Technology inflections : digital signal processing

[Editor’s note: Gene Frantz’s post is copied here as a guest post to allow the guest posts on this topic to be gathered in one place. Check out Gene’s other posts at On the Fringe with Gene Frantz.]

I just read the EDN article by Robert Cravotta on “Recognizing technology’s inflections.” We have talked about this at length over the last several months which leads me to add to his thoughts in this blog.

I have been in the middle of technology inflections for the last 30 years. In that period of time I have experienced multiple inflections. Many were, as Robert suggests, better noticed in hindsight than they are predicted. But I have had the privilege of both being blindsided by the inflections and also being able to predict them. In fact, my job at TI today is to predict the inflections and then figure out how to exploit them. With that in mind, let me give some examples of some of the inflection points I have participated in. But as I do, please remember that I am viewing this from the world of digital signal processing.

The first one I experienced was the introduction of the digital signal processor in the early 1980s. We, at TI, had developed our first one as the voice of a spelling learning aid – the Speak & Spell. That began a thought in the industry of, “if DSP can be applied to a toy, what else can it do?” That success spawned our first general purpose DSP, the TMS32010. Our focus was to build a better speech processing engine for synthesis, recognition and vocoding. We very quickly found that our customers were not using it for speech processing as much as they were using it for modems, hard disk drives and 3D graphic engines. (Yes, this is one of the symptoms of a technology inflection.) Our response was twofold: 1. change our marketing pitch and 2. architect the next generation device to do modems, hard disk drives and 3D graphics.

This same experience occurred when we introduced the TMS320C30, our first floating point DSP. We were warned that the DSP community would not adopt floating point as it was considered cheating – a DSP coder worth his salt could do it in assembly language with fixed point math. It surprised us as many new DSP coders began to use it, and at the same time, many old-school users converted. I did a study of our customers a year or so after we introduced the C30 and found something surprising: it seems that I could not find any applications in which only one C30 was being used; they were multiprocessing applications. We were certain that we had found another inflection point, but weren’t sure what the technology was that created it. Was it because the C30 was floating point, or because it was a C engine, or was it because it was 32 bits? Or, was it because of all three? By the way, I still don’t know the answer to this inflection point.

After three generations of DSPs of which the market blindsided us with how they used them and what new opportunities they created, I began to feel comfortable to begin predicting technology inflections. My measuring stick was simple – anytime we could introduce a DSP with significantly more performance than the present offering, an inflection point occurred. For example, in the mid 1990s I began to visit a lot of video compression companies that were springing up. I was a bit confused as it seemed that the problem of video compression was mature. So I asked one of the companies why the new interest in video compression was happening. I got an interesting answer. He said “your latest offerings have given enough performance improvement for us to go back and re-think every assumption involved with doing video compression. We now have enough performance to do many things that were not possible until now.” Well, that sounded like an inflection point to me.

When we introduced the TMS320C64x, I was asked to predict what new capabilities would be created with the new technology – yes, my first opportunity to predict a technology inflection. My answer was, as it should be, simple. “I can suggest some of the opportunities we already know that will take advantage of the higher performance and improve their solution. But those aren’t the interesting ones. The ones I am excited about are the ones that I can’t think of, which our system designer customers will blindside us with.” That has been my story ever since when I see an inflection point on the horizon. The good news is that it’s easier to see them coming than to figure out whether anyone will take advantage of them. That is where hindsight comes into play.

One last comment: We are on the cusp of a major technology inflection point. I’m getting really excited about it.

I’ll stop here as I have already rambled too long. But it would be nice to hear if anyone agrees with me on the ability to predict the technology inflections.

Gene Frantz - TI Principal Fellow, Futurist and Business Development Manager, DSP – Texas Instruments