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.

The power system in plain English¶

9 September 2026. Everything else on this site describes the power system in the wiring diagram's own terms. This page explains what is actually going on, in ordinary language, for someone who wants to understand the machine rather than look up a wire number.

The one-sentence version¶

400 volts comes in the door, and everything else in the machine is made from it — bigger things run straight off it, smaller things get it stepped down by transformers, and the two low-voltage DC supplies are what let the controller decide anything.

Four voltages, and why each one exists¶

Think of them as four separate worlds. They rarely touch, and where they do touch is deliberate.

Voltage	What it does	Why it exists
400 V, three phase	Muscle. The vacuum pump, the drill motor, the three servo drives	Only three-phase moves this much power efficiently
220 V	The carousel drive, and the socket the PC plugs into	The Panasonic indexer is a 220 V machine, so it gets its own transformer
110 V	Contactor coils, the two cooling fans	An old and sensible convention: control circuits at a lower voltage than the motors they switch
24 V DC	🔴 Nerves. Every sensor, every valve, every lamp, and everything the controller reads or drives	Low enough to be safe to work on, high enough to push current down a long cable

⭐ The useful mental model: 400 V and 110 V do the work, 24 V does the thinking. The retrofit only touches the 24 V world. Nothing in the Mesa build goes anywhere near the 400 V side.

The tree, from the wall to a valve¶

   400 V 3-phase, 60 A
         |
     -QM1 main isolator (80 A, door-interlocked)
         |
    +----+---------+-----------+------------+
    |              |           |            |
  MOTORS        -T1 3kVA    -T2 700VA    -T3 1500VA
  drills 1.5kW   400->220     400->110      400->220  --> the PC socket
  pump  7.8kW       |         400->18       400->24 AC --> the OLD CONTROLLER
  servos X/Y/Z      |           |   |
  (400 V direct)  Panasonic   fans |
                  carousel         |
                                   |
                              -G2 bridge rectifier
                              + 22000 uF capacitor
                                   |
                             rough 24 V DC  (wires 7 and 8)
                                   |
                            rails 8, 8A, 8B, 8D, 2A
                                   |
                       EVERY valve, lamp and coil on the machine

       -T3's 220 V ---> -G3 Cabur switching supply
                             |
                     clean 24 V DC, 6 A  (014 positive, 12 zero)
                             |
                    the servo drives' logic

🔴 The bit that confuses everyone: there are TWO 24 V DC supplies¶

They look the same on a meter and they do completely different jobs.

	`-G3`, the Cabur	`-G2`, the bridge
What it is	A modern switching power supply, like a laptop charger	A transformer winding, four diodes and a very big capacitor
Fed from	220 V	18 V AC
Output	Clean, regulated, exactly 24.0 V, 6 A	Rough, unregulated, about 25 V unloaded — sags under load and has ripple
Wires	`014` positive, `12` zero	`7` and `8`
Feeds	The servo drives' logic	🔴 Every solenoid, lamp and contactor coil on the machine

⭐ Why two? Because they suit different jobs. Electronics want clean power — a servo drive's logic would be upset by ripple. Solenoids do not care at all; they just want current, and a rectifier with a fat capacitor is cheaper and tougher than a regulated supply. So the builder used the cheap rough one for the muscle-work and the clean one for the electronics.

✅ This is the single most useful thing to understand about the machine's power. Almost every "which supply does that come from" question resolves once you ask "is it electronics, or is it a coil?"

Why did the old controller take 24 V AC?¶

A fair question, since everything else low-voltage on this machine is DC.

What is certain: it was designed that way. The device list reads -CNC. 025TEC0022 POSIZIONATORE M68K, 24 Vca — the controller is specified as a 24 V AC device, so it is not a wiring choice somebody made on site. Tecnos built it to be fed AC, gave it a dedicated transformer winding (-T3's 0-24 V, 250 VA) and its own fuse.

🔴 The likely reason is in the axis connector¶

The Tecnos sends the servo drives a velocity command of ±10 V — CONN.X pin 8, V-REF, referenced to pin 15. Note the ± : it has to produce negative volts.

⭐ You cannot make a negative rail from a single +24 V DC supply without extra circuitry. From an AC winding you get one almost for free — a bridge rectifier and two capacitors in series, with the junction as 0 V, gives you +V and −V straight away.

So the AC feed is most likely there because the analog section needs a split supply. The board also needs 5 V for the 68000 processor, and making 5 V, +12 V and −12 V from one AC winding is a standard, cheap arrangement.

⚠️ Reasoning, not a reading. The Tecnos board's own schematic is not in the wiring diagrams — those cover the machine, not the controller's insides. It is checkable, though: if the board carries a bridge rectifier and two large capacitors in series, that is a split supply and the explanation holds.

Two smaller reasons that also fit¶


Isolation	A dedicated winding means the controller's power has one job. Feed a board from a shared DC rail and every contactor coil collapsing and every solenoid kicking back arrives at the electronics
It was normal for the era	Controllers of this age commonly took AC and rectified it themselves. It keeps the transformer outside the box — cooler, cheaper board — and one design works on 50 or 60 Hz anywhere

⭐ The machine has both eras of practice in it: -G2 is a bridge and a capacitor, the old way; -G3 is a switching supply, the newer way. The Tecnos taking AC belongs with -G2.

✅ And it is why the AC feed has no successor¶

Mesa cards take DC and make their own rails. The 7I77U generates its own ±10 V analog output from its 5 V section, so nothing in the new build needs a negative supply brought in from outside.

🔴 That is the whole reason wires 9 and 10 simply go away rather than being replaced by something.

What actually happens when the machine clamps a panel¶

Follow one action all the way through. This is the whole system in one story.


1	You press the pedal. The pedal is a dry contact — it has no power of its own, it just joins two wires
2	It connects rail 14 (a 24 V control rail) to a wire that runs back to an input module. The module sees 24 V and reports "pressed"
3	The controller reads that, decides to clamp, and switches an output
4	The output drives a relay coil inside an output module — 18.5 mA, tiny
5	That relay's contact closes, connecting rail 8A to a solenoid valve out on the machine
6	The valve opens, air moves, the panel clamps

🔴 Notice steps 4 and 5 are separate. The controller never touches the valve. It closes a tiny relay, and the relay connects a completely different power rail to the valve. That is why the modules exist, and it is why the retrofit is simpler than it sounds: LinuxCNC only has to do step 4.

The rails, and the safety built into them¶

Rail 8 from -G2 does not go straight to the valves. It passes through relay contacts first, which split it into gated rails:

Rail	Only live when…	Feeds
`8A`	🔴 the controller says it is healthy (`CNC OK`)	The clamps, the Y datum stops, the zone 1 lamp
`8B`	the spindle contactor is in the right state	The drills, the saw, the router ram
`8D`	—	The tool release
`2A`	always (whenever the auxiliaries are on)	The drill motor, the lubrication pump, the router motor

⭐ This is real safety, built in hardware, and it does not depend on software at all. If the controller stops saying "I am healthy", rail 8A dies and the clamps cannot operate — even if something told them to.

✅ And 2A being ungated is deliberate: the lube pump runs whenever the machine has power, because a ball screw that stops being oiled does not complain until it is too late.

🔴 This survives the retrofit untouched. The gating is downstream of the modules, in the cabinet. LinuxCNC inherits it whether it knows about it or not — which is why it must drive CNC OK, or half the outputs will do nothing and look broken.

Earthing, and one thing worth knowing¶

Two earth bars — one in the cabinet, one on the machine — joined by three 6 mm² conductors. Every panel, door, upright and connector shell bonds back to them.

🔴 And the 24 V DC zero volts is bonded to earth, at the Housing, in one ferrule at CN5P pin 5 where a blue wire and a green/yellow wire are crimped together.

In plain terms: "0 V" and "earth" are the same thing on this machine.

What that means
✅ Good for noise	The 0 V has a firm, defined potential instead of floating around
⚠️ Worth respecting	A 24 V wire touching the frame is a short circuit, not a harmless touch. There is a real return path through the steel

What the retrofit changes, and what it does not¶


✅ Untouched	400 V, the motors, the drives' power, 110 V, 220 V, `-G2` and every rail it feeds, all the valves, the gating, the earthing
🔴 Changes	One thing. The little relay coils in the output modules stop being driven by the old controller and start being driven by the Mesa cards
Goes away	The 24 V AC feed on wires 9 and 10 — it only ever powered the old controller
New	The Mesa cards take 24 V DC from `-G3`, the same supply already feeding the servo drives

⭐ Put plainly: the retrofit swaps the brain and leaves the body alone. The muscle, the wiring, the valves and the safety gating all stay exactly as the machine was built.

power-distribution.md — the same system in the wiring diagram's own terms
housing-power.md — the detail on feeding the Mesa cards, and the checks to do first
safety-and-enable-chain.md — how CNC OK gates rail 8A