Input channels can be single-ended (SE) or differential (DI), and many boards allow you to choose between the two types. Differential inputs offer noise immunity (common mode rejection), and can improve accuracy when long cables, low-level input voltages (less than 1 V full-scale), or high-resolution converters (greater than 16 bits) are used, or when input signals are at different ground potentials.
A third type of input, known as pseudo-differential, offers enhanced, common-mode rejection in designs with a single, external common ground. This type is typically used with external signal conditioning.
The accuracy of a data acquisition board is defined by how closely the binary code matches the true value of the incoming or outgoing analog signal. Yet, manufacturers' specifications rarely tell the whole story. Several different methods are used to specify this important measurement. One common specification is system error, stated as a percent of full-scale range. It typically includes all sources of error—analog noise, system nonlinearities, and reference variation.
The major component of accuracy, and its limiting factor, is resolution. Stated in bits, resolution determines the number of counts or binary numbers used to represent the analog signal. Twelve-bit converters, for example, divide ranges into 4096 parts. This translates into digital code, which tracks the analog signal to within 0.024% of the range. More bits yield exponentially higher resolution.
The limiting factor is the least significant bit (LSB). The LSB is the smallest change in the analog signal which can be represented by digital code. LSBs are specified both as both a percent of range and the smallest voltage change that can be resolved on a particular range.
Total harmonic distortion (THD) is the ratio of the sum of the harmonics of the fundamental frequency of the input signal to the fundamental frequency itself. THD is a good indicator of the quality of the circuit design. A high THD measurement, for example, may indicate a flawed analog-input design.
When evaluating a data acquisition board, it is critical to understand the relationship between resolution and accuracy. Resolution is simply one factor affecting accuracy. Many board manufacturers' specifications, however, use these terms interchangeably.
The throughput of a board—specified in megasamples per second (Msamples/s) or kilosamples per second (ksamples/s)—is a crucial measurement for high-speed applications such as audio, radar, and destructive testing. If multiple ADCs are used on a single board, the specified throughput represents the sum total of the individual converter throughputs. For example, to sample four channels at 40 ksamples/s each, you need a throughput of at least 160 ksamples/s. Delta-sigma (Δ-Σ) ADCs are an exception, since each channel has its own ADC.
Aliasing can be avoided by following the Nyquist Theorem guideline (sampling an input signal at least twice as fast as the input's highest frequency component). When the input's frequency content is unknown, many users sample at the highest frequency available, or use a low-pass filter to remove very high frequencies.
ADC throughput is determined by three elements: conversion time (the time needed to do actual conversion), acquisition time (the time needed by associated acquisition circuitry—the multiplexer, amplifier, and sample and hold—to acquire a signal accurately), and transfer time (the time needed to transfer data from the board to system memory). Normally, a board first acquires a signal, then converts it. Some high-speed boards increase throughput by overlapping the acquisition time on one sample with the analog-to-digital conversion time of the previous sample, in effect handling two signals at one time.
Most analog output circuits have a separate DAC and data buffer for each channel. The major portion of DAC throughput is settling time—the time needed to reach rated accuracy after receiving an output change. Settling time varies proportionately with the size of the output change, and is specified in microseconds.
To provide a cleaner analog output signal, some boards use a low-pass reconstruction filter. A key consideration is the capacitive drive capability when using a typical cable with 30-pF/ft. capacitance load.
Clocks, Triggers, Etc.
To perform multiple conversions at precisely defined time intervals, many data acquisition boards come equipped with one or more pacer clock circuits. Pacer clock circuits are triggered by either a software instruction or a digital pulse (or analog voltage) at the board's connector.
Many boards also contain general purpose counter/ timer circuits. These consist of several counters and a frequency source. While many manufacturers specify the number of counter/ timers available on the board, make sure these devices are not dedicated to the clocking and triggering of the analog-to-digital and digital-to-analog subsystems.
Conceived and designed as a way to give peripheral components high-bandwidth access to a host processor in a PC, the Peripheral Component Interconnect (PCI) bus meets the demanding data rates of today's computer peripherals. It does this by providing a 132 Mbytes/s (theoretical), 95 Mbytes/s (typical) burst-rate highway. In replacing the 3- to 5-Mbytes/s ISA bus, this improvement provides important benefits to data acquisition technology, enabling simultaneous operation of subsystems on the board at full specified rates.
In addition, PCI data acquisition boards can feed acquired data directly to the PC's memory, eliminating the need for on-board memory and the resulting gaps in data. Furthermore, PCI boards are auto-configured upon installation to match the system's resources. Manual configuration of jumpers or DIP switches is a thing of the past.
Matching application requirements to data acquisition board features is a fairly straightforward process. But how do you know if the specifications will match real-world performance? Will the board be reliable over time? These questions can be answered by closely examining the manufacturer's specifications.
Accuracy is the mainstay of many a board manufacturer's advertising message. But with signal and data conversion rates on the rise, it is no longer enough to measure performance by specifying a board's dc or time-domain specifications (such as relative accuracy). For an analog-input subsystem, relative accuracy is determined by the maximum deviation from the theoretical value of the full range. This is considered a dc or time-domain specification, because it is typically measured at very slow speeds, using a single analog input channel.
Although these conditions may cover some applications, most multifunction data acquisition boards monitor a variety of analog inputs at rates from 1 to 1-million readings per second. At higher speeds and multiple-channel acquisitions, the relative-accuracy specification may no longer be a good indicator of the performance of a data acquisition board.
For example, in multichannel systems, a multiplexer is often used to switch from channel to channel. Assessing the accuracy from measurements made on just one channel ignores the errors caused by the input channel's settling time.
When frequency-domain performance is characterized, you can be sure that the acquired data will be as accurate as you need it to be. The effective number of bits (ENOB) specification is the way to clearly convey a data acquisition board's ac performance. Combining all critical, real-world performance considerations such as accuracy, settling time, and dynamic performance in one easy-to-comprehend specification, ENOB specifies the overall accuracy of the analog-to-digital transfer function. With ENOB, you can evaluate how closely a digitized output sine wave matches the ideal.
Figure 2 shows an ENOB measurement for the 12-bit DT3010 high-speed PCI data acquisition board. In this case, the board achieves a nearly ideal ENOB of 11.6 bits with a 10-kHz sine wave input.
Two good indicators of quality and reliability are the FCC and CE certification standards. These certifications are more than a formality required by law; they indicate to a buyer that a board has met certain standards and will perform robustly in a real-world environment.
A final indicator of reliable performance can be found in the manufacturer's specifications: maximum input voltage. This voltage should be well above the normal voltage range of the board. If it is not, and an over-voltage condition occurs, the board may be damaged. A more subtle specification, the power-off overvoltage, specifies what happens when the input signals are present while the system power is turned off.
Planning For Future Changes
A data acquisition system should meet your needs today and provide flexibility for the future, while protecting your investment in hardware and software. Common upgrades to a data acquisition system are changing and adding boards. If the application software is not designed with an open systems approach, adding new boards can cause expensive, time-consuming reprogramming. Making sure your application programming interface (API) is hardware-independent will allow you to change boards with no, or just minor, reprogramming. One standard available is Data Translation's DT-Open Layers. All Data Translation software, for example the Dat Acq SDK (software development kit), is developed under this standard.
Also, make sure your data acquisition software will support multiple Windows operating systems. What will happen when you load Windows 98, and how will the manufacturer of the data acquisition software support that upgrade? The answers to these questions will tell you how difficult it will be to migrate from one of the early versions of Windows to either Windows 98 or Windows NT.
Another hardware issue involves the use of interrupts. Traditionally, computer peripherals request the host CPU's attention via the hardware interrupts on the CPU. However, the number of peripheral components on a typical system (modem, scanner, CD-ROM drive, etc.) has increased to the point where it frequently exceeds the fifteen interrupts available. To address this problem, Data Translation and some other board manufacturers have stopped using hardware interrupts on new boards, achieving the same function with a software feature.
Planning a data acquisition system for your demanding needs can be achieved. Keep your application in mind when choosing a data acquisition board. Look for some simple indicators of accurate, robust performance, and choose software designed for future upgradability.