There seems to be a presumption that just because the industry migrated from bipolar to MOS that this also implies that EE students mostly or only learned digital design.
Analog design is becoming more, not less important -- but that doesn't mean bipolar is making a comeback. There are very good reasons why the early versions of the classic Gray & Meyer textbook from the 70s & 80s mostly discussed bipolar circuits, while later editions had a much greater emphasis on MOS circuits.
MOS does not equate to "digital". The industry still needs and still employs circuit designers -- but economics strongly prefer that those circuit designers avoid bipolar transistors and find a way to accomplish the assigned task using only NMOS & PMOS, and integrated R's & C's of varying quality and accuracy.
If you think bipolar engineers are hard to find out of college, consider vacuum tube engineers. (In fact the Air Force sponsered a grad school program thirty years ago already.) And yes, vacuum tubes are still best for high power.
Peter, I couldn't agree more. Further, it would seem that LT has not figured out what those of with a lot of real design experience have - that bipolar post-saturation sweeping of base regions is totally annoying for high speed design, whereas CMOS linear designs only are dealing with Miller capacitance for the desaturation transient. I wonder, then, if Mr. Dobkin's desire for better bipolar designers might actually hide a wish for reinventing the physics of the transistor itself.
The schools are not to blame. The basics of bipolar design could be taught in a one or two semester class, but a company could easily do the instruction itself, over a 6 month period, after employing a graduate with good basic skills and the will to try something "new" in design. The rest is just experience which takes time. It would be nice if one could higher a new graduate that would come with 8-10 years of analog design experience but that is not likely--wishful thinking.
The other point is that the engineer needs to think about employment options. How many firms want a bipolar engineer specialist? I think the engineer would find out the number of companies are very very limited. That make employment risk higher and with limited options to move that makes the bipolar career option choice even riskier.
One of the problems: students are looking forward to working after four years, and about the only activity that you have a chance of getting good at in that amount of time, unless you are quite brilliant, is digital, or at least hardware design focused on digital applications, i.e., numerical/symbolic domain apps. And since the bulk of that sort of hardware is MOS that will be the concentration as well.
Although low threshold MOS is not trivial, bipolars are considerably more complex and subtle, and you simply don't acquire that design skill set within the four-year EE curriculum. On the other hand, if you can afford to stay on for a Master's, you've got a chance. Linger for a PhD and you risk extreme specialization and lack of hands-on skills.
Companies are faced with an additional problem if they accept the notion that the entering BSEE employee is unprepared, as many times she or he will get the experience on the job, then leave to take a better one. Of course management can be famously inept in administering rewards, which many are liable to believe are solely financial. The rare technical manager who was once doing the design work may recall his or her motivations, the creativity and excitement of the engineering activity.
@cd2012, your point is spot on. I can't tell you how many (OK, it's 100 %) of executives I've talked to over the years (and the era really doesn't matter), who make that same complaint.
Some add that universities need to do a better job at evolving curricula; but others say they expect to train even the best fresh-outs just because their business is specialized.
And some others don't want academia to get more specialized; they want professors to emphasize the basics of electronics so that every engineer they turn out has the fundamentals down cold.
I think they're all excuses but for what, great big problem, I don't know.
It could be that no one can see outside his or her own business cul-de-sac. Someday, I'm hoping to run into the executive who says "You know, we get great engineers out of school. Sure we have to train them in our ways and culture, but they come here prepared to be trained and to think differently and solve problems. It's great."
Yesterday, eetimes had an article about 2000 employees (presumably mostly digital designers) being laid off at AMD. In this article a CTO complains that newly graduated engineers aren't properly trained in a technology popular 20 years ago that allows them to work at one of maybe 4 companies.
An engineering degree should give someone the fundamentals (linear system theory, electromagnetics, etc.), and she can learn about specific technologies on the job. However, companies want schools to produce technicians who they can dispose of when technology or business conditions change. Retraining isn't considered as an alternative.
It's no wonder that engineering isn't regarded as a true profession like law or medicine, despite being as mentally challenging (in my opinion).
Two orthogonal topics I see discussed here. One is that we need bipolar CMOS devices, to handle the increasing speeds of digital chips. The other is that engineers have to consider themselves as system designers, not just wigit designers.
The first point is probably well taken. There's no doubt that the more subject matter you have to cram into a 4 or 5 year EE degree, the more such older topics as bipolar transistor circuit design will be summarized. That's hard to avoid. You trust that engineers who will find themselves working in those areas will get up to speed on their own, very much like they always have to do with new, emerging trends.
Weren't we always taught that you go to school to learn HOW to learn? Anything you learn is school will be hopelessly outdated in 10 years anyway.
I don't buy the second complaint at all, though. This one:
"Your engineers are making, selling and supporting specific parts, but the very best will look at a customer's board and think how they can sweep up the components around their part into an integrated, higher-value device the design cycle. It's not a common talent."
Every individual course you take in school is clearly just to learn that one building block. It should be obvious to any student that ultimately he'll be putting these blocks together, to develop systems. Even if he might not get a chance to do so at school (which actually you often do get to do, as class projects at the very least).
I simply do not believe that this isn't common knowledge among engineers or engineering students.
Blog Doing Math in FPGAs Tom Burke 15 comments For a recent project, I explored doing "real" (that is, non-integer) math on a Spartan 3 FPGA. FPGAs, by their nature, do integer math. That is, there's no floating-point ...