feature article
Subscribe Now

In Memoriam: James Boddie, DSP Pioneer

Starting with the development of the world’s first single-chip DSP, the Bell Labs DSP1, Jim Boddie devoted his long career to expanding the world of digital signal processing. His early work in developing signal-processing algorithms on minicomputers with attached array processors led him to cutting-edge signal-processing research at Bell Labs. That work resulted in the development of the DSP1, which AT&T successfully deployed in the 5ESS electronic switch. After the DSP1 project, Boddie held a variety of positions at Bell Labs, AT&T Microelectronics, Lucent Technologies, Agere Systems, and finally StarCore, where he directed the development of increasingly powerful DSP chips. Boddie passed away on December 2, 2024, nearly 20 years after he retired.

Boddie grew up in Tallahassee, a small town in Alabama. He attended Auburn University and graduated in 1971 with a BSEE degree. He then received an MIT fellowship to study for his master’s degree, which he earned in 1973. While at MIT, Boddie worked at the university’s first interdepartmental academic research center, the Research Laboratory of Electronics. His research focused on the psychoacoustics associated with localizing sound sources in 3D space. Boddie then returned to Auburn University in 1973 to work on his PhD.

During his PhD work, which focused on speech recognition, Boddie found that he enjoyed teaching at the university level, and he began preparing for a life in academia. Then, he got a phone call from Jim Flanagan, who was the head of the acoustics research department at Bell Labs. Flanagan offered Boddie a slot in the Bell Labs post-doc program, and he accepted the offer. After earning his PhD in 1976, Boddie started working for Bell Labs. His early work there focused on “de-reverberation,” which removed echoes from sound captured by microphones placed in non-ideal locations to make it sound more natural. Boddie developed the digital signal processing techniques needed to massage the reverberations out of the captured sound using a refrigerator-sized array processor controlled by a Data General Eclipse minicomputer, because single-chip DSPs did not yet exist.

During this post-doc work, Boddie received an offer to join Bell Labs as a full-time member of the technical staff. He accepted the offer, forgot about his plans for a teaching career, and quickly decided to make a career at Bell Labs. At Bell Labs, Boddie soon met Dan Stanzione, who managed a microprocessor applications group at Bell Labs. He offered Boddie the opportunity to work on a new project that was just getting off the ground: a single-chip DSP.

AT&T’s need for a DSP chip grew out of the company’s extensive use of passive analog filters in its telephone switching equipment. The analog filters were built with resistors, capacitors and inductors, and equipment designers were perpetually faced with the need to fit more filters on a board, to reduce the cost of the filters, and to make the filters more efficient. Switched-capacitor filtering technology grew out of this research.

During the latter part of the 1970s, telecommunications was going digital. AT&T’s 4ESS long distance switching system was already being deployed and there was an effort to move digital technology into local switching through the 5ESS switching system, which was under development. Bell Labs was already developing microprocessors during this period, but these early microprocessors did not have the throughput needed for digital signal processing. It was clear that a DSP could become a universal signal-processing element, capable of implementing a variety of filters, equalizers, tone generators, and tone decoders, if a suitable processor could be developed.

After one memorable lunch with Dan Stanzione to discuss DSP challenges, Bell Labs research fellow John Thompson developed the architecture for a pipelined processor that would be fast enough to use as a DSP. Stanzione put together a DSP study group to develop a recommendation for the development of a single-chip DSP. The project was approved in 1977, and work on the chip design started in early 1978.

The DSP development work was done mostly by hand because AT&T had few EDA tools for chip design at the time. Boddie designed the DSP’s ALU logic, including the all-important 16×20-bit pipelined multiplier and 40-bit accumulator. The DSP was developed using 4.5μm design rules. Design rule checking was done by hand and by running a wafer through the fab to see if the chip worked. When laid out, the ALU consumed 25 percent of the DSP chip’s real estate. The DSP wafer had 137 possible die and some early wafers produced zero usable devices.

 

AT&T DSP1 die photo, designed using 4.55μm design rules. The ALU designed by Jim Boddie consumed 25 percent of the die. Image credit: Computer History Museum

Eventually the design problems were solved and the Bell Labs DSP1 was the result. Sampling of the DSP1 within AT&T started in 1979. The DSP1 was never offered commercially outside of AT&T. That same year, the DSP1 development team described their work in a paper presented at the 1980 International Solid State Circuits Conference. At that same conference, Takao Nishitani’s DSP development team from NEC (now part of Renesas) delivered a paper describing the NEC μPD7720 DSP. The DSP era had begun.

Boddie continued working on DSPs and DSP technology for another quarter century. Although he started that work at Bell Laboratories, his work moved to AT&T Microelectronics, then to Lucent Technologies, and finally to Agere Systems as Bell Labs itself moved. Along the way, Boddie became a Bell Labs fellow. During this time, Boddie oversaw the birth of six DSP architectural generations, including the follow-on DSP20 – a die-shrink version of the 20-bit DSP1 designed with 2.75μm design rules, the floating-point DSP32, and the fixed-point DSP16.

In 1998, Boddie became the founding Executive Director of the StarCore Joint Design Center in Atlanta, Georgia. StarCore started in 1998 as a joint venture between Lucent (later Agere) and Motorola. Its mission was to develop DSP IP cores that the partner companies could implement in chips. Infineon joined the StarCore venture in 2002. NXP eventually acquired the StarCore DSP line when it merged with Freescale in 2015. (Freescale was spun out of Motorola’s Semiconductor Products Sector in 2004.) When the StarCore Design Center moved to Austin in 2006, Boddie retired with an immense DSP legacy.

DSPs were once a hot commodity on the IC market. They rescued Texas Instruments from oblivion during the 1980s when the TTL market began to falter. However, single-chip DSPs are no longer the hot new thing, although the need for digital signal processing continues to grow. By the late 1990s, TI, Motorola, and Philips had developed monster DSP processors with VLIW architectures, multiple multiplier/accumulators, and additional function units for special operations such as bit swizzling. However, development of bigger and more powerful standalone DSP chips came to an abrupt halt when a competing chip technology veered out of nowhere and blindsided the DSP vendors. FPGAs crashed the single-chip DSP party at the turn of the millennium when they started to incorporate hardware multiplier/accumulators (MACs).

Most DSP algorithms contain a lot of inherent parallelism that multiple MAC units can exploit. The first FPGA device family to incorporate fast hardware multipliers was the Xilinx Virtex-II FPGA family. In July, 2001, Xilinx announced that it had already shipped $1 million worth of Virtex-II XC2V6000 FPGAs, each with 144 hardened, on-chip 18×18-bit multipliers. That FPGA, the first to incorporate hardware multipliers, could already outperform any single-chip DSP that existed at the time. Today, FPGAs incorporate as many as several thousand hardware multipliers in the form of DSP blocks embedded in the FPGA’s programmable logic fabric. Although the era of single-chip DSPs may have ended, Boddie’s DSP legacy is still in full use.

References

Computer History Museum, “AT&T DSP1 Oral History Panel: Daniel Stanzione, Richard Pederson, and James Boddie,” January 16, 2015

James Boddie, “A Brief History of AT&T’s First Digital Signal Processor,” IEEE Solid State Circuits Magazine, Spring 2017, pp 14-18

Dan Stanzione, “DSP Architecture Innovation,” IEEE Solid State Circuits Magazine, Spring 2017, pp 19-20

Victor B Lawrence and Bryan Ackland, “John S Thompson,” IEEE Solid State Circuits Magazine, Spring 2017, pp 21-24

5 thoughts on “In Memoriam: James Boddie, DSP Pioneer”

  1. Thanks for this article! In 1998, I used a 64-bit MIPS processor as a DSP, assembly-coded to run a few frequency component DFT as ADC samples arrived every 10 uS. Not calculate after all ADC samples are in memory, but calculate while ADC samples are arriving. As you state, FPGAs did not yet have MACs. Existing DSPs assumed 16-bit ADC data. I was using a new 20-bit audio ADC, so needed more MAC bits.

  2. Good article, Steve.
    On a personal note, I was privileged to work with Jim Boddie at Bell Labs for 7 years, as a DSP architect and colleague. He was a kind mentor, always thoughtful and straightforward, even when we argued. He is fondly remembered by many of us.
    DSPs are no longer standalone chips, but DSP technology is pervasive. I think this fine outcome resulted in part from the rise of FPGAs, as you discussed, but also from the rise of DSP-capable CPUs as well as DSP IP.
    Early DSPs required specialized ISAs to meet performance goals. However, as CPUs gained performance, DSP ISA extensions (e.g. SSE), enabled them to take on most DSP applications. This works where the system otherwise requires a costly CPU – and the CPU can spare some cycles for its DSP tasks. Elsewhere, cost and power-sensitive applications can integrate a DSP core (with greater or lesser specialization) into its SoC.

    1. Thanks for the personal remembrance and the additional analysis, Pat Hays. I agree with your analysis. DSP is far from gone. It simply became symbiotic with FPGAs and CPUs.

Leave a Reply

featured blogs
Feb 19, 2025
Silicon Labs is expanding our portfolio as the industry-leading supplier of energy harvesting and power-optimized solutions for IoT....
Feb 19, 2025
In a world where truth can be stranger than fiction, it's a joy to see history presented with such authenticity'”whether on the page or the screen...

featured chalk talk

Power Modules and Why You Should Use Them in Your Next Power Design
In this episode of Chalk Talk, Amelia Dalton and Christine Chacko from Texas Instruments explore a variety of power module package technologies, examine the many ways that power modules can help save on total design solution cost, and the unique benefits that Texas Instruments power modules can bring to your next design.
Aug 22, 2024
43,127 views