The Path to MFG Vision: Part 1 of 2


When we bring on-board a new customer, they are really subscribing to our Vision. This Vision has its roots back in the eighties and nineties. I decided to summarise here the path I took and how it has led us to where we are today and to the “Vision” in MFG Vision, which I will share with you in part 2. Hopefully you will find the journey interesting. Maybe you also have experiences to share of coming up with novel ways to fix recurring systematic problems. There always were and still are so many opportunities in this industry for engineers of all grades to make a real impact.

Starting out as a Test Development Engineer

When working in a fairly large semiconductor IDM as a test development engineer back in the late eighties and early nineties, yield problems would inevitably come up in production after I released a product, problems that required correlations between fab (it was 3 inch fab initially!) and final test. One time, the product engineering manager actually sat down with me on a tester to watch the performance of a key lot which was supposed to have a fix. This was enjoyable (he was smart and funny) but not really practical as an ongoing solution and just a tad stressful at 11PM at night! The fab was just across the corridor from the test floor but information took a long time to flow across that corridor!


So I thought, well the tester itself is also fairly smart: can the tester’s CPU not generate the statistics in real time and spin them out in a report when the lot was done? Within a couple of years that idea was implemented for practically all testers in my division, both in wafer sort and final test. The summary stats were fed into a database which allowed reports and correlation with fab parameters. We called the system “GRIPS” or “Generated in Realtime Instant Product Statistics,” a good name which also I suppose helped it catch on. I even presented an ITC paper on this topic in 1993 in Baltimore, which was great fun.

The Problem

The fundamental problem with the real time approach to GRIPS was if you wanted to get additional statistics or add filters, then a lot of software across multiple testers had to be updated. That could take months. Then someone might come up with a new requirement and the loop would have to happen again. Life is too short!!

The Solution: Server-side

So we decided to process all final datalogs to get the GRIPS statistics. On some testers the real time GRIPS statistics were still being used by production but were not needed for final summaries any more. This “server side” solution was better and more flexible as the data analysis software was centralized now and more easily maintained and updated. Complex formats such as STDF and others were processed fairly robustly.

The Web Arrives!

With the growth of the use of the intranet in the company in the mid to late nineties, instead of just generating a statistics table, we started generating static HTML pages from the datalogs with all kinds of charts and tables. These looked great and could easily be shared across the division. An engineer could even have a link to such a page emailed to them when a lot finished.

I will continue this story next week with what happened in the early years of 2000 and why I left the above company and started MFG Vision Ltd in 2005. Comments are welcome below.

John O’Donnell


MFG Vision Ltd