IPTV thoughts from CES: for SoC design cost may not be the key
Conceptually television over Internet protocol, aka IPTV, is a simple problem. All you have to do is packetize the compressed MPEG4/H.264 bit stream, send it over the Internet, reassemble it on the other end and feed it into an otherwise-conventional set-top box or digital TV receiver. What could possibly go wrong?
A lot can go wrong, according to a Consumer Electronics Show conversation with Rainer Hoffmann, president and CEO of Micronas, a semiconductor company that is increasingly turning its attention to the IPTV problem. While Micronas’s German operations focus on picture-level processing and control for flat-panel displays, the US operation has been assigned the IP problem.
Hoffmann says that today, the personal computer is the source of most IPTV signals, and this has made the design problem so complex—with the interactions of CoDec hardware, software and the underlying PC architecture–that it may be slowing the emergence of the whole medium. “It’s almost impossible for OEM’s to evaluate a design now,” Hoffmann said. “I know of one OEM with a huge R/D lab that did in-depth quantitative evaluations on a number of ATSC demodulators, and they were unable to reach a conclusion. They ended up making a decision by waving a hand-held antenna around in the air while watching the screen.”
Nor have set-top boxes been an easier nut for design teams to crack. “I have Comcast at home,” Hoffmann related. “My new digital set-top box crashed, so I took it in for service. There was a huge line of people, all with their new set-top boxes under their arms. Now there’s something wrong with this situation,” Hoffmann continued, “when you have a top set-top box manufacturer, using highly-rated silicon, and the result is dead on arrival in the field.”
The addition of IP won’t make the problem any easier. “It’s very difficult to make IPTV look good on a high-definition display,” Hoffmann said. But the problems don’t start with internet protocol—they are embedded in the bit stream already. “Actually, broadcast digital TV is a bigger problem right now,” Hoffmann warned. “There are completely legal bit streams that result in broken macroblocks. It’s going to take a combination of demodulator algorithms, error recovery procedures and picture-processing algorithms to make this look good.”
And that won’t be easy in the current market environment. “Today on a product development BestBuy gives you a ship date, and everything else is almost irrelevant,” Hoffmann lamented. “This way of doing development leads to disaster.” Even working with an existing platform, he said, it is a challenge to integrate the right functional blocks and to verify the resulting system. Part of the challenge is just interoperability—there are so many assumptions in interpretations of the specifications in digital TV that just meeting the specs doesn’t imply interoperability. You have to test against other systems.
In this uncertain environment, Hoffmann emphasizes building observability and debug aids into the silicon. That seems counter-intuitive in a consumer market that is Draconian about cost, but Hoffmann insists that there is a growing demand from system integrators to spend silicon on debug capability. “These OEMs have been through hell with previous designs,” he said. “They appreciate the effort to help with system debug.”















