Working documentation
This page is generated from the project's working notes. It records a real retrofit in progress, corrections included. Anything marked 🔴 or ⚠️ is unresolved or carries a caveat - read those before acting.
Serial port: PC to Tecnos link¶
Answered 31 August 2026, from config files on the rp5800. Read-only; nothing was changed.
This closes the highest-value outstanding item in HANDOVER.md. The link is on COM1 at 19200 baud, protocol ACKNACK. One detail (the framing bits) is not written down anywhere and still has to be observed — see "The one gap" below.
The answer¶
| Setting | Value | Where it comes from |
|---|---|---|
| COM port | COM1 | Porta COM=1 |
| Baud rate | 19200 | Baud rate=19200 |
| Protocol | ACKNACK | Protocollo=ACKNACK |
| Data bits | not recorded | see below |
| Parity | not recorded | see below |
| Stop bits | not recorded | see below |
| Flow control | none at the driver level; ACK/NACK handshaking is done inside the protocol | Protocollo=ACKNACK |
Where the real setting lives¶
C:\Windows\Mw310.ini — not in the C:\Mw310 folder. This is the main HMI's own configuration file, in the old Windows 3.x style of keeping app INIs in the Windows directory. Last modified 16 November 2021.
The relevant block:
[Configurazione]
Lingua=ENGLISH
File PLC=MW310.BIN
Baud rate=19200
Protocollo=ACKNACK
Porta COM=1
Risoluzione DAC=12 bit
Modo assi=N.C.
Baud rate AUX=38400
Protocollo AUX=XONXOFF
Porta COM AUX=2
COM1 Syncro=ACKNACK
B.R. COM1 Syncro=19200
TipoCN=1
Note COM1 Syncro=ACKNACK and B.R. COM1 Syncro=19200 repeat the same answer for the Syncro (motion controller) channel specifically. Two independent keys agree, so this is solid.
A second port is in use¶
Porta COM AUX=2 at 38400 baud, XON/XOFF. COM2 is configured as an auxiliary channel. Both Protocollo linea and Protocollo Laser are set to >DISABLED<, so nothing is actively using COM2 for line or laser control. Windows reports two physical ports (COM1, COM2, both "Communications Port" — onboard, not USB adaptors), which matches the settled finding in CLAUDE.md §5.3 that the link is the onboard rear port.
When tapping, tap COM1 and leave COM2 alone.
Do not be misled by the other config files¶
Three other files in the tree also name a port and baud, and they say something different about the protocol:
| File | Port | Baud | Protocol | What it is |
|---|---|---|---|---|
C:\Windows\Mw310.ini |
1 | 19200 | ACKNACK | the main HMI — this is the one that matters |
Mw310\Sv\SV.INI |
1 | 19200 | TCBIN | sv.exe service/diagnostic tool |
Mw310\Sv\PLCEDI.INI |
1 | 19200 | TCBIN | PLCedi.exe PLC editor |
Mw310\Sv\PROFI.INI |
1 | — | — | Profi.exe parameter tool |
Mw310\Sv\TXFLASH.INI |
1 | — | — | TXFLASH.EXE firmware loader |
All agree on port 1, which is the important part. They disagree on protocol because the service tools negotiate a different framing (TCBIN, Tecnos binary) with the same controller. Mw310\Sv\SV32.log shows the handshake doing protocol autodetection:
Controllo presenza di ComDrv... (check the comms driver is loaded)
Richiesta Handle non indirizzato... (request unaddressed handle)
Riconoscimento protocollo seriale... (detect serial protocol)
Protocollo seriale TCBIN riconosciuto. (TCBIN detected)
Ricezione Configurazione Syncro... (receive Syncro configuration)
Richiesta Handle indirizzati... (request addressed handles)
Lettura file Simboli... (read symbol file)
Richiesta Inizializzazione a Syncro... (request Syncro init)
Richiesta Parametri Macchina a Syncro... (request machine parameters)
That sequence is a useful map of what to expect at the start of a capture: the controller answers a protocol probe first, then sends its configuration, then its machine parameters. The machine parameters are read back over the wire, which means a capture of a cold start should contain a full parameter dump — worth grabbing as capture number one.
Mw310\Comdrv.ini is a red herring. It lists every port COM1–COM10 with BaudRate=2 and Protocollo=0, an untouched default table indexed by code rather than literal values. It tells us nothing about the live link. It is dated 10 August 2026 only because something opened it recently.
The one gap: data bits, parity, stop bits¶
These are not recorded in any config file, and they are not in the Windows registry either (HKCU\Control Panel\Ports and the serial driver's Parameters key are both empty). Comdrv.exe sets the framing programmatically when it opens the port, and it contains no readable strings naming parity or data bits.
So the framing must be observed, not looked up. Two ways, in order of preference:
- Get it from the tap itself. When
hub4combridgesCNCB0to the real COM1, it can be told to mirror the framing the application requests rather than forcing its own. Log what MW310 asks for when it opens the port — that is the answer, and it costs nothing extra. - Watch the port open. A passive port monitor on COM1 while MW310 starts will report the
SetCommStateparameters directly.
Until then, 8-N-1 is the reasonable first guess for a 19200 ACKNACK link of this vintage, but treat it as a guess. If the tap produces garbage at 8-N-1, try 7-E-1 before assuming the baud rate is wrong.
🔴 The cable itself - from the drawing, 2 September 2026¶
COLLEGAMENTI SERIALI was finally located: SGM section, pages 11 and 12, not in the 40-sheet Q.E. section at all. Four serial cables are drawn.
Cable A - PC to Tecnos. This is the one the tap goes on.¶
SERIALE SYNCRO-P.C., Tecnos CAV90240, Masterwood 025TEC0028.
| Connectors | DB-9 female at both ends |
| Wiring | pin 2 to pin 3, pin 3 to pin 2, pin 5 to pin 5 |
| Shield | CALZA bonded to both shells |
| Protection | Two 13 V suppressor diodes, type F-126, across the data lines |
A plain three-wire null modem with no handshake lines. RTS, CTS, DTR, DSR and DCD are all unused, which is consistent with ACKNACK being a software-level protocol.
What this means for the tap:
- A passive listening tap is trivial and completely safe: bridge TX and RX to the receive lines of two spare ports with a common ground. Nothing in the link can detect it, and no handshake can be disturbed.
- The
com0com+hub4comapproach inCLAUDE.md§7.3 still works, but a passive Y-tap is simpler and needs no change toMw310.iniat all - no rollback risk. - Keep any replacement or extension cable short, or carry the suppressor diodes over.
The other three cables¶
| Name | Part | Connectors | From / to | |
|---|---|---|---|---|
| B | SERIALE ARMADIO-P.C. |
CAV90420 / 025TEC0030 |
DB-25 female to DB-9 female, 3-wire, shield, 2 suppressor diodes | Cabinet PULSANTIERA (right side of the cabinet, VISTA LATERALE DESTRA ARMADIO) to the PC |
| C | SERIALE PC-YASKAWA |
- | DB-9 male to DB-9 female, shield to shells | 3CN on the Yaskawa Sigma drives, X-Y-Z |
| D | SERIALE PC-PANASONIC |
- | mini-DIN to DB-9 female, shield both ends | CN SER on the Panasonic drive |
⚠️ The exact DB-25 pin numbers on cable B are at the limit of the scan. The standard mapping (25-pin 2/3/7 to 9-pin 3/2/5) is the expected one but has not been read off definitively. It does not matter for the tap, which is on cable A.
🔴 The drives have serial service ports¶
Cable C connects a PC to 3CN on each SGDH. That is the digital-operator / SigmaWin port.
⚠️ Corrected 2 Sep (Grant): 3CN is not a generic serial port. It needs Yaskawa's own software (SigmaWin / SigmaWin+ or the Sigma II parameter tool) and a compatible cable. It is not a case of pointing a terminal emulator at it.
That said, two things make it worth a short investigation:
- The cable existed. TEA built one (
SERIALE PC-YASKAWA, DB-9M to DB-9F, SGM page 11). It may still be with the machine's spares or in the cabinet. - The software may be on the rp5800. Only
C:\Mw310was inventoried. Search the whole disk, or the 30 Aug image, forSigmaWin,Sigma,Yaskawa,JUSP. Masterwood's commissioning engineer would have needed it.
If both turn up, the full Pn dump (Pn000.1, Pn100-Pn103, Pn201-Pn203, Pn300, Pn50A/B/E/F) can be captured as a file with no transcription errors. If not, the front panel still works and Pn000.1 alone is five minutes.
⚠️ SigmaWin writes as well as reads. Take the dump before touching anything, and stay read-only.
The Panasonic has the same arrangement on CN SER via cable D.
The PC's two ports¶
The PC (P.C./CN12) has COM1 and COM2. Per C:\Windows\Mw310.ini, COM1 is the Syncro link at 19200 ACKNACK (cable A) and COM2 is the auxiliary at 38400 XON/XOFF (cable B, to the cabinet pushbutton station - possibly the barcode reader, given EN_BARCODE = 2 in pargen.prm). Cables C and D are service cables, plugged in as needed.
Handwritten annotations¶
Page 11's Monoscheda illustration carries a technician's pencil notes against the front-panel LEDs: VIN, 0V, +05, +150 VOLT, -150, 0 OUT, 12E. That matches the physical panel legend (V IN, +5, +15 V OUT, -15, 12, E) - they are supply rail indicators, useful for a quick health check when the rack goes back in.
What this unblocks¶
The tap plan in CLAUDE.md §7.3 can now be set up concretely:
- Real port: COM1, 19200, framing to be confirmed on first open
- Create the
com0compair, repointC:\Windows\Mw310.ini→Porta COM=at the virtual port (this is the single line to change, and the single line to change back) - Bridge the other half to COM1 with
hub4comat 19200 - Rollback is one number in one file. Record that the original value is
1before touching it.
Do this on the cloned SSD, not the current disk — see HANDOVER.md item 3.
Bonus finding, relevant to the servo decision¶
The same [Configurazione] block contains:
Risoluzione DAC=12 bit
Risoluzione DAC 2=12 bit ... through ... Risoluzione DAC 8=12 bit
Modo assi=N.C.
The controller is configured for eight 12-bit DAC channels. A DAC resolution setting only means something if the controller is commanding the drives with an analog voltage. This is independent supporting evidence for the reading in CLAUDE.md §5.2 that the SGDH drives are in analog velocity mode, and it points at Option B (add a Mesa 7I77) rather than Option A.
It does not settle it on its own — the DAC could be configured and unused — so still read Pn000.1 on the drives and check CN1 pins 5/6 as planned. But go in expecting to find analog.
12-bit over a ±10 V range is roughly 4.9 mV per count. The Mesa 7I77 is also 12-bit, so nothing is lost by matching it.
Related¶
- mw310-recon — the wider inventory this came out of