28 Jan
2006
28 Jan
'06
4:51 a.m.
so by your reckoning not all of the asics are readily
reproduceable. What about the gate array on the
graphics board (lets pretend there wasnt a primitive
ttl prototype to guide)? What about the 7220 itself?
Now something like that is bound to have loads of
docs.
I think there's enough published data on the 7220, and enough working
samples around to be able to design a replacement without copying the
existing IC. It would be a lot of work, sure, the 7220 is not a simple
chip.
As regards the other ASICs in a QX10, the problem is that some of them
contain considerable analogue circuitry, and certainly couldn't be
replaced by an FPGA (IIRC the mains chopper transistor is in one of the
PSU hybrids). But I thinkit would be possible to work out a replacement.
The disk drive ASICs are harder in that I have no idea what the exact
functions of them is. That is something I intend to look at fairly soon,
actually. See what the pins do during a seek, etc.
In general there are 4 classes to consider :
1) Well-documented, simple. For example a 555 timer. It should be
possible to make an exact functional replacement from the docs alone.
2) Undocumented, simple. For example a programmed PAL. Simple enough that
given a working example of the chip and maybe the schematic it's used in,
you could work out a series of tests to identify the internal logic and
make some other device that would do the same thing.
3) Documented, complicated. For example the Z80. Again you'd not need to
actually inspect the chip. Thre are Z80-a-likes around (e.g. as programs
for FPGAs) that seem to have been designed based on the behaviour of a
real Z80.
4) Undocumetned, complicated. A large ASIC, ULA, etc. Since yoy've almost
certainly got hidden state, you can't always work out what's going on
from outside. At this point you may well have to
consdier removing the
encapsulation.
-tony