Serial and the modem
Interfaces
Status: Solved. Both connectors and their signals are read from the Technical Manual’s own §4.2/§4.3 tables and confirmed against schematic sheet 7, which carries the DE-9P pin numbers through to the on-board P1 header unchanged. Two things are still open, both named in place below: the physical orientation of the 2×5 P1 header on the board, and which part the RXD receiver is.
P1 is the MODEM port — a DE-9P at TTL levels, not RS-232, which is what makes
it the clean attachment point for a modern serial device: TXD and MCB are
5 V TTL, RXD is a bipolar input (negative for mark), MSB is an
open-collector modem-status bit, CTS is the only hardware flow-control line,
RI is the ring indicator, and pin 7 is a real +12 V supply — the Astec
rail is rated 2.02 A, and pin 7 reaches it through R21 (22 Ω), a resistor
the keyboard branch shares. P2 is the
RS-232 port — a DB-25S wired as RS-232C-compatible DTE, with CB, CC and
CF held at +5 V because the machine does not use them. Both connectors are the
same 6850 ACIA, at status 2A00H / data 2A01H.
The rest of the document is the pinout evidence (§4.2, §4.3 and the schematic-confirmed P1 header), the design notes the WiFi modem (#56) has to satisfy, the baud-rate options the archive documents (#21), and the 6850’s framing byte in the BIOS.
Source: Osborne 1 Technical Manual, ch. 4 “Osborne 1 Interface Design” —
§4.2 SERIAL RS232 INTERFACE, §4.3 MODEM — archived in-repo at
research/osborne1/modem/o1techman-interface-extract.pdf (printed pages 13–22),
with schematic sheet 7 of research/osborne1/OCC1_1A2011-00_Schem_RevE.pdf for
the on-board header.
This is the authoritative connector/pinout reference for the WiFi-modem work
(#21, extended by #56) and for the severable modem module (#56 §E / COMM PAC
variant). The memory-map and ACIA register detail live in
docs/o1-memory-io-map.md §”6850 ACIA detail”; this document adds only the
physical connector information that doc does not carry.
4.2 RS-232 serial interface (P2)
- Configured as RS-232C-compatible DTE, but “certain of the RS-232C signals are held at +5 volts since they are not needed to control the Osborne 1.” A 6850 ACIA drives the port.
- Port addresses: status
2A00H, data2A01H(shadow mode). Status bits (Fig 4.2):RDRF,TDRE,DCD,CTS,FE,OVRN,PE,IRQ— see the 6850 datasheet for full semantics.
DB-25S pinout (Fig 4.2.2)
| DB-25 | RS-232 | Definition |
|---|---|---|
| 1 | AA | Frame ground (optional) |
| 2 | BA | Transmitted data (low = 1) |
| 3 | BB | Received data (low = 1) |
| 4 | CA | Request to send (high or no connection enables) |
| 5 | CB | Clear to send (always high on OCC 1) |
| 6 | CC | Data set ready (always high on OCC 1) |
| 7 | AB | Signal ground |
| 8 | CF | Received line signal detected (always high) |
| 20 | CD | Data terminal ready (high or no connection enables) |
Pins 9–19, 21–25: no connection. Only the listed pins are wired.
4.3 MODEM interface (P1) — TTL levels, not RS-232
“The modem and RS-232 interfaces are basically one and the same. In addition to the serial port, TTL-level signals may be directly input into the 6850 ACIA using the modem port connection.”
The MODEM jack is the clean way to attach a TTL-serial device (an ESP32 WiFi
modem, our module) straight onto the ACIA without an RS-232 level shifter.
Read/write it via the CP/M IOBYTE function, same as the serial port.
4.3.1 Signal direction
| Osborne 1 | dir | Modem device |
|---|---|---|
| GND | — | GND |
| TXDATA | → | TXDATA |
| RXDATA | ← | RXDATA |
| MSB | ← | MSB |
| CTS | ← | CTS |
| MCB | → | MCB |
| +12 | → | POWER IN |
| RI | ← | CD |
4.3.2 DE-9P pinout (“standard numbering of the DE-9P connector”)
| Pin | Signal | Definition |
|---|---|---|
| 1 | GND | Signal ground |
| 2 | TXD | Transmitted data — TTL logic, 1 = high |
| 3 | — | Not used |
| 4 | MSB | Modem status bit — open collector, 50 µA sink = inactive |
| 5 | CTS | Clear to send |
| 6 | RXD | Receive data — bipolar input, −0.5 V…−10 V = 1 |
| 7 | +12 V | +12.6 V DC from the 2.02 A rail, through R21 (22 Ω); the Comm-Pac manual labels it VDD |
| 8 | MCB | Modem control bit — TTL, low suppresses output |
| 9 | RI | Ring indicator — TTL, high-to-low sets flag |
Design notes for #56
- Levels are asymmetric, and the receive direction is not free. TXD and MCB
are 5 V TTL because the 6850 and the PIA are +5 V parts; the ESP32-S2’s
inputs are not 5 V tolerant (
VDD + 0.3), so both need a translator. CTS, RI and MSB leave the module at TTL thresholds; MSB is open-collector and needs an open-drain driver. RXD is a bipolar input (−0.5…−10 V = mark): on sheet 7 the RXDATA line enters a receiver with zener clamp CR1 (1N5231B, 5.1 V) — a comparator-style front-end, not a TTL gate. So driving RXD with 0–3.3 V leaves the mark at or below threshold, and the module must swing negative for a clean mark. A MAX3232 on 3.3 V (its own charge pump, ±5.5 V) is the cheap answer, and the negative swing still wants confirming on real hardware — see the open item below. The receiver’s part number is open:docs/o1-mainboard-schematic.mdsheet 10 lists UE3 as a 1489 with three LM3900 comparators at UD4, while a pin-level read of the same region returned UE3 as an LM1458. It does not change the requirement (negative for mark), but it changes the input impedance, so re-read that region at native resolution before fixing the drive current. -
P1 pin 7 is the
+12Vnet throughR21(22 Ω), and that resistor is the limit on this port. The Technical Manual’s §4.3.2 table says it in words — “Connected to power supply through 22 ohms” — and sheet 7 shows the arrangement:R21hangs off the+12Vrail and its lower end is a junction that feeds both this port’s pin 7 andJ6 → P4-19, the keyboard branch. The two loads therefore share one 22 Ω resistor.A 22 Ω feed is a hard current limit, so this is a low-current supply. At 500 mA it would drop 11 V, which no 12 V rail survives; at 100 mA the drop is 2.2 V and about 10 V reaches the load. A period Bell-103 modem with its own DAA draws tens of mA, which is what makes a 22 Ω feed possible at all — and the Comm-Pac’s own manual quotes no supply current anywhere (its specification page gives the modulation, the carrier frequencies and the line levels, and nothing about current). So the port’s budget is the one #56 already assumes: small, and shared with the keyboard adapter. The eleven-lamp panel is sized at 2–3 mA per lamp for exactly this reason.
- MSB (pin 4) is open-collector and MCB (pin 8) low suppresses output — together these are the modem’s “hook”/enable handshake. RI (pin 9) falling edge sets the ring flag; the ESP32 can drive it to signal an incoming “call” (AT RING) to the O1.
- CTS is the only hardware flow-control line brought out (pin 5).
On-board P1 header — schematic-confirmed
Schematic sheet 7 (SERIAL I/O, OCC1_1A2011-00_Schem_RevE.pdf) draws the P1
(MODEM) connector on the right edge with the DE-9P pin numbers carried
through unchanged — there is no renumbering between the case DE-9 and the
on-board P1 header:
| P1 pin | Net | P1 pin | Net |
|---|---|---|---|
| 1 | GND | 2 | TXDATA |
| 4 | MSB | 5 | CTS |
| 7 | +12V via R21 22Ω — see Design notes for #56 | 6 | RXDATA |
| 8 | MCB | 9 | RI |
| 3 | (NC) |
Physically P1 is a 2×5 IDC male header on the board, cabled to the case DE-9P with a straight ribbon (1:1, no conductor swap). A straight IDC crimp interleaves the DE-9 rows, so on the 2×5 the signals land as: row A (1-2-3-4-5) = DE-9 1,3,5,7,9 = GND, NC, CTS, +12V, RI and row B (6-7-8-9-10) = DE-9 2,4,6,8,— = TXDATA, MSB, RXDATA, MCB, NC.
The schematic confirms the signal→pin mapping (the table above). The physical orientation of the 2×5 header on the board (which end is pin 1) is a layout detail not in the schematic — confirm against the PCB silkscreen / a real board before committing a PCB footprint. Same class of in-hardware validation as #47.
Also on sheet 7 (for the record): the ACIA baud clock comes from UC3
(LS161) dividing a 4USEC/2USEC tap (jumper J1 selects the rate), and
MODEM RI routes to the video-PIA UC15 CA2 input (sheet 5) — consistent
with docs/o1-memory-io-map.md.
Baud-rate options the archive documents (#21)
The stock machine gives 300/1200 (J1 in) or 600/2400 (J1 out), and the factory
documents 2400/9600 as a modification — the jumper and divider detail is in
docs/o1-memory-io-map.md §”6850 ACIA detail”. Two further things come from
The Portable Companion #8 (July 1983).
The factory modification, as the Companion cites it
The issue’s “The CRT Connection” box (PDF p. 87, printed 85) tells a user who
wants an 80-column terminal to “set your computer for 9600 baud using Osbaud or
the method shown on page 22 in the Technical Manual.” That page is the
manual’s Figure 4.4 “Baud Rate Hardware Modification” — the LS161/LS163 at
C3, pin 2 moved, J2 fitted, J1 “DON’T CARE” — quoted in
docs/o1-memory-io-map.md §”6850 ACIA detail”. So the Companion’s
pre-SCREEN-PAC answer to “how do I get 80 columns?” was the serial port plus
the factory’s own baud mod, not a video board.
The CRT recipe itself
- Set the machine for 9600 baud (as above).
- Connect a standard serial cable between the terminal and the computer. The cable is only pins 1, 2, 3 and 7, straight through (1–1, 2–2, 3–3, 7–7).
- Set the terminal’s DIP switches to 1 start bit, 1 stop bit, 8 data bits, no parity, full duplex.
STAT CON: = CRT:to make the terminal the console.- “The end result is a 80-column, 9600 baud refreshed, block cursor screen.” The procedure is “untested with terminals other than Televideo” — the reviewer’s was a Televideo 920C.
Advent Products “Osbaud” — a clock substitution, not a bus interposer
Reviewed by Brad Baldwin in the same issue (PDF p. 87, printed 85; extract
at
research/technical-articles/TPC-Issue08-1983-pp87-osbaud-baud-rate-generator.pdf).
$59.95, 16 DIP-switch-selectable baud rates, used by the reviewer at 110
(an ASR33 Teletype), 9600 (the Televideo 920C) and 19200 (a $10,000
high-speed band printer), with mainframe up/downloads at high speed. It is
“a small auxiliary circuit board … soldered to the main logic board in the
area behind the serial port“, with the DIP switches reached through a small
slot cut in the venting grill and a self-adhesive rate chart stuck to the
keyboard. Advent even supplies the .050” Allen wrench for the brightness and
control knobs, and the package covers both case colours (blue/gray and
black/brown). Installation wants a Phillips screwdriver, an Xacto knife to cut
across a circuit trace, and a low-wattage soldering iron — and, like any such
product, voids the warranty.
The four solder points, decoded
“Osbaud is soldered to pins 4 and 12 of the 6850 UART, pin 7 of the
74LS32, and pin 7 of the 74LS04.” Read against the schematic
(docs/o1-mainboard-schematic.md sheet 10, and §”6850 ACIA detail” in the
memory map):
| Solder point | Is | Why |
|---|---|---|
| 6850 pin 4 | TxCLK — strapped on the board to pin 3 (RxCLK) | the clock input: one wire drives both halves of the ACIA |
| 6850 pin 12 | Vcc | +5 V, taken from the ACIA’s own supply pin |
| 74LS32 pin 7 | GND | ground |
| 74LS04 pin 7 | GND | ground — and this one is UB3, the inverter inside the ACIA’s own baud divider |
So OSBAUD does not tap the bus, needs no address, and does not use the LS161’s
socket: it replaces the ACIA’s clock, powered and grounded from the ACIA’s
immediate neighbourhood. That makes the Xacto knife unambiguous — the trace to
cut is the one carrying the factory clock (UC3’s QD) to the ACIA’s clock pins,
so that OSBAUD’s oscillator is the only driver on that net.
What the design has to be
The review gives the connections and the rate count, not a schematic, so the rest is inference — but tightly constrained:
- The ACIA still divides by 16 or 64 internally (the factory’s Figure 4.4 is captioned “SPEED SELECTED W/IN 6850”), so OSBAUD’s clock is 16× the marked rate on the ÷16 setting — 800 Hz … 307.2 kHz across 50 … 19200 baud.
- Sixteen rates off a DIP code means sixteen selectable taps, but a single
binary chain from one crystal cannot produce the odd rates of the standard
16-rate list (110, 134.5, 2000), so the divider has to be
preset/programmable — the same trick the factory divider next door uses,
its LS161 being a preset-load counter (
RCO→ inverter →/LD). - The three rates the reviewer names (110, 9600, 19200) all sit inside the
classic 50/75/110/134.5/150/300/600/1200/1800/2000/2400/3600/4800/7200/9600/
19200 set — the same list Nuevo’s optional generator advertises (“16 baud
rates from 50 to 19.2K”,
docs/o1-upgrade-boards.md).
Open: the crystal and divider topology, whether the ACIA is left in ÷16 mode, and where the 74LS32 sits on the board (the LS04 is UB3 on the serial sheet; a 74LS32 near the serial port has not been located on the schematic yet). The OSBAUD instruction sheet — which the review calls thorough — or a photo of a fitted board would settle all three; worth watching for on the second-hand market.
Framing — the 6850 control byte in the BIOS (#21, #56)
Baud rate is only half of what a modem personality has to present. The other half
is the frame — data bits, stop bits and parity — and OCC documented how to
change it. “Programming the serial port — modifying the serial port for
scientific test devices”, signed by the Osborne Computer Corporation System
Technical Support Group, in The Portable Companion #7 (June 1983, PDF
pp. 56–57 = printed 54–55; extract
research/technical-articles/TPC-Issue07-1983-pp56-57-programming-the-serial-port.pdf)
is a worked BIOS patch:
- The factory default is 8-bit data, 1 stop bit, no parity.
- The change is a single control byte written into the BIOS image: build a
relocatable CP/M (
MOVCPM 60 *for a 1.3 BIOS,MOVCPM 59 *for 1.4),SAVE 39 CPM60.COM,DDT CPM60.COM, thenS1FE9and put the code in at1FE9H/1FEAH, andSYSGENit to the target disk. So the ACIA’s control-register init byte lives at BIOS offset1FE9H— a second, independent handle on the ACIA beyond the2A00H/2A01Hport pair, and the address to watch for #21/#56. - Figure 1 is a “Table of 6850 control codes” covering all 16
combinations at 1200 and 300 baud — e.g. 8-N-1 @1200 =
55H, 8-N-1 @300 =56H, 7-E-1 @1200 =49H, 8-N-2 @1200 =51H; arrows mark the factory default. Read as the 6850’s control register (word-select bits 2–4, clock-divide bits 0–1) the values are exactly right, which is a useful cross-check on thedocs/o1-memory-io-map.md§”6850 ACIA detail” claim that the ACIA divides by 16 or 64 — the55H/56Hdefault sits in the ÷16 field. - Figure 2 is a which-BIOS-do-I-have test:
DDT,DA900, and look at the third byte after the490—C3 xx E5is a 1.3 BIOS,C3 xx E1a 1.4.
The article does not touch the baud divider, so it is complementary to the Hogan/Figure 4.4 rate modifications above: rate from the divider, frame from the BIOS control byte. For the module (#56) both matter — a WiFi modem that only ever presents 300/1200 8-N-1 will fail on a machine whose owner has patched the frame for a lab instrument, which is exactly the case the article exists for.