Koby’s Kaos

While I never witnessed one, “old timers” at processor design sites would tell me about ancient times when they would close the company cafeteria and spread out the charts of a new chip on the floor. Then the engineers and designers would pace along the electron paths discussing how or “if” the real thing would actually work.

These days, even if we trained our engineers as triathlon marathoners, I don’t think people alone are up to the journey. By the time they followed the logic halfway they might not remember where they started—like demanding total recall of 19^th century trekkers along the Oregon Trail. (And how many would take the Donner cutoff? No, let’s not go there.)

We have to trust our tools to be able to tell toll roads from tar pits, and recognize when they really have seen that cactus before. Even tools, however, are based on understanding and experience. Therefore they are optimized, like the sources they draw from, for specific needs and cases.

Thus each step of IP design, ASIC design, FPGA emulators, and package and board (and, in reality the whole device) designs will have specific imperatives. Each of these sub designs may be maxed out at their stage only to fizzle out when they come together.

Given infinite time and money and endurance, most anything can be fixed, but such expensive solutions are rarely practical. Hence the need to fix future faults as they go along, optimizing the end product as well as that specific stage. Some of the principles (say “low power” requirements) may be easy to agree on at the start, but others will be much more subtle.

Yet tools depend on statistics as well as theory, and as our robotic designers begin pacing an as yet infinite cafeteria floor, they are going to be faced by huge gaps requiring leaps of intuition, as well as the wisdom to unravel the Gordian knots of conflicting solutions they pull into existence behind them. (Yes, I know, the sword—which worked for Alexander at least until he got to India.)

How many dimensions of thought will future design require? 3D is now obvious (in concept at least) but what about 4D? Using time as a design dimension may make vertical functions possible in some cases while shunted off in others due to predictable parasitic effects. Master designed-in traffic cops then shuffle traffic patterns for best performance. Then what might the 5^th or 6^th dimension of design be? Can paradigm shifts come from (not in) computer intelligence? And where will the experience come from that will be necessary to make the tools to make it practical on your desktop?

Thomas Edison depended on failures to make a success. As long as we communicate, and share results whether intra-department or in broad initiatives, every serious attempt brings answers that much closer.

When fact, fiction and history weave together, the fabric supports supposition, if not exact science. I started reading Eric Flint’s 1632 series a few months back. It starts with science fantasy, smacking a West Virginian coal mining town down into the devastation of 1632 Germany and the Thirty Years War. The science fiction part begins with contemporary ideals and technology concepts trying to jump start 17^th Century technology, if only to stay alive. It’s long been my complaint that we assume anyone from olden times (and implicitly from anywhere else) to be or have been just plain stupid. The contributions of many writers to extending the series has spawned a cottage industry of examining social and technological change. (Okay, not all that deeply, but we’re talking Loony Tunes here, not Descartes.)

The fact part of my thoughts is more recent. Spawned by reading Ed Sperling’s interview with Bijal Vakil, a partner at White & Case and an expert in intellectual property, on the straining patent system and its effect on Silicon Valley innovation (Trouble in the Patent Office, http://chipdesignmag.com/lpd/blog/2011/01/13/trouble-in-the-patent-office)

The great good of the patent concept is that it encourages the flow of information, the exploitation of inventions, and promotes the creation of new technology while still protecting the rights of the originators. Thus new technology does not become a family or guild secret, which may wither and die from too small a base, if it’s not stolen to begin with.

But with the unresponsiveness and legal sniping of the present system, it seems we may be teetering away from the free flow of information that I think our civilization is based on. With effective protection, just getting the word out can be effective marketing. Developing rapid fire guns in the 17th century is very different from spinning out new chip IP in the 21st. Technology changes in IP can be very subtle and a new concept may or may not borrow from hundreds of other thoughts, each with a jealous and litigious patent holder.

Under the strain, three years is now typical for a patent process to be completed. This is an industry where three months may be the life of a new product. Getting the word out on your exploits can set you up for piracy and frivolous lawsuits rather than success. Some companies even are said to be avoiding getting patents. Are we ready for ox carts, pitchforks, and the Dark Ages—again? Bijal tracks some legal and common sense solutions. But the Inquisition will be in the details.

“Design for Yield” is more than a manufacturing mantra. It’s a philosophy we can gain from continually. It’s an invitation to check whether we’re actually achieving something beyond keeping our heads down—whether the cosmic hammer is being wielded by work, family or self-loathing.

Back in my student days, when I was working for college money, whether in an office or at the starch mill (Argo Corn Starch uber alles), I dreaded the evil eggheads with stopwatches and T&M surveys. Didn’t matter if I was stacking bags on a pallet or filling in a train log, I thought I was already doing my best, and I didn’t like having the spiffy clean guys in white suits and clipboards treating me like another stamped part. The “work harder, longer, faster” results may have briefly improved throughput, but it made life worse and dragged other measures down.

What I did appreciate was actually learning how to be more effective—like when the forklift driver or Archie Green, the graveyard shift foreman, came over and showed the (once) skinny college kid a better way to work. Many little improvements, such as pushing the 100 lbs. sacks into position as they were still falling weightlessly from the chute, became positive habits that made me a better clerk, loader, cleaner or writer. We got the product out faster and got injured less. (Besides just getting a breather, the person-to-person respect they showed helped me want to do better for them.) Dallas Cox, my work partner’s exasperated comment, “You throw the bag; you don’t let the bag throw you” became a life mantra.

So I’m seriously biased every time I see a Design for Yield initiative. (A warm glow in my chest teetering between heart-warming and heartburn.) They need to be enablements to excel, not snide fountains of criticism. Well-designed DFY habits will take more thought and a more thorough examination of materials and procedures, but they CAN save time and cost each time you do it NOW and every link down the chain (and if not well thought out and investigated, change for change’s sake will Murphy’s Law backfire the life of the chip).

Real DFY needs to gather wisdom from the “factory floor” and the support from the CEO down. Not an easy combination. Just like DFY for life takes a self-examination of where you are going, why, and who is counting on you. Good luck for 2011! (grin)