Jan 27, 2014

Guru #4, Carl Nelson

Someone asked me once if I was related to Carl Nelson. No, not that I know of. I’ve since done a bit more research into my ancestry and no, I am not related to Carl Nelson.

Carl Nelson is one of the five “gurus” in the early Linear Technology advertisement. The caption under his image says “Creator of the industry standard sample and hold, the first high current regulator, holder of numerous patents and is Linear’s Bipolar Design Engineering Manager.” The sample and hold amplifier is LF398, the high current regulator is LM323 and Carl was granted 10 patents at National Semiconductor, and 32 more after joining Linear Technology. I asked Carl about the name-recognition or “rock star” status that occurred back then compared to now. He said, “We enjoyed a time period where an individual could make multiple big contributions to the technology, because not much had been done yet. Now, a lot of the work goes toward incremental improvements, so the splash factor is smaller. Plus, projects are getting more complex and work is done with teams. There are still plenty of young studs walking the halls at LTC, and they seem to be having as much fun as I did, so some things are the same.”

So what were Carl’s footsteps on the way to Linear Technology?

Departing from the theme of “Analog Footsteps”, Carl did not got to MIT or Stanford and didn’t work for Fairchild. Carl grew up in Homer, Alaska – before it was a state. It was a tiny little town and Carl was class president of his high school (in a class of 17 kids). But don’t put Carl into any stereotype. Around 1959, Carl Nelson and Jimmy Vollintine hatched a plot to scare the bejesus out of the younger grade school kids waiting in the hallway to get their vaccinations. Jimmy screamed in the outer part of the office after getting his shot, and Carl pretended to drag the unconscious body out of the office and down the hall past all the little kids. There was pandemonium. Twenty five years later Carl met the nurse in charge at Millie Nelson’s house in Palm Springs, CA. It turned out that the nurse never knew what had caused the kids to panic, and she was still mad...

I asked Carl about the inspiration for his career in analog design.

“My standard cracker-barrel story is that I made an oscilloscope out of a Mason jar in high school. That’s totally untrue, but I did fiddle around with ham radios a little, actually designed a tube amplifier stage, and built a snazzy little AM radio from a kit. I learned a valuable lesson from the kit – one mistake and electrons is just junk. The radio wouldn't work, even after I spent hours and hours going over the instructions trying to find a mistake. My dad worked as a flight controller, so he took the kit to the airfield electronics maintenance guy, who eventually found the problem. There were two 5-pin terminal strips – the kind you wrap wire around to make solder connections. One had the 2nd pin from the left as a ground pin, and the other had the 3rd pin as ground. I had reversed them. I still have the radio and keep promising myself I will restore it.”

“As a little kid, I used to try to sneak into the electronics room at the airport when dad was on duty. It was crammed with all kinds of large glowing, humming, clicking electronics with lots of knobs, dials, and meters. It was always hot in there – a welcome change from the Alaska outdoors. But it was also probably pretty dangerous, so I rarely succeeded in staying long.”

Being from a small town, Carl preferred to go to a small college. He settled on Northrop Institute of Technology in southern California. NIT doesn’t have the name recognition of MIT or Stanford, or quite the rich history. NIT was founded by Jack Northrop of Northrop Aviation, originally as the Northrop Aeronautical Institute in 1946, to train aeronautical engineers and mechanics to work at Northrop. That reminds me of my own alma mater, General Motors Institute (now Kettering University). It was renamed NIT in 1953 when it became independent from the aviation company. It closed in 1993 for financial reasons, although a portion was sold to Rice Aviation. Rice then created what is now the Northrop Rice Advanced Institute of Technology which continues to operate in the old Northrop Institute of Technology facilities.

Carl graduated as an electrical engineer in 1967. But even there, he spoke his mind.

“I was nearly kicked out of NIT after I wrote a very inflammatory lab report. We were studying clipping and clamping circuits using diodes, capacitors, and resistors. They gave us function generators along with diagrams about how to hook them up, and waveforms to expect. Problem was, most of the circuits failed to operate as expected. The other kids wrote these long explanations trying to justify the bogus results, but I knew something had to be very wrong behind the scenes. So I played around until it suddenly became apparent that the function generators had cap-coupled outputs. That made them useless for studying clipping and clamping circuits. The big problem was that the lab instructors had created the lab material, complete with 'scope photos, from something that didn't work. So they must have falsified the data at some point, making it look like it was supposed to. I did something to get a DC coupled output, got the right results and wrote the report. But then I added a long diatribe about ethics, morals, and whatever other crap a 19 year old can dream up.  The professor called me into his office and told me my heart was in the right place, but I needed to fix my mouth…”

His first job was at Continental Device Corporation in research with the promise that he might be able to join their tiny linear IC design group. Continental was a spin-off from the semiconductor division of Hughes Aircraft. Most of us forget that the semiconductor industry was more than just Bell Labs, then Shockley, then Fairchild, then boom! Of course there were transistor companies that failed to make the transition to integrated circuits. But there were also some other significant companies with significant spin-offs. Texas Instruments and Motorola are two who survived and never had any spin-off companies eclipse them. But Hughes Aircraft had a few spin-off companies that did fairly well before blending with the other Fairchildren. Pacific Semiconductor led to TRW. Ed Baldwin, the first general manager at Fairchild, was actually from Hughes and he left to join Rheem Semiconductor with other Hughes defectors. A quote from Hughes in 1956 foretold the future, “Within four years the semiconductor industry will increase fivefold, resulting in extensive expansion of silicon and germanium production, according to Hughes Product Group of Hughes Aircraft.” Joseph S. O'Flaherty was the manager of the semiconductor group at Hughes and he left to start Continental Device Corporation in 1960.

I found a notice in a Cal Poly newsletter from 1968 that is likely quite similar to the one Carl responded to:

Pete Johnson, industrial relations manager, will interview seniors in EE, EL, IE and ME and Physics and Chem for positions as engineering management trainees and research engineers.
Continental Device Corp. is one of the leading silicon semi-conductor manufacturers, with plants at Hawthorne, California; Mooresville, North Carolina; and Hong Kong, with a complete line of silicon planar semi-conductors that range from high conductance switching diodes and UHF tuner transistors to high power devices.

Carl shared an early experience from Continental. “One of my first projects was to design and build a 30,000V aluminum vapor deposition machine. Looking back, I'm lucky to have survived… I was walking down the hall one day with a Dewar flask of liquid nitrogen for the deposition machine, and as the head lab technician passed me, he said, "What's in there?", followed by plunging his hand into the flask. He jerked his hand out just as fast as it went in, commented "have to do that fast", and walked on down the hall. Scared the crap out of me, but I guess that was the point.”

“Continental had designed a 15V monolithic three terminal regulator. I took one look at that and my fate was sealed. I made suggestions for improving it, and promptly got transferred to the fledgling linear IC group. I became aware of the goings-on of Widlar and Dobkin at National, and read everything they published. Continental was bought out by Teledyne, who had also bought Amelco, and I was transferred to Mountain View to work at Amelco.”

Continental would help fund the founding of Semtech and O'Flaherty remained on the board of directors for many years.

Amelco is an interesting side story. Amelco is a spin-off from Fairchild, founded by four of the eight original Fairchild founders: Hoerni, Last, Roberts and Kleiner in 1961. Jean Hoerni, of course, invented the planar process which totally changed the industry. Financial backing was arranged by Arthur Rock with Teledyne Corporation. Hoerni would then go on and found Union Carbide in 1964 and then Intersil in 1967.

Ted Malcolm recalled this story about working in Mountain View:

“Working at Continental Device, a semiconductor spin off from Hughes Semi, in 1961 thru 1965. When VP Jim Hines called a mgt. meeting and announced that we were no longer to gather at a local watering hole for lunch or after work. The reason given was because of what he had learned, about Fairchild, during a recent visit up north to the Wagon Wheel restaurant. At that time Fairchild was the enemy up north and the "wheel" was their local watering hole. We were using the alloy process with 1 1/8 wafers which did not yield many DO7 size die. The planar process began to be used after obtaining a license from Fairchild.”

“Being interviewed by plant mgr. Fred Bialeck in 1965 and being shown what he referred to as the most important piece of paper in his office. On it were the names of his key people and the areas of the factory where they had experience. No wonder they called it ‘the academy’.”

For Carl, National Semiconductor was the center of the analog world.

“I second sourced some of the National stuff and worked on proprietary designs, but secretly dreamed of working at National. Then one day I got a phone call out of the blue from Dobkin, who had somehow heard of me, and wanted to meet over lunch. I can still see and smell the huge mushroom salad I was eating when he asked me if was interested in working at National. Amelco sent the president in to talk me out of leaving, but he was the third president in 18 months, and the Amelco ship was sinking, so I waved goodbye and never looked back.”

“I remember Dobkin showing me to my office, pointing at the phone, and telling me I would be fielding most of the customer calls on National linear ICs. Not long after, the Army called, wanting to know if National would design a timer for the Grasshopper anti-personal mine such that the bomb would explode horizontally right at waist height after it jumped out of the ground. I told them to take their bomb and shove it. When I told Dobby what transpired, he stared at me for a second, then said ‘OK’.”

“We designed whatever the hell we wanted, and got it right most of the time. I got to correct Dobkin and Widlar from time to time, so we fed off each other to make stuff happen, but in my mind those two remained firmly on an imaginary pedestal.”

In 1981, Dobkin, Widlar, Swanson, Hollins and Welling left National to start Linear Technology. The split was not amicable and lawsuits were plentiful. Carl was still at National.

“I actually got Dobby's job when he left National. He told me that legally he couldn't recruit me, and I told him I wanted to try his job as director of ALIC. But National was going through major changes at that time, throwing huge amounts of money into emerging digital markets. They made quite a few decisions that made life in the linear groups much less fun. The final straw for me was a layoff caused as much by management error as the economy.”

Carl was not one for obfuscation, always managed to speak directly in his data sheets. From the LM323 data sheet: “New circuit design and processing techniques are used to provide the high output current without sacrificing the regulation characteristics of lower current devices. … The 3 amp regulator is virtually blowout proof.”

And from his seminal switching regulator application note, AN19:

“WARNING Before reading this section, be aware that the intent of the author is not to insult, but rather to relate in an attention-getting manner a list of goofs that, in many cases, he personally has had to own up to.”

1. Transformer Wired Backwards Those dots indicate polarity, not smashed flies.

5. Fred’s Inductors (Or Transformers) Inductors are not like lawn mowers. If you want to borrow the one out of Fred’s drawer, make sure it’s the right value for your application.

9. 60Hz Diodes The LT1070 will eat 1N914 and 1N4001 diodes, and not even burp.

12. Didn’t Read the Datasheet Then you shall have no pie.

I asked Carl about the LT1070 development and AN19.

“The 1070 was the result of going back to basics. I spent many days playing with all the possible ways inductors, diodes, and capacitors could be connected to create a switching power supply. I was mostly reinventing the wheel, but it gave me a good grasp of the principals involved and a great appreciation of the symmetry inherent in topologies. I was pretty sure customers wanted the simplest possible design so they could understand it and get it up and running quickly. It didn't take long to realize that five pins was the minimum needed, and that matched up with the 5-pin TO-3 and TO-220 packages, so it was a no-brainer. Then lots of work went into making the design work over as wide a range of input and output conditions as possible. Next was adding extra functions to as many pins as possible so users could play if they wanted to.”

“AN19 was born because I suspected (and marketing confirmed) that many companies needed to use in-house design engineers who had little or no switching power supply design experience. If I could write down everything I had learned, I assumed it would encourage more companies to try their hand at in-house designs as opposed to modules, etc. The project quickly started raising more questions than it answered, and I knew it would consume me at work. So I shifted most of the effort to home, staying up until the wee hours of the morning night after night for months. There was no practical way to verify equations as they were developed, so I spent a lot of time doing cross checks and sanity checks. I began to realize how difficult the design process was going to be for designers tossed into the fray, so I tried to maintain a sense of humor from time to time - hence the 'you shall get no pie' quote.”

Carl said his career choices were based on fun and glory, not necessarily in that order. He certainly had fun. Look up any of the posts about office pranks with the likes of Jim Williams and Bob Dobkin, and you will find Carl well represented. Based on his circuits, he clearly earned his “guru” status. But don’t stereotype him as a power guy, either.

“My best patent was on the LM35 at National - the centigrade temp sensor. I was mulling about bandgap theory one day and because I love the power of symmetry, it occurred to me that the magic of adding Vbe and delta Vbe to get a zero TC reference, and having zero TC always come out to 1.25V, might have a symmetrical equivalent. In other words, if you subtract them to get a strong linear drift and you do it in just the right way, the result might be a theoretically accurate temperature sensor. It turned out to be true – trimming the delta Vbe portion correctly makes both the slope error of the sensor and the offset error go to zero at the same time. So you can trim the chips once at 25C at wafer sort and get a very accurate centigrade temperature sensor.”

Carl retired from Linear Technology and moved back to Alaska. He’s still doing some circuit design, but we can’t talk about it yet.


SEC News Digest, Issue No. 67-79, April 26. 1967 http://www.sec.gov/news/digest/1967/dig042667.pdf

SEMI Oral History Interview: Monte M. Toole, Interviewed by Craig Addison, SEMI

I remember when, Ted Malcolm, Oct. 4, 2007 http://corphist.computerhistory.org/corphist/view.php?s=stories&id=347&PHPSESSID=c3f5124ac78ffafa90924b8343726678 

Semiconductor Boom's Just Starting
Hughes sees fivefold increase in silicon and germanium by 1960; announces new diode and rectifier
Chem. Eng. News, 1956, 34 (50), pp 6042–6048
Publication Date: December 10, 1956

Thursday February 8, 1968 Cal Poly

High school yearbook, [http://homerhighak.com/Homer_High/Welcome.html ]

Jan 21, 2014

Guru #5, Tom Redfern

Of the five “gurus” in the early Linear Technology advertisement, Tom Redfern was the one I knew best from my days at National Semiconductor. But I didn’t know him as a “guru” – just as an incredibly smart, effervescent and really nice guy – always with a big smile. So I had to ask him how he ended up on that particular advertisement.

Tom Redfern did his graduate work at Stanford University in Control Theory and by chance did a lot of work with A/Ds and D/As. After his masters degree, he went to work at Garrett Corporation (aerospace) and managed a circuit design group and also did work with A/Ds and D/As. There he worked on an all-digital computer for the F14 Tomcat. “While at Garrett I used military grade µA709's in a jet engine fuel control application. We were on allocation because Fairchild could not supply the demand and we paid $75 each for them. Can you imagine! Fairchild was printing money in those days (circa 1967),” per Tom. The contract, for the F14 computer, was awarded to American Micro-Systems (AMI) for the custom MOS chip set. He became enamored with MOS semiconductors and joined AMI as a field applications engineer in Southern California. At the time the only two companies capable of manufacturing reliable MOS devices were AMI and General Instruments. Only metal gate P-MOS was available in those days. No N-MOS and consequently no CMOS.

Garrett started a new MOS company, Garrett Micro-Circuit Corp. (GMCC) in Rancho Bernardo, CA (San Diego area). Tom joined as the Applications Manager. Originally AMI had more business than they could handle and GMCC would get the business they could not address. By the time GMCC got their factory up and running, business had slowed and AMI did not need this extra capacity. There were lawsuits and bad feelings and eventually Garrett recovered their investment by selling the fab to Burroughs. (Garrett is now Honeywell International.) When GMCC was sold to the Burroughs Corp, Tom left for National Semiconductor (1971) as a Field Applications Engineer covering Orange County to San Diego. After one year he moved to the headquarters as the MOS Product Marketing Manager. He had various jobs at National ending up working as the Data Acquisition Design Manager. At the time, there were three analog groups under Vice President Bob Swanson at National: ALIC (Advanced Linear Integrated Circuits) headed by Bob Dobkin, CLIC (Consumer Linear Integrated Circuits) headed by Tim Isbell, and SLIC (Standard Linear Integrated Circuits) headed by Jim Solomon. Tom worked for Solomon. Along the way, he amassed several fundamental patents for MOS circuits; many focused on A/D's.

Tom noted that while working for Jim Solomon, he was given the task of designing a twelve bit A/D. “Consensus was, at the time, this required 15 volt supplies and bipolar circuit technology. I convinced Jim that I could do it at 5V in CMOS. The rest is history and that pretty much established my ‘guru’ status in CMOS.”

In 1981, Swanson and Dobkin left National to start up Linear Technology Corp. Around then Tom left and joined Intersil which was very strong in CMOS. Tom worked for Dave Fullagar, who reported to Jack Gifford. Immediately, Dobby lured him over to Linear.

“I was employee #22, I think. I had left National to join Intersil. After a week there I got a call from Dobkin about joining LTC. I said ‘I never thought of that because I thought you guys were strictly focused on bipolar.’ He said investors were insisting LTC do CMOS as well and they needed someone to start that effort. I signed up and we worked in rented space on Bernardo Ave in Mountain View (by Central Expressway and 237). Who was there that I remember: Bob Swanson, Bob Dobkin, George Erdi, Carl Nelson, Brent Welling, Brian Hollins, Wadie Khaddar and Una Brown.” (Bob Widlar was also there, but living in Mexico.)
“I worked with Wadie to define the first CMOS process. It turns out this had a very positive effect on LTC's bipolar process. LTC could not afford to put two fabs in place so everything was processed in the same line. Because MOS is very sensitive to surface contamination the whole process had to meet MOS requirements. This meant the bipolar circuits had very good noise performance.”
 Tom did a significant amount of work on the CMOS process at Linear Technology:
“One of my many jobs at National was I headed up a group responsible for process characterization and device modeling. Device models were, of course, used with the circuit simulation programs. This experience along with my education gave me a very solid background in device physics. This was of great value as a designer and helped define the process we finally came up with. My philosophy was to define and characterize the process thoroughly. Test chips on every wafer allowed us to measure and control critical process parameters. If a design did not work right, and the process met its specifications, then it was a design problem. The process guys loved this because they only had to have one process and they just needed to keep it in control."
As I recall Wadie was the process engineer/manager I worked with.”
At one point MOS Design, Mask Design, CAD and Applications reported directly to Tom. He noted that Linear Technology has always been very design centric. He said the operator had a list of design engineers and which products they'd done. If a customer asked to speak with a design engineer about their product they were transferred to that designer.
“The ‘guru’ thing, in my opinion, is a bunch of hype. There are a lot of really bright people in the world who are just as smart and creative as any of us.”
Ultimately, Tom left in 1989 and eventually moved back to National Semiconductor.  That’s where I met him, around 1993 perhaps.

In 1997, Tom was awarded National Semiconductor Corporation's first Fellow Award, given to the technologist who has consistently shown outstanding contributions. The Fellow was National's highest technical position. In the announcement, Tom was credited for eleven patents, eight while working at National, and he won National's "Most Outstanding Patent of the Year" award in 1979 for his charge balance techniques used in the manufacture of analog to digital converters. In 1987, he was recognized by Electronic Products magazine's "Product of the Year" award for his one-chip data acquisition system.

An article in the April 13, 1995 issue of Electronic Design News (EDN), "On the Shoulders of Giants," recognized Tom Redfern as a genius in his field. "When we contemplate powerful desktop computers..., we tend to think of the geniuses who made such wonders possible. In information theory, great thinkers such as Von Neumann and Wiener come to mind. In linear technology, we have Dobkin, Philbrick, Redfern, and Widlar."

Whether Tom accepts the mantle of “guru” or “genius”, he’s in good company.

Jan 12, 2014

A Personal Analog Computer Story

EAI-680 Analog Computer

During my undergraduate years, I took many mechanical engineering courses. In one course, we were modeling an automotive suspension (springs, shocks and the rotating mass of the wheels going over a conveniently sinusoidal road surface). Capacitors and resistors were used in the model, of course. Then we went to the lab and our task was to properly connect wires to appropriate banana jacks on this big patch panel to set up the simulation. The patch panel was to be connected to an old computer, I was told. I realize now that was an analog computer. The computer itself didn’t work, however, so the professor merely checked our panel to see if it was set up correctly. Not the rewarding feedback it could have been.

Today, none of my classmates seem to remember this. The photo that most closely matches my imagination is this EAI-680 analog computer. Except the panels were a bit smaller – slightly more than one-foot square based on my fading memory.

This was the mid-1980s; I’m not so old that my college needed analog computers (although they retired the punch-card programmers the year before I got there). Analog computers like this were transistorized, but they evolved directly from vacuum tube models. In searching for the type of computer I used, I came across this lone record of a “personal” analog computer from 1963 which I’ll quote directly.

The idea of a personal computer in 1963 is enchanting (and way ahead of its time)! The Pastoriza Personal Analog computer was designed to be used by students at the Case Institute of Technology (which later merged with Western Reserve University in 1967 to form Case Western Reserve University). These computers were issued to 200 students in late 1962 or early 1963 by the school for use in their linear systems course. The students were part of a study to compare students who used slide rules with those who had their own analog computer. The computer was designed by Dr. James Reswick, who was then Head of the Case Engineering Design Center, James Pastoriza (who manufactured them and for whom they were named) and George Philbrick who was already well known as one of the founders of the analog computer systems.
The computer was battery powered and built into a small case so that it was easily portable. They consisted of several modular units that could be configured however the student wanted. 
Dr. Reswick demonstrating the personal analog computer

Reswick had been at MIT and is now best known for his work in rehabilitation engineering. Philbrick, of course, is a legend in the analog industry. Pastoriza worked for Philbrick at one point. Rumor has it that Philbrick actually gave Pastoriza the company that became Pastoriza Electronics. In 1969, Analog Devices bought out Pastoriza Electronics to get into the modular analog-to-digital converter business.

Everything is connected in this business. You don't have to go back very far.