<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>
Date   : Sat, 18 Dec 2010 20:58:50 +1100
From   : awilliams@... (Alan Williams)
Subject: Harddisc fakery using a microcontroller - back on

Rick said:

Excuse me - but if you use an SD card (instead of, say, a flash chip),
can't you just say "write this sector here"? It *ought* to be optimised
to write a sequence of sectors, but we don't have that luxury.

Alan said:
Yes that's exactly how you do it but only once!
Like the old UVEPROMS that are in your Beeb, you can write anywhere, but
only once.  In order to reprogram (a 0 to a 1 to be precise) you have to
erase with UV light.  The Flash rom am29F010 (I think) which was
discussed here a month or so ago can be erased in 16K chunks, just fine
for a BBC ROM Image. 

The SD card is perfectly happy to let you write sectors at random, but
once done you have to arrange an erase in order to do it again.  The
erase blocks are typically a reasonable number of sectors long, 8, 16,
32 etc  This is why its painful.

Rick Said:
> for use with a Filestore rather than a BBC that's not all that
possible.

Indeed, that's why I'm looking at a dirt-simple SCSI->SD hack. It seems
the easiest way, as opposed to major firmware patches. It'd be different
if we had the sources, but so much of the disassembly is "unknown
function" and suchlike. What is safe to yank out? What isn't? Can we
even rearrange the code? There's blanks in EPROM - does this serve a
purpose in live mode? Well, it must as firmware and operational memory
are all together in the RAM. But what/where/how? There's just so many
questions that a different (hardware) approach seems simpler.

Alan said:
Agreed, except that the semantics of SD card operation are making it
look a lot less than simple.

Rick said:
If I ever get my A5000 going again, I might try wiring up the parallel
and I/O podule user ports to be a SCSI interface, and see if I can put
together some code to make it work. How this goes should tell me if an
AVR is viable, basically, as glue logic, and what needs to be done to
achieve this.

Alan said:
Look for an Atomwide parallel to scsi interface.  Or a Trantor PC one
and port the inevitable Linux drivers.

Rick said:
Just a damn shame the user port podule doesn't fit into the RiscPC.

Alan said:
yep.  Pity there are bugs in 6522 used in most of them which stop them
working in a A540 too.
<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>