that's all relevant of course, if it's a big
production house, producing thousand or millions of
boards. But the outifits that make these retro-fits
(LOL LOL no pun intended) are done in a garage more
then likely. So that being said, just use individual
ic's. Or perhaps it's alot about protecting their
investment. But I would also have to say a
considerable amount of development time goes into
designing something that way.
Why doesn't someone just draw up PLANS to build this
stuff, and sell that? I'd buy it for sure (well, if it
was for something groovy I owned...).
--- Ethan Dicks <ethan.dicks at gmail.com> wrote:
On 12/8/06, Chris M <chrism3667 at yahoo.com>
wrote:
--- Lee Davison <lee at geekdot.com>
wrote:
> Two 8 bit buffers, two 8 bit latches and one
GAL20V8
> or similar.
> The GAL isn't essential but saves four or more
> packages to do the address decoding.
Note the "or more" here... it's not hard to imagine
reasonable
situations where one GAL saves 6 or 8 popcorn TTL
parts, especially
where you start using latched outputs.
So 10 discrete ic's. Why all the fuss then?
And
why
bother with the GAL at all if it only saves 4
chips?
I really have to wonder why these developers go
to
all the trouble they do if that's all it
takes (I
would have guessed at least 20).
There are two major reasons why a hardware developer
would use a GAL...
1) board space/interconnections are "expensive"
When one is hand-wiring a project, replacing 4 16-20
pin parts with a
single 18 or 20 pin part, that's a significant
savings of effort.
Equally if one is trying to fit in a small space, 4
DIPs is a huge
amount of room these days (and there are PLCC GALs,
allowing a lot of
DIP circuity to fit in one cm^2).
2) GALs are easy to change
Practically speaking, folks don't want to rewire a
board to change an
address setting, so you end up with address
comparators, jumpers, etc.
With a GAL, for infrequent changes (once or twice
over the life of
the product, not once or twice per year), you can
arrange the logic so
that there are no jumpers, but the GAL still does
internal address
compares, etc. The Spare Time Gizmos Elf2000 does
this - if you don't
like what devices are selected by which combination
of N-lines, etc.,
take the (provided) GAL source, change it, recompile
it with WinCUPL,
then burn a GAL to your liking. They are re-usable,
so you don't even
have to buy a spare unless you'd like to do a quick
swap back.
The downside, of course, is that as the
hobbyist-end-user, you are
somewhat out in the cold if you don't own a GAL
programmer. They can
easily run to hundreds of dollars for basic ones,
and, unlike an old
4K EPROM, they are not trivial to make programmers
for from scratch.
I have an older GAL programmer, so I don't mind
GAL-based designs.
The only thing that's eluded me lately is
programming "C rev" Lattice
GAL22V10s. I've had no problems with burning older
Lattice GALs or
other sizes of new Lattice GALs, so it's been an
inconvenience, but
not a show stopper. OTOH, I'm willing to look
around for a newer
programmer and consider an upgrade. I'm not going
to give up GALs
entirely, so I'll find some solution when the need
arises. Others
(those not already posessing GAL programmers) might
find the jump to
be too expensive to justify.
-ethan
____________________________________________________________________________________
Yahoo! Music Unlimited
Access over 1 million songs.