21 Sep
2008
21 Sep
'08
9:12 a.m.
On 9/21/08 5:24 PM, "Alexis" <thrashbarg at kaput.homeunix.org
<http://www.classiccmp.org/mailman/listinfo/cctalk> > wrote:
On Sun, 21 Sep 2008 09:37:51 pm Andrew Lynch wrote:
things
I have a working prototype on my bench and can
read and write sectors and
format tracks.
I never worked out how to format tracks on the 8272. It's one of the that puts me off using them. Could you describe how
it's done?
The other thing that puts me off is the data clock separator. How are you
doing this?
Look at the code for the CompuPro Disk 1 interface. I don't have to code for
my CBIOS handy but I recall that you have to establish a 6-byte command
buffer (with the right parameters), establish the interleave (if formatting
with one) and then loop through all of the tracks sequentially, sending the
format command for each track. I used a hardware interleave so that my CBIOS
didn't have to do it.
The command buffer for my CP/M system looks like this. The fields come right
from the 8272 manual.
FMTDAT .DB F_FMT ; format command
.DB 00H ;HDS,DS1,DS0
.DB NFIELD ;N (0, 1, 2, 3) 0=128-byte sectors
.DB FSC ;Sector count (26,15,8) =1ah
.DB 01BH ;SPECIAL FORMAT GPL.
.DB 0E5H ;D (FILLER BYTE)
Rich
--
Rich Cini
Collector of Classic Computers
Build Master and lead engineer, Altair32 Emulator
http://www.altair32.com
http://www.classiccmp.org/cini
------REPLY------
Hi Alexis, Rich,
Thanks Rich! Yes, Rich has the right command sequence but you also need to
send the C, H, R, N to the i8272 for each sector during the execution phase.
It is not very intuitive but I got it working last night on my test bench
and confirmed it is working with Dave Dunfield's IMD program (Thanks Dave!)
and the Catweasel with CW2DMK (Thanks Tim!). For some odd reason it also
requires a leading $00 to format correctly which I don't understand either.
Basically, it is like writing a sector but you feed it sector metadata as it
goes around the track formatting sectors. Note, during execution phase it
is not like feeding the MSR/data registers like during the command/result
phase. The whole NEC765/i8272 process is kind of screwy IMO and I am
struggling to get my brain around it. I have sample code if you'd like to
see it but it is an awful mess right now.
As for the data separator, I am using an FDC9229 although you could use a
FDC9239 I think. The chips are still available but not exactly common. You
have to look but you can find them. I have some pointers if you'd like.
Actually, Alexis, you'd be the perfect guy for building your own Z80 home
brew FDC as you already have done most of this on your other projects. You
did an awesome job on your home brew 8080 a while back and your other
projects have been great too. Your experience would be a lot of help to the
project in putting out a design that not only works but is reliable and
flexible.
I'd like to support as many hobbyists as possible so I am trying to preserve
the hardware "hooks" to support 8" drives, etc. Due to the complexity of
this project it could really use another set of eyes to check the design.
There are many subtle issues with FDCs that require real hardware
experience.
One final note, the i8272 data sheet for the FORMAT TRACK command is rather
misleading or at least incomplete IMO. I prefer the SMC FDC765 datasheet
found on the BitSavers.org site (Thanks Al!). It also has data sheets for
hard to find units like the FDC9216, FDC9229, FDC9239, etc. Warning: 45MB
FILE!
http://bitsavers.org/pdf/standardMicrosystems/_dataBooks/1985_StandardMicros
ystems.pdf
Thanks and have a nice day!
Andrew Lynch
21 Sep
21 Sep
6:52 p.m.
On Sun, 21 Sep 2008 10:42:46 pm Andrew Lynch wrote:
Hi Alexis, Rich,
As for the data separator, I am using an FDC9229 although you could use a
FDC9239 I think. The chips are still available but not exactly common.
You have to look but you can find them. I have some pointers if you'd
like.
I'd like to see if the data separator on the FDC-1 for the SYM-1 would work on
the 765. The FDC-1 uses a FD179x controller and it appears it uses the same
data clock and raw data inputs as the 765. I've built the FDC-1 for my SYM-1
last week and it works a treat, in both FM and MFM modes and on 720k 3.5"
disks. I used an SMC FDC1797 from an old word processor. The board it came
off had a 9229 on it so I think there's a good chance the circuit will work.
The decoder is made up of a 74LS197 (clock divider), a dual 74LS74 flip-flop,
both used, a 74LS163 counter and an inverter. There are also some open
collector NAND gates to select the clock rate for either 8" or 5.25" drives.
It'll use more individual IC packages, but they're *much* easier to find.
The circuits are, of course, at 6502.org:
http://6502.org/users/dallas/SYM/SYMDOS2/fdc_sheet1.pdf
http://6502.org/users/dallas/SYM/SYMDOS2/fdc_sheet2.pdf
Thanks for that link to the SMC datasheet too. I'll give it a good looking
over when it's downloaded. I was never 100% clear on the format procedure
given in the Intel datasheet, so hopefully the SMC one will be a little
better!
Thanks,
Alexis.
9:02 p.m.
Alexis wrote:
The FDC-1 uses a FD179x controller and it appears it
uses the same
data clock and raw data inputs as the 765.
Not quite. I don't recall the details of the difference, but I think it
may have just been the polarity of one of the signals. The 9229 and
9239 had a configuration input to select between 179x and 765 modes.
The 9216 would directly connect to one, and needed a small amount of
logic (maybe just an inverter) for the other.
The decoder is made up of a 74LS197 (clock divider), a
dual 74LS74 flip-flop,
both used, a 74LS163 counter and an inverter. There are also some open
collector NAND gates to select the clock rate for either 8" or 5.25" drives.
It'll use more individual IC packages, but they're *much* easier to find.
Sure, but it doesn't sound like it's a very good data separator. The
good ones have a PLL (either analog or digital), because it is necessary
to track speed variations, not just of the drive that you're using to
read a disk, but also of the drive that wrote it. Non-PLL data
separators work OK when the disk is both written and read under optimal
conditions, but are unreliable otherwise.
The 9216, 9229 or 9239 are *highly* recommended, as they have a good
digital data separator. The 9229 and 9239 also contain write precomp
logic. The 9239 uses higher resolution timing for its PLL, so it may
perform better.
7:21 p.m.
------REPLY------
Hi Alexis, Rich,
Thanks Rich! Yes, Rich has the right command sequence but you also need to
send the C, H, R, N to the i8272 for each sector during the execution phase.
It is not very intuitive but I got it working last night on my test bench
and confirmed it is working with Dave Dunfield's IMD program (Thanks Dave!)
and the Catweasel with CW2DMK (Thanks Tim!). For some odd reason it also
requires a leading $00 to format correctly which I don't understand either.
Basically, it is like writing a sector but you feed it sector metadata as it
goes around the track formatting sectors. Note, during execution phase it
is not like feeding the MSR/data registers like during the command/result
phase. The whole NEC765/i8272 process is kind of screwy IMO and I am
struggling to get my brain around it. I have sample code if you'd like to
see it but it is an awful mess right now.
As for the data separator, I am using an FDC9229 although you could use a
FDC9239 I think. The chips are still available but not exactly common. You
have to look but you can find them. I have some pointers if you'd like.
Actually, Alexis, you'd be the perfect guy for building your own Z80 home
brew FDC as you already have done most of this on your other projects. You
did an awesome job on your home brew 8080 a while back and your other
projects have been great too. Your experience would be a lot of help to the
project in putting out a design that not only works but is reliable and
flexible.
I could post the entire format code if you want but there is about 50 lines
of code dedicated to properly formatting the CHRN data for the 8272. I
actually grossly oversimplified the process for the sake of brevity, and I
agree that it's needlessly complex. OTOH, the sector read/write is fairly
simple for sequential sectors -- like when loading CP/M. You just tell it
the starting sector data and the number of "additional" sectors to read and
let it loose. Works like a charm.
Rich
--
Rich Cini
Collector of Classic Computers
Build Master and lead engineer, Altair32 Emulator
http://www.altair32.com
http://www.classiccmp.org/cini
5783
days inactive
5784
days old
3 comments
4 participants
participants (4)
-
eric＠brouhaha.com -
lynchaj＠yahoo.com -
rcini＠optonline.net -
thrashbarg＠kaput.homeunix.org