At 10:00 AM 12/13/2008, Olin Lathrop wrote:
>Xiaofan Chen wrote:
> >>> Even for the more broad SMPS supply world
> >>> actually designing an SMPS is becoming less and less necessary, in
> >>> many cases
> >>> just use the online tools to select which chip is best, and follow
> >>> the reference design.
> >>
> >> For some level or performance, cost, efficiency, and board space
> >> this is true, but not for all.
> >
> > Unfortunately the performance, cost, efficiency, and board space
> > are often issues involved in my world.
>
>Exactly.  It seems every power supply is different in the range of input
>voltages, output voltage, amount of ripple tolerated, output current,
>inefficiency tolerated, and of course the usual tradeoffs between cost,
>board space, cooling requirements, standby power, etc, etc, etc.  Then
>startup is a whole other issue that off the shelf simple PWM chips are
>usually bad at or tricky to make work right over a wide input voltage range.
>Too many casual SMPS designers aren't even aware there are special issues
>there.
>
>Anybody that thinks they don't really need to know about switching power
>supplies to plunk one on the board from the app note doesn't belong there
>and is headed for trouble.  The guy who designed the original product for a
>local startup company had this attitude.  He grabbed a "simple switcher"
>from National (great marketing, but their switchers aren't more or less
>simple than others) and copied the circuit from some app note.  Both the
>inductor and FET were always hot, usually just about where you couldn't
>touch them for more than a second, even though the whole unit only took
>about 5W.  I was told that when people pointed this out to him he would only
>respond that the circuit was right out of the app note and it therefore
>couldn't be wrong.  He was fired and I was brought in.

As an engineer who has been involved with switching power supplies 
for 35 years (they are very different now!) I can only agree. I spend 
as much time designing the SM regulators on a PowerPC based CPU board 
as I do in the entire PowerPC part, which is what I am doing right now.

One thing that does annoy me somewhat is the lack of decent 
simulation models for many switchers. Linear Tech do a good job, but 
National, as an example, do a really lousy job. But I end up being 
conflicted because all too often the National part is much more 
economical but the lT Part has a good model.

There is far more deep theory in SM power supply design than any 
other common aspect of modern electronics. I work on everything from 
1V @ 1W or less, to 150kV at 100kW every day. So any applicant for a 
job in my group that says "oh switchers are simple" doesn't tend to last long!

John 

-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist