By Gene Frantz, TI fellow and DSP applications manager
Though most signals in nature are analog, today's computers require digital data to operate efficiently. That's why digital signal processing is emerging as the key technology of 1995 and beyond.
For such applications as speech recognition and full-motion video-applications requiring inputs and outputs from nature's analog world-digital signal processing is essential. Digital signal processing filters out noise and strips away extraneous information that can result in the well-known phenomenon, garbage-in-garbage-out.
The question is, how can we best accomplish digital signal processing? Do we need dedicated devices, known as digital signal processors (DSPs), or are there other ways to do the job?
A short answer is that digital signal processing can be accomplished by a variety of devices, not just by DSPs. The best device depends on the design requirements and end-user needs of any particular system.
Is real-time processing important? Does the user expect the system to accomplish more than one thing at a time? Are the applications computation intensive? If the answers to these and similar questions are "Yes," then DSPs are the most effective, cost-efficient solutions.
Why? Because DSPs are designed specifically to prepare data for processing and to accept the results. They perform mathematically intense algorithms with amazing speed, and they do it in real time. When coupled with microprocessors in an integrated system, they facilitate multi-tasking by relieving the core processor of these functions.
The special-purpose design of the DSP is the key to its effectiveness. DSPs available today share five important characteristics:
What do these characteristics mean to system designers and to end users? For designers, they mean more performance, greater flexibility, better power management and lower cost. For end users, they mean added features and system response that look and feel like the world they know.
As an example, look at one popular type of application: Full-motion video. Designers wrestle with the problems of moving the huge amount of data required-about 40 Mbits per second for a standard television picture-through 64-Kbit-per-second buses. Consumers expect output that looks just like the video they are accustomed to from TV and movies. DSPs offer the only cost-effective solution.
DSPs can achieve the compression and decompression needed to overcome design problems. And they can do it without overburdening the microprocessor. As a result, a consumer can watch a TV-quality picture on the computer screen while also running a spreadsheet or updating a menu, as you have seen in recent ads.
Several kinds of devices,general purpose processors, for example,may manage the compression and decompression, though without multiple bus architecture, that is problematic. To do so, though, they must run flat out. Other applications, such as a spreadsheet, screech to a halt.
Just as important, a microprocessor running full bore generates tremendous amounts of heat. And when it overheats, the I/O timing fails. Obviously, this creates a condition that is not satisfactory either to designers or to consumers.
Cost is another consideration. Typically, the price of a DSP is no more than a fraction of the price of a general purpose processor. In some cases, the difference approaches an order of magnitude.
And of course, there are questions of power consumption. Today, 3-V DSPs are widely available, and the trend is toward devices that require 2 volts, 1 volt and even less. Power management features, such as sleep modes, drop power consumption to the microAmp range.
As multimedia computing, video conferencing, digital cellular telephones and similar systems proliferate, designers will need better, faster less expensive ways of accomplishing digital signal processing. For most applications, they will find their best solutions in digital signal processors.
January 1995, vol. 12, no.1
TI Home
Search
Semiconductor Home