Last edited by Ray Depew, 19 Aug 2013
But while I might have been good at fixing other people's mistakes, it wasn't a smart thing to do or a good place to put my time. You see, the people who created the things, even if they did it wrong, got all the public recognition for it, which got them bonuses, promotions, raises, and so on. I got recognition privately, from the few people who really knew what happened, but that never translated into promotions, raises or bonuses – or job security.
I've learned my lesson now. I'm one of the greatest fix-it men in the business. But I don't work anonymously anymore, and I don't work for peanuts. If you're interested, contact me and we'll talk about it.
Of course, that's not what I called it back then. I really wanted to be a forest ranger, a musician or a teacher. But I also wanted to make enough money to raise a family on one income. I figured out early that teachers, musicians and forest rangers didn't bring home enough money to afford that luxury.
One of the best learning experiences our teachers gave us in 8th grade was a career-search unit. We all chose careers that we liked and for which we had shown some aptitude. We researched the careers and interviewed professionals in our chosen careers. After briefly flirting with both medicine and law, I settled on a career in chemistry. This was partly because I thought chemists made a lot of money, partly because of all the pretty colors, and partly because chemists got to make cool stuff. Especially stuff that goes "boom".
(This many years later, I realize I should have stuck with medicine or law.)
The teachers tried to find a chemist for me and the other embryonic chemists to interview, but they were not successful. (I know that this was strange, because we were within easy commuting distance of Amoco Labs, Bell Labs, Battelle Labs, Fermi National Lab, and Argonne National Lab.) So they brought in a chemical engineer instead. He told us that, if we became chemical engineers, then within 5 years of graduating, we'd all be in management positions, and the chemists would be working for us. We believed him. I was converted to chemical engineering.
I enjoyed chemistry, mathematics and physical sciences in middle school and high school. With my high school science teacher's encouragement, my senior chemistry project was "How to Build a Bomb", something that probably wouldn't go over well in today's world. I built a mockup of a thermite bomb, a mockup of a plain old gunpowder bomb, and a TNT device. I researched the synthesis of TNT, and sadly informed Mr. Byers that we wouldn't be able to demonstrate it in the classroom. I suggested that I put together some plausible-looking glassware and rubber tubing and pretend that I was making it. He pulled out a crate of exotic glassware from an old paper-mill laboratory, and asked, "Can you use any of this?"
For my project, I demonstrated how to combine saltpeter, sulfur and charcoal and pack it in a glass bottle with an igniter. I showed how to grind up aluminum foil and rusty nails in a blender and pour the pulverized mixture into a magnesium casing with a cap-gun fuse. Then I demonstrated how to make TNT. I had a very complex setup, with colored liquids boiling in flasks, and tubes running all over, and an output tube disappearing into a fume hood – from which I pulled a beaker of crushed corn flakes and announced that the reaction was complete.
Mr. Byers was hiding behind the lab benches on the other side of the classroom. I still don't know whether that was theatrics or whether he thought I had really made something.
BYU's main research focus was on fossil-fuel-energy related topics, such as coal slurries, coal gasification, combustion research and catalysis. I did undergraduate research in two areas: building and operating a test apparatus to quantify the rheological behavior of coal slurries; and developing and running a Runge-Kutta computer simulation of exhaust gas concentrations in a monolithic catalyst.
Most of the ChemE graduates at that time went into petroleum or petrochemicals. Sometime in my junior year, I came to the realization that none of that fossil fuel stuff interested me. I couldn't see myself standing on top of a catalytic cracker at 3 o'clock in the morning, in a blizzard, trying to figure out why the darned thing didn't work.
What really interested me was semiconductor processing: how integrated circuits were made. While other schools like Stanford and Cornell had strong semiconductor processing curricula, this was not one of BYU's strong points. Their chemical engineering department had nothing to offer other than the core curriculum. For some reason, all the IC fabrication courses were in the Electrical Engineering department.
So I wrote my own option, and got professors from both EE and ChemE to endorse it, and I ended up with a BS in chemical engineering, with an emphasis in semiconductor processing. We actually made CMOS inverters using BYU's own IC fabrication facility. I took my new engineering degree to Hewlett Packard.
It was an experience I'll never forget, a 30-year-old man going to class with all these 20-year-old whiz kids. I was constantly in awe of them, because they were so much smarter than me, and I had to study extra hard just to keep up with them. Imagine our amusement when I discovered that they felt the same way about me!
Most of them thought I was the same age as they were. One day I was working with a classmate on a difficult homework assignment, and he noticed the family portrait on my bookshelf. He asked, "Are those your brothers and sisters?"
I said, "No, those are my children."
He pointed to my wife and said, "Then who's that?"
"Oh, that's my wife."
He gasped. "How old are you?" he asked. When I told him, his eyes went big and round, and he gasped again. "I should be calling you 'Mr. Depew!'" he exclaimed.
I laughed and said, "Don't you dare!"
I got to meet some great men at Stanford, professors and scholars whose names I had seen on book covers and on seminal articles in their fields. I was stunned when these men treated me, a professional of ten years but with only a BS degree, as one of their peers. (One of them even offered to buy, from me, a computer simulation that I had written for a homework assignment.) In 2005, as part of the 50th anniversary celebration of their Honors Cooperative Program, Stanford published a series of articles about some of the program's graduates. One of the articles was about my experience. Unfortunately, the Internet is not forever, at least not always, and my story is no longer at the SCPD website. However, the Wayback Machine has a copy of it.
Like I said, I got into this for the pretty colors. And as a plating engineer, I was surrounded by pretty colors. There's the sky-blue of an acidic copper solution, the deep mountain-lake blue of an alkaline electroless copper solution, and the deep indigo of an ammoniacal copper etch solution. Then there are the green of the nickel plating solution, the dangerous yellow of the cyanide gold solution, and the funny violet-brown of the colloidal tin-palladium catalyst.
The colors of metallic copper themselves are fascinating. First, there's the ruddy color of freshly electroplated copper, smooth and shiny. Less well-known is the delicate, salmon-pink matte of electroless copper, a thin metallic deposit that uses chemistry, not electricity, to provide the electrons that pull the copper out of solution.
An electroless copper plating bath is inherently unstable. It's a witch's brew of formaldehyde, sodium hydroxide, EDTA and copper ions in an unstable state. The copper will plate out on anything immersed in the bath, including the sides of the tank and even eddies or concentration differences in the bath itself. The trick in controlling an electroless copper bath is to keep it balanced on the edge of the knife, so it will plate out quickly, but only on the right surfaces – like a PC board.
One of the most fascinating things to watch is the death, the self-destruction, of an electroless copper bath. When the solution gets contaminated with catalyst, dust or just too much copper, then the copper spontaneously precipitates. In less than ten minutes, you can watch 180 gallons of solution turn from a healthy blue to the green of monovalent copper, then to a milky gray as the precipitation begins, and finally to a dilute red that fades away leaving a tank of clear liquid with a layer of copper flakes or granules on the bottom.
Either business dropped off, or we did our jobs so well that we engineered ourselves right out of work. The parts plating shop next door needed some help, so we hired ourselves out part-time to parts plating. Between the PC shop and parts plating, I got to work with some great managers, including Irv Hawley, Reed Ogden, Al Willits and Jean Faurot.
Al, my immediate manager, was terribly overworked and rarely at his desk. I was extremely self-motivated, and would find out what needed to be done, and write it up as a project. I would report my progress to him via memo or email, and he would respond every three weeks, or sometimes less frequently. It was in large part because of his hands-off attitude that we were able to be so productive.
Jean was an engineer-turned-manager who was quiet, competent, and yet always unsure of her abilities. Her parts plating shop had gotten several difficult assigments, and she depended on her engineers' confidence and enthusiasm to reassure her when schedules or technical results looked to be in jeopardy.
One of the greatest things I did in parts plating was to develop a process to gold-plate the inside of a four- to twelve-inch long copper tube, less than a half-inch in inside diameter, with near-perfect uniformity. I did it by doing what today is called "thinking outside the box". It was supposed to be impossible, but my process could do it time after time. (The phrase "they said it couldn't be done" is literally true in this case.) I wanted to patent my process, but then our competitors would know how we did it, so HP chose to keep it a trade secret instead.
A few months after my process went into production, somebody showed me an article about how a NASA contractor chrome-plated the insides of the giant fuel lines for the Space Shuttle. The similarity between their process and mine was uncanny, though purely coincidental.
My gold-plated tubes were used to calibrate microwave measurement equipment, which was used to measure waveguides, which were used to make and adjust the radars on U.S. Marines' helicopters. I was very proud of what I had done, even though very few people would ever see or touch those tubes.
One of the stranger things my colleagues and I did in that PC shop was more detective work than engineering. We had discovered a discrepancy between the amount of gold salts requisitioned for our plating baths and the amount of gold plated out. Since gold is so valuable, we had to routinely account for every ounce of gold we requisitioned. The chemical analyses were performed by a Ph.D. chemist in the parts plating shop's analytical lab, and the chemist was accusing us of sloppy work or sloppy accounting. We did mass balances on all the gold plating tanks in the PC shop. We did chemical analyses of the gold baths and of the rinse tanks that followed them. We did fire assays of the plated metal. We measured the efficiency of the electroplating operation. We found out that a large part of the gold salts requisitioned for our tanks never even made it into the tanks. It appeared that the Ph.D. was the one with the sloppy practices.
Sometime after I transfered to Corvallis, the chemist quit abruptly and fled the country. As Jean went through his files after his departure, she discovered that he had been cooking the books for years, and stealing gold salts from HP for a business that he and his brother were running on the side.
Less than a decade later, ICO had expanded to cover nine buildings on the Corvallis site (six of which they built), plus fabrication and assembly operations in Singapore, Scotland and South America. Maybe even Canada and Mexico. By then I had moved on, but it was fun to have been in on the ground floor of something that ended up taking over the world, even to the point of paying the entire company's light bills during hard times.
I was hired to move the operation's electroforming process from R&D into full production – a "manufacturing development engineer". Instead of trying to get the plated metal to stick to the substrate, as I'd been doing for six years, we wanted to get the thin metal foils to detach from the substrate. Again I was in the pretty colors of electrochemistry. Even today, I enjoy being able to open a printer, pull out the cartridge, show people the little gold square and tell them, "See that? I made that!"
I got to work with many sharp people in ICO. My mentor was Paul H. McClelland, a true wizard, and I got to play mentor to a bright young man named Rio Rivas, who has since made many important contributions to HP's printer business and had a successful career.
One of the more interesting things I did at ICO was design and build a computer-controlled, biaxial, hydraulic bulge, ductility tester. The bulge tester has several advantages over the traditional Instron tester, especially for testing thin metal foils such as the electroformed foils we were creating. The bulge tester also measures in two axes, while the Instron measures in only one axis. Like my gold-plating invention, the Duck Tester was kept proprietary to HP, but its invention and use is one small reason for the worldwide success of HP's inkjet printers.
I finally got to work in semiconductor processing.
My greatest achievement is nothing glamorous or spectacular. The deposition process used an old, conveyorized, sputter-down system. The process never had yields consistently above 80%. Many engineers had gone through, gotten yields temporarily above 80%, and then left (or given up). The process required a lot of babysitting.
I used Statistical Process Control (SPC) to fix it once and for all. I went through the process with the operators and technicians, identifying the defect categories and their root causes, and identifying the process parameters that would affect those defects – the "knobs" to control the defects. Then we set up control charts to dictate when, and how much, operators and technicians were allowed to adjust the process – to "fiddle with the knobs," as it were.
I got the process yield into the high 80s, and it stayed there long after I left. In fact, the engineer who replaced me seldom worked on the sputter dep process because the operators and techs had it under control.
I got some more wet chemistry experience, too. One bench, used for acid cleaning of the metal shields used in the deposition chambers, had hosted one runaway reaction too many. Normally the nitric acid quietly dissolves the thin layers deposited on the shielding, but occasionally a contaminated shield will cause the acid to heat up, causing a runaway reaction with massive clouds of brown nitrogen-oxide gas and enough heat to melt the plastic lab bench almost to the point of failure.
I got the mandate to specify and purchase a new lab bench, one that would contain the clouds of gas, stand up to the intense heat and quench the runaway reaction. After three design iterations, we had the thing built. It was, like most safety equipment, never really needed while I was there.
However, after I left Corvallis, I got a report that they had finally gotten another runaway reaction. The operators had stood inches from the bench and watched the billowing clouds swirl up the exhaust plenum, safely contained behind the plexiglass doors, while the heat-activated "dump door" flushed the hot nitric away from the shielding, and both the inner tank and the outer tank contained it with nary a blister. My safety bench had worked as advertised.
One of the chips I was responsible for was a color scanner chip, which included three 8-bit half-flash ADCs and four 8-bit DACs. Manufacturing defects caused lots of "stuck-at" problems with the ADCs and DACs, and the test we had was written for a pre-production version of the chip. I fixed the test and made it production-worthy, then set about to find a way to isolate the defects causing the "stuck-at" failures. I created an Excel-based simulator that could model the ADCs and DACs, mimicking their behavior perfectly. This allowed us to determine which bit was stuck, and whether it was stuck at 1, stuck at 0, or exhibiting some voltage-limited behavior. Knowing this information would tell us where on the chip to search for the defect.
Then I got the chance to do some serious project management: moving a new family of VLSI testers from R&D (beta test, actually) to production status. I had to coordinate the efforts of three groups of about ten people each, none of which had direct-report lines to me, and to make sure that no detail of the project was overlooked. (This time, I wasn't ignored.) We ended up with sixteen of these five-million-dollar testers on the test floor, arranged in three rows. Each tester had a million-dollar wafer prober mated to it, and all of this equipment had my imprint on it.
This was where I got introduced to LabView, one of the slickest pieces of software I've ever seen. None of our testers used LabView, but a lot of the equipment in ETest and the FA lab used it. What a fantastic way to perform data acquisition and control, and to build customized virtual instruments. I sure could have used LabView on that Duck Tester in Oregon. Years later, when I worked at NREL for a summer, I found that practically every research project in NREL ran on LabView.
All of these problems happening at once made them look like one huge, interconnected problem. It was a complicated mess. I teamed up with a sharp, young engineer named Heather Stickler, and the two of us finally got to the bottom of all of it.
The Opens problem was due to the fact that the probe card was only making intermittent contact with the chips. Development engineers had chosen to economize on a $5 million tester purchase by not spending $250,000 on hard docking tooling. Without the hard docking, the massive, counterbalanced test head floated above the wafer, with only a tenuous phsyical contact between the probe card and the table holding the wafer. (We discovered this one night when I leaned on the test head in despair and the Opens test suddenly started passing.) Heather and I purchased and installed one set of hard docking tooling. Now the test head, when docked, was rigidly connected to the wafer prober, independent of the probe card. The opens problem went away.
We traced the rebooting problem to a custom test function that the development engineers had written. The tester OS did not include a built-in test function that the engineers needed, so they had "rolled their own." This custom test function depended on a single line of code in the tester OS, and when the manufacturer upgraded the OS, that single line of code disappeared. So when our custom test function tried to execute that nonexistent line of code, the tester gave up and rebooted. The manufacturer didn't know that we depended on that piece of code, and wasn't willing to put it back in the OS, so we replaced the custom test function with something else.
The middle-of-the-night problem was the most controversial. We tracked it down to a change in a test suite, made remotely by a development engineer one evening while he was attending a conference on the east coast, and not tested prior to its implementation in the production test suite. Since the development engineers had a history of making these kinds of changes, Heather and I pulled a palace coup. We reorganized the test suite directories on the server, put production tests under strict revision control, and changed ownerships and passwords so that the development engineers could no longer mess with released tests.
We took some heat for this from the engineers and their managers, but we never had an incident like Black October again.
Yields dropped dramatically and inexplicably for one product. When we consulted with the design engineers about it, they said that nothing had changed on their product, and it must be a fab problem. We couldn't find any evidence of a change in the fab process. Under pressure from design and sales engineers, we approved the low-yielding lots, and off they went. Then, a few months later, the customer returned $1.8 million worth of defective parts. As we dug deeper into the problem, we found that the defective lots were the first lots from a minor mask revision. The engineers had played games with the revision numbers, so test operators had used the old test revision, and we had ended up shipping defective parts.
I had previously previously lobbied for product revision control at test, because number games like this always caused us problems at test that required engineer intervention. Now we had the data to show just how big a problem the number games caused. I recommended again that we implement strict revision control at test as well as in the fab, and this eliminated any more $1.8M warranty return episodes.
It seemed like programming or software engineering always intruded into my work (see my Programmer page). Here are a couple of the larger projects that I did.
Our testers all ran on HP-UX, and they were all programmed in C and C++. A team of engineers in another division had written a new driver to interface Electroglas probers with our tester. I was anxious to implement the new driver, but the development engineer in charge of it at our site kept saying that he was having problems with it and that it wasn't ready.
The department manager asked my opinion of it once, and I was reluctant to badmouth my co-worker. The manager, a very direct and decisive type, said "Come on, put your foot in it." So I did. I told him that the task wasn't as hard as my co-worker was making it, and that he had turned a one-week project into a three-month fiasco. He gave the project to me, and I got the driver debugged and released to production in less than a week.
In the days before Java and Python, one of my predecessors had developed a tool that automatically queried two different databases, calculated trends in product yield, defect density, line return rate and field failure rate, formatted the data in text and graphs, and printed it out with beautiful fonts and graphics. The user invoked it with a one-line command that included inputs to customize it according to customer, product line and time period. It was a combination of sed, awk, ksh, calc, SQL queries and troff. It broke down and had to be debugged and rewritten every month.
At the time, we were converting all our documents from troff/eroff to Interleaf (now called QuickSilver). Interleaf was written in Lisp, and users could write powerful macros using Lisp. So I learned Lisp, and replaced that report generation tool with one much more robust, with a GUI front-end, and written entirely in Lisp, including the SQL queries. Today, you can be sure that I would use Java or something else.
Because of the success of my Lisp project, I acquired responsibility for several other software projects, where I used awk or Visual Basic to perform data analysis and output the results to text files, Web pages or automatically generated email messages. And because of the success of my drivers project, I acquired responsibility for the software end of other machines, such as a laser RAM repair system and a package lead robotic inspection and repair system.
I also engaged in some Web programming, where I used scripting languages to generate interactive webpages which allowed users to enter data, submit reports, request information and log transactions.
---
Because of my work on the test floor, I was the highest-rated engineer in our IC Test and Assembly departments, and a well-known and sought-after expert in R&D and Product Design.
Once again I was doing project management. However, because of the scope of the enterprise, I concentrated on the technical aspects of overseas manufacture: maintaining product quality; reviewing, analyzing and reporting test results; and introducing new products into the fab shops.
One of my responsibilities was to collect data on the performance of our suppliers in the areas of Technology, Quality, Responsiveness and Delivery, grade each supplier on their performance, and present my findings in a semiannual report to their CEO and other executives, at their headquarters. One of our suppliers was in Singapore. I was kind of nervous the first couple of times, but eventually I got very confident. My confidence and my honesty allowed me to present both good news and bad news effectively.
The department grew quickly in three years. We were dealing with four different foundries in the Far East and one in Idaho. Our department's charter and my job description required me to be an expert in many different areas. Sometimes I felt like one of those plate-spinning guys on the Ed Sullivan show: I just made sure that the plates kept spinning on their sticks and didn't fall down. I was so busy keeping the plates spinning that I didn't see a whole new crowd of plate spinners, waiting in the wings for me to take my exit.
In retrospect, I should have left much earlier. I had been having entrepreneurial leanings for a long time, and although I could see that all manufacturing jobs were moving overseas, I deluded myself like so many other people, telling myself that my job was secure. I should have followed those entrepreneurial promptings and set myself up in business years earlier.
As it was, I took some of the severance pay and set myself up as Unique Engineering Solutions LLC. I sold my software engineering and programming services, covering all levels from embedded programming in assembly to web-based applications in Java. I also did some technical writing – anything to make a buck.
At the same time as I was operating Unique Engineering Solutions, in 2004 I started another company with six other laid-off tech workers. (In the two years from March 2003 to about June 2005, approximately 2400 tech workers in Northern Colorado were laid off. The job picture remained bleak until around January 2007.) We called ourselves Horsetooth Technologies, and our product or service was 3D computer simulation and visualization. Our primary tool was a game engine from GarageGames, called the Torque Game Engine. This powerful piece of software machinery allowed us to do three-dimensional simulation and visualization of all sorts of real-world stuff.
I was the technical spearhead for the group. I was the one who found the technical tools we needed to use. I dug up the books and other training resources for those tools. I created and conducted training classes for the other members and showed them how to use the tools.
I was also the guy who led the storyboarding and did most of the prototyping. Since I was familiar with everyone's capabilities and limitations, I became the de facto program manager, project manager, and head cheerleader.
Unfortunately, our sales and marketing guy wasn't very good at sales or marketing, and most of the technical people weren't as motivated to succeed as they needed to be. With no sales, no clients, and no prospects, after a while everyone lost interest and Horsetooth Technologies just faded away. Samples of the only work we did (mostly my work) can be found at http://ray.datech-net.com/HTI_Gallery.html.
Unique Engineering Solutions still exists. I'm working on some quiet, confidential projects.
First, I got a temporary position as a firmware engineer at DisplayTech, a local company that had been bought by Micron Technologies (and, as of August 2012, has been sold to Citizen Finetech Miyota Co. ). Both DisplayTech and Micron were great companies to work for, with exciting product lines, exciting business plans, and a dynamic and enthusiastic work force. My project turned out very well, and we were looking for a way to turn my temporary position into a permanent one.
Then, just days before Micron/DisplayTech could make their move, I was recruited for a full-time engineering position with a small company I'd done some work for two years earlier. I joined RLE Technologies, a small and successful engineering company. The company was so small that my title, Design Engineer, only described a small part of what I did. Like DisplayTech/Micron, RLE had an exciting product line, exciting business plans, and a dynamic and enthusiastic work force. They were a family-owned business, and they treated their employees like family.
America stands tall, not on mega-corporations like General Electric and Bank of America and others that are "too big to fail", but on thousands of small businesses like RLE Technologies. I enjoyed working at RLE more than I had enjoyed working anywhere else – except for Micron/DisplayTech. I'd have to say that RLE and Micron/DisplayTech are tied.
|
Created by Ray Depew, 09 Feb 2002
Last edited by Ray Depew, 19 Aug 2013 |
|
|
|