ROM variants and third-party ROMs
The machine
Status: Solved as the ROM-identity authority. Every image this project holds is
identified by CRC32 over its raw bytes plus size, cross-checked against MAME’s
driver, and tools/gen_rom_table.py re-derives every CRC in the tables below from
the bytes on every build — a document/bytes disagreement fails it. What is not
settled is interpretation rather than identity: four stock revisions (BIOS A, 1.2,
1.2.1, 1.3) have never been dumped, so a machine fitted with one cannot be
identified at all; one third-party label may be wrong; and one image’s
host-independence is unconfirmed. Those are the open questions at the end.
The answer. The machine has two ROM sockets — UD11, the 2732 monitor, and UA15, the 2716 character generator — and “the Osborne 1 ROM” is a family, because both were replaced or re-programmed by the upgrade vendors. What we hold is:
- three dumped stock monitor revisions — 1.4, 1.43 and 1.44 — plus a rev-B
image built by Osborne themselves (
OCCROM0B.COMon the OCC source disk, a vendor build artifact rather than a chip dump); - two character generators that really are different typefaces — 93 of 128 glyphs differ — and five names for one of them;
- third-party replacements: OZROM 1E (IEEE-488 removed, video cut to 3 KiB),
Nuevo 1.51, and the SWP co-processor’s 8088 firmware; and an 8 KiB
daughterboard image holding OCC 1.44 and OZROM 1E with
A12selecting; - four stock revisions we do not have: BIOS A, 1.2, 1.2.1 and 1.3 — MAME
carries them as
NO_DUMP.
Those CRCs are what the personality gate in firmware/rom-ident/ (#86) is built
on, together with the byte-level properties in Personality-relevant properties
below.
Issues: #70 (the inventory), #44 (the two character generators), #86 (the
personality gate built on it), #72 (the ROM source), #205 (the 1.3 reconstruction,
which is upstream’s gain and not ours), #222 (which NO_DUMP slot the rev-B image
belongs in — MAME’s call), #74 (the two images whose source is still unrecorded).
Vocabulary:
rev Bhere is the Osborne 1’s monitor ROM (9/12/81) — a 1979-era ROM part, not a board revision. Board revisions belong to the boards; see [hardware/README.md](https://github.com/Myrcurial/o1next/blob/main/hardware/README.md).
The two sockets, and what a socket tap can read:
| Socket | Package | Size | Role |
|---|---|---|---|
| UD11 | 2732 (24-pin) | 4 KiB | Monitor / BIOS — the Z80 code that boots the machine |
| UA15 | 2716 (24-pin) | 2 KiB | Character generator (the 6×9 font the video logic reads) |
Method
Every image is identified by CRC32 over the raw bytes plus size. CRC32 is what MAME’s ROM definitions and the archive indices both publish, so it is the only key that cross-references cleanly. Two independent sources agreeing on a CRC is treated as positive identification; a CRC agreeing under two different names is treated as a finding, not an error.
Stock monitor ROM (UD11, 4 KiB)
| Rev | Our copy | CRC32 | MAME name | MAME BIOS slot |
|---|---|---|---|---|
| A (1981) | — | — | osba.bin |
vera — NO_DUMP |
| 1.2 | — | — | osb12.bin |
ver12 — NO_DUMP |
| 1.2.1 | — | — | osb121.bin |
ver121 — NO_DUMP |
| 1.3 | — (source extracted: research/source-code/ROM-13-source/) |
— | osb13.bin |
ver13 — NO_DUMP; reconstructible in principle from the source disk research/source-code/OSROM13.IMD, blocked on ACT 3.5E — docs/rom-13-reconstruction.md (#205) |
| 1.4 | tools/emulator-setup/roms/rev1.40.ud11 |
3d966335 |
rev1.40.ud11 |
ver14 |
| 1.43 | research/roms/os1-143.rom |
91a48e3c |
rev1.43.ud11 |
ver143 |
| 1.44 | research/roms/os1-144.rom |
c0596b14 |
3a10082-00rev-e.ud11 |
ver144 |
A rev-B image, from the OCC source disk (#205). research/source-code/OCCROM.IMD
carries OCCROM0B.COM — a vendor-built 4096-byte monitor ROM, rev B, dated
9/12/81 — together with its source (occrom0b.asm plus nine modules) and its
full listing (OCCROM0B.PRN). CRC32 f5f697c7, SHA1
212b7992db91225909f47972799eeb878d9ed806. It is not a chip dump: it is
Osborne’s own build output, shipped on their own source disk. It is consistent
with both the source and the listing on that disk, and it boots under
mame osborne1, printing OSBORNE SYSTEM ONE.M / REV B 227 and
INSERT DISK IN DRIVE ... AND PRESS RETURN.
Its CRC is not one of the three known dumps, and rev B (9/12/81) sits between
BIOS A and 1.3 — so it is a candidate for the osba.bin / osb12.bin /
osb121.bin NO_DUMP slots, and which one is MAME’s call, not ours. That
question is filed upstream as #222. Two cautions: it is a vendor build artifact
rather than a dump, and under MAME’s osborne1 video model the banner shows two
stray Ls — rev B emits ESC+letter video escapes where 1.3/1.44 use single
control bytes, and the model does not decode the older form (recorded, not
explained). Extract it with tools/research/o1imd8.py + cpmcp; see
docs/rom-13-reconstruction.md §6.1.
Corroboration: BrettHallen’s ROM/OCC1_v140.BIN, OCC1_v143.BIN and
OCC1_v144.BIN are byte-identical to the three we hold (same CRCs), and
research/source-code/BIOS-144-source/rom144.asm is the original ACT80 source
for 1.44 (#72).
Two facts about this table:
- MAME’s “ver144” is the part
3a10082-00Rev E, not a file named “1.44”. The revision and the part number are two names for one image. - BIOS A, 1.2, 1.2.1 and 1.3 have never been dumped (MAME carries them as
NO_DUMPplaceholders). We hold the ROM 1.3 source disk, so 1.3 is potentially reconstructible — but not for us: we build from the 1.44 source, which is in-repo and now proven to reassemble to the shipped part (#203). Reconstructing 1.3 would fill an upstream gap, so it is filed as a side quest, #205, and nothing in this project depends on it.
Character generator (UA15, 2 KiB)
There are two distinct images in circulation, and which one a machine has tracks the BIOS revision:
| Image | CRC32 | Paired with | Our copies |
|---|---|---|---|
| Early | 5297c109 |
BIOS A … 1.4 | tools/emulator-setup/roms/char.ua15 |
| Late | 6c1eab0d |
BIOS 1.43, 1.44, and every ScreenPac / Nuevo machine | research/roms/os1-vid8.rom, research/roms/os1vid80.rom, research/roms/OCC1_ScreenPac_RevA.BIN |
MAME’s driver carries the early image five times, once per BIOS slot 0–4, with the comment “this is CHRROM from v1.4 BIOS MB”, and the late image for slots 5 and 6 (1.43 / 1.44). That is the evidence for the pairing, and it is the “early font / late font” distinction that #44 asks about — both variants are captured in this repo.
The socket itself (UA15, on mainboard sheet 4 of 9 — see
o1-mainboard-schematic.md §4a) is wired 2732-style: pin 21 is strapped to
+5 V, and /CE (18) and /OE (20) are grounded. Both variants are 2 KiB 2716s,
so either drops straight in with no strapping change, and whichever is fitted
is what a socket tap reads.
How the font is stored
The layout is not glyph-contiguous, and it is not obvious from the dump — it
took MAME’s driver to settle it (src/mame/osborne/osborne1.cpp):
u8 const gfx = m_p_chargen[(ra << 7) | (chr & 0x7f)];
so the byte for scanline ra of character chr is at rom[ra * 128 + chr]:
all 128 characters’ row 0 are contiguous, then all of row 1, and so on. Ten rows
× 128 characters = 1280 bytes, and the remaining 768 bytes of the 2 KiB are
0xFF. Each row is one byte, bit 7 = leftmost pixel, and the cell is 8×10.
(Reading it as 128 glyphs of 16 bytes, or as 8 bytes per glyph, produces plausible-looking noise and no recognisable letters; MAME’s driver had it in one line.)
That layout matters to the video personality (#24): a socket tap reads the font
as rom[ra * 128 + chr], and the Nuevo ROM adds an underline by forcing row 9
to 0xFF when bit 7 of the character code is set.
The two fonts really are different — 93 glyphs of 128
Comparing the two images glyph by glyph (tools/romfont.py decodes both):
| Differing characters | |
|---|---|
| 0x00–0x1F graphics set | 0 of 32 — byte-identical |
| 0x20–0x7E printable ASCII | 90 of 95 |
| 0x7F and the rest | 3 |
Only three printable characters are shared: space, the apostrophe and the backtick. So this is not a tweak — the late font is a different typeface, and the graphics set was carried over unchanged. That settles the open question below, and it means “which font is fitted” is a visible difference, not a subtle one.
Finding: one late image, five names
The bytes 6c1eab0d appear under five different names across three sources:
| Source | Name |
|---|---|
| Maslin’s archive | os1-vid8.rom |
| Maslin’s archive | os1vid80.rom |
| BrettHallen | OCC1_char.BIN |
| BrettHallen | OCC1_ScreenPac_RevA.BIN (“Screen Pac BIOS Rev. A”) |
| MAME | 7a3007-00.ud15 and character_generator_6-29-84.14 |
So the stock late character generator, the ScreenPac’s 2 KiB ROM and the
Nuevo-era date-stamped character generator are the same 2 KiB image. The
ScreenPac’s 80/104-column capability therefore does not live in this ROM, and
MAME models the ScreenPac as a clone of osborne1 with no ROM differences at
all (osborne1sp is a COMP of osborne1, while osborne1nv is a separate
machine with its own ROMs). That matches the picture that the Nuevo 80-column
implementation is a different design from the OCC one.
Third-party ROMs
| ROM | Size | CRC32 | Our copy | What it is |
|---|---|---|---|---|
| Micro Management OZROM 1E | 4 KiB | ba1f6131 |
research/roms/OZROM-1E/OZROM_1E.BIN |
Third-party replacement monitor ROM, © 1984 Micro Management. Strings: OZROM 1E #00187, (c) 1984 Micro Management, Slip disk in drive and press RETURN. Manual in the same folder. |
| Nuevo Electronics Rev 1.51 | 4 KiB | 298da402 |
tools/emulator-setup/roms/monrom-rev1.51-12.ud11 |
Nuevo’s replacement monitor ROM. Strings: OSBORNE 1 COMPUTER, ROM Rev 1.51, COPYRIGHT 1983, OSBORNE COMPUTER CORP., COPYRIGHT 1984, NUEVO ELECTRONICS CORP. — i.e. an OCC ROM with Nuevo’s DD/video patches. MAME name: monrom-rev1.51-12.ud11. |
| SWP co-processor | 4 KiB | 948556db |
research/roms/swp-p88.rom |
The SWP co-processor board’s own ROM. This is 8088 code, not Z80 (8C C8 8E D8 = MOV AX,CS / MOV DS,AX), and it carries the WD179x disk-BIOS error table (Clock Error, Late DMA, ID CRC Error, … Type C to CANCEL, R to RETRY, I to IGNORE error). Maslin’s index describes it as “SWP Plus-88 card for Kaypro 2/4/10”. The board and interposer card are common to every host machine and only the driver varies, so this is the 8088-side firmware of the CoPower-88 family — #3/#4/#5. |
| OCC 1.44 + OZROM combined | 8 KiB | edc31846 |
not copied | BrettHallen’s 2764 image for the 2732→2764 daughterboard, holding OCC 1.44 and OZROM 1E with A12 selecting between them. Recorded here for completeness. |
Personality-relevant properties
Some interposer personalities are only compatible with some of these ROMs, so the firmware needs to know more than the CRC. The facts below are what the personality gate is built on (#86), each recorded with its evidence — the manual text or the bytes at the vector table.
| ROM | IEEE-488 driver | Video window | Evidence |
|---|---|---|---|
| OCC 1.4, 1.43 | present | 4 KiB (F000-FFFF) |
013F-0156 holds eight JP vectors (c3 36 09 …), and the printer entry points 0187-0190 are JPs. Byte-identical between 1.4 and 1.43 |
| OCC 1.44 | present | 4 KiB | Same structure with shifted targets (c3 49 09 …), and the implementation is in source we hold: rom144.asm ;IEEE-488 INTERFACE. line 2606 (o1-rom-source.md) |
| OZROM 1E | removed | 3 KiB (F000-FBFF) |
Manual (OZROM_1E_Manual.ocr.txt), verbatim: “The OZROM 1E does not support the IEEE-488 functions”; “013F-0156 WERE IEEE-488 ROUTINES / CALLING THESE FORMER JUMPS RESULTS IN A CONTROLLED SYSTEM CRASH (NO DAMAGE TO DISK DATA)”; “IEEE-488 AS IOBYTE DEVICE 3 NOW NULL DEVICE”. Confirmed in the bytes: 013F-0156 is ef f3 00 f7 … — not jumps — and 0187-0190 is f6 ff c9 f6 ff c9 af c9 c9 c9 = OR 0FFH; RET ×2 then XOR A; RET; RET, exactly the documented printer stubs. Video: “only 3K of RAM is used for the video (F000-FBFF hex) since the last 1K is used for the redefinable keyboard and extra function keys” |
| Nuevo 1.51 | present, routines unverified | unknown | The vector block is intact (013F-0156 = eight JPs, 0187-0190 = JPs), so Nuevo kept the entry points — but nothing shows the routines behind them are unpatched. Not usable for Drive C until read (#86) |
| OCC 1.44 + OZROM (8 KiB) | depends on A12 | — | Daughterboard image; which half is live is a hardware strap, so a probe cannot tell from the CRC alone |
Three consequences, all of which the gate encodes:
- Drive C (#6, #22, #84) and the virtual printer (#19) both need the IEEE-488
driver, so both are refused under OZROM 1E. The printer is the less obvious
of the two: the ROM’s printer entry points
0187-0190are precisely what OZROM stubs out. - The video personality (#24) must cope with a 3 KiB window under OZROM 1E,
where
FC00-FFFFbelongs to OZROM’s keyboard features. - An unknown ROM must not be assumed good. BIOS A/1.2/1.2.1/1.3 have never been dumped, so a machine fitted with one is unidentifiable — the gate refuses rather than guesses.
MAME cross-reference
MAME carries three Osborne 1 machines; the relationship between them is itself evidence about the upgrade boards:
| MAME machine | Year | Parent | ROM set |
|---|---|---|---|
osborne1 |
1981 | — | BIOS A/1.2/1.2.1/1.3 (NO_DUMP) + 1.4 + 1.43 + 1.44, chargen early + late |
osborne1sp “Osborne-1 with SCREEN-PAC” |
1983 | osborne1 |
no ROM differences — a clone, so the ScreenPac is modelled as hardware only |
osborne1nv “Osborne-1 (Nuevo Video)” |
1984 | osborne1 |
monrom-rev1.51-12.ud11 + 7a3007-00.ud15 + character_generator_6-29-84.14 — its own BIOS and chargen |
tools/check_o1_map.sh already selects a driver from the boot image’s
52-/80-/104- prefix (osborne1 / osborne1nv / osborne1sp), so this
table is also the map from video mode to ROM set.
Open questions, and the one that is closed
- Do the two character generators actually differ in the font? Yes, and
substantially. 93 of 128 glyphs differ, and the 32 graphics characters are
byte-identical (see “The two fonts really are different” above). Only space,
apostrophe and backtick are shared.
tools/romfont.pydecodes either image to a font, so the comparison can be repeated. - What is BrettHallen’s “Screen Pac BIOS Rev. A”? It is byte-identical to
the stock late character generator, so the label may be wrong — or the
ScreenPac genuinely ships an unmodified char-gen and does its 80-column work
entirely in hardware. Settle against the ScreenPac install and service pages
(
research/screenpac/,research/osborne1/). Affects #24. - ROM 1.3: the source is extracted, the ROM is not certified — and the
blocker is a missing assembler (#205). We hold the 1.3 source disk
(
research/source-code/OSROM13.IMD, commentrom 1.3, cbios .3 osborne source 12/16/81) while MAME carries 1.3 asNO_DUMP. The disk is an 8-inch 77-track FM/MFM disk with CP/M skew 3, and it has been read: the 19 files are inresearch/source-code/ROM-13-source/. The source uses theDBE/DCEpseudo-ops, which only ACT 3.5E has — everyACT80.COMin the local archive is 3.0/3.2 — so a byte-exact reconstruction needs an assembler we do not have. What those pseudo-ops emit is recovered (from the shipped rev-B ROM onOCCROM.IMD), and a 1.3 image can be built with a mechanical source rewrite, but it is not certified and must not be presented as the real ROM. Full account, including the control build that decides it:docs/rom-13-reconstruction.md. The project itself has no use for a 1.3 ROM — everything we build is built from the 1.44 source — so this is upstream’s gain, not ours, and nothing here depends on it. - Confirm
swp-p88.romis host-independent. The working assumption is that the co-processor board and interposer are common and only the host driver differs, which makes this the CoPower-88 ROM rather than a Kaypro-only artefact. A disassembly would confirm it and simultaneously yield the host mailbox protocol — #3/#4/#5.
Provenance
- Stock monitor ROMs
os1-143.rom,os1-144.rom, and the character generatorsos1-vid8.rom/os1vid80.rom— Don Maslin’s ROM archive on retroarchive.org. OZROM-1E/*,OCC1_ScreenPac_RevA.BIN— github.com/BrettHallen/Osborne_1ROM/(the OZROM manual is his cleaned-up copy).tools/emulator-setup/roms/*— the MAMEosborne1/osborne1nvROM sets, matching the CRCs MAME’s driver publishes.- ROM 1.44 source —
github.com/BrettHallen/Osborne_1ROM/Source_Code_v144/; the OCC engineering source disks are from bitsavers. - Cross-reference data — MAME
src/mame/osborne/osborne1.cpp(ROM_START( osborne1 ),ROM_START( osborne1nv ), and theCOMP()lines).
See research/README.md for the per-file archive provenance, and #74 for the
two images whose source is still unrecorded.