0.1 The idea  (Page 3/3)

When an engineer at TI had an idea for a new product that could be valuable to TI, but for various reasons there was no funding available through the OST program, the IDEA program gave another path to funding. The IDEA program assumed that the engineer wasn’t ready to give up for lack of funding – in other words the engineer was far too passionate to let it go away. To get funded the engineer agreed to do the work on their own time with assurances to the boss that it wouldn’t interfere with the assigned work load. The funding would take care of any need that would require a small amount of money to buy things, or to contract outside expertise to complete the project. Completion was defined to be a feasibility demonstration of the product concept to management. Then if accepted, the project would be put in the OST process to get further funding.

The process for obtaining IDEA funding was simply to present the idea to any of the IDEA representatives. These representatives were chosen from the technical staff and were generally senior technologists who could better understand the technical risks. An IDEA representative could fund any project up to $25,000. There was very little paperwork required to make this funding happen.

Paul presented the Spelling Bee idea to Ralph and he agreed to provide the $25,000 needed to demonstrate feasibility of the Spelling Bee.

There was still a lot of work to be done to show feasibility of the product idea. The first step was to choose a method of creating speech with an IC. There was a brand new concept for synthesizing speech using a mathematical model of the human vocal tract. This mathematical model used a 10 pole recursive filter, specifically it was known as “Linear Predictive Coding” with ten poles (LPC-10). Further LPC-10 was an analysis and synthesis system. The speech code would be captured using a professional speaker in a studio. The spoken words would be processed by a mini-computer in non-real-time. The resulting data set was then hand edited by a human editor to remove the errors that were caused either in the analysis/synthesis process or by how the model was implemented in the IC. I will go in to more detail on how the synthesizer worked in a later chapter. But this should be enough for us to continue on with the story of how we demonstrated the concept. It was important that we could demonstrate acceptable synthetic speech with a data rate of about 1200 bits per second.

Now with resources and a basic path forward, Richard Wiggins went to work on developing a computer simulator of the Spelling Bee. The demonstration would be shown on a mini-computer in TI’s speech research lab with all of the spoken words synthesized in non-real time and played back for the demonstration in what would be the product's output form. I’ll cover this in more detail in the next chapter.

The daunting issue that we were facing remained: we were developing a product targeted at a cost significantly less then the then state of the art mini-computer to be used in demonstration. At the same time it needed to be small enough for a child to play with, rather than that of a mini-computer in a large room. But neither of these issues seemed to bother any of us as we prepared for the feasibility demonstration.

Figure 5 shows the original design to cost. Paul envisioned a two chip system: the first chip having both the system controller (TMS1000 based) and the speech synthesizer, and the second chip a 256 thousand bit ROM. The total cost of the finished product would be $12.65 with a recommended selling price of $29.95. As you will see later in the book, we missed the design to cost goals by a bit.

References

1. http://datamath.org/Edu/Professor-76.htm
2. http://datamath.org/Speech/SpeaknMath.htm