Digital signal processing (DSP) chips certainly are not new to computing. Manufacturers have been incorporating DSP technology into a wide variety of products for several years.
Many engineers think of DSP chips only in conjunction with fast Fourrier transforms and audio/visual applications. This thinking fails to appreciate the great strides that these chips have made in recent years.
For nearly all embedded applications, a DSP chip offers the most efficient means to achieve a desired level of processing power. It is most effective because it provides the simplest hardware design, lowest power usage, and most compact program code.
Today, DSP chips are used extensively for voice-band applications such as speech compression, processing, and recognition. For example, they can be found in telephone answering machines, voice mail system pagers, dictation equipment, personal digital assistants, cellular phones, and cellular radios. Chips from suppliers such as Analog Devices, Motorola, and Texas Instruments pervade these commercial applications.
Now, the ubiquitous DSP chip is finding its way into instrumentation and industrial applications. It is being used for high-speed numeric applications, data filtering, intelligent alarming, and control including motor and motion.
A DSP chip has the same basic components as a general-purpose CPU, but it is optimized for high-speed processing of data streams, such as applications that are highly numerical and repetitive and operate in real time. Since this describes the typical data acquisition application, it’s not surprising that DSP chips work in conjunction with analog-to-digital converters and digital-to-analog converters.
Forward-looking data acquisition suppliers know that they can substantially improve the performance of their data acquisition products by equipping them with a DSP chip. But, just what does DSP really mean to you? When buying a DSP-equipped data acquisition board, what additional functionality should you expect?
An important reason for incorporating a DSP chip into a data acquisition product is to free the PC from the need to handle functions such as:
Scan and control.
Memory management.
Data management.
Communication bus management.
When the DSP chip performs these functions, your PC is free to perform other tasks more efficiently. It can take advantage of Windows NT and Windows 95 to better perform functions such as data analysis and reporting while the chip takes care of data acquisition tasks.
The DSP chip also is beneficial to the software engineer. The capability to control real-time aspects of an application, commonplace under DOS, is difficult with Windows 3.1, 95, and NT.
DSP rectifies this situation. Now, software engineers can program high-speed numeric-processing and control applications at the machine level—not in the PC but in the DSP chip. This capability becomes very important, particularly for OEMs and systems integrators who program the chips for embedded applications.
Not all suppliers implement DSP technology as extensively or as efficiently as possible for data acquisition tasks. Here’s a checklist of features a DSP-equipped data acquisition board should provide.
Throughput
DSP technology allows a board to assume more signal-processing functions by off-loading the PC and enabling faster data acquisition throughput. For example, newer DSP boards accommodate 330-kHz throughput at $31/point. This figure is based on a 64-channel board priced at $2,000 and represents a cost savings of 25% to 50% when compared to older DSP-equipped data acquisition boards.
Bus Mastering
For continuous high-speed transfers to PC memory, the ISA bus uses direct memory access (DMA) transfer operations. Accordingly, most plug-in board suppliers perform data-transfer tasks using dual-channel DMA.
While the PCI bus specification does not define any specific technique for DMA, it does allow a bus master card to control the bus and move data streams of any length directly to computer memory. For this reason, PCI bus suppliers transfer data to memory with bus mastering.
Some, but not all, data acquisition vendors incorporate bus mastering into peripheral component interconnect (PCI) boards. Without bus mastering, PCI bus board performance cannot be guaranteed.
Other vendors incorporate bus mastering by using proprietary application- specific integrated circuits (ASICs) rather than commercially available DSP chips. A word of caution here: A commercially available DSP chip provides more benefits for data acquisition than does a proprietary ASIC.
Commercially available DSP chips are more widely used and better documented, which facilitates custom programming for OEM and other embedded applications. Usually, it also is easier to understand the subtleties of these chips. If you ever had to debug a proprietary product, you can vouch for the value of working with open and commercially available platforms.
Triggering Options
Because a DSP-equipped data acquisition board does not rely on the PC to enable its triggering, it can provide more sophisticated multichannel triggering.
Simultaneous Analog-to-Digital, Digital-to-Analog, and Digital I/O
Similarly, the independence of the DSP-equipped board from the PC better enables it to handle multiple concurrent I/O operations such as simultaneous analog-to-digital converters, digital-to-analog converters, and digital input/output.
Aggregate Sampling
Aggregate sampling refers to the maximum throughput of a data acquisition board divided by the number of channels being scanned. A DSP-based PCI board with a throughout of 333 kS/s should scan 16 analog inputs at 21 kS/s each and 64 analog inputs at 5 kS/s each.
Some PCI bus boards, however, exhibit a degradation of aggregate sampling rates as the number of analog inputs rises. Make sure that the boards you select are not handicapped in this fashion.
OEM Advantages
DSP-based PCI boards provide many significant advantages to OEMs who incorporate data acquisition functionality into a larger system for resale. Such boards may be required to perform intelligent alarming, fuzzy logic, or extensive data filtering independent of the PC. OEMs will appreciate a DSP-equipped board that handles these tasks and accommodates advanced application programming.
Conclusion
For a DSP-based data acquisition board to be truly useful, it must be designed so that the DSP chip provides maximum benefits in terms of throughput, bus mastering, advanced triggering options, simultaneous operations, aggregate sampling, and programmability. No two DSP-equipped PCI boards are alike. Carefully examine all the available boards to find the one that offers the best combination of price and performance.
About the Author
Fred Glow is the marketing manager for the Keithley MetraByte product line. Previously, he was affiliated with Automation Research and Foxboro. Mr. Glow holds a B.S. degree in engineering from Lehigh University and an M.S. degree in engineering management from Northeastern University. Keithley Instruments, 28775 Aurora Rd., Cleveland, OH 44139-1891, (440) 498-2948.
Sidebar
DSP Operation: The Motorola 56300 Family
One DSP chip used in many data acquisition applications, including on Keithley’s new SmartDAQ™ PCI Bus Data Acquisition Boards, is from Motorola’s 56300 family. According to an independent testing source, this family contains several of the fastest DSP chips on the market today. Each 56300 is optimized for data acquisition applications and, with its intensive processing power, performs the work normally done by a Pentium PC.
Essentially a specialized RISC processor, a DSP chip has a narrow command set but can execute each command more efficiently. For example, a typical chip can execute an arithmetic operation and place the result in a new memory address, all with a single command.
With Motorola’s 56300 family, this command occupies only one clock cycle. As a result, when performing operations for which it is optimized, the DSP chip can execute with one command and one clock cycle what typically would require several commands and many clock cycles for a general-purpose CPU to accomplish.
The DSP chip achieves performance of 66 MIPS (million instructions per second) when operating at a 66-MHz clock speed and 100 MIPS when operating at a 100-MHz clock speed. These frequencies are derived from a lower base clock rate of 16 MHz by means of a highly flexible phase lock loop that reduces system cost, lowers emitted radiation, and allows you to configure the best power/speed optimization.
Ultimately, what really matters is not the hardware features but the software structure of the DSP. A DSP is a pipeline optimized for the most rapid execution of commands, and the command architecture will determine the efficiency of the DSP chip, both in terms of operation and development.
Copyright 1997 Nelson Publishing Inc.
November 1997
|
|