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.

Housing power — can the Cabur DC replace the 24 V AC?¶

9 September 2026. Grant's question, in full:

"There are two 24 V power supplies in the cabinet. One outputs DC and the other (T3) AC. The Housing currently has 24 V AC coming in on wires 9 and 10 with wire 12 looking like an earth. There is possibly a rectifier in the Housing to create DC for the output. That rectifier is on the Tecnos board and will be removed. Will bringing power from the Cabur DC supply cause any issues?"

The short answer¶

✅ No issue in principle, and it is the right move. But the investigation turns up three things: one place where the drawings disagree with what you are seeing, one supply you had not counted, and one interlock that will silently kill every output if it is missed.

There are THREE DC/AC sources, not two¶

	Source	Output	Feeds
`-T3`	1500 VA transformer, `0-24 V` 250 VA secondary	24 V AC	`-F13` (6 A GL) → wires 9 and 10 → `CN5P` pins 1-2 → the Monoscheda only
`-G3`	Cabur `CS6 24/90-264`, fed `N`/`L` from the 220 V winding	24 V DC, 6 A regulated	wires `014` (+) and `12` (−) → the servo drives' `24V`/`0V`
🔴 `-G2`	Zaniboni bridge off `T2`'s 18 V winding, `-R1` 1 kΩ, `-C1` 22000 µF	unregulated DC	wires 7 and 8 → rails `8`, `8A`, `8B`, `8D`, `2A` → every solenoid, lamp and contactor coil on the machine

🔴 -G2 is the one that was not in the question, and it is the one doing the heavy lifting. An 18 V winding into a bridge with a 22000 µF reservoir sits around 25 V DC unloaded and sags under load — which is exactly the "24 V" that operates the machine.

🔴 The output path is TWO stages, and the two supplies never meet¶

This is the crux, and it makes the answer much simpler than it looks.

STAGE 1 — the coil          STAGE 2 — the contact
Monoscheda TD62783   ---->  UF module relay COIL
(in the Housing)            18.5 mA, from the Housing's own rail
                                   |
                                   | relay contact
                                   v
                            rail 8A / 8B / 8D / 2A  ---->  solenoid, lamp, contactor
                            (from -G2, in the cabinet)

✅ The UF modules are interposing relays. Their coils are driven from the Housing; their contacts switch the -G2 rails out to the machine. io-map.md records the rails per module, and Grant measured the coil at 18.5 mA.

✅ So what actually changes is only stage 1¶

	Today	After
Coil supply	24 V AC in on 9/10, rectified on the Tecnos board	🔴 Mesa `VFIELD`, from `-G3`
Coil return	ribbon ways 25/26 to the board's rectifier negative	ribbon ways 25/26 to Mesa `GROUND`
Contact side	rails `8A`/`8B`/`8D`/`2A` from `-G2`	✅ unchanged, untouched

Everything downstream of the module contacts stays exactly as built — the valves, the lamps, the contactor coils, and the rail gating. Nothing in the retrofit goes near -G2.

🔴 Three things to deal with¶

1. ✅ SETTLED 9 September — wire 12 is the DC 0 V, and it is BONDED TO EARTH¶

Grant, at the connector: "Wire 12 is blue. It is bonded with the green wire with the yellow trace in a ferrule that is then in position 5 in the CN5P connector."

🔴 That is a deliberate bonding point. Blue is the DC-negative convention and matches the drawings' wire 12; green/yellow is protective earth. Somebody crimped them into one ferrule, so -G3's 0 V is tied to protective earth, in the Housing, at CN5P pin 5.

`CN5P`	Wire	Colour
pin 1	9	red	24 V AC from `-T3`
pin 2	10	red	24 V AC from `-T3`
pins 3, 4	—	—	empty
pin 5	🔴 12 + PE	blue + green/yellow, one red ferrule	`-G3` 0 V bonded to earth

✅ Three things follow, two of them good¶

The Cabur's 0 V is already at the Housing. ✅ Only 014, the positive, needs bringing in. The return is there.

The -G2 bridging worry largely evaporates. If -G2's negative is earthed anywhere in the cabinet, the two supplies have been bonded through earth since the machine was built, and putting the module coils on -G3 changes nothing. ⚠️ Still worth one check: if -G2's wires 7/8 read open to earth, it is floating, and the coil connection would become its first earth reference.

🔴 The Mesa field circuit will be earth-referenced, not floating. Mesa GROUND and ribbon ways 25/26 sit at earth potential. Usually good — quieter, defined potential — but a field conductor touching the frame is a fault path, not a harmless touch.

🔴 And one thing to be careful about before `CN5P` comes off¶

That ferrule may be the Housing's protective earth. If the green/yellow at pin 5 is how the enclosure is earthed, unplugging CN5P without replacing it leaves the metalwork unearthed. It is bolted to the backplate, but bolted-through-paint is not a protective earth.

Before removing the connector: confirm the Housing has an independent earth — a separate green/yellow to a chassis stud, or proven continuity from the enclosure to the earth bar. If pin 5 is the only path, run a dedicated earth conductor first, not after.

The original reasoning, superseded¶

⚠️ Wire 12 is almost certainly NOT an earth

The drawings say wire 12 is the 0 V return of -G3 — the Cabur's negative — with 014 as the positive. power-distribution.md confirms it from SGM page 6: CONN.X pin 7 carries 014 to -UX:24V, and 12 lands on the drive's 0V.

✅ And the earth conductors on this machine are G-numbered — G1, G31-G53 on sheet 1. A bare 12 is not in that series.

🔴 If wire 12 at the Housing really is -G3's 0 V, then half the job is already done — the Cabur negative is present and only 014 needs bringing in.

Worth one measurement before assuming either way: with everything off, check continuity from that wire to the -G3 0V terminal, and separately to the earth bar. It is a two-minute test that decides whether you run one new conductor or two.

⚠️ Do not treat it as an earth and bond it. If it is the DC 0 V and it gets earthed, the Cabur's return is now referenced to the frame — which may be fine, may be what the builder intended, or may create a path nobody planned. Establish what it is first.

2. 🔴 The coil common must not bridge `-G3` to `-G2`¶

The one real electrical risk. Today the module coils sit on a rail derived from -T3's isolated 24 V winding. Tomorrow they sit on -G3. If a module ties its coil common internally to its contact common, connecting the coils to -G3 would tie -G3's negative to -G2's negative — two separately derived supplies bonded through the modules.

⚠️ That is not automatically harmful, but it is not something to discover by energising it.

The test, with the machine off and a meter on resistance:

Measure between	Expect	If not
Ribbon way 25/26 and wire `12` (`-G3` 0 V)	open	The coil common is already tied to `-G3` — good, use it
Ribbon way 25/26 and wire `7`/`8` (`-G2`)	🔴 open	The coil and contact sides share a common. Stop and think before feeding coils from `-G3`
Ribbon way 25/26 and earth	open	Same caution

✅ The likely answer is all three open, because interposing relays exist precisely to keep the two sides apart. But it is the one measurement that would turn a clean job into a puzzling one.

3. 🔴 Rail `8A` is gated by `CNC OK` — the outputs die silently without it¶

From safety-and-enable-chain.md: KA and KA2 contacts gate rail 8 into rail 8A (sheet 11), and KA is driven by OF.1, CNC OK.

🔴 So if LinuxCNC does not drive OF.1, rail 8A never comes alive — and UF1's outputs (both zones' clamps, both Y datum stops, the zone 1 lamp) will do nothing at all, even with the Mesa correctly driving their coils and the relays audibly clicking.

⚠️ That is a failure mode that looks like a wiring fault and is not one. The relay pulls in, the contact closes, and there is no voltage on the other side of it.

Rail	Gate	Feeds
`8A`	🔴 `KA` — `CNC OK`	`UF1` (`OF.1`-`OF.8`), `UF5` `AUS35-37`
`8B`	`KAMF` — spindle contactor	`UF2`, `UF3` (`OF.9`-`OF.24`)
`8D`	—	`AUS40`
`2A`	ungated	`KM101` drill motor, `AUS38` lube, `OF.48` router motor

✅ The 2A outputs work without CNC OK — which is why the lube pump runs whenever the auxiliaries are on.

🔴 Capacity — what `-G3` already supplies is NOT established¶

⚠️ An earlier version of this page said "~1 A of 6 A, no second supply". That rested on an assumption that was never checked — that -G3 feeds only the three servo drives. It is not established, and it is probably wrong.

What is confirmed¶

Load	Evidence
Three servo drives' `+24V-IN`	`CONN.X/Y/Z` pin 7 carries `014` to `-UX:24V` etc. SGM page 6. Tens of mA — opto commons

That is the entire confirmed list. The power sheets were read for the transformers and the supplies; where 014 and 12 fan out afterwards was never traced, because the dig was following I/O rather than power.

🔴 What is probably on it, and would change the arithmetic¶

Suspected load	Why it matters
🔴 Rail `14`, the 24 V control rail	It is the source side of every dry-contact input on the machine — start buttons, the hold chain, the photocells, the home switches. `io-map.md` shows inputs wired `-X3 14` → contact → `E1nn`. If rail `14` comes from `-G3`, every field input's supply current is already on it
Seven Cabur NPN-PNP converters	`-OPT1` and `-OPT11`-`-OPT16`, sheets 5.2 and 20.3-20.8. Each needs a supply
A "24 Vdc sensor supply"	Named on a connector schedule. Proximity switches and photocells draw real current, unlike dry contacts

⚠️ 14 and 014 are different wire numbers on this machine, so rail 14 may equally come from -G2 or elsewhere. Nothing read so far settles it, and it is the single fact that decides whether -G3 has 5.9 A spare or considerably less.

✅ The answer is a measurement, not more document reading¶

machine-layout.md already names the place: X1A distributes rails 7, 8, 8A, 8B, 12, 14, 14A, 15, 16, 2, 3 with jumper bars linking each group, and is "the single most useful place in the cabinet to measure any control rail."

	Measurement	Tells you
1	🔴 Current out of `-G3` on `014`, machine idle and then running	The real headroom, directly. Better than any paper trace. Clamp meter, or in series at the supply terminal
2	Voltage rail `14` to rail `12` at `X1A`	✅ A regulated supply and a bridge-plus-capacitor look different: `-G3` holds ~24.0 V steady; `-G2` sits nearer 25 V unloaded, sags under load and carries ripple. That distinguishes them without tracing a wire
3	Continuity, `-G3` `+` terminal to rail `14`, machine off	Settles it outright

What the retrofit adds, whatever the answer¶


41 module relay coils at 18.5 mA, all on	~0.8 A
44 Mesa sinking inputs at 1-2 mA	~0.1 A
Mesa `VIN`, three cards at ~25 mA	~0.1 A
Added	~1 A

✅ The addition is small and well characterised. ⚠️ What is not characterised is what it is being added to.

🔴 And note the retrofit removes a load as well as adding one: the module coils currently draw their ~0.8 A from the Housing's rectified AC, not from -G3. So this is a transfer of about 0.8 A from -T3 onto -G3, plus about 0.2 A of genuinely new load.

✅ If measurement 1 shows -G3 sitting under about 3 A with the machine running, there is nothing to think about. If it is closer to 5 A, a separate DIN-rail supply for the Mesa field power is cheap and removes the question — and the cards are happy on any 5-28 V source.

✅ The 7I76EU can take unregulated 8-28 V on TB3 pins 21/22 and feed 5 V to the 7I77U over the DB25, so the whole card stack runs off whichever rail is chosen.

⚠️ VFIELD must connect directly to the DC source — Mesa's rule, on all three cards. No switch, breaker or relay contact in that circuit, because field power has to ramp in under 10 V/ms. A fuse is fine.

🔴 So do not be tempted to gate VFIELD through KA to reproduce the CNC OK interlock. That interlock already exists downstream, on the -G2 rails, and it stays there untouched.

What the 24 V AC was doing, and whether anything else needs it¶


The Monoscheda	`-CNC. 025TEC0022 POSIZIONATORE M68K, 24 Vca`. Goes with the board
The output-coil rail	Rectified on the board, per Grant. Replaced by Mesa `VFIELD`
⚠️ `FUSE 1` and `FUSE 2`, 1 A, on `FILI VIOLA`	On the `CN5P` panel, sheet 35.1. Purpose unrecorded. Check what these feed before cutting the AC
`CN5P` pins 3, 4, 5	Drawn but not obviously used

⚠️ -F13 is a 6 A GL fuse on a 250 VA winding. Once the AC feed is dead, that fuse and its wiring are spare — label or remove wires 9 and 10 rather than leaving a live 24 V AC pair loose in the Housing.

The checks, in order¶

	Check	Why
1	Is wire `12` at the Housing the `-G3` 0 V or an earth?	Decides whether you run one new conductor or two
2	🔴 Is the module coil common isolated from the `-G2` rails and from earth?	The one real electrical risk
3	What do `FUSE 1` / `FUSE 2` on the violet wires feed?	The only unaccounted load on the AC
4	Confirm `KA` still gates rail `8A`	So the `CNC OK` dependency is understood before it bites
5	🔴 Measure the standing current out of `-G3`	The headroom question. What it already supplies is not established

✅ If 1 and 2 come back as expected, bringing -G3 into the Housing for Mesa VIN and VFIELD is straightforward and nothing downstream changes.

power-distribution.md — the transformers, supplies and earthing
safety-and-enable-chain.md — how CNC OK gates rail 8A
io-map.md — which rail feeds which output module
mesa-allocation.md — where the coils land on the Mesa