I was sitting in a restaurant in New Hope on a bright Winter day chatting about some technologies that have evolved dramatically in the past several decades and was asked about the term tape-out.
New Hope, Pennsylvania is one of those places that has become something of a tourist trap on nice weekends. Stores come and go with a few enduring to give the place a bit of flavor, but you get the impression this place is hiding something.
I've found glass blowers, sculptors, a variety of painters, a precision telescope builder, steam engine restoration experts, musicians, lyricists, dress designers, architects and an truly amazing workshop...
If you have the time, arrange to visit George Nakashima's workshop. George was one of the leaders of 20th century furniture design and, although he died in 1990, his family keeps the workshop going and it continues to turn out furniture as fine art. This is where Steve Jobs finally decided to buy his chairs and tables after years of contemplation.
There are several artisan woodworkers at work and the setting is decidedly old world. But you are struck by the care and thoughtfulness that permeates the place. Selection of the wood is a borderline religious experience for the people I met. They will tell you about a specific piece of wood, where it came from and how they intend to honor it.
Seriously elegant pieces.
I am often struck by the importance and complexity of material in design. Function, form and beauty all strongly depend on the proper choice. Dave and I "know" that carbon fiber is wrong for any bike we would ride - the "feel" of a proper lugged steel frame is entirely different and wonderful for the types of riding we do.
A fundamental principle of design is that it is the art of compromise and the choice of materials is an important part of that. There are hundreds of materials that can a used for building a working bike ranging from cardboard, bamboo, and several different hardwoods. More commonly steel, stainless steel, aluminum alloys, titanium, or any one of a large number of carbon/resin combinations are used.
So even if you choose a steel frame it turns out there are several dozen types of steel, each with its own distinctive characteristics. The special frame that was built for Colleen required a very exotic stainless steel - Reynolds 953 - one that costs considerably more than carbon fiber and is difficult to fabricate, but just the ticket to solve some frame resonance problems that didn't show up in the CAD modeling tools.1
Once material selections have been made construction techniques are important. Some techniques, like those used a table from George Nakashima's workshop or a making a fine cello, are celebrated as serious art, but even common household objects represent complex choices and construction techniques.
Back to the lunch in New Hope... The conversation had focused on 3D printing
My normal observation about 3D printing is that the material selection is terrible compared to those conventional builders have at their disposal - the situation is not unlike a chef being limited to white flour, water, and mustard. She can make lovely prototypes and create designs that would be impossible with normal techniques, but the final result isn't exactly appetizing. I'm bullish on the technique for certain applications, but not for general use anytime in the near future.
But it struck me I had been ignoring a very important form of 3D printing - one that has already changed the world. Integrated circuit fabrication.
When thinking of the development of technology over the centuries I tend to consider major themes and drivers. Things like transportation, improvements in power densities and energy conversion efficiencies, and the ability to write fine lines and spaces.
Lithography is an important subset of writing lines and spaces and I was lucky enough to have become deeply invovled some of the R&D associated with it when I joined Bell Laboratories. While IC fabrication is a very technical process, it is conceptually simple. You shine light at a silicon wafer through a patterned mask. The silicon is coated with a material that exposes where light strikes it and a chemical process transfers the pattern to the silicon. While this is subtractive, a series of layers are built up forming a thin three dimensional structure.2 Although usually not considered 3D printing, it should qualify.
Many layers are used with very specific material choices. Up to thirty masks may be required for an advanced processing technique. Large sums are spent on material science, applied physics and several other areas with the progress is recognized as Moore's Law.
This ability to write fine lines and spaces had progressed for centuries driving our capability to increase the precession of mechanical objects and dramatically improve the reproduction of text and images on paper and electronic screens. Photolithography is merely one branch of a much larger trend.
My first involvement with photolithography came when 2.25 micron technologies were used in production chips and 1.5 in development.3 I was interested in the part of the process where information became physical - mask making.
The process has strict rules, as do most fabrication processes, but there is enormous flexibility in the capabilities of the circuits that can be produced. The number of possible combinations of simple logic gates is, for all intents and purposes, practically infinite.4
And that brings us back to tape-out.
When I first heard the term I though it meant writing a digital representation of the design to digital tape which would be used to tell the mask making machines exactly what to write. It turned out the term had been in use for some time. At Bell Labs and other places it described the process of making a photomask. A huge enlargement of each mask level would be from black line tape or cutting rubylith on a table or wall. These were photographed and reduced in the form of a photomask. Somehow the term tape stuck and has found its way down to us over the years.
These early circuits only had a small number of gates. Well before my time came the first really successful microprocessor - the Intel 4004 with its 2,300 transistors and 10 micron half pitches. Now we call carry billions of transistors around in our pockets embedded in these rather serious networked computers we call smartphones.5
Some say tape-out came well before early ICs - that it was used to describe the photolithographic process to make printed circuit boards.
This 3D printing technique based on photolithography has changed our world beyond what seemed possible in the 60s when it was first tried. What we call 3D printing simultaneously solves some hard prototyping problems, but falls short in many others.6 This tends to be the rule of many technologies - they find their niches and new ones that we hadn't envisioned while falling short of our original dreams. We are selective creatures who tend to search for patterns. The technologies we remember are those that open entirely new areas. Making bets is a tricky matter.
One of the tasks I'm asked to do as part of my work is assist in making these bets. Everyone has their own approaches. I find having a deep grounding in some areas through experience creating the technologies or in the fundamentals behind them is important, but not enough. Increasingly it is important to understand the social element.
These developments seem obvious in retrospect, but are very difficult to create or predict. They are fundamentally multi-disciplinary - something our education system seems to be moving away from and something that only a few companies have internalized. I've seen it in the most peculiar places ranging from a workshop that builds tables to a storyteller that makes movies that take years to create. From a curious consumer electronics company to a massive program that put a human on the moon. It even exists in workgroups in and outside of companies as well as a few very rare individuals. A commonality is it is often difficult to easily categorize these organizations as they connect many dots and, as a result, find interesting new patterns. Much more to say, but this is the theme of this blog so more will be said...
Oh - and tape-out again. I've heard the term applied to what is considered 3D printing in the past week. Chalk it up to the inertia of language.
1 This is an important point. Although computer aided design tools have become very sophisticated, they can't handle the variety of compromise in design that is given them. In bicycles handling characteristics are important and current CAD tools can't perform the level of analysis to understand the dynamics of a bicycle with a specific rider and road. There needs to be some grounding with reality and test "mules" - sometimes instrumented - are constructed, wind tunnel tests conducted, and so on. This process can be expensive and time consuming, but it is sometimes essential to produce something that is great.
2 The thickness is generally under a half micron at the silicon level and perhaps five to ten microns when the interconnect layers are added.
You can get a sense of the additive nature of the layers created by the successive masks by considering this simple device - a simple CMOS op-amp. The layers are polysilicon (red), a metal layer (blue), N-doped silicon (green), P-doped silicon (brown) and the X's are connections that cross layers. (The design is from a public domain digital design course.)
3 Design technologies usually refer to the half pitch distance between two identical features in a repeating array. The dominant production technologies are based on CMOS processing and a series of generations has been identified. Current leading edge production is at about 0.22 microns (Intel's Ivy Bridge processor line for example) and the next generation will be about 0.14 microns.
The cost of fabrication lines has risen dramatically with each succeeding generation. When I started in the early 80s a small fab line could be constructed for $10 to $15 million and a leading edge facility perhaps three times as much. The "steppers" that expose the silicon (they step from chip to chip in a large array on a wafer) cost about $1 million then. Steppers for 0.13 microns probably cost $100M each and the fab lines probably will cost more than $5B a pop. The number of companies that can participate has dropped dramatically over the years and the scale of produced has increased even more dramatically to match an enormous demand.
4 In addition to the functionality of a fixed chip there are FPGAs - field programmable gate arrays - where chips can be customized by engineers after manufacture for specific applications.
5 I remember bringing the first transistor into our house. I was seven years old and had used some Christmas money and my savings to buy a five transistor AM radio. I stopped counting after I got my first calculator - a HP-35. Now there are over a half trillion transistors under our roof. The mind wobbles.
6 Sometimes it is important to consider the "job we have hired something to do" .. in the case of 3D printing it is usually assumed to be extreme customization and/or local fabrication. These need to be very carefully considered in their own context and the context of the technology. If the goal is mass customization there are other segments of the economy that will be impacted in large ways well before 3D printing becomes a mass market reality. Included in these are custom specification and fabrication of clothing. Way too demanding for 3D printing, but addressable using other techniques.
This used some "wheat berries" - whole grains of wheat. We have a lot of Kamut berries (I recommend Bob's Red Mill as a source). The instructions all tell you to soak them overnight, but I find you can get away with simmering them in water or stock for about 45 minutes. It leaves you with a grain with a very nice texture. Experiment and find what you like.
Kamut and Brussels Sprouts
° 180g Kamut or other wheat berries
° 700g vegetable broth
° 450g trimmed and quartered brussels sprouts
° 1 large shallot, chopped
° 2 tbl olive oil
° 50 g walnut pieces
° 1 tbl lemon juice from a fresh lemon
° zest from the lemon
° salt and pepper to taste
° Add the both and wheat berries to a pan, bring to a boil and reduce heat to a simmer. Cover and cook for 50 minutes, drain and set aside
° Heat the over to 425°F
° Toss the brussels sprouts with the shallots and 1 tbl of olive oil in a big bowl. Transfer to a baking sheet and bake about 20 or 25 minutes until browned. Turn them over at about 15 minutes.
° Remove from the oven and add the walnut pieces (you can throw them on a couple of minutes before you take out the sprouts if you like them toasty)
° Whisk the lemon juice, zest, 1 tbl olive oil, and salt and pepper in a big bowl. Add the sprouts and wheat berries and toss them to combine.