On 12/04/2010 07:23 PM, Andrew Lynch wrote:
SCSI to IDE
Alexandre Souza - Listas pu1bzz.listas at
gmail.com
<mailto:cctalk%40classiccmp.org?Subject=Re%3A%20SCSI%20to%20IDE&In-Reply-To=
%3C04B9CB3C0CBB4169A8660B24014CB7FB%40portajara%3E>
Sat Dec 4 14:50:26 CST 2010
[snip]
Time to define DOWN TO EARTH specs and begin working :)
Greetings from Brazil,
Alexandre SOuza
[snip]
Alexandre,
The work on a SCSI to IDE and SD bridge has already started. I have a
schematic, parts list, and PCB layout ready for prototype board up on the
N8VEM wiki. Check out the thread on
vintage-computer.com forum. It has the
Z80 controller, Z53C80, and IDE plus the SD interface. An 8 MHz Z80 will
push the data plenty fast enough for early microcomputers. We've seen it
first hand with XT-IDE board, the N8VEM DiskIO board, and the S-100 IDE
interface. Connect an IDE to CF adapter and it will be a fast drive on most
any vintage system.
Low cost commercially available SCSI to IDE bridges are already available
for those with Ultra-SCSI and later so there is no need to help them. I
think the need is for the older SCSI format which *classic* computers use,
not the fancy modern stuff. An Ultra-SCSI interface does no good for an
early microcomputer with a SCSI-1 interface which I think should be the
"classic computer" audience since they need a SCSI to IDE bridge the most.
If memory serves the real problem is wide scsi and narrow scsi don't mix.
The fast and slow scsi bus parts work fine if it's fast on a slow host but
slow on a fast host was problematic. However, most are trying to solve
the slow (relatively) host to anything and that can be fast so long as it's
not wide. But wait we are not done as the host may be limited in the
drive sizes (and expected ID string) they will work with.
I know one example of that, AmproLB+ the bios supplied only worked
with two scsi bridges (only with 10mb drives) or a oddball Sugart
ST125N or some such. anything else required a new BIOS and rewriting
drive install/initialize tool. MY SB180 with the SCSI adaptor board
had the oddball requirement that the target drive/adaptor had to have
256byte sectors (not as common). Some SCSI bridge boards did that
but not all and I've not found a drive with SCSI on it that could.
This means a SCSI to anything board has to have the ability to:
A) look like a known drive (supply drive ID and all).
B) allow partitioning the IDE (or whatever) to match the
replacement drive and identify as multiple LUNs aif
needed/required.
C) match sector size of obbball systems (since IDE and most
CF and SD tech are 512byte sector) that means blocking
and deblocking are required.
D) Operate in CHS or LBA as needed for sector addressing
on the SCSI side.
E) Depending on the size/age of the IDE drive also do CHS of LBA
Translations. (CF and SD are LBA).
This sounds more practical. Parallel IDE can easily do 33mB/S
and faster. SD from actual use is not fast but cheap.
The limiting factor is the Z80 with no dma as the best you can
get to with that is maybe 200KB/s and this is as good or better than
the old SCSI to MFM cards.
The mailing list is called "CCTALK" for a
reason. That's what's pretty much
the only thing done here. I agree with your sentiment -- I to would like to
see less "talk" and more "do". Don't listen to the naysayers or
those who
set unrealistically high expectations but refuse to actually do anything
except complain. I've been offering to help design a KiCAD schematic and/or
PCB board and/or prototype PCBs based on *anyones* design since the start of
this thread and have only gotten flamed and ignored. Maybe you'll have
better luck.
One of the pet issues I have is hearing I want... fast... cheap.
The rules are the same in 1960 as now, Good, fast, cheap
Pick two only. Fast is rarely cheap. In this case fast complicates
the parts used, bus layout and even impacts software.
PS, if you are planning to offer a PCB or kit, I
advise you not use SMT or
you'll be assembling nearly all of them yourself. A lot of hobbyists
especially those with diminished vision struggle with even DIP/PLCC
soldering and SMT parts will just make you the assembler as well as the
designer.
It's worse than that. Even with through hole parts, people hire out
K1 and K2 (ham transceivers) kit assembly! Those that can pay do
and those that can't and are unskilled, just want. SMT parts
are not that hard and the vision thing is a red herring as thats what
magnifiers are for. I've built lots of SMT stuff and my reading
glasses are +2.0 and getting worse. Not a problem as a good
+5 magnifier makes it happen. I do down to 0403 and TQFP to
64pins. Those require good soldering tools.
Me, this is NOT a project I might be interested in as I have a boat
load of SCSI drives and only a few machines that use SCSI
(all of them are microVAXes). So I likely have a lifetime supply.
Allison