Fast forward to 2002, I met this fellow who had ordered and built a Microbox, and still had it lying around. I immediately bought it from him.
|
|
|
It works!
Back in the good old days, there was a little shop called the Electronics Supermarket just around the corner from the Greenmarket Square. They sold all kinds of nifty stuph I couldn't afford, I remember they had an AIM 65 on display.
Anyway, I bought an ASCII keyboard from them, model 756C. It's made by Carter Keyboards, here are the docs I got with it Page 1, 2, 3 and 4. This is functionally identical to the George Risk Industries datasheet I pulled off the web. I suspect they (GRI) just rubbed the badging off a bit and flogged the thing as their own.
These keyboards (well, the KR2376 encoder chip) need -12V in addition to the +5V supply and they do have a site for a DC/DC converter. I removed an ethernet DC/DC converter and hacked it onto that site, it wasn't pin-compatible. The Microbox provides -12V on the keyboard connector, I suspect the 756 or keyboards based on the same chip was pretty much standard all over the world for a decade at least.
The schematic from Carter (or equivalent) is somewhat lacking. It doesn't show the power pins, for example -- +5V on pin 1, GND on pin 17, -12V on pin 18. Here's the KR-2376 datasheet, it's not a huge improvement -- power consumption can be up to 150mW if it gets cold but no mention as to how much of that's on 5V and how much is on -12V. Not that it matters, it's peanuts either way.
Note that the keyboard matrix is shown in the documentation is not correct for the KR-2376-150 that I have on my 756 keyboard.
Sebastian Kienzl's interface does all the hard work in a PIC. Or you can do it the hard way.
|
After much deliberation I decided to build my Microbox II into a
Hammond Sloped Front Desktop Enclosure,
RS
Components 868-5312. At 431 x 280 x 89mm, the
Microbox PCB fits very nicely, with room for a small power supply
and of course the 756 keyboard.
|
|
Starting with the MON09 V4.5 sources from the FLEX library disk, I modified things to work with as09 under DOS (DOSBox). The resulting binary is identical to the one on the library disk.
Alternatively you can use FLEXemu and ASEMB on the original disk image (which you will need to pipe through FloppyMaint to get FLEXemu to read it). It's slower than as09 though. Also take note that this creates a FLEX binary which is not binary, it consists of "0x02 address length data" records.
There's also LWTOOLS which is much faster under windows, but the binary it creates differs from the as09 / ASEMB binary. Specifically:
A7 A4 STA 0,Y (as09 / ASEMB) A7 20 STA 0,Y (LWTOOLS)These are equivalent, the former is no offset, Y register direct and the latter is 5-bit offset, Y register direct, offset = 0.
Caveat: if you try to get LWTOOLS to emit a binary you get the bytes in source file order, i.e. the stuff that's supposed to go at $F000 will be at the start of your binary and the $E000 stuff will be in the middle. Make it spit out S-Records and convert that to a binary.
I should maybe also mention asm6809, I tried it but it expects a different input file format. All comments need a ; for example. I think it can be made to work with a bit of effort.
I needed a way to get my code transferred to the Microbox. Because of the way MON09 works, in that it reads two characters, parses them and passes control to the code to handle that command (as opposed to for example MVME101BUG, which takes a two character command + CR) I could add a command "S1" which then reads the rest of the S-Record line and stores it in memory.
In what used to be COMTABLE:
; WRM 2022-01-15 Add S1
FCC 'S1'
FDB S1 ; At end of $E000-$EFFF block
and after "SB":
S1 JSR BYTE ; Get one byte in A (S Record length) (preserves B)
DECA ; Payload is three bytes less than record length
DECA ; because of the address
DECA ; and checksum
PSHS A
JSR BADDR ; Get two byte destination address in X (Destroys A and B)
PULS B ; Count now in B
S1LOOP JSR BYTE ; Get byte (preserves B)
STA ,X+ ; store it
DECB
BNE S1LOOP
JSR BYTE ; Get checksum (and ignore)
JMP CONTRL
This being the 21st century, the printer is hanging off CAT5 on the other side of the house. So I can use what used to be the printer port to interface to a "20481 / HD44780" type 4x20 LCD in 4-wire mode, and maybe hang a few LEDs off the remaining pins.
This still being the 21st century, I decided to go with a single 3 1/2" floppy. And that's just because I want to, it's easier to use a GoTek.
Now modern 3 1/2& floppy drives support HD, i.e. 1.44 Meg disks. These run at twice the clock rate (using FM, the data rate is half the clock rate, using MFM they're the same). So you can do 125kbps FM (clock is 250kHz) or 250kbps MFM (clock is still 250kHz) using DD disks (no hole) and 250kbps FM or 500kbps MFM (clock is 500kHz) using HD disks (with the hole opposite the write protect hole). The hole tells the drive what the data rate is (The 5 1/4& world is much less simple. You need to tell the drive what kind of disk to expect, and the disk runs at 300rpm for SD/DD and at 360rpm for HD. OR at 360rpm all the time but then you need to feed data to it at 300kbps in DD mode).
Anyway, neither the 1770 or the 6809 can do 500kbps, so let's stick to the point. 300rpm and 125kbps gives you (60/300) * 125 = 25 kilobits = 3125 bytes per track (a track is one side, a cylinder is both sides) in FM mode and 6250 bytes per track in MFM. These are raw bytes, they get wrapped in track and sector headers and footers and you get to use what's left.
Page 17 of the WD1770 datasheet shows what it expects a track to look like. But one would want compatibility with other FDCs as well. So I decided to reverse-engineer some existing code. This proved to be a bit of a mistake. The code claims to do one thing and then end up doing another, leading to confusion on my part. Either the authors had an unintentional bug which just so happened to make things work, or they were lazy and left dead-end code, well-commented dead-end code, in there to confuse people like me. The code builds a sector header and then overwrites most of it with the first track header. The WD1770 doesn't need a sector header but the code is there, leading me to think that the header is needed for other controllers maybe, except as mentioned it then gets overwritten anyway.
I've documented what the double sided newdisk utility for the Microbox (NEWDISK2.TXT from library disk 49 does if you look at the code quickly, as well as what it actually does.
And as an aside: the aim of all of this is to update the MON-09 built-in, 40-track, single-density format routine to 80 tracks double-density (I am not going to bother with single-density. There might be someone out there using 3 1/2" disks in single-density mode, sucks to be them).
Single Density
WD1770 Datasheet | NEWDISK2 | NEWDISK2 | MON09 |
Format | What it says | What it does | |
No. of Hex value | No. of Hex value | No. of Hex value | No. of Hex value |
bytes to write | bytes to write | bytes to write | bytes to write | Info
----------------------+----------------------+----------------------+----------------------+--------------------------------
40 FF | 30 FF | | | Post index Gap (Gaps 1+2)
| 6 00 | | 6 00 | Pll lock-up
| 1 FC | | 1 FC | Index address mark
| | | 7 FF | MON09-specific gap
/ 6 00 | / 6 00 | / 6 00 | / 6 00 | Pll lock-up
| 1 FE | | 1 FE | | 1 FE | | 1 FE | ID address mark
| 1 | | 1 | | 1 | | 1 | Track number
| 1 | | 1 00 | | 1 00 | | 1 00 | Side number
| 1 | | 1 | | 1 | | 1 | Sector number
| 1 | | 1 01 | | 1 01 | | 1 01 | Sector length
| 1 F7 | | 1 F7 | | 1 F7 | | 1 F7 | Writes 2 CRC bytes
| 11 FF | | 11 FF | | 11 FF | | 11 FF | Sector ID / Data Gap (Gap 3a)
| 6 00 | | 6 00 | | 6 00 | | 6 00 | Pll lock-up (Gap 3b)
| 1 FB | | 1 FB | | 1 FB | | 1 FB | Data address mark
| | | 2 xx | | 2 xx | | 2 xx | Track/Sector link (FLEX)
| 128/256/512/... | | 254 FE | | 254 FE | | 254 00 | Data
| 1 F7 | | 1 F7 | | 1 F7 | | 1 F7 | Writes 2 CRC bytes
\ 10 FF | \ 12 FF | \ 12 FF | \ 14 FF | Post data Gap (Gap 4)
369 approx FF | FF | FF | 256 FF | Runout Gap. Write until
| | | | next index hole.
40 + n(31+data+12) | 37 + 10*(31+256+14) | 10*(31+256+14) | 14+10*(31+256+16) | Bytes per track
| 3047 | 3010 | 3044 | With 10 x 256 byte sectors
| 78 | 115 | 81 | Bytes slack
Double Density
WD1770 Datasheet | NEWDISK2 | NEWDISK2 |
Format | What it says | What it does |
No. of Hex value | No. of Hex value | No. of Hex value |
bytes to write | bytes to write | bytes to write | Info
----------------------+----------------------+----------------------+--------------------------------
60 4E | 60 4E | | Post-index Gap (Gaps 1+2)
| 12 00 | | Pll lock-up
| 3 F6 | | 3 synchronisation bytes
| 1 FC | | Index address mark
/ 12 00 | / 12 00 | / 12 00 | Pll lock-up
| 3 F5 | | 3 F5 | | 3 F5 | 3 synchronisation bytes
| 1 FE | | 1 FE | | 1 FE | ID address mark
| 1 | | 1 | | 1 | Track number
| 1 | | 1 | | 1 | Side number
| 1 | | 1 | | 1 | Sector number
| 1 | | 1 01 | | 1 01 | Sector length
| 1 F7 | | 1 F7 | | 1 F7 | Writes 2 CRC bytes
| 22 4E | | 22 4E | | 22 4E | Sector ID / Data Gap (Gap 3a)
| 12 00 | | 12 00 | | 12 00 | /Write splice time (Gap 3b)
| | | | | | \Pll lock-up time
| 3 F5 | | 3 F5 | | 3 F5 | 3 synchronisation bytes
| 1 FB | | 1 FB | | 1 FB | Data address mark
| | | 2 xx | | 2 xx | Track/Sector link (FLEX)
| n | | 254 FE | | 254 FE | Data
| 1 F7 | | 1 F7 | | 1 F7 | Writes 2 CRC bytes
\ 24 4E | \ 24 4E | \ 24 4E | Post Data Gap (Gap 4)
668 approx 4E | 4E | 4E | Runout Gap. Write until
| | | next index hole.
70 + n(60+data+26) | 76 + 18*(60+256+26) | 18*(60+256+26) | Bytes per track
| 6232 | 6156 | With 18 x 256 byte sectors
| 18 | 94 | Bytes slack
The 18 bytes slack in what NEWDISK2 seemed to do in DD mode is what then started me investigating. Two FUFU list members sent me NEWDISK5 and NEWDISK3 and from that I finally twigged. The codebase exists to be used with 8", 5 1/4" and other disk formats. You just tweak the defines at the beginning of the code, the code then clears the workspace, sets up a track header, and based on your defines overwrites the track header again. When they ported newdisk to the Microbox, they removed the now-redundant comments but not the now-redundant code.
Another funny I noticed is that the side is set to 0 for both sides of SD disks. The reason for this is
That's what I would call an SWTPC "quirk". On the DC-4 controller the 1797 side select output is connected via an inverter to the DDEN pin. So commands with side=0 select single density and those with side=1 select double density. Since you cannot stop a 1797 from checking for the correct side, the side bytes in the sector address fields must be 0 on both sides for single density and 1 on both sides for double density. On the DC-4 the floppy disk side is selected by one of the bits in the drive select register.
OK, so now I feel I know enough to write my own newdisk.
Finding 1: You need a post-index gap. Trying to emulate what NEWDISK2 does did not work for me (in DD, didn't test SD). I added the 60 x 4E (but not the 00 F6 FC sequence) and that seems to work.
Finding 2: NEWDISK2 writes the buffer out, testing for DRQ but not for BUSY. If the drive rotates too fast, BUSY will drop, DRQ won't happen, and the system hangs. MON09 tests for BUSY, but in DD mode this only works at high (2MHz) speed.
Write loop from NEWDISK2. This works at 1MHz, double-density.
LDB #2
LDX #WORK
L0583 BITB <WD1770COM
BEQ L0583 ; S1 = DRQ = 0
LDA ,X+
STA <WD1770DAT
CMPX TEMP1 ; End of buffer
BNE L0583
Write loop from MON09. This works at 2MHz, double-density, but fails at 1MHz.
LDX #WORK
L0583 LDA <WD1770COM
BITA #$01 ; S0 = BUSY
BEQ L0590 ; BUSY = 0, exit
BITA #$02 ; S1 = DRQ
BEQ L0583 ; DRQ = 0, loop
LDA ,X+
STA <WD1770DAT
CMPX TEMP1
BNE L0583
L0590 ...
Over the years Dave Dunfield wrote stacks of good code and released a substantial amount of it into the public domain. His Runtime Aid for Interactive Debugging can be found as part of the N8VEM CUBIX release.
I removed the command-line code at the beginning and changed GETCHR, RDCON and WRCON to use MON-09 calls.
Raid> D C052,C071
C052 17 03 4E ..N LBSR $C3A3
C055 52 R FCB
C056 61 a FCB
C057 69 64 id ROL $04,S
C059 3E > FCB
C05A 20 00 . BRA $C05C
C05C 17 03 5B ..[ LBSR $C3BA
C05F 17 03 67 ..g LBSR $C3C9
C062 8E C0 7B ..{ LDX #$C07B
C065 A1 80 .. CMPA ,X+
C067 27 0E '. BEQ $C077
C069 30 02 0. LEAX $02,X
C06B 6D 84 m. TST ,X
C06D 26 F6 &. BNE $C065
C06F 17 03 31 ..1 LBSR $C3A3
Raid>
Nice! A mini-disassembler like my Apple ][ has.
hits since 2020-10-06.