A Little Semiconductor Related History

[mawyatt]

I also prefer to actually calculate delicate circuits, instead of blindly trusting simulations.

Very important point!!

When we created the first Graduate Courses in RFIC technology way back as a adjunct prof, we emphasized that students should be able to crudely sketch the waveforms of circuit under study and figure out bias points BEFORE jumping in front of Cadence. As expected this fell on mostly deaf ears, so we often created homework/exams that would cause significant stress if directly thrown at Cadence without regard to a rudimentary understanding of how the circuit worked!!

Devious yes, but it did help a few students (we hired them!!) to think before jumping in front of Cadence!!

Best,

[harerod]

An interesting little anecdote about hot ICs. ... Motorola...

How should the poor guy have known? At one time my Atari ST sported a 16MHz-rated 68000, overclocked to 24MHz, instead of the original 8MHz CPU. The general impression was that Motorola stuff was rather sturdy.

[harerod]

...
Devious yes, but it did help a few students (we hired them!!) to think before jumping in front of Cadence!!
...

We have a saying in German, which goes "Nur die Harten kommen in den Garten." A English rendering may be "This is Sparta!!!".
While not fun at the time, rigid training helps a lot down the road. Says the engineer 2nd class...

[coppice]

Quote from: jwet on Today at 05:13:44

    They have a somewhat unusual program in the Engineering and Architecture/Design schools where each student alternates semesters with internships in relevant industries.  It makes for some well rounded new graduates.

How well does that idea work in the US? They call them sandwich degrees in the UK, and they have a long history, but a very patchy history. The needs of industry and the needs of the universities and students don't consistently align very well. I know Hong Kong's HKUST silicon design degree program gets a number of people into internships in US design teams, so some people seem to make it work internationally.

Sandwich degree, ey?
I hold one of those. Places called "University of Applied Sciences" (ger. "Fachhochschule", "FH") would hand out diplomas after 8+ semesters. Two of those, 3rd and 6th, would be spent as interns. During internships, a much reduced number of lectures would be attended.
Due to Germany's rather decentralized education system, we would see big differences between institutions. Furthermore, the general consensus was that the FH graduates were second class engineers, in the sense that they lacked the theoretical (math) background that was part of "Technical University" (ger. "Technische Universität", "TU") training. However, the experience from the internships was sought for many medium grade positions. Conversely, payment was lower and career opportunities were limited, especially with the larger employers.
Things have changed since the 1990's. Although I painfully see my limitions compared to TU graduates of my generation, these days math skills seem to be lower in general. A couple of years ago I found myself explaining Laplace- and z-transform to TU Munich graduates (one doctor and a couple of masters in electrotechnics). I also prefer to actually calculate delicate circuits, instead of blindly trusting simulations.

In the UK sandwich degrees have been a mixed bag. I haven't looked at them for years, so I don't know how they may have changed. They come in what is termed thick and thin sandwiches. A thick sandwich is basically just a 3 or 4 year degree from the point of view of the university, with a year in industry before and a year in industry afterwards. There were people at top tier universities on this type of program. In return for some commitment to work you can get some sponsorship from the industrial partner.

Thin sandwiches are specialised degrees, because the timing of the teaching and industrial placement sections don't align with the normal degree teaching semesters. These were more often seen at the lower ranked institutions, but not exclusively. I've known people who did thin sandwich degrees who were taught to a high standard.

Judging from experience with recent applicants, a skill like "silicon design" listed in the CV would mostly imply having seen a certain software tool during education.

The biggest problem with CVs is the very common type that lists "I worked on this", "I worked on that", without saying if they were the project lead or the tea boy/girl.

Every academic I have spoken to in recent years complained about there no longer being any real lab work in university programs now, due to cost consideration and weird health and safety clampdowns. Some of these have been the mavericks who still do some, and constantly fight to keep the lab work going. Others have just secumbed to pressure, and do pseudo lab work with matlab. or other software tools. Things like MCU programming is probably an exception. Still mostly software work, but cheap and with such benign hardware the health and safety people don't create obstacles.

[tggzzz]

I also prefer to actually calculate delicate circuits, instead of blindly trusting simulations.

Very important point!!

When we created the first Graduate Courses in RFIC technology way back as a adjunct prof, we emphasized that students should be able to crudely sketch the waveforms of circuit under study and figure out bias points BEFORE jumping in front of Cadence. As expected this fell on mostly deaf ears, so we often created homework/exams that would cause significant stress if directly thrown at Cadence without regard to a rudimentary understanding of how the circuit worked!!

Precisely.

Such reliance on numbers without theory intuition and experience  was a pet hate of my father. He only begrudgingly got an HP85 very late on in his career, for just that reason.

His attitude was good. One of his last papers was predicting how a PWR would react if its bottom ceased to exist. He didn't simulate, but used the equations (validated with experiments) of how water flows down the plughole in a bath. He concluded there would be time to scram it before the core warped and prevented insertion of control rods.

[harerod]

One Wire History
...
This is a little more than I know about this history.  I hope harerod is pleased with the result.  This will be the last of these for a bit- I have to get my Taxes together for Uncle Sam (and my wife!).  I'll continue to follow this and may pipe in but these long stories are surprisingly time consuming.  Thanks for your time and attention.  I may post more content after April 15......

jwet, thank you very much for taking the time to write that essay. Let's continue this story, when you time permits.
I have written several 1wire-master implementations on different platforms. I have also wrotten different slave implementations for custom sensors or as replacement, due to issues in the supply lines (most recently due to Chipageddon...). All this makes me wonder how big the 1wire logic is on a DS18x20? I assume that this is implemented by a state machine, not a CPU? How many transistors might that take?

[jwet]

Harrod, it's certainly just logic, I'll pulse Dias on this.

Also on sandwich degrees, UC extended the program to 5 yrs to give same class time.  My son studied industrial d3sign there and has done well in his career, he's 35.  I think it's especially good for industrial design and architecture.

Dallas hired a lot apps and business people out of UC.  For designers and very technical people, the tended towards big 10 schools in the Midwest- Illinois, Michigan, Purdue, Indiana, Iowa, Ohio, etc.  Maxim was big on MIT, Berkeley and Stanford grads for IC designers but there was a smatter of everything- and some other favorites- Caltech, Worcester, Illinois (My alma mater) and Iowa (our Last CEO's alma mater) along with some brits like Fullagar (Trinity I believe).

Stay tuned on your question
.

[Njk]

The 88k was such a weird project. Does anyone outside Motorola know about the 77k? That was a really interesting design, but I'm not sure if anyone outside Motorola ever heard of it before priorities changed, and it was dropped.

It was a century old company, no wonder there were many interesting projects of various success and publicity. For instance, the MCore CPU architecture. It was designed for low power applications, was capable to operate using a power voltage of a bit more than 2V. I heard about the demo where an MCore MCU was running powered by improvised lemon battery. Symphony digital radio, the first SDR chipset for car radio makers. The RF front-end was still of analog type but all the IF processing was done by 56xxx DSP. Time Processing Unit (TPU), a complex proprietary sub-system like TI's PRUCC but more difficult to understand so it required longer learning curve. Still in production, AFAIK. There were also a controversial initiatives like pushing big endianess or attempts to model the SW quality management process using that for semiconductors (M-gates system, CMM/CMMI levels, etc.). The quality folks were a large and powerful clan where diversity was thriving so there was a good career opportunity for women. And there were some funny traditions. It was mandatory for every employee on Fridays to report upstairs about his personal achievements during this week. The CEO could not report to upstairs so he distributed his weekly report to every employee, completing the circuit.

[coppice]

The 88k was such a weird project. Does anyone outside Motorola know about the 77k? That was a really interesting design, but I'm not sure if anyone outside Motorola ever heard of it before priorities changed, and it was dropped.

It was a century old company, no wonder there were many interesting projects of various success and publicity. For instance, the MCore CPU architecture. It was designed for low power applications, was capable to operate using a power voltage of a bit more than 2V. I heard about the demo where an MCore MCU was running powered by improvised lemon battery. Symphony digital radio, the first SDR chipset for car radio makers. The RF front-end was still of analog type but all the IF processing was done by 56xxx DSP. Time Processing Unit (TPU), a complex proprietary sub-system like TI's PRUCC but more difficult to understand so it required longer learning curve. Still in production, AFAIK. There were also a controversial initiatives like pushing big endianess or attempts to model the SW quality management process using that for semiconductors (M-gates system, CMM/CMMI levels, etc.). The quality folks were a large and powerful clan where diversity was thriving so there was a good career opportunity for women. And there were some funny traditions. It was mandatory for every employee on Fridays to report upstairs about his personal achievements during this week. The CEO could not report to upstairs so he distributed his weekly report to every employee, completing the circuit.

Apart from the 88k, I think all the things you mentioned were very pragmatic and made money. The 88k was set up for failure.

We were never required to report upwards on a Friday.

There are always career opportunities for women in engineering. Its so rare to find a CV from a woman for a seriously technical electronics job, that's not pure software, that everyone tries to fit her skills to their needs, however badly they actually fit. CVs from women are so rare they get passed around for everyone to look at, and ponder over, like anything rare. Motorola did have more women in senior engineering positions than most companies, however few that might be. I reported to one.

They did go overboard for the CMM/CMMI stuff at one point. They were very aware, as stated in presentations. that getting to level 5 was a tradeoff. Somehow they still kept trying to get to level 5 across the board. It all seemed a very messed up way of trying to improve their capabilities. Seeking perfection has a bad habit of harming genuine improvement. They also had a huge 10x speedup push, to get designs out the door faster. That was such an ambitious goal when presented in a simplistic way it really put people off. Thought through, its easy to see how you can have some long slow complex projects running in parallel with fast turn around spins, and manage a much better average turn around time by ensuring those fast spins get to market fast. However, presenting people with a simplistic 10x goal up front turned most of them off so badly they were not listening to the details.

[David Hess]

An interesting little anecdote about hot ICs. ... Motorola...

How should the poor guy have known? At one time my Atari ST sported a 16MHz-rated 68000, overclocked to 24MHz, instead of the original 8MHz CPU. The general impression was that Motorola stuff was rather sturdy.

My understanding is that this relates to how operating temperature affects reliable operating life.  The standard for reliable operating life from the big names was something like 20 years, but if you were willing to sacrifice this, then higher power is possible allowing higher clock speeds.

[harerod]

...The standard for reliable operating life from the big names was something like 20 years, but if you were willing to sacrifice this, then higher power is possible allowing higher clock speeds.

Would I do that today? No. However, back in 1990 it was just impressive how that 68k handled the overclocking.

[Njk]

Apart from the 88k, I think all the things you mentioned were very pragmatic and made money. The 88k was set up for failure.

I'd jumped on the board of SPS in late nineties and had been working for the company for the next decade and half. Never heard about the 88k. Perhaps the time was wrong because the project is too old. Anyway only a big organization can afford the luxury of designing something for loss. It's not always clear if and when and for whom the thing will be profitable. The lone cowboy approach (take the venture capital service and start running around like a deer wounded in the ass from the shotgun, to entertain the sponsor) is not sustainable for everyone. Remember, IBM PC was not created in that way.

We were never required to report upwards on a Friday.

I'd travelled a lot at the expense of the shareholders, but newer been in the UK. There must be many peculiarities at the locations like Basingstoke. The guys who were sent there all characterized the island as a place where everything is wrong. No one was brave enough e.g. to rent a car there. But I don't think it's protected from the corporate spam. Do you remember the Chris Galvin's weekly reports?

BTW my daughter and her husband just received a formal job offer from the UK and passed the exam. Good luck for them.

There are always career opportunities for women in engineering. Its so rare to find a CV from a woman for a seriously technical electronics job, that's not pure software, that everyone tries to fit her skills to their needs, however badly they actually fit.

I do remember one Indian woman in senior engineering position, Padmasree Warrior. Another known spammer. She was very active, little Nikki Haley. But to no avail for her, so she eventually moved to Qualcomm and became CTO there. I heard many tales about women in EE (e.g. about the mother of the mother board at IBM) but my personal experience is different. At a times, my workload had tripled because of them. Known solution to the problem is to kick them upstairs ASAP.

They did go overboard for the CMM/CMMI stuff at one point.

Absolutely. But in the end, we've good memorabilia

[coppice]

I'd travelled a lot at the expense of the shareholders, but newer been in the UK. There must be many peculiarities at the locations like Basingstoke. The guys who were sent there all characterized the island as a place where everything is wrong. No one was brave enough e.g. to rent a car there. But I don't think it's protected from the corporate spam. Do you remember the Chris Galvin's weekly reports?

I once joked at a dinner of Motorola people from various parts of the business that the biggest career limiter anyone could have in Motorola was travel sickness, which made them all laugh heartily. So I assume every division had big travel budgets. I never worked for Motorola in the UK. I work in HK, and left in 1994.

[mawyatt]

When we had our small research company that got acquired in 2006, on our board was an ex Motorola VP that headed up the Iridium project. He told stories about how they decided to build the satellites on the automotive production line to save costs. He said they could produce the satellites for ~$10M each, which was well below what a full space qualified satellite would have cost. As previously mentioned they also used the commercial Power PC as the main on board CPU.

At the Smithsonian you can see the original Motorola engineering notebooks describing the Iridium concept, like a cell phone concept except the cell towers are moving overhead and handing off at ~28,000KPH, so your speed and position is almost stationary relative to the LEO sats. The Smithsonian also has the Iridium satellite on display.

Best,

[David Hess]

...The standard for reliable operating life from the big names was something like 20 years, but if you were willing to sacrifice this, then higher power is possible allowing higher clock speeds.

Would I do that today? No. However, back in 1990 it was just impressive how that 68k handled the overclocking.

Not too long ago there was a jump in rated CPU junction temperatures to 80C instead of 50C.  The increase in GPU junction temperatures has been even higher, with nVidia having had some big thermal problems in the past couple of years resulting in lawsuits.  So they are doing it for you.

I suspect this was achieved by lowering the reliable operating life down to 5 years for "consumer" hardware.