How We See CE

I'm Chief Technical Officer for Digital Entertainment Products. I joined Texas Instruments in 1989 as a systems engineer wide-eyed about the opportunity to join a small start-up group, ironically in the memory products division, working on a new processor architecture for improved definition TV.We had plans to grow with the worldwide transition to HDTV as standards finalized over the next couple of years. Soon after that, it became clear that the US standard for HDTV was going to be reset so it could be introduced using digital video instead of analog technology.The project I was working on moved primarily to Japan and continued evolving for several generations, providing advanced display processing solutions.

With the ramp down of the US-based team, I was walked literally across the street, between buildings at the TI Houston facility, and joined a new media processor architecture development team as one of the first software engineers.I got a chance to influence the architecture through focused benchmarking, refining the architecture concepts with the senior architects on the team. The resulting device was introduced in 1994 as the TMS320C8x multimedia video processor.It included a RISC processor, multiple SIMD DSPs optimized for pixel processing, and a sophisticated DMA engine optimized for video and graphics, including various features for supporting flexible linking of 2D transfers and various types of graphics operations such as transparency and fill values.One of my first tasks as part of the architecture team was to implement real-time H.261 video encoding and decoding.

We eventually offered a single-chip solution for videophone that integrated real-time H.261 encoding and decoding, G.711/G.722/G.728 voice codecs, and H.242/H.230 system control stacks on a single device.This allowed me to gain experience with heterogeneous RISC/DSP processors on a single device as well as different models of distributing parallel processing across DSPs.Each DSP processor at the time had only 2K bytes of instruction cache, and I still remember painfully optimizing every single line of code to come up with partitions that could stay resident across Macroblocks in the 2K memory.

The videophone market didn’t take off as quickly as we expected and in the late 90’s, I had the chance to take the SIMD experience I had on the C8x and integrate some of those capabilities in the TMS320C64x ISA extensions to the existing C6000 VLIW architecture TI had introduced in 1997.The CPU development also included changes in megamodule design including support for multi-level cache to allow us to extend the DSP clock rate from the initial 600 MHz launch to over 1 GHz on some devices.

In 2003, we introduced the TMS320DM64x digital media processor that included the C64x ISA in a device focused on the broad IP video market.This device enabled the first embedded products implementing H.264 Main Profile decoding as well as WMV9 encoding and decoding.It seeded innovative products in areas of video security with advanced analytics, IP set-top boxes with videophone capability, place shifting, HD video conferencing, and video infrastructure that have since led to multiple product lines.

The next step was architecting an SoC video platform that TI introduced as DaVinci™ technology. DaVinci offers a platform with multiple device variants focused on different vertical markets but with openness enabling smaller emerging companies to develop innovative products using higher integration than had typically been available on video processors for the broad market.

Over the last few years, I have continued to dive deeper into vertically-focused architectures in the digital entertainment space, including SoCs for digital TV, portable media, digital still cameras and set top boxes.While many of the core processing elements are similar across these markets, the tradeoffs on quality versus power, cost and area are different.A key challenge is leveraging as much re-use in software and system architectures while still driving the needed level of customization for each market.

The number of products integrating digital video should explode over the next decade, much like we’ve already seen with digital audio.I’m actively involved in emerging technologies that will be crucial for this explosion, such as real-time video transcoding for the connected home, improved error handling in video codecs, and constantly improving image capture quality and video display processing. One of the rewards of my work is getting to see cool new products introduced that are fun to use and enjoy.