Get email delivery of the Cadence blog featured here
This week it is CDNLive Japan on Friday July 20th. I will be there so obviously this will be my latest trip to Japan...but we will start by looking at my first trip to Japan.
The first trip I made to Japan was in 1983. This was very early. If you have been in semiconductors or EDA all your career, you've probably had at least one trip to Japan. But almost certainly it wasn't earlier than 1983. I went to Japan before anyone I knew had been to Japan.
Usually, a trip to Japan involves going to Tokyo, but I only went to Osaka. In 1983, the new Osaka airport, Kansei, which is on a man-made island, had not been created. I flew in and out of the old Osaka airport.
To understand why I had to go to Japan, you have to remember that 1983 didn't just predate the internet, it largely pre-dated computer networking. I worked for VLSI Technology (VLSI). It was the era when the first video games boomed. The games were stored on "cartridges", which were plastic cases you inserted into a slot on the console. Inside was a read-only-memory (ROM) that contained the game code.
VLSI in that era had a secret business model. We positioned ourselves as a foundry, but that business was just getting started. Today, we'd call it an ASIC business since foundry means pure manufacturing now, but that word hadn't been invented. By definition, most of the designs hadn't gone into volume production, which would take a year or more from when the design started. Furthermore, until we finished developing our first suite of design tools, most of the designs hadn't even started. And anyway, our fab was not ready.
But we had a guilty secret as to where we made the money that we were investing in EDA and ASIC technology. We sold millions of those video game ROMs to the video game manufacturers like Mattel and Coleco. We didn't make a lot of noise about it, I don't remember any press releases, and it was not widely known.
We had a small, but really skilled, ROM design team. Because our fab in San Jose was not up and running by then, so we actually subcontracted the manufacture to Rohm in Osaka. However, we had to keep what was going on secret. We didn't want Rohm to know who were the end-customers they were manufacturing for, and we didn't want the video-game manufacturers to know who we were subcontracting to. Obviously, we didn't want to get cut out of the deal, and eventually, we planned to use the wafer volume to jump start our own fab when it was ready, as both a technology and business driver, which we eventually did.
A mask-programmed ROM is programmed with...duh...a single mask. The design of that mask was done somewhat manually before I got there, but I got tasked with setting up a system that would automate the process. We would receive the codes as EPROMs, which were what the customers' programmers would use during development. When development was complete, they'd take the EPROMs out of the sockets and then get them to us.
The first thing I had to do was suck the ROM data out of the EPROMs. This required creating an interface to an EPROM programming machine to download from its terminal interface (RS-232), which was the only interface it had apart from a manual keyboard and LED readout.
Then my code would process the ROM data and write a magnetic tape, known as the PG-tape (pattern generator), for the mask-making machine to make the special mask for manufacture. This was still the era when tapeout involved actual magnetic tapes. The PG-tape would then be used to make a mask, and then a courier would have to hand-carry the mask to Japan. I forget the exact state of computer networks at the time, I remember that dial-up modems were 1200 baud. Anyway, it wasn't practical to copy megabyte sized files to Japan so that we could generate the mask in Osaka, so we were stuck with expensive couriers.
I pointed out that it was silly to run a program to take a few thousand bytes of data, turn them into a few megabytes, make a mask, hand-carry the mask halfway around the world, and generally waste a couple of days and a lot of money. Even with a 1200 baud dial-up modem it would only take a few minutes to transfer the ROM code to Japan. We could then run my program in Japan to create the PG-tape, make the mask in Japan, and then go straight into production. The only question was where in Japan to run my program, since we didn't yet have any of our own offices in Japan. Anyway, this was the pre-workstation era when everything ran on VAX mainframes, and another complication was that whatever computer we used needed a magnetic tape drive. The obvious answer was to run on Rohm's own mainframe (which was an IBM 360, and had tape drives). Once they agreed with the idea, I was rapidly on a plane to Japan with a couple of colleagues to get my software up and running on Rohm's mainframe. I don't recall now why, but I'd written the original program in Fortran, rather than Mainsail, which was the language we used for all EDA development. Or perhaps I rewrote it into Fortran so that it was in a form we could compile and run on a mainframe.
I assumed we would just use an interactive terminal but Rohm was still in the dark ages and used punched cards. The card-reader was in the computer room, and as a non-employee, I was not allowed in the computer room. So a task that should have taken a couple of hours took most of the day, but a Rohm software engineer and I got my program up and running by evening. We tested it by transferring a code using a modem, although from across the room rather than from the US, and then creating a PG-tape.
From then on, we didn't use couriers, and we could get a ROM into production a day or two faster.
Despite not being allowed in the computer room for security reasons, I did get my first tour of a working fab. Even then I found it amusing that I wasn't allowed to run cards through a card reader but I was allowed in a fab full of dangerous chemicals, and valuable wafers being manufactured. I had to wear a bunny suit and go through the air shower and across the sticky mats. This was the old style fab, where the equipment had a front and a back. The front was in the clean room, and the back was in a maintenance corridor. The fab was on an upper floor of the building, with a floor above and below used for air handling. In a modern fab, the equipment is completely enclosed, and there is almost nobody required to be in the fab at all since the wafer handling is all automated.
But it was still the era when a lot of 12 wafers in a plastic cassette would be hand carried between machines. Lot is a technical term, not just a way of saying several. This wasn't the modern era of automated wafer handling. It wasn't even the intermediate era of SMIF pods that kept the wafers inside a sealed box that attached to each piece of equipment (SMIF stands for Standard Mechanical InterFace). In 1983, the wafers were bathed in the air in the fab. The floor and ceiling were perforated, and a laminar flow of air went from ceiling to floor and was then recirculated through filters. The airflow wasn't very noticeable, I think it was about one foot per second. But at one foot per second, the air in the room was completely changed every 10 seconds.
The only other fab I ever went in was VLSI's own San Jose fab. For a period, we allowed any employee to go on a fab tour on Friday mornings. You just had to sign up. That was probably in about 1985, so for the last 30 years of my career in and around the semiconductor industry, I've never got back into a fab. I've tried in Taiwan and was told "as the Japanese would say, <breath drawn in between teeth> that would be difficult." I wouldn't have had a clue during this 1983 visit, but I've been enough since to know that when the Japanese say something would be difficult, it means not a chance.
When I went to Japan in 1983, the exchange rate was about 240¥ to the dollar. Everything seemed very cheap. For years, the exchange rate has been about 105¥/$ so you the purchasing power today for your dollars is only about 40% of what it was back then. Japan certainly doesn't feel cheap any more.
The next installment: The First Japanse Food I Ate was in Japan.
Sign up for Sunday Brunch, the weekly Breakfast Bytes email.