Corrections¶
Things this project got wrong and later fixed. Kept in one place so nobody — including a future session reading the notes — re-derives an answer that has already been shown to be wrong.
Most of these were caught by Grant looking at the actual machine. That is the pattern worth noticing: the paper record is excellent, and it still lost to a physical look about half the time.
Changed the design¶
| Was | Is | How it surfaced |
|---|---|---|
| Axis loom connector is female at the Housing | 🔴 Male. The panel end is female — that is the half to buy | Physical check. The device list had said MASCHIO all along |
| Land the encoder screen at the Housing | 🔴 Already landed at the drive, on the drive's frame-ground pin. A second bond would create the ground loop the screen exists to prevent | Re-reading the connector schedule |
| LinuxCNC will not see the axis limit switches | 🔴 It does. Three inputs are per-axis overtravel and arrive through the ribbon. The drive's own overtravel path is separate and additional | Reading the symbolic map properly |
| The controller reads a carousel encoder | 🔴 It does not. That connector is the carousel drive's own motor encoder. The controller gets position from binary bits | Grant's instinct, confirmed in the device list |
| The servo enables need sinking outputs or interposing relays | 🔴 Neither. The analog card's enable outputs are floating isolated switches — a dry contact does not care which way current flows | Reading the card documentation |
| Panel-mount adds four connections, a slot adds none | 🔴 Misleading. The ribbon ends in a latching header either way; the breakout board is itself a component. Counted properly, the panel route has fewer joints | Grant pointed out the flaw |
Changed the numbers¶
| Was | Is |
|---|---|
| 46 inputs and 42 outputs | 41 and 41. Those were symbol counts — one input carries six symbols. The output shortage was 9, not 10 |
| Ribbon pitch 2.24 mm | 2.54 mm. "22.4 mm over 10 pins" is nine gaps, not ten. 2.24 mm is not a standard pitch and does not exist |
| One fuseholder on the panel | Two |
| Ferrules for the 28 AWG conductors | 0.08 mm² ferrules are hard to source. Fold the conductor double instead — and never tin it, solder cold-flows and the joint loosens |
Changed the understanding¶
| Was | Is |
|---|---|
| The summary-fault input is the E-stop chain | It is a summary fault fed by seven breakers, three drive alarms, the carousel relay and the inverter |
| The drives are self-homing | They are not. The home switches come back to the Housing on a pin that has no drive destination, and in speed mode these drives have no homing function at all |
| The drive serial port is a generic serial route | It needs the manufacturer's own software and cable |
| The spindle pot might be the speed source | It is an override in the controller's path. The S word will control spindle speed |
| P-clips will anchor the ribbons | Wrong part for flat cable — and then the requirement was dropped entirely |
The big one¶
"We have no records of the driver connectors."
The records existed. They were in a 40-sheet drawing, a 14-page device list and a 12-page connector schedule that had not been read. Reading all 66 pages produced the complete axis pinout, the full drive interface, all 82 I/O points, the homing method, four hardware interlocks invisible in the software, and the spare capacity inventory.
The planned work — opening both ends of the servo connectors to record pin-to-pin — turned out to be unnecessary. Nothing had to be unplugged.