Monday, July 04, 2005

My First Patents

Some people ask me about my inventions. So here's the story of my first two inventions, at least the first two to be patented. I'm lumping two together, because the second invention was an improvement on the first one, and because the second invention was the first to be patented, and vice versa. First, a little historic background..

The U.S. Army started using digital communication long before the commercial world, because only digital communication could be safely encrypted. A voice signal was sampled 8000 times per second, and each sample converted into 8 bits, converting the voice into a stream of 64000 bits per second. To minimize the number of radios or cables, 12 (or more) voice signals would typically be multiplexed (merged) into one signal, so that one radio or cable could carry 12 voice signals at once. The company I worked for (ITT) made radios, cable modems, and multiplexers for the Army.

A 12-channel (12 voices) multiplexer would arrange the data in 'frames', at 8000 frames per second. Since the frame rate equaled the sampling rate, each frame contained one sample from each voice channel (signal) -- 12 samples in all, 8 bits per sample, or 96 bits per frame. So a received stream of bits could be divided into 96-bit frames, the frames divided into 8-bit samples, and the samples sent to separate circuits that ultimately reached 12 different soldiers, one of which was the communications operator.

If a radio or cable modem was turned on, or had recovered from an outage, it wouldn't generally be starting at the beginning of the frame. The circuits needed a way to discover where the frame began, else those 12 soldiers might all get the wrong bits, and that would be very confusing. So they 'stole' the last bit of the frame, which was the last bit of the sample for the last channel, for a marker (called a 'synch bit') to identify the 'edge' of the frame. The 'synch bit' was zero and one on alternate frames -- an easy pattern to recognize. That left only 7 bits for each sample used by the last channel, the one used by the communications operator, degrading his voice quality, so he had to say "What was that again?" more often than the other soldiers.

A 'frame synchronization' circuit was used to find the synch bits, correcting the multiplexer's timing so that it would start at the beginning of each frame. From an arbitrary start, it would count off every 96th bit and check if it looked like a synch bit, meaning that it matched a 10101010... pattern. If it matched, it would check one frame later to verify that it wasn't an 'accidental' match; but if it didn't match, it would slip the timing by counting 97 bits (instead of 96) to the next potential synch bit.

There was a need to make the synchronization procedure faster so that communication could get started faster, and restarted faster when there was an outage. This would also make the communication less vulnerable to enemy jammers.

My first invention made the frame synchronization twice as fast, at a cost of about one more 'flipflop' in the circuit. The second invention made it even faster, using more flipflops. After many experiments, I found that the second speed-up was proportional to the square root of the number of additional flipflops. So the cost/benefits were:

1 flipflop -- 2 times faster
1+4 flipflops -- 2x2 times faster
1+9 flipflops -- 2x3 times faster
1+16 flipflops -- 2x4 times faster
1+25 flipflops -- 2x5 times faster (5 = square root of 25)

I thought the square root relationship was strange and mysterious. It illustrates the fact that inventions are generally half bright-idea and half discovery.

The first invention allowed the next bit to be examined after a mismatch -- a delay of 1 bit rather than 97 bits. The second invention anticipated the timing slips, examining the next several bits before they become the current candidate for synch bit. Later inventions dealt with the problem of noise (bit errors). These inventions helped ITT get more contracts.

Patent 3,597,539 - issued 8-3-71
Patent 3,594,502 - issued 7-20-71 - links to USPTO

Later, I was asked to sign papers when rights to use these patents were sold to various countries: Brazil, Canada, Denmark, France, Netherlands, India, Italy, Mexico, Sweden, Russia, South Africa, and Belgium. It seemed strange to sign papers in languages that I couldn't read, although there were English copies. The ones for Russia had the most paper and the most signatures. They even double-notarized some of the documents -- they didn't trust us! Years later, they must have changed the procedures, because they stopped asking me to sign such documents. I didn't really have a choice, anyway.

I guess I should esplain that when I joined ITT, I had to sign a document giving them full rights to any inventions arising from my work for them. So that's why I didn't have any choice about signing the papers. The only time ITT didn't get full rights was when the Contracts Department goofed, and one of my inventions became the possession of the U.S. Air Force.

The patent protection rights only last 17 years, so these patents have been in the public domain since 1988. And you can't get full-text copies from the US Patent Office web site, because their database only has patents issued since 1976.

No comments: