26 Jun
2012
26 Jun
'12
1:44 p.m.
On Jun 26, 2012, at 1:37 PM, Chuck Guzis wrote:
On 26 Jun 2012 at 12:49, Dave McGuire wrote:
I might think a PIC32 would be a bit more efficient (depending on the
ARM used); don't most MIPS architectures have a bit more horespower
than most ARM architectures? Mileage may vary, of course.
In any case, having thought on it a bit, it would probably be fairly
simple to do the actual data streaming with a very cheap FPGA (you
don't need much onboard RAM to store a ring buffer of bits for the
current cylinder) and a tiny micro for housekeeping and loading
data between FPGA and bulk storage. Store a cylinder back in when
the step command comes, perhaps?
With such a setup, you could get away with a very low-performance
micro. The FPGA would likely run about $15 in single quantities.
If anyone is interested in working on such a solution, I may have
time to dedicate to it soon, but I have no hardware which talks
MFM or RLL, nor working drives to compare to. Still, one can do
lots without the actual devices.
- Dave
I would think that a bit of analog hardware
could generate the
signals to mimic flux transitions. At that point, the microcontroller's only
high-speed-requiring job would be generating the encoding, which isn't
tough for a fast one.
No analog needed at all (at least not in the pure sense)--the signals
are RS422-differential type that requires some transceivers not
usually part of a modern microcontroller. The control signals are OC
TTL levels, just like a floppy.
Using an 8MHz AVR to generate floppy MFM worked just fine, so an
80MHz PIC32 or ARM should be up to the job--I'm using the former,
simply because it's a bit easier.