Editor's Notes—Volume 39, Number 1, 2005

40 YEARS OF REAL WORLD SIGNAL PROCESSING
Forty years ago, Ray Stata and Matt Lorber opened the doors of Analog Devices for business, offering a line of high-performance operational amplifiers. We’ve survived and prospered beyond their fondest expectations, and are still rarin’ to go. In celebration of that anniversary, Analog Dialogue’s four print installments will each be devoted to one of our major technologies. We start with digital signal processing (DSP).

ANALOG SIGNAL PROCESSING GOES DIGITAL
In 1986, a new—and apparently unlikely—contender entered the young field of digital-signal-processor manufacturing—then dominated by TI, the colossus of “Speak & Spell,”—with a single-chip DSP, our ADSP-2100. As we celebrate our 40th year in the business of components for signal processing, it seems worthwhile to reproduce here our editorial comments that accompanied the introduction of the first Analog Devices DSP in these pages (Analog Dialogue 20-2, 1986):

“Microprocessor?” we hear you ask. “Isn’t it a bit unseemly for a nice ‘Analog’
IC company to be designing a microprocessor? (What could be more
digital?)”

Good question.

Our objective has always been to design and manufacture cost-effective components that are key elements of the signal path for processing real-world (i.e., analog) data and for which performance is maximized and errors minimized.

The signal path? Real-world data almost always starts out as analog (i.e., parallel, non-numeric) variables, which are measured by sensors that provide analog electrical signals—voltage and current. The signals must be accurately and speedily amplified, conditioned (almost always in parallel) and converted to digital for processing. Once in digital form, they must be processed rapidly. Often they again wind up as analog signals.

Key elements of the signal path may include preamplifiers, analog signal processors, data converters—to and from digital—and, when the signal is in digital form, a digital processor. Inadequacy in any one of the key elements—amplifier, analog processor, data converter, or microprocessor—can cause poor performance of the overall system.

Obstacles in the signal path include noise, drift, nonlinearity, and measurement lag at the analog stages, similar obstacles in conversion—and throughput delays in digital processing, often because of the lack of parallelism in von Neumann architectures.

Throughout our history, ADI’s role in the signal path has been to initiate new products (or product lines) when dissatisfied with the performance and cost-effectiveness of what’s available (which is often limited to user-assembled kludges, when nothing else is available). At this point in time, we (and our worthy competitors) have virtually eliminated the user-assembled amplifier, signal conditioner, and data converter, by designing and marketing families of high-performance, cost-effective products.

We have always been dissatisfied with the cost, power dissipation, and slow throughput in the digital domain; this concern led to our pioneering development of CMOS multipliers and other digital signal-processing ICs (note that because we were
already familiar with analog multipliers, digital multipliers became just another analog signal-processing tool). Note also our commitment to signal processing—not payroll, desktop publishing, or order-handling products).

And our dissatisfaction with insufficient throughput in DSP processors led to the
design of the ADSP-2100, which stresses the use of that analog characteristic,
parallelism, to minimize instruction cycles, whether in processing, data transfer, or interrupt handling. It’s neat! We invite you to read about it.

Since that time, such names as SHARC®, TigerSHARC®, Blackfin®, EZ Kit, and VisualDSP++® have become household words, as they remove barriers whenever DSPs are considered.

Dan Sheingold [dan.sheingold@analog.com]

FROM NUMBER CRUNCHING TO MULTIMEDIA
In the early days of digital signal processing, the ADSP-2100 single-chip microprocessor was typically used for applications that required high-speed numeric processing. Integrating a 16-bit arithmetic-logic unit (ALU), 16-bit multiplier-accumulator (MAC), 16-bit shifter, two data-address generators, and a program sequencer, it used external memory for program and data storage. Operating at 8 MHz, it dissipated 600 mW. In a single clock cycle it could: generate the next program address; fetch the next instruction; perform one or two data moves; update one or two data address pointers; and perform a computational operation.

Over the intervening twenty years, digital signal processors have gotten smaller, faster, less expensive, more powerful, and more efficient—and they integrate up to 24 Mbits of on-chip memory. Even more important, perhaps, are the host of peripherals that can be found on modern embedded processors. The ADSP-BF537 Blackfin processor, for example, includes an IEEE 802.3-compliant 10/100 Ethernet medium access controller, Controller Area Network (CAN) 2.0B interface, parallel peripheral interface (PPI) supporting ITU-R 656 video data formats, and dual-channel, full-duplex synchronous ports (SPORT) supporting eight stereo I2S channels. The ADSP-21367 SHARC processor’s digital audio interface (DAI) includes an S/PDIF digital audio receiver/transmitter, 8-channel sample-rate converter, sixteen pulse-width modulators, four PLL clock generators, eight serial ports, and ROM-based audio decoder and post-processor algorithms. The ADSP-TS201 TigerSHARC processor includes an 8-Gbps 64-bit external port, 14-channel direct memory-access (DMA) controller, and four 8-Gbps bidirectional link ports. Together they provide unparalleled interface capabilities without the use of any additional external glue logic.

Processing power and peripherals have created opportunities for digital signal processors in diverse applications—including professional audio mixing consoles, always-on cell-phone coverage, home-theater surround sound, fingerprint recognition, network music players, wireless video, satellite radio, and 3D motion tracking. Some of these are described below. Details about these applications and many more can be found at http://www.analog.com/processors/news/customerstories.

The TigerSHARC processor is the only processor capable of implementing a software-defined digital baseband for 3G base stations, allowing the same platform to be easily adapted for use in multiple regions—and to be easily upgraded to support new capabilities. The TigerSHARC processor is also the first to implement an all-software physical layer for IEEE 802.16 WiMAX broadband wireless modems. Its best-in-class I/O bandwidth and scalable architecture allow OEMs to differentiate their products through advanced techniques, such as smart antennas using space-time coding and adaptive beam-forming.

The 32-bit floating-point SHARC processor has the necessary speed and efficiency to handle the complex post-processing algorithms required to deliver 6.1 discrete channels of surround sound from any audio material, allowing listeners to take full advantage of their home-theater speaker systems, even when listening to VHS tapes, FM radio broadcasts, or stereo music CDs. The SHARC processor’s digital audio interface, large memory array, and VisualDSP++ graphical system design and development environment combine to allow manufacturers to base multiple products with various I/O requirements on a single hardware design, fully leveraging their design time and development costs.

Blackfin processors provide both control functions and multimedia processing capabilities, enabling diversity receivers to operate in harsh weather and low light conditions. Providing fast information transfer, these receivers allow soldiers, police officers, and firefighters in the field to exchange audio, video, and data from sources such as cameras, microphones, and global-positioning systems (GPS)—increasing personnel safety in environments that are subject to high levels of interference. The low power consumption and dynamic power management inherent in Blackfin processors is crucial for their successful use in compact, portable, battery-powered equipment.

Scott Wayne [scott.wayne@analog.com]