The day I started at Nat Semi in Swindon I felt it was wrong.
The company culture was so different to what I'd known. Also I was
now in the Marketing group, not Engineering - this puts a different
shine on things! My job was as European Applications Manager for a
product group, Advanced Peripherals. This had been sold to me as
having some creative content but in reality it was just another way
of selling - there's nothing wrong with that, it was just not part
of my culture. I did learn an awful lot about electronics systems
for hard disk drives, Ethernet, DRAM control
and graphics which was interesting, but not
really what I had in mind.
I played a small part in the development of Ethernet though: on a vist to ICL (Britain's once-great computer company) I met Dr Alan Flatman who had pioneered 'thin' Ethernet - before the days of twisted pair!. He said that what Ethernet needed to be successful was a cheap repeater. One of my engineers, Andy Moorwood, took this on board as a project; he successfully pushed it through and eventually moved to Silicon Valley where it became reality, and, I think, a great success.
I did get to travel, and it was on a grand tour of Europe in 1986 that we arrived in Copenhagen to deliver a series of lectures. The buzz was that a massive radiation cloud had been detected. It was the Chernobyl incident.
In 1987 National bought Fairchild for a song, a move with huge historic significance because Fairchild was the birthplace of so many Silicon Valley startups, not least National - Charlie Sporck was one of the 'Fairchildren'. Fairchild had not been able to compete with National in the arena of analogue standard parts, so they had the idea of focussing on the disk drive market; they soon saw that customized parts were needed, and so the concept of CLASIC was born - Customized Linear ASIC.
I was asked to set up a design centre in Europe (first in Swindon, we later added engineers in Germany) which was certainly appealing. We grew with the addition of a design engineer and the inevitable marketeer, bought expensive '386 PCs (with maths coprocessors!) and got stuck in. Our charismatic product line director visited the UK and succeeded in selling our skills and "vast library" to IBM - we were to design two ICs for a new disk drive, controllers for the spindle motor and the actuator (i.e. read/write head motor).
CLASIC had one mainline process, a fairly advanced bipolar technology called LFAST, which was then still running in Fairchild's fab in Mountain View, California. It also had a CMOS process which ran in some other distant fab, but most of our designs were to be in bipolar. We were different from the norm at the time in that we used PCs for design and simulation - a Silvar Lisco schematic tool repackaged by IBM as CIEDS and MicroSim's PSpice for simulation. Layout in the early days was done back in California using ?; later we moved to a PC package called ICED.
A big bipolar IC - photo courtesy PP photography
The concept was to develop a cell library as we went along, which could be reused for new designs. But the reality of analogue, and this is still true today, is that each and every design is unique, and forcing the use of a cell library will not result in the desired function or performance.
Our first two designs worked well enough - as with most analogue ICs one or two mask revisions were required to achieve a reasonable performance and yield. On one of the chips an op amp had a rather larger input offset than was expected; nobody in our group could explain this and eventually the designer was able to talk to Bob Pease, one of the greats of analogue design. He quickly spotted that the mirror transistors in the first stage had not been laid out "common centroid" - they were not truly symmetrical. We made the change and it fixed the problem. We learned a lot of skills on the fly, including the design of asynchronous state machines. The chips went into production at quite large volumes for several years, then one day, in the IBM manner, production just stopped.
After these early successes our confidence was high and we embarked on more designs, each having its own challenges. Many did not go into production in the intended volumes, and so did not bring in adequate revenues. This led to a huge increase in the target revenues for each project.
One interesting field we got into was telecoms, particularly with Siemens in Berlin. I did a dial pulse translator and then a 2.048 Mb/s receiver. This required the development of an adaptive level comparator which was new to me then but has been very useful subsequently in RFID tag circuits (where I developed a really neat switched capacitor peak detector).
Shortly after the fall of the Berlin Wall, Siemens took over a factory in the "east". I made a visit and we stayed in a superb hotel on the Baltic coast - I'm sure this hotel was never used by the average citizen of East Germany! The country still had loads of Russians living there; what with bizarre Trabants everywhere it was quite an unusual sight!
I later used this telecoms knowledge to design a multi-channel combined driver and receiver for 2.048 Mb/s transmission systems (see photo). This was to be used in SDH add-drop multiplexers, with a high level of redundancy to provide fault tolerance and also "hot" card insertion. This in turn needed the design of input and output circuits that would tolerate floating or shorted supplies - quite difficult in bipolar where you get parasitic diodes all over the place! This chip took a fair number of revisions to get right and a number of new hurdles had to be jumped - a new package was developed to keep the die temperature down, and a we had to develop new test techniques to verify the drivers' pulse spec. However it did end up making a very significant contribution to the company's coffers.
After great upheavals in the company, when all marketing functions were "relocated" to Germany, our turn to close down finally came in 1994. Thanks guys.