Jul 28, 2014

Widlar's U.S. Air Force Performance Review

Bob Widlar's brother sent me a copy of the Air Force's official review of Bob's service. The actual document should eventually be available at the Computer History Museum. Bob, by all accounts in later years, was a difficult person to manage. So I was anxious to see how he was perceived in such a structured organization. He received superior marks except for the category of "getting along with others" - apparently it wasn't always harmonious.

Below are the written comments:

Strengths: A/2C Widlar is perhaps the outstanding electronics technician assigned to Course ABR33130. His background in the field of electronics as a civilian and his present endeavours in off duty educational pursuits serve to put him "head and shoulders" above the average technicians. He demonstrates a willingness to assist his fellow instructors, who looks to him for guidance on complex electronics problems. Airman Widlar is not satisfied with mediocrity in his efforts and constantly strives for perfection. He has an above average ability to use clear concise words to express himself.
Recommended Improvement Areas: Airman Widlar, in the past, has tended to dramatize his frustrations at inefficiencies that exist, to a point that he creates an impression of immaturity. He has improved greatly in this area, toning down his approach to problems, and has evidenced a willingness to accept that which he cannot control. No further recurrence of this comment are anticipated.
Facts and Specific Achievements: Airman Widlar was assigned to the Nuclear Weapons Branch 28 October 1958. On graduation from Instructor Training School he was assigned as an instructor in the electronic blocks of Course ABR33130A. The critique comments from the classes he taught were full of praise for the outstanding manner in which he presented the course material and expressed gratitude for the extra effort that he expended in assisting the weaker students. Due to his electronics background and outstanding knowledge, he was assigned duty as a course writer. He has produced clear, concise work which has required a minimum amount of editing. He has devoted a great deal of time in working out solutions for problem areas in the course and submitting suggestions on methods of presenting technical material. In this vein, he revised, in part, the method of teaching in the electronic phases by producing simplified schematics and developmental schematics with the text material. This has aided our new instructors a great deal in presenting this material. Some of the material he has provided required original effort. With a minimum of assistance, Airman Widlar produced training projects on the MC 627/628 and the T-187, although no adequate reference material other than schematics and test procedures were available. His efforts were commended by the Sandia Corporation Representative, Mr. Mark Elich, as being outstanding. Airman Widlar is now preparing an "In Service Training Course" for the Nuclear Weapons Branch, on transistors. Adequate technical material was not found on base. He, on his own volition, went to the Denver Public Library to accomplish the necessary research.
Airman Widlar has accomplished more in the way of self-improvement than any airman I have had the occasion to meet. Since entering the service he has acquired 26 semester hours of college credit. He is at present taking 16 semester hours during off duty time at Colorado University. In accomplishing this, he has never permitted his job performance to go below a superior level. Airman Widlar has also been instrumental in initiating and accomplishing the draft work on many visual training aids used in Course ABR33130. These have proven superior for use in this course to those which were commercially procured.
Suggested Assignments: Airman Widlar can be most effectively used by the Air Force in his present assignment. Should he desire, at the end of enlistment, to make the Air Force a career, I recommend that he be given further training by the Air Force and also that he be assigned to ARDC as a scientific aide.

Jul 16, 2014

Route 128, Data Converters and Hybrid Systems

Route 128 around Boston is home to many great analog stories. Before analog integrated circuits, companies in this area made board-level products. Stephen Ohr wrote a wonderful article back in Y2K. For years the region’s biggest electronic firms made minicomputers –Digital Equipment Corp. (DEC), Data General, Prime Computer and Apollo, etc. As a reporter, Stephen called on data converter makers – Datel, Micro Networks, Analog Devices, Data Translation, Hybrid Systems and Teledyne Philbrick. Apart from the hybrid circuits produced for military customers, the analog signal-conditioning circuits, data converters and integrated data-acquisition systems were all board-level products, sometimes potted in epoxy, for industrial control manufacturers that would use a minicomputer as a host. As you know, hybrid circuits were bare-chip module circuits on ceramic substrates with thick-film interconnects, packaged in hermetically-sealed gold tubs.

One name that disappeared was Hybrid Systems, founded in 1965 by Sam Wilensky and Don Bruck (and others?). It had a strong lineage to MIT and possibly some influence from Philbrick. Brian Lowe gave me some insight into the early days:

Sam Wilensky was the catalyst for everything at Hybrid Systems, he was a great guy, tall and slim, confident and friendly. He had had aspirations to be an astronaut but said he was disqualified because he was too tall. He had a Ph.D. in Nuclear Engineering from MIT and I believe he knew Jim Williams there. He was one of the founders, with Don Bruck, of the company. He was always positive and upbeat.
Sam had an Apple II in 1983 and liked to run Flight Simulator on it during lunch. One day I asked him why he didn't go get a pilot's license and fly for real. He replied that, statistically, if he flew a small plane over a period of ten years, he had a 100% chance of being involved in a crash, so this was his way of flying safely. Sam has an entry in Jim Williams' first compendium Analog Circuit Design, Art, Science and Personalities.
As a mixed signal manufacturer, Hybrid Systems was a competitor to Analog Devices, Analogic, Burr Brown, Micro Networks and others, especially in the military market. Starting in 1965 they manufactured data conversion modules from discrete components, then moved to hybrid chip-and-wire, then ultimately ICs. Hybrid Systems was the "first second source" of the AD574 12 bit A/D converter you mention in one of your blog posts. The HS574 was done as a two chip hybrid, one bipolar chip with a voltage reference, comparator and op amps, one CMOS chip with SAR, MDAC switches, output buffer logic, and a thick or thin film network.
Sam had pioneered the decoded DAC structure where the first 3 or 4 MSBs were split into 7 or 15 equally weighted currents, with the remainder of the ladder being R-2R. This increased the circuit real estate but reduced the glitch energy of the MSB switch event by a factor of 8 or 16, which cut settling time. Also, it decreased the sensitivity of output change when trimming the resistor ladder, which made manufacturing easier (and thus increased yield). This and more are described in his chapter of Analog Circuit Design, Art, Science and Personalities.
The DAC + SAR chip was manufactured as a custom circuit by Micro Power Systems in their proprietary Moly gate CMOS process. The resistor network was manufactured by Hybrid Systems in an internal facility that made NiCr and TaN thin film resistors. The manufacturing floor had a room full of manual laser trim systems consisting of stereo microscopes with lasers attached, poised above X-Y tables, with test sockets attached to custom built relay switched test jigs attached to HP 9825 Desktop Calculator via HPIB.
The bipolar chip was designed by Jerry Collings, a consultant to MPS who had originally been a control systems specialist. He had moved out of Silicon Valley to, as I recall, San Luis Obispo, and commuted to and from the Valley on Monday morning and Thursday night. I believe the bipolar chip was also fabbed by MPS.
An interesting side bar passed along to me by Sam – in 1976 Hybrid Systems introduced, through its subsidiary Audio Pulse, the first digital audio delay product, the Audio Pulse One. Designed using bucket brigade device (BBD) chips as delay elements, it sampled audio with an A/D, passed the digital data through various BBD loops, reassembled the audio for presentation through stereo rear channel speakers. It simulated a concert hall experience at about the same time that quadrophonic audio systems were entering the market.
During R&D of the One, a "golden ear" audiophile and frequent magazine author and reviewer Peter Mitchell was hired as a professional listener to evaluate the sound. The recirculation of the various BBD paths was rerouted numerous times to find the best sounding reverberation, and Mr. Mitchell would evaluate and say whether it sounded better or worse than before. Blending of various samples of delayed signal was done in such a way that the delayed signals were statistically uncorrelated with each other, providing a more realistic blend of delay.
At one point, Mr. Mitchell stopped the audition and told them one of the channels was slew rate limited and could not accurately follow the audio. The engineers looked at each other skeptically. They were using µA741 op amps, which are rated at 0.5V/µsec. They did some evaluation measurements and determined that Peter was correct, replaced the op amps with something faster and, after an audition, Mr. Mitchell was happy and the engineers were amazed and no longer skeptical of his golden ears.
The product was, according to Sam, so successful that it caused a serious cash flow problem for Hybrid Systems due to inventory, tooling and manufacturing start-up costs. They ended up selling the business and it never was a huge market success. A few years ago I bought one on eBay but due to time and other constraints have yet to hook it up in my listening room.

I re-read Sam’s chapter in the 1991 Analog Circuit Design book, “Reflections of a Dinosaur”. He provides a nice discussion of DAC topologies leading up to the Hybrid Systems’ DAC371-8. There is much to love in this chapter. He said in the 1950s the top-of-the-line Chevrolet and a year at a private university cost about the same – $2000 – and implied that it was still true (in 1991). I think it’s still true today. He uses the classic line, “It will be left as an exercise for the student to show that…” He also substituted diodes for transistors because they were smaller and cheaper (this was for a discrete circuit on a PCB – through-hole, of course). Today, on silicon, we use transistors configured as diodes because it’s so easy and there’s no cost or size penalty of any significance. And he gives some timeless advice:

“You should keep up to date on recent developments and not be afraid to research how a particular function was implemented in the past. You can benefit from the accomplishments and the mistakes of others. Fight the NIH (Not Invented Here) attitude and improve on the work of others with your own original ideas.”
He also relayed a useful distinction between digital and analog engineers from Don Bruck, but let’s finish the story.

In the mid 1980s, Hybrid Systems made a transition into "monolithics" according to Jim Donegan, former Sipex Chairman and CEO. Because of IC integration, it was only a matter of time before every hybrid circuit became a monolithic device. "It was easier to move into ICs," he recalled. "You could live in the hybrid arena for three or four years – until it became a monolithic," Donegan said.

My friend Joe Sousa co-designed, their version of the AD574 with Jeff Van Auken (now at Vicor).

In 1986, Hybrid Systems merged with a Harris spinout, Data Linear, and changed its name to Sipex – meaning Signal Processing Excellence. “95% of its sales were to the Pentagon and all its products were hybrid circuits,” Donegan said. Two years later it acquired a BiCMOS supplier, Barvon, and inherited a position in the market for 1488/1489 RS-232 line drivers/receivers. “Sipex's business is now (in the year 2000) 95% in commercial markets, with a third of that in communications, another third in data communications, and the rest in consumer devices,” Donegan said in Ohr’s article.

In 2002, Sipex formally moved their headquarters from the Route 128 area to Silicon Valley. "With the previously announced exiting of the hybrid business, the Billerica test operation no longer made sense," said Walid Maghribi, the CEO who followed Donegan. Frank DiPietro, CFO and VP of finance at Sipex, said “the company's headquarters has effectively been in Milpitas for some time and today's announcement just makes it official.” In 2007, Sipex merged with Exar Corporation “with Sipex surviving the merger as a wholly-owned subsidiary of Exar” according to Exar’s website. I find that choice of words, “surviving the merger,” interesting. Whether it is good or bad, it is no longer a Route 128 story.


http://www.eetimes.com/document.asp?doc_id=1224816 “The remaking of the 128 Circle”, Stephan Ohr, EETimes, 11/9/2000

http://www.bizjournals.com/boston/blog/mass-high-tech/2002/10/sipex-to-pare-billerica-operations-move.html “Sipex to pare Billerica operations, move headquarters to Calif.”, Matthew French

Analog Circuit Design, Art, Science, and Personalities, 1991, edited by Jim Williams

Email from Brian Lowe (check out his current work at http://belleson.com/ )


Please add comments below or send me an email if you have things to add to this story or corrections that need to be made.  Thanks!