“Necessity, Computational Models and Competition,” Intertrust’s CTO Describes How Tech Innovation Really Happens


“Throwing transistors at something always helps, but there is another view.” Reflecting on the 50th anniversary of Moore’s Law, Intertrust’s Executive Vice President and CTO David Maher recently gave a very personal talk on another perspective on what drives innovation in the technology industry.

For a bit of background, Mr. Maher came to Intertrust via Bell Labs, where he was the chief scientist for AT&T’s Secure Communications System department and also a Bell Fellow. Based on his experience at AT&T, Maher talked about when in World War II President Franklin Roosevelt and Prime Minister Winston Churchill wanted to securely talk to each other at any time. This directly led to today’s mobile market, where most anyone can cheaply and conveniently telephone anyone else around the world.

Of course, in World War II, transistor-based integrated circuits weren’t available. For secure calls, “the Allies used a scrambler which had a pretty bad voice quality,” Maher says. Not what you wanted the leaders of the free world to use, knowing that nuances in conversations are so important. In this most decidedly analog age, the realization that a digital voice system was needed for a secure, yet high-quality voice communication system led to today’s digital voice communications systems.

To understand how this happened, it first helps to know that digital voice codecs used in modern telephone systems are not your actual voice. “When you speak on a mobile phone….. it’s making a real time model of your vocal tract and how you produce sounds,” Maher says. The voice codec does this every 20 milliseconds and the voice model is sent to the other phone, which reproduces the same model.

Back in World War II, the only digital voice codec available was at Bell Labs. It used pulse code modulation (PCM) but had never been deployed. Bell Labs also had access to vocoder technology (for recreating voice) that was demonstrated in the 1939 World’s Fair; however, it also hadn’t been deployed. Based on these technologies, after the order was placed in 1942, “15 months later a bunch of engineers came up with 40 racks of equipment that was able to digitize voice and transmit it 10,000 miles,” Maher says. All this was done without using transistors or integrated circuits. In this case, it was necessity and computational models that drove this technological innovation, not the availability of more powerful processors.

“Pin Drop” Competition

The AT&T STU-III secure telephone that Maher and the Bell Labs team developed.  Photo:  Mark Pellegrini (CC BY-SA 2.5)

The AT&T STU-III secure telephone that Maher and the Bell Labs team developed. 

Photo:  Mark Pellegrini (CC BY-SA 2.5)

Fast forward a bit over 40 years to the 1980s. AT&T’s competitor Sprint was gaining market traction with their “pin drop” commercials, promising clear voice service over the telephone. It was marketing fluff without any real technical backing, but AT&T had to respond. They looked to resurrect the World War II digital voice technology for their commercial telephony operations.

This activity also helped Maher, then working with a team at Bell Labs, fulfill the request of another President, this time President Reagan. A request had come out of the White House for a secure high-quality voice system that “needed to interoperate with everything,” Maher says. This was required so the White House could securely talk with anyone throughout the world. AT&T had a new vocoder technology called code-excited linear prediction (CELP), but “the problem was it was 500 times too slow even when ran on a super computer.”

To solve this, Maher said “the moral was not so much transistors but to look at the model and make sure the computation was the right one.” Maher and his team turned to digital signal processers (DSPs) that were more suited to the digital filtering need to make CELP work. “DSPs were pretty well developed. AT&T was developing DSPs for the network and they were pretty cheap.” By a quirk of AT&T’s accounting, however, the team ended up using DSPs from Texas Instruments.

There was another issue as well. To fulfill the secrecy rules of the National Security Agency (NSA), the Bell Labs team had to use long encryption keys. “A neat thing at the time was we could call up anyone who was an expert and find the latest tech,” At the time, Sandia Labs was operated by Bell Labs and they had a systolic array model already coded up which fit the bill. With this technology, Maher and the Bell Labs team were able to come up with a highly secure good voice quality digital voice communications terminal running on five DSPs. This was also the beginning of CELP adoption. “Today, everyone uses CELP. It’s the official voice codec for MP4,” Maher says.

Tackling the Internet Privacy Conundrum

Maher’s experience demonstrated how factors other than Moore’s Law, namely necessity, computational models and competition, drove technological innovation. Now, in the 21st century, he is focused on another challenge. “The Bell Labs of the 21st century is the Internet and Silicon Valley.” In the 21st century, we are now faced with a not-so-distant future where trillions of hyper-connected sensors, down to bots traveling through people’s bodies, will be transmitting all sorts of information over high-speed wireless networks,.

Much of the commercial activity on the Internet so far has really been about manipulating individuals to react to commercial messages. With this, Maher has been considering a thought experiment about how space aliens might perceive the Earth in a few years. With billions of people enmeshed in a network of trillions of devices, “will they see a passive network of human nodes reacting to stimuli controlled by special interests?” From a straight forward requirement of getting two people to communicate securely in the 1940s to the hyper-connected world of today’s Internet, the goal remains the same, making sure humans can trust the privacy and security of the networks that connect them.

Comment