A Brief History of DECtalk (*)
Around September of 1981 I attended a meeting at Digital Equipment Corporation in which Dr. Dennis Klatt of MIT gave a short talk and played a tape on a small portable tape recorder of some synthesis work he was doing. He was looking for someone interested in doing something with this technology. I was in the Terminals Engineering Advanced Development Group at the time and after discussions with my boss, Walt Tetschner, decided to pursue this technology. In November of that year Dennis advised me that another company was interested in a non-exclusive license and asked if there was sufficient interest within Digital to put them off for a while. I told him yes, there was enough interest.
In January of 1982 an agreement was signed and the project was started with the addition of Martin Minow to the newly formed group. The first task was to get the code running on a PDP-11 connected through a parallel I/O port to a hand-built board containing a signal processing chip. We didn't have a name for the project at that time. Dennis called his system KLATTALK. Martin came up with the name of CLATTER at the time. What did not exist with the software Dennis licensed to us was letter-to-sound rules. His system at MIT used rules developed by Sharon Hunnicutt. We did the obvious and bought nonexclusive rights to her letter-to sound rules in January of 1982. Tony Vitale joined us initially as a contractor to write better letter-to-sound rules for DECtalk.
With a bit of luck I ran into one of the best engineers I have ever had the pleasure to work with, Dave Conroy. Dave became the project engineer for DECtalk. Unfortunately, I can't remember all the people's names and when they joined the group. Rich Ellison was a young and very talented engineer who joined the group. Over time the team grew as we prepared to launch the first DECtalk.
At the end of the project we did Design Maturity Testing. We discovered that we had a failure occurring once every ten thousand hours on average. The frequency of errors was so low that we had to work deductively to find the problem. We worked 100-plus hours a week for several months before isolating the problem. It turned out to be two different chip design problems suffering from the "Great synchronizer problem." The manufacturer of one of the failing chips fixed the problem by changing the specification for the chip. It said that the interrupt pin was asynchronous. Erase the A from asynchronous and you are good to go.
The first DECtalk shipped! Very shortly thereafter a customer was able to break the system. We had missed the worst case nesting of the stack by three bytes. We were able to free up three bytes but it forced us to replace all the ROM code in all the units.
DECtalk continues up till the present time. It is now a part of Force Computers Inc. which retains the right to use the name DECtalk.
* - DECtalk is a registered trademark of Force Computers Inc.