Jeffrey Chee said:
After joining the work force, I find that those analogue
stuff, circuits analysis stuff are not as useful or important as I
thought. Is that true?
For a very large number of electrical engineering students, yes, this is
true. As Kevin already mentioned, there's far more need for people who can
write software, do a little digital design, etc. than there is for people
who can do 'real' analog design. In fact, I'd even venture that a lot of
'engineering' jobs could be done by modestly talented individuals with far
less than a four (or five!) year degree, just as successfully. In the U.S.
these days it's almost taken for granted that everyone is 'supposed' to go
to college... so colleges oblige and fill up the years with classes that may
or may not be relevant to what 90+% of their students will end up needing in
industry.
I see many things are done by programming. MCU, DSP, FPGA and Visual
Basic, C++. I see some people don't really border much about circuits
and mostly don't really know much about it, or not border to go
further.
Keep in mind that there are some very challenging problems within each of
those areas -- digital signal processing can become quite involved,
especially if you're the guy implementing low level subroutines and have to
understand what's actually going on. FPGAs are often a prime candidate for
very high speed signal processing, and of course high-level software is used
to tie it all together (see Ray Andraka's web site,
http://www.andraka.com/,
for a good example of some very impressive DSP work in FPGAs). Although
many people programming in VB or VC++ are just implementing starightfotward
applications, there certainly is the person who gets to implement programs
that handle, e.g., machine vision (more signal processing!), high end
databases (some of those guys who wrote MS SQL Server are pretty sharp),
circuit simulators (there's of course Kevin -- and Nagel et al. who started
SPICE way back when at Berkeley), electromagnetics simulators (as complex as
you'd like it to be), and even just compilers (highly optimizing compilers
are tricky to get right).
I'd venture that a lot of the people who are tackling those more challenging
problems with software could have done pretty well as circuits designers
too -- it's just that there's far more 'interesting' problems to solve than
anyone can do in one lifetime, coupled with the reality of where one can
find or create a job.
I am interested in circuits and electrical signal because I think it
is the most basic element that connects. But looking at the industry,
more functions can only implemented by having Know How on MCU and
other programmable stuff. And that doesn't really need very deep
understanding on circuits, as the supplier and vendor could provide
help, application notes and others. So the value that we can add is
programming, I assume.
Is that true? What do you think and do you have any doubts as I do?
Here's a little secret that isn't advertised too prominently: A lot of the
guys who used to connect up all those discrete circuits at the board level
became chip designers. There are tons of people at National Semiconductor,
Analog Devices, Maxim, etc. who never touch a soldering iron, yet they're
very much doing analog circuit design.
As far as advice goes -- I think it's important to sit around with discrete
circuits and connect them up and see how they behave. While it's still on a
breadboard that you can poke around at, it provides more confidence that
everything is really working the way it's supposed to (at least for me it
does... I built this huge breadboard once of a large digital circuit design,
got the thing to work, and since then have had the confidence to never
breadboarded another digital circuit -- I just got straight to laying out
the PCB). There's a certain amount of faith involved in doing a chip design
that's only ever been simulated in SPICE and spending the tens of thousands
of dollars to have it made and hoping that it'll work the same way in
reality as it did in the simulator. This is more faith than is required
with digital designs in FPGAs or standard cell ASICs -- if your VHDL or
Verilog or whatever simulates correctly there and meets timing, you're about
99.99% assured the chip will work, since some ANALOG GUY sat around and
designed and characterized all the gates within the FPGA to come up with the
timing models.
'Board-level' analog design is a small niche -- and it's becoming smaller.
For any application where the expected volume is large, someone (an analog
person!) will design a chip to do the job (e.g., look at all the single
chip... radios, audio amplifiers, micronctrollers with gobs of I/O and
analog linterfacing, network interface ICs, etc. -- devices that used to
require tons of discrete components). My impression is that many of the
board-level analog designers hanging out in this newsgroup tend to work at
Universities, labs, or other research centers where novel, high-performance
analog designs are necessary that you can't buy off the shelf -- but the
expected volume for the product is low. These guys are worth their weight
in gold, I imagine, since I think it's easier to take a really good
board-level analog guy and make him a chip designer than it is the other way
around -- especially for items such as RF circuitry, where suddenly
distributed effects become very significant!
And as long as I'm on my soapbox here (aka, 'sure to annoy everyone')...
I've often felt that the decline in interest in designing and building your
own amateur radio devices is partially due to the fact that it is so
difficult to compete with on-chip solutions with board-level design. There
are many exceptions to see (see
http://www.elecraft.com/k2_page.htm for a
beautiful product); I'm just speaking generally. I'm confident that a lot
of the RF designers in the amateur radio community could have designed their
own cell phone-like systems (yes, I know about autopathces, but this isn't
as sophisticated), IEEE 802.11b wireless networking systems (packet radio
doesn't cut it), etc. years ago, yet it tended not to happen because the
costs involved in a board-level design were prohibitive compared to the cost
of an on-chip design (...there aren't many ICs in a wireless network card,
nor a cell phone any more!).
Well... that's my opinion... now I need to get back to work... which for
today is... MCU programming.
If you're interested in analog design,
however, by all means pursue it -- there is plenty of industry demand for
good analog designers, and even always have the option to fall back to
software or digital design if you can't immediately find anything in analog
you'd like to do.
---Joel Kolstad
