14 Nov
2011
14 Nov
'11
8:12 p.m.
On Nov 14, 2011, at 4:05 PM, Tony Duell wrote:
Sure, I've had that too. A lot of the problem is that the FPGA synthesizers try to
infer what you mean in the HDL as registers and lookup tables, and sometimes your code
varies from their inference template just enough that you get the wrong hardware.
It's a pain, about as much as a compiler generating incorrect machine code (see the
m88k gcc saga). And of course the simulators live in a world full of binary signals and
finite time granularity where two things can actually happen at the exact same instant,
which of course FPGAs do not. There's more of an art to relating simulation results
to the real world than a lot of people suspect.
Note that availability is not my problem; it's cost. I could scrounge them from dead
boards, I guess, but a) I hate cannibalizing what could be working hardware unless it was
total garbage in the first place and b) I loathe desoldering. It's the worst, even
with the nicest solder suckers and braid and flux I can find.
True. And in quantity, that wouldn't be too bad, pricewise. I might see about
cooking up a few dozen and offering them up (to get volume discounts), if anyone has
suggestions for models.
- Dave
Eh? I _loathe_ simulation. I find nothing enjoyable
about 'designing' on
a computer at all.
For almost everything I've ever designed or reparied I just use my brain.
I have nop problems understnading hopw such things work. WHen I need a
bit of assistance, I use a pocket calcualtor (seriously [1]). I've never
needed anythhing more.
Different strokes for different folks, I suppose. For the stuff I can solder together
(like my 68K board), I like doing it on paper. Simulating it in HDL just feels like
cheating. :-) However, for most large-scale FPGA stuff, after I've
"simulated" the smaller blocks on paper I tend to take it to HDL and thus HDL
simulation as a next step to make sure I didn't make any dumb mistakes (and I often
do).
The bit I enjoy is actually buidign the hardware.
'Getting my hands
dirty' if you like. SOldering in the chips, wiring them up, etc. And as I
said before, I'll do what I enjoy when it's for my hobby.
I enjoy that as well, but I'm primarily limited by budget (and space) these days.
Once we've moved into a larger place, it might be a different matter.
bigger
headache than simulation, because unlike simulation and a board
full of TTL chips, you can't go in and probe specific nodes without
rebuilding the design.
That is one thing I dislike about FPGAs, certainly. On the other hand,
I've had problems iwth (expensive commeraical) simualtrs that either fail
to show glitches when they do exist, show glitches where the can't
possibly exist, or generally have problems. That said, I
do FPGAs for a living and I enjoy working on them from a
hobbyist perspective, but largely because I don't have a huge library of
TTL parts at my disposal (and it's expensive to acquire them all at
It's amazing what you can still easily get. Alas (foer me) many of the
more obscrue and intersting ones are no logner manufactured, but ther are
plenty about. And it's aamzing what you can do with just gates and
D-types, both of whcih are easy to get new. The
availability of nice FPGA eval boards (like the Terasic ones,
which I absolutely love) make it attractive to me. I'm still working on
my 68000 SBC, but it's mostly in concept only until I can find a decent
source of wire wrap boards (the kind with traces, not just bare
phenolic) around here.
I guess you couldalways design the sort of board you need and eitehr etch
it or get a PCB house to make it for you. And then wire-wrap on top of
that.