Slashdot Mirror
Is Microprocessor/Controller Design Dead?
blanchae asks: "I work for a Canadian post-secondary institute and I have been scouring the web job sites, newspapers and newsgroups for career adds for microprocessor/controller based electronic designers at the technology level (2 years training). We are re-evaluating our curriculum and are looking at the job market as one way of warranting specialization training to existing programs. There's lots of career adds for embedded controller designers with University degrees but not a thing for technology level microprocessor/controller design. It is very puzzling. So the question is: Is microprocessor/controller design dead? Has it moved offshore? Is it off the radar and mainly in small upstart companies (5 to 25 employees) that hire word of mouth and not through the big corporate media methods?"
Maybe it's only mostly dead. Remember, there's a big difference between mostly dead and all dead. Mostly dead is slightly alive. With all dead, well, there's only one thing you can do....
You work for a Canadian post-secondary institute and it is very puzzling and you turn to Slashdot? God help our post-secondary institutes.
Don't you think 2 years is perhaps not enough time to have someone be competent at something as complex as microcontroller design? A 2-year degree is generally associated with technicians/technologists that are not hired for design work.
No. Actually the microprocessor or controller design business has at least a 20 years to survive according to Gordon Moore's projection (Google Moore's Law) and ITRS(http://public.itrs.net) projections. This is true for technology level designers too. You are probably mistaken by the fact that large companies do not seek only 4-year graduates. You must try Intel; TI, IBM or etc. You will see that the business still needs lots of people. I heard that there are 2 year technology graduates in TI from a friend who conducted an internship there. Don~t get discouraged it is just customary for them to put it that way. There are lots of opportunities for technology graduates in the industry. The thing is you should try big companies.
There's plenty of room at the bottom! Richard P. Feynmann
just far too hard for anyone with a two year degree (and for most people with bachelors degrees)
At the bare minimum, to be able design even a relatively simple chip you need the following classes:
1.5 years physics (mechanics, em/wave, and quantum)
3 years math (calc 1, calc 2, multivariable calc, diff eq, linear algebra, stats)
3 years electronics (intro to electronics, digital logic, basic design i.e. intro to hdl, analog signal processing, solid state devices, advanced design) 1 year CS (CSI/II)
Anyone capable of covering that much material, in addition to general school requirements, in two years destroyed their college admission exams and already has a good scholarship to a 4 year school (where they can get the degree in 2 years if they really want).
DeVry lied to you about your earning potential.
(There's others who've hired programmers and try very hard to keep them away from their embedded designs, since your typical CompSci grad thinks a MB of compiled code is compact!)
So, I'd suggest equipping your students with the sort of skills that will get them a foot in the door of companies doing embedded design, and suggest that they get a couple of cheap design kits for popular MCUs to play with in their own time.
Firstly, you're looking in exactly the wrong order. If you were looking for a research assistant, would you ask a student you know, or would you put an ad in the local college paper? Hopefully, you'd choose a student you know. You have a good idea of who they are, their work ethic, etc. And you won't have to somehow sort through the flood of applications you'd receive in a newspaper posting. So a newspaper is exactly the last place to look for most jobs, unless the employer is hoping to find the cheapest among several qualified applicatants. I don't believe Intel actually places ads in papers; some places prefer that you take interest in their company and seek them out instead. If you want to match your cirriculum to employers needs, I'd hope your "institute" has a few industry connections, since this is often a good avenue to your students actually getting a job. These are the people you need to talk to.
Secondly, 2 years training to design microprocessors? What exactly would they be doing, that only takes two years to go from high school education, to mastered enough to be productive? Programming microcontroller devices, maybe. Designing them in today's market takes a knowledge of what's been done in the past, and ways one might improve them. The industry is simply too competitive to accept the kinds of mistakes and inefficiencies a novice would make when multiplied by a large scale production run. A 4 year degree is a good start, nothing more. Many of the largest chip design places have internal education to address academic cirriculum shortcomings. These would also be good people to talk to.
Finally, what do you think qualifies as a distinction between a microcontroller and an embedded system? I'd say not much. 386's are being used more often now, in places where DOS or Linux can do far more than a PIC traditionally does. And if you're seeing so many postings for embedded systems, remember that a number of these projects are likely for US military applications, and non-US citizens, like Canadians, are usually unemployable in that field as a security precaution. If this still seems fruitful, why not adjust your cirriculum to match the demand you see right now?
I Browse at +4 Flamebait
Open Source Sysadmin
Ah, sweet BCIT. I worked on embedded systems development (Not exactly chip design mind you), and I delt with the challenge adequately. Another friend, same school, same company ended up being one of their most proficient developers. Just because you don't get the entirety of the education to be entry 'qualified' doesn't mean you're incapable of ever picking it up. Given the chance, many can perform quite well above their current educational level.
With that said, I think 2 years experience would be a challenging task. Then again, they don't necessarily hire entry level chip developers. They could start you off with more remedial jobs and make you work in-house a while.
The question really is, are there -any- chip manufacturers still around in the Americas? Well, there seems to be a few big-houses still around, and if you're really questioning to keep the program or not, why not ask these companies personally? Maybe you could even arange career seminars with soon-to-be graduates?
Bye!
In response to other postings, we do have industry contacts but you must appreciate that when an educational institute comes knocking asking for information, the priority on answering is way down on the list of things to do like emptying the garbage can...
I agree that word of mouth is a common method of finding suitable employees and that's what I did when I was looking to hire employees when I was in industry. The issue is how to track the "word of mouth" career offerings?
Slashdot is on the pulse of technology and seems like a quick and dirty method of acquiring data. How much value is put on the data is dependant on the quality and quantity.
It's not dofficult to work out what happened. I started out in this career path long ago, straight out of high school. Back then, I was designing Z-80 based computer systems... Later, I went on to MCS-48 and MCS-51 based designs as well as flirting with x86 and 68xx(x) architectures at times.
I was pretty good at it. My success ratio exceeded 90% throughout my career. And I was a one-man engineering lab... From design (including PLDs) through fabrication, prototyping and production following successful prototyping. Many of my successful projects were valued in the millions of dollars to the companies I worked for, back in the 80's.
But there's not many people with those sort of skills, so over time, employers couldn't get the skills cheaply and stopped advertising for those people. They turned to PCs to perform jobs that would normally be performed on micro's or found other ways to do things. Most design work became an offshoot to inhouse production teams and never really became a critical business component.
And when someone did have an engineer with those skills, they tended to undervalue them. I worked for many employers as a part of their churn. I replaced a cheap engineer and they weren't prepared to pay extra for the skills I brought to the position. Not all employers can see the value of someone who can design a modem out of three 20pin PALs or produce an engine management unit if all they want is someone to design their latest pinpad.
The lack of people who could cut perfect code in assembly language and manually route circuits more efficiently than the auto-routing algorythms of the day became less important as circuit design apps got better and processors got fast enough that high level languages could be used instead of low level languages.
So more people came in to the industry, but lacked the skills. Employers worked around it by asking less of them, but that diluted the products and so in turn diluted the value of such engineers to their employers.
Universities and technical education centres simply couldn't produce the skills in people coming up. This further diluted the available skills resources.
And no matter that you can get away with this 95% of the time, the other 5% of the time, you need the low-level skills. Otherwise your success ratio tend to drop below 50%.
It got to the point where the average wage earner made around $40K per year, and Electronic Engineers in my city (Major capital city) would average about $35K.
So I usually left after a while, chasing salary increases with other companies, when the ones I worked for didn't want to pay.. Until one day I realised I could make more for my family just by doing basic low-level tech work as a PC assembler. So I threw away my old skills and became another tech on the production line.
Everyone else I knew - people who designed their own home PCs from the chips up only ten years earlier did the same... they became miners, postmen, builders. The work was less stressful, less hours and paid better. Some stayed. The lucky ones found companies that looked after them. It was rare.
But now with only the diluted skills left in the marketplace, employers had a problem. I would speak to old employers who seemed suprised that their new projects were failing and no engineers were left. They wondered why it took a 386 processor and six weeks of C development to develop and debug a replacement to a keypad that cost $300 per unit to make that replaced one I designed for them from concept to prototype using $30 of parts with an MCS51 just six years earlier.
So Business got out of that industry also. No engineering skills means it's not viable business. So they got into PC software development or similar related industries and just dropped that line of revenue from their business model.
So, No new low-level skills, No engineers with the skills available to fix the problem and no positions because companies let this source of revenue die out.
And the industry disappeared.
Enjoy science fiction? "Turing Evolved" - AI, Mecha, Androids and rail-gun battles. What more could you want?
I licked quite a bit of the microcontroller-based embedded design, and from what I saw, only amateurs, and only most clueless of them use separate processor and controller. In the past it was making sense. Nowadays the market is saturated with microcontrollers that carry enormous amounts of extra hardware on chip and a hour with soldering iron spent on including a dedicated controller chip in the project can be easily avoided by a hour of browsing the catalogues for derivative that has that controller on-chip. Price increases are often negligible. Speeds are amazing.
5 : 75MHz 64M addressable, ethernet, 1w, spi, CAN, 3x RS232, 8x bidi 8-bit GPIO, IP stack plus UDP, TCP, DHCP, ICMP, TFTP, IGMP in ROM, Wake-On-Lan, watchdog, clocks, and God knows what more.
www.fairchildsemi.com/products/micro/ - SOIC-8 package, the size of an optocoupler - 8 pins, thingy would fit on the nail of your pinky, whole, with surface-mount pins. 64 bytes of RAM, 1-2K of program eprom, 64 bytes of data eprom, clocks, power monitoring, wake-up on any pin, 6 GPIO lines, eeprom writing, watchdog, serial output generator, sleep mode, idle mode, oscillator, and quite a few other goodies.
On the other end of the scale: http://www.maxim-ic.com/quick_view2.cfm/qv_pk/453
Add to that DSPs which are quite specific but achieve speeds higher than newest pentiums and athlons in their tasks (and often carry some "extra", add PC for heavyweight number-crunching and user interaction and you see:
Controllers are dead. Microcontroller is way better because it allows for just the same on the hardware side, while vastly simplifying the interface side. With your current knowledge you should catch up and learn microcontroller-based design pretty fast.
Anagram("United States of America") == "Dine out, taste a Mac, fries"
There's a lot of IC development done here in north america (and even here in vancouver). Admitidly a lot of the fabrication and rest is done overseas for cost and to be honest, poor environmental regulations. Chips are quite a concoction of nasty shit.
For some reason lots of people also forget fpga development. At SFU we got into FPGA and ISA/PCI design in our second year (although not anymore since they started babying the curriculum to "double the opportunity"). We still do our VLSI class in 4th year though. Desinging chips is a lot more than merely designing logic.
----
Go canucks, habs, and sens!
There are basically three categories of microprocessor design. And armed with an electronic degree, one can decide which area to focus on:
1. CUSTOM BOARD INTEGRATOR
3. PROGRAMMERS (both HW and Software)
3. THE MICROPROCESSOR DESIGNER
All areas entail different stages (and thus different skills). They basically cover requirements, designs, coding, integration, testing and maintenance. It is entirely possible to have a lifelong career in just within one of those stages, particularly test and maintenance.
Even so, each area utilize different skill sets.
1. Lowest man on the totem pole (but still well-paid) is the custom board integrator. Involves research and selection of hardware components using interchangeable interfaces (i.e., PCI, PCI-X, LVDS, Rocket I/O, VME and lesser known interfaces such as USB, Firewire, Parallel and serial). Testing of each HW components (not to get bad capacitors) are a non-trivial effort. Most low-budget company skimp these component testings. Nevertheless, it entails buildup around THE microprocessor.
2. Midway is the programmer. VHDL, Synoptic and many other custom hardware programming languages which tends to be chipset-specific. Bulk of the job market are in this category.
3. The elite is THE microprocessor designer. Intel, AMD, IBM, Motorola, Hitachi, Fujitsu, Xilinx and many others make uses of M-Designers. Most of them tend to be cultivated from within each company. Much research material have to be digested and assimulated to be able to design one of today's complexity. Best and easiest break into this arena is startup company, successful or not.
I suggest, for a startup university department, you shoot for #2 as the majority of your curriculums. This ensures that these skill-sets are transportable to either #1 or #3, depending on how good they grasp the elementary logics.
[finally, something on Slashdot that I can comment intelligently about]
I work for a small (6 EE, 10 ME, 20 Industrial designers) design firm. Small and large companies come to us for all kinds of design work, some of it is development or improvement of EE designs.
We're always doing some sort of microprocessor/controller design, as well as CPLD and FPGA programmable logic. Pretty much every job we do incorporates one or more micros. In the past three years, I've used PICs, TI's MSP430, Freescales MC9S12, Atmels and probably a couple more. Development is done on PCs, running something like a Metroworks IDE. Sometimes we use an embedded OS like uCos, sometimes not.
At least from where I sit, microprocessors are still very much relevant. I'm currently working on an embedded controller for a mechanical system -- two motors, limit switches, temperature sensors and two serial communication ports to other controllers not built by us. There's analog and digital interface design, the micro is a Freescale 9S12, power supply regulators and more. Lots of fun!
I wonder about this, and worry a little bit. These companies certainly need people with those skills, so ... would society benefit from a return to some form of indentured servitude? Perhaps if companies had protected their image over the past 30 years instead of letting hotshot MBA's slit their cash cows' throats and ride it into the ground, screwing all their customers in the process, then the stable companies could be trusted to provide a lifelong career for someone who chooses to learn these skills. I think that in the current environment, there aren't enough jobs to entice someone to get the necessary training. Turnover in skilled disciplines -- from both the employer's and employee's sides -- is way too high to justify the kind of dedication it takes to learn to do these things well.
It would be nice to have enough faith in the long-term plans of a company that, eg, when IBM or Ford Motor Group needs someone who can do this, an employee could be sent to school for 2-4 years with a reasonable expectation of some long-term benefits.
Lately, it seems like you need 10-15 years of experience just to be an asset rather than a liability in some fields. So why would an employer hire a college graduate for a reasonable salary, when the chances are next-to-nothing that this person will work for them long enough to contribute to the company? And who will guarantee that some new MBA won't fire him for some stupid reason? I once lost a job because some middle-manager decided that being "late" to work was defined as punching in more than 3 minutes after your scheduled time, and if you were late more than 8 times a year, you should be fired.
Back to the core of the topic: It's the question of Freddy Fastfingers, the coder who can churn out functional code super-fast, but for every hour of his work, the company invests 2-3 hours of manpower fixing, explaining, or otherwise ameliorating the effects of solveable flaws in his code. Does he even deserve to have a job? Probably not. The question is, is it reasonable for the company to nurse his career for 10 years until he's learned his way around in his field, or should they find a way to do his job with less-skilled labor, using tools that (while overpriced and underperforming) aren't filled with amateurish, glaring bugs?
Employers can't trust Employees to stick with their company, and Employees can't trust Employers not to fire them. It's a vicious cycle, and it's destroyed much of what made this country a leader in high technology in the first place.
True science means that when you re-evaluate the evidence, you re-evaluate your faith.
There's a lot of these chips selling, so somebody must be buying them. Have you tried putting your question to the local offices of the chip companies?
here's thoughts:
1) companies advertise for 4 year engineers and higher, they simply don't see the need to advertise for 2 year certificate level applicants. you need to train your people in networking to get around that. also provide placement service.
2) microcontroller work is good experience, and there is a lot out there, but my company works with SOAC - system on a chip - level embedded systems. these are complete, very high performance, large computer systems targeted to DSP applications, in my case video compression. it is far more sophisticated than a microcontroller system. yet the skills of working with software to controll interrupt controlled on chip I/O devices are the key.
3) hardware design is handled by highly skilled designers who are working with GHz signals and very high density components. the need for 2 year level applicants is in software.
4) it will be crucial for your 2 year level applicants to be well versed in the basic vocabulary of Computer Science (data structures and algorithms) and well experienced in embedded software. as far as I am concerned, experience designing hardware around a microcontroller is excellent way to increase software savy.
all in all, I don't think the hardware experience is the sell, except insofar as it bolsters the software skills.
I'm a recent college grad and I took the embedded course they offered here as my main design course. We worked with Atmel AT91 (ARM) using eCos as the embedded OS. In the course we not only worked on the microprocessor architecture, but also interfaced it with hardware that we designed and built ourselves (not just your run of the mill "read the ambient temperature and do something about it" projects, but stuff like building signal boosters, RF controllers, ethernet controllers, etc). Needless to say, it was a fairly taxing course. I really enjoyed the class and took a lot out of it so much that I decided to make it my career path. Companies that I have interviewed with loved the experience I got out of the course saying most other candidates having taken some embedded course never go as far as making the onboard LEDs blink in some special pattern or displaying text on an LCD. The issue I found with looking for the right embedded job was it seemed like being proficient in design was merely a secondary qualification. Most companies working with embedded systems expect candidates to be specialized, if not at least familiar, with specific technologies. Look at Motorola's job postings. They not only expect some embedded experience, but also experience with stuff like TDMI, RF, or sometimes even TCP/IP. Unfortunately I didn't land my ideal embedded systems design job. In fact, the company I now work with hired me because of my embedded experience, but the system they're using that was embedded-based is being converted to be run from a full-scale computer system. I can't say for sure whether or not companies are moving away from microprocessors, but as with any industry, demands for certain resources (in this case, processing power/capability) are increasing. Microprocessors may one day not be sufficient for large-scale projects.
I'm general manager for an embedded design house.
:)
Dead? What? In a day and age when everything around you has a uC of some sort in it? Now are a lot of those consumer products being designed in Elbonia? Of course, but still, there's a lot going on. This year I've worked on projects using small (say PIC like) microcontrollers in:
house arrest system
sports watch
in store kiosk (touch screen controller, the brain of the kiosk is an embedded x86 running Linux)
cargo container status device
Of course we've also used embedded x86/pentium class machines, ARM, Coldfire...
There are huge volumes of microcontrollers shipped from 4-bit (just like those obsolete mainframes making obsolete money for obsolete banks) to 32-bit and up. Not to mention DSPs. But I think the days of super simple designs, say replacing a mechanical timer on a washer, are long gone. Sure you may be replacing a mechanical timer, but you've got to understand so much more, like being able to run on nanowatts or making the device intrinsically safe.
Unless somebody has a lot of talent and a lot of experience, we hire the two-year types as bench techs. Four years for an engineering position. Things are so fast and competitive that the engineer also has to be something of a project manager. A strong understanding of hardware, firmware, and software is important. Yes, I think there's a huge difference between the mindset needed to write good firmware and good software. To write good firmware you've got to understand what's going on in the hardware. Plus you've got to be able to understand that yes sir it is indeed possible to toggle a single bit without doing an operation on a whole register!
The people who claim "ah, it's dying because for only a few dollars/cents/credits more you can have a whole embedded super-whoopAMDIntelMot128bit blah blah" don't understand capitalism. 1) if it costs more, um, it costs more, and somebody will do it cheaper, this we call competition, and blast it whether or not Adam Smith is right it happens, and 2) if I want to monitor a real world event or spin a motor or some such I can do it with a PIC/AVR/8051 faster than you can get your makefile working right just the core of your OS. Time is money, at least in my business. The right tool for the job. And sometimes a $0.35 PIC is the right tool. But because of the competition and current needs in the field, the project that $0.35 PIC goes in to probably needs to be designed by someone with more than two years of tech school. The code may be easy, but understanding enough about battery chemistry, thermal concerns, and other deeper issues means that making the whole thing run for three years from one set of batteries while on a ship somewhere in the Pacific...you see what I mean, it's no longer just "hey designer write some code to spin a motor".
It seems that a lot of what we do are projects that are either "extreme" one way or another -- environmental, regulatory, power/life requirements -- or are a small piece of a much larger and more complex system. Again, something that requires a higher level of thought and understanding.
We've seen more and more applications for the smaller and cheaper super tiny 6 and 8 pin micros. It's absolutely mindboggling how many places these things end up, and low margin but high volume works. Moore's law not only means more transistors on the same amount of real estate...it also means you can get the same number of transistors as you did before for much much cheaper.