On 11/15/2011 9:00 PM, ben wrote:
When I tried FPGA's I tended to design the
hardware first.
Simulation does not help, when the design does not fit in the FPGA
you are using.
True, but your compilation tools _will_ tell you that.
I think you're missing the point. The problems start when part of the
design will not (and isn't suppoed to) fit in any FPGA (or even any FPGA
supported by the tools you're using).
Perhaps you want to have a significant amount of memory. Perhaps so much
that you cna't get an FGPA that will hold the logic _and the memory_. No
problem in the real hardware, you conenct a memory chip to the FPGA. But
a lot of FPGA_specific simulators can't handle that.
It gets worse if the external memory is somplex to talk to, like many
forms of DRAM. Trying to simulte a DRAM _inside an FPGA_ and have it fial
i nthe right ways if the timing is wrong, just so you can simulate the
DRAM controller you were designing in the first place is fraught with
problems.
Then there's the time you wnat to link your PFGA circuit to some external
devies. Suppose you're designing a disk controller for a PDP11, anf you
want it to use DMA and interrupys. I hardlly think you should try ot
desing a UUnibus or Qbvus arbitor in the simulator just so you can test
your deisng. And yet many, many, simulators can't handle tie idea of 'If
you get this signal form my circuit, then assert this one 100ns later,
then wait for my cirucit to assert this...).
Or if you have an external device that accepts and produces a bitstream
of some kind. Simulating that is always 'fun'.
Somebody already mentioned genrating video signals in an FGPA. I don';t
know about you, but I can't look at the waveforms going into the DACs
(which is waht a ismulator is most likely to show) and see what the image
will look like.
Of course anythign weith ADCs or DACs is goign to be a problem.
-tony