5 Jan
2012
5 Jan
'12
8:06 p.m.
On 4 Jan 2012 at 22:27, allison wrote:
As can a state machine. You change the progrtammaing of the state ROM (or
whatever you're using).
My choice was a 2816(2k EEPROM) , 74374(D latch)
and 555(osc). State
machine trumps CPU for dumb task. Took an hour to explain it, likely
more to teach him to program a EEprom.
Nowadays, I'm not so sure. An 8-pin MCU is dirt-cheap and can be
easily adjusted to change the algorithm. I think it was ED that had an article not too long ago that asked if an MCU or a
555 was better
for new projects.
The question came down to the need for programming of the MCU, so the
555 still had an edge.
Ther;'es also the issue of reliability. It's true that a microcontroller
is more reliable than the same thing built out of simple logic chips --
if you need all the functionailty of the microcontroller. But my
experience suggests that more complex ICs are noticeably less reliable
than simple ones, and that a boaud of 3 or 4 TTL packages can well be
more reliable an the same thing done with just a single-chip microcontroller.
In my case there;s no constast. I can have the 555 version built and
running before I've even fired up a compute rto urn the MCU tools.
My last application that I would have normally solved with a couple
of one-shots was solved with an 8-pin PIC--I could design the thing
with a few more smarts and self-calibration.
I'd probably hav eused a state machine, but then I don;t know what the
problem was.
So choices aren't always so clear-cut, particularly in today's cheap
silicon days. It's already coming down to which is cheaper--an FPGA
or an MCU.
I am rapidly gettign the feeling that just as few people seem to be able
ot program any more, very few people can design hardware any more....
I remmeebr a conversation I had some yeas ago THegroup was deisnging a
device which involved some high-speed analiguye and ECL circuity and thus
had ot go on a 16 layer, arefull-made (read : mega-expensive) PCB. There
was one bit of the desing that handn't been completed, It came down to
producing a logic signal from 3 other logic signals (I think all F TTL).
Thing is, they wanteto get the PCB off to the manufcaturer ASAP, without
finising that part of the desigbn.
I glanced at the problem and said 'That's trivial. I can do it in one
IC'. One of the others's replied 'A PROM? A PAL? are they fast enough? Do
we have the programmer for them'. 'No', I said, 'Just a plain TTL chip
with no programming needed'.
OK, what was the chip I suggested? (if did get used, BTW, and worked
prefectly).
-tony