#voron #voroncascade #cnc #electronics #wiring #history
# Overview
The history of [[Voron Cascade Electronics Wiring]]. It was moved out of that note on 2026-11-16, when the note was condensed at Erik's request, so that the wiring note reads as the current design. Nothing here is needed to wire the machine. It holds, as they stood that day (wiring Rev O, Layout J):
- every decision entry in full, including the eleven that the wiring note now carries as a few lines each (reversed, replaced, or only naming a revision letter);
- the TODO items that were done and struck through;
- the wiring note's changelog in full, 2026-11-08 to 2026-11-16.
Nothing is reworded. Links that pointed at a section of the wiring note are re-pointed so that they still work from here. The whole wiring note as it stood before it was condensed is `media/cascade-wiring-note-2026-11-16-before-condensing.txt`, byte for byte.
New decisions and changelog lines go in the wiring note, not here.
# Decisions
Newest first, as in the wiring note on 2026-11-16.
## 2026-11-15: One grounding block on the rear rail, at block 11 (Layout J)
Erik's call: two rail-bonding blocks on the rear rail are one more than the job needs. One 7641K81 goes back to stock and a plain green 7641K51 takes its place, and the remaining grounding block moves from the left-hand end of the PE group to the right-hand end. The group is now blocks 5–10 green, jumpered, and block 11 the 7641K81, linked to block 10 by a 12 AWG wire because the jumper bar still does not fit it. The row is still 15 blocks and nothing moves. What lands where follows from that, and it is this note's allocation, not part of Erik's instruction: the incoming earth, the plate stud, the frame lug and every device earth land on the six green blocks, one conductor per screw, so they meet on the jumper bar and none of them depends on the wire link. The link only bonds the rail, and the spare screw is the second one on block 11. Two wires change blocks: the lower-rail feed goes to block 6 and the gantry jumper to block 7. This reverses the decision of 2026-11-13 below, and is the alternative that entry names. ==It needs one jumper bar that spans the six green blocks.== The layout is at J for it and now has its own note, [[Voron Cascade Electronics Layout]]; the wiring stays at Rev O.
## 2026-11-15: Rev O
Erik's call. Rev O is Rev N as the third independent review saw it (the package of 2026-11-14) plus what was decided on 2026-11-15, each item with its own entry here. One change to the wiring: the door relay's contact is out of KM1's coil circuit, so the door acts through LinuxCNC only, and only while the spindle is commanded on (rungs 4 and 7, and one wire at K5). One accepted limit: a FlexiHAL that answers the Pi but ignores its commands, for which the E-stop button is the stop. And the third review's corrections. The letter moves in this note, the Configuration note, the Bench Checks note, the three drawings and the Layout I page. Rev O has not been reviewed: the next step is the bench, not a fourth paper review. What still gates wiring is in the hold notice at the top.
## 2026-11-15: The door acts through LinuxCNC only
Erik's decision, after the third review. What he wants from the door switches is a stop when the door is opened while the machine is cutting. With the spindle off he wants to jog the axes and touch off with a 3D probe with the door open, which the hardware door channel did not allow. So the door relay's contact comes out of KM1's coil circuit: KM1 A2 moves from K5 14 to K5 A2, the 0 V side, and K5 stays as the relay that reads the two switches and drives the DOOR input. LinuxCNC does the rest. Door open with the spindle commanded on, which includes a spindle start with the door already open: the VFD's run command drops, the feed stops, the program is aborted and the spindle command cleared; the machine stays on, drive power stays up and the axes stay homed; nothing restarts when the door shuts. If the spindle is still commanded on 1 s later, the E-stop latch trips as a backup. Door open with the spindle off: nothing happens. The alternatives were a hardware cut of the spindle alone, which needs more parts and wiring, and leaving the door as it was. **What it gives up:** the door no longer removes drive power or the valve's feed in hardware; the stop depends on LinuxCNC, the Pi, the SPI link, the DOOR input and Modbus; and the axes are live with the door open. The door was already described as a convenience interlock and not a safety function (S2), and it is now a narrower one. In the simulator the door test passes 23 of 23 and the scripted run 82 of 82 ([[Voron Cascade Electrical Configuration#Simulator run (2026-11-13)]]). Not reviewed, and not run on the machine. ==For the bench: DOOR polarity, how long the spindle takes to stop, the VFD's stop mode, the backup, and whether QtDragon's own HAL files clash with the three pins the door uses ([[Voron Cascade Bench Checks]], part C).== This change is what made the revision Rev O (entry above).
## 2026-11-15: Third review: one accepted limit, corrections, and the bench next
The third independent review, of the Rev N package of 2026-11-14 (`Cascade-Electrical-Review-RevN-v2`), did not pass it. Of the second review's six findings it closed one (N04), closed three with a problem (N02, N03, N05) and left two open (N01, its blocker, and N06). Of the first review's 25 findings that were still open it closed one (HW-06), closed three with a problem (S4, SW-05, SW-06) and left 21 open, nearly all of them waiting for the bench or for parts. It added two blockers (R3-01, R3-02), two majors (R3-03, R3-04) and four minors (R3-05 to R3-08). Like both earlier reviews it found the hardware E-stop channel intact: the E-stop button cuts drive power whatever the software does.
All three blockers sit under one fault: a FlexiHAL that answers the Pi's reads but ignores its commands. Both new ones were reproduced in the simulator before anything was decided. R3-01: every click on the screen's E-stop reset restarts the SPI link for 0.75 s, refused or not, so clicks a second apart or less keep the board's watchdog quiet and K2 on. R3-02: until the board lets K2 go, about 1.3 s after a trip, anything that restores KM1's coil circuit brings drive power back. The statement made here on 2026-11-14 that KM1 stayed out in 15 of 15 such trips came from a script that looked only at the end of each run, and is withdrawn. Erik's decisions:
1. **The fault is an accepted limit.** Nobody has named a way it can happen; the present link arrangement is the fourth, and each of the last three had a hole found within a day. The HAL's link logic stays as it is. If the board stops taking commands, the E-stop button is the stop. The link stop and the firmware's timeout are a backstop that acts about 1.3 s after a trip, and only if nobody clicks reset in that time. The reviewer's position is that N01 is not closed. Two further patches were weighed and not made: a lockout that keeps the link off for longer than the firmware's timeout after a trip (every reset refused for about 2 s on a healthy machine, and the KM1 window stays), and a following error tripping the latch (it broke the first reset when tried).
2. **The door.** The review's other design question led to the entry above.
3. **Corrections, with no design change.** The withdrawn claims ("KM1 stayed out, 15 of 15", "cannot come back by itself", "every trip ends with the outputs off about 1.3 s later"). The reset check is described as what it detects, a status contact reading "in", and not as a weld detector (R3-03). The simulator's latency figure is withdrawn and its "every output off" assertion widened to all ten outputs (R3-04, R3-05). Quick double presses are described as they are: two AIR presses both act, and only a quick second CYCLE START can be lost (R3-06). HALT has its own input circuit, so the 117 mW figure is not its (R3-07). Stale lines in Layout I and the cable schedule (R3-08). And three of this note's own errors that the reviewer caught in passing: a 240 V feed is no fix for a 110 V-input VFD; KM1's DC coil is 6 W in the data sheet's table where this note assumed 9 W; and the gates before wiring are more than two.
4. **The bench is next, not a fourth paper review.** The review's own list shows that what is left is bench work: the Aux rework, then the powered checks on the FlexiHAL alone, where the firmware's real timeout and reset behaviour can be measured. Parked for that phase: the firmware's separate reset input from the Pi (`flexi.PRU-reset`), which might give a way to switch the board off that does not depend on the board taking a command.
Still open and unchanged: the list at the end of the second review's entry below. The door change has had no independent review.
## 2026-11-14: The SPI link follows the E-stop latch
Found while writing the response for the next review. The afternoon's fix for the second review's blocker (N01) stopped the SPI link 0.5 s after LinuxCNC's own E-stop, by tying it to `iocontrol.0.user-enable-out`. LinuxCNC leaves that signal on when the E-stop comes from outside, so after a door trip, an E-stop-button trip or KM1 dropping out the link never stopped. In the simulator, with the board taking no commands, K2 then stayed on at the board and the board kept stepping; shutting the door, or POWER ON after the E-stop button, brought KM1 back in with the axis running and LinuxCNC in E-stop (9 of 9). Erik approved the change: the link runs while the E-stop latch is OK and for 0.5 s after it trips, whatever tripped it, and a reset click on the screen starts it again. The same trips then end with the board's outputs off 1.21 to 1.30 s later and KM1 staying out (15 of 15: **withdrawn 2026-11-15**, because the script looked only at the end of each run; see the third review's entry above), and the scripted run passed 72 of 72 ([[Voron Cascade Electrical Configuration#Simulator run (2026-11-13)]]). Nothing in the wiring changes. **Not covered:** a board that stops taking commands when nothing trips. A feed hold then does not reach it; LinuxCNC reports a following error and goes to machine-off, and the axis runs on with the drives powered until an E-stop is pressed. No ordinary fault is known that stops commands and leaves the board answering, so this is a missing backstop, as N01 was. Tripping the latch on a following error was tried and made the first reset after such a stop fail, so it was left out. A side effect: after any trip, not only the on-screen E-stop, `halshow` shows the inputs' last values until the next reset.
## 2026-11-14: The 48 V −V bond stays 18 AWG
Erik accepted this as a deviation from the 12 AWG bonding rule, after the second review pointed out that the bonding table and the wire labels disagreed (EL03, D1). The one 48 V −V → PE bond is 18 AWG white/blue, in a 2 × 1.0 mm² twin ferrule with U4's −V feed on the NDR's second −V terminal, and lands on the lower-rail grounding block: what the wire labels already had. Its job is to tie the 48 V rail to earth. A short from the 48 V rail to the chassis returns through it, and the NDR limits that to 10.5–13 A and shuts down after about 3 s, which 18 AWG carries. What the deviation gives up: the bond is not sized to carry a mains fault onto the 48 V wiring until the branch breaker clears, which is what the 12 AWG rule is for. The 18 AWG driver feeds are no heavier. The alternative was a 12 AWG bond on a terminal of its own, which the NDR's two −V screws leave no room for without a distribution block. The 24 V 0 V bond is not affected: it has an XT3 screw to itself. ==Still to check: that the twin ferrule seats in the NDR's terminals ([[Voron Cascade Bench Checks]]).==
## 2026-11-14: Second review: three HAL changes, and what stays open
The second independent review, of Rev N, did not pass it. Of the first review's 45 findings it counted 20 closed, 3 closed with a problem and 22 not closed: 13 of the 22 wait for the bench or for parts, as this note already said, 7 were statements in the notes or drawings that were wrong or claimed too much, and 2 were questions of design. It added one blocker and five majors, most of them about how much the simulator run had really shown. Erik approved all five proposals the same day.
1. **The link stops 0.5 s after LinuxCNC's own E-stop (N01).** With the link kept up for good, a board that stopped taking writes while still answering reads would have held K2 and the last step frequency indefinitely, because the firmware's watchdog counts reads. The firmware's timeout is the backstop again. As approved it followed LinuxCNC's enable output, which covers only the on-screen E-stop; corrected that evening (see "The SPI link follows the E-stop latch" above).
2. **KM1 dropping out trips the E-stop latch (S5, N02).** A dip in the 24 V that drops KM1 but not K3 can no longer bring drive power back by itself, and a short dropout cannot slip past the un-home line. (2026-11-15: true as long as the board takes the command that opens K2; review R3-02.)
3. **The latch cannot be reset while KM1's status contact reads "in" (EL07).** A dead contact or wire, or a welded KM1 or K2, now shows up as a refused reset. (2026-11-15: what it shows is a status contact reading "in". It is not a weld detector; review R3-03.) The alternative was to accept the contact's limits as they stood.
4. **Two presses of one button less than 0.25 s apart count as one (N06).** Documented, not changed. (2026-11-15: that holds only for a second CYCLE START. Two quick AIR presses both act; review R3-06.)
5. **Corrections.** The simulator scripts and board model, and the statements the review quoted from the notes and drawings.
The simulator was rerun on the HAL built from the Configuration note ([[Voron Cascade Electrical Configuration#Simulator run (2026-11-13)]]). Not changed, and still open: the mains budget (EL01), the bonding sizes (EL03; the 48 V −V bond was accepted as it is later the same day, entry above), the inlet (EL04), contactor duty (EL05), the status contact's rating (EL07: now detected, not cured), fuse and contact coordination (EL08), the unrated button contact (EL09), the FlexiHAL's input resistors (HW-04) and the bench items. The three HAL changes have had no independent review.
## 2026-11-14: H100 on the lower rail, on a printed DIN mount
Erik's call, made after the Rev N review bundle was cut: the lower rail runs on under the VFD to a 352, and the H100 hangs on it on a printed mount he is designing, instead of on two M5 screws tapped into the plate with 5 mm spacers. Layout I had drawn this as an option. The VFD's body stays where it was (a 257–347, h 55.6–205.6), so nothing around it moves; it stands 7.5 mm plus the mount's thickness off the plate, which leaves 42 mm less that thickness to the door. The rail becomes one 330.5 mm piece with a fourth screw, H4, at a 340 on the gantry upright's T-slot. What goes away: the two tapped VFD holes (one of which had to thread between two M8 bolt heads), the spacers, the VFD marking jig, and the second earth path through the lower-left bolt, so the VFD's PE terminal wire is now its only earth. The alternative was to stay with the bolted mounting of 2026-11-13.
## 2026-11-14: Rev N
Rev N is the revision that goes to the second independent review. It changes nothing by itself: it collects what happened after Rev M was approved on 2026-11-12, each item with its own entry here. From the parts and the CAD: the PE group linked by wire, the second grounding block and the 119 mm rear rail, Layout I with the relays in K2 K3 K4 K5 order, the lower-rail holes at a 40 / 140 / 240, the VFD 5 mm lower, the lower-rail grounding block with the drivers at 31 mm pitch, and the two marking jigs. From the review: no DC branch fuses and F1 / F2 as they are (EL11, EL08), and the documentation fixes D2, D3, HW-07, HW-08, SW-10 and SW-12. From the simulator run of the HAL: the axes are un-homed whenever drive power drops, and the SPI link stays up through an on-screen E-stop. From the FlexiHAL on the bench: the Aux pull-up resistors are fitted and are to be removed (HW-01), the input commons are common, so W24 is six cores, and the pin order and terminal names are read from the board. What still gates wiring is in the hold notice at the top.
## 2026-11-14: Remove the FlexiHAL's Aux pull-up resistors (review HW-01)
Measured on Erik's board (PCB revision A7), unpowered: `Aux +` to `Aux −` reads 1.00 kΩ on all four Aux outputs and open on Mist, and R81 and R82 are marked `01B`, the 1 % code for 1.00 kΩ. That is the published A7 schematic: R56, R77, R81 and R82 run from the Aux supply to each output, in parallel with the load, and with an output on at 24 V each dissipates 0.58 W in a 0.1 W part. K2 on Aux0 is on whenever the machine is out of E-stop. Erik's call: remove the four resistors. The Aux drivers then match the Mist, Flood and Spindle drivers, which use the same FET and diode with no parallel resistor, and the wiring, the drawings and the HAL stay as they are. The alternative needed no soldering: K2 on the Flood output and the two lamps on Aux at 5 V (P17 moved, lamp resistors about 330 Ω), with the HAL's latch output moved from `AUX0` to `COOLANT`. Wiring stays on hold until the rework is done and all four outputs read open.
## 2026-11-13: The SPI link stays up through an on-screen E-stop
Expatria's sample HAL, which Rev M had kept, lets LinuxCNC's E-stop switch the SPI link off (`iocontrol.0.user-enable-out` → `flexi.SPI-enable`). The board then carries on with the last packet it received until its own timeout. In the simulator that held K2 in for 0.63 to 0.72 s after an on-screen E-stop, and in 1 of 40 on-screen E-stops made during a 3000 mm/min move the board kept stepping at full speed for 0.55 s. Erik approved the change: `flexi.SPI-enable` is tied true, and `user-enable-out` is ANDed with the link status into the E-stop latch. K2's command and the stop then reach the board in the same servo cycle: no overrun in 40 simulated on-screen E-stops. Nothing in the wiring changes. What it gives up is the sample's side effect that every on-screen E-stop also forced all board outputs off through that timeout; they now go off because the HAL commands them off, and the timeout remains the backstop for a dead LinuxCNC, Pi or cable. It departs from Expatria's chain, so it has to be re-applied if a newer sample is ever taken, and it needs bench proof like the rest. **Changed 2026-11-14:** with the link up for good nothing dropped K2 if the board stopped taking commands, and "no overrun in 40" was measured after the stop call had returned, so it showed less than it said (second review, N01 and N03). The link now stops 0.5 s after any trip of the E-stop latch; see the 2026-11-14 entries.
## 2026-11-13: Un-home on loss of drive power
The simulator run of the Rev M HAL showed that LinuxCNC often left the axes marked homed after a door or E-stop-button trip, and always when KM1 dropped by itself, although the drives had lost power every time. LinuxCNC un-homes `VOLATILE_HOME` joints only when it runs its own E-stop or machine-off, and a hardware trip usually disables motion through `motion.enable` first. Erik approved one more HAL line: KM1's status contact, not inverted, drives `halui.joint.N.unhome` for all three joints (`power-lost`). In the simulator that un-homed the axes in 30 of 30 trips. The alternative was a written rule to re-home after every trip. The line needs no wiring, but it hangs on the same contact as the machine-on gate: a broken wire in the Bend pair reads as power present and defeats both (rung 6). **Added 2026-11-14:** the un-home line alone could miss a dropout shorter than halui's 20 ms poll, so KM1 dropping out now trips the E-stop latch as well, and a Bend pair that reads "in" with K2 open blocks the reset (second review, N02 and EL07; see the 2026-11-14 entry).
## 2026-11-13: Lower-rail grounding block, drivers at 31 mm pitch
The lower plate rail gets its own 7641K81 at a 248–254, between the SDR and the VFD, and the three drivers sit at 31 mm pitch instead of 33 to make room for it, as Layout I draws them. The 48 V −V bond lands on that block. The alternative was to keep the 33 mm pitch, bond the rail through its masked footprint with a paint-piercing washer under one rail screw, and put the −V bond on the PE group's spare screw. Erik's call: the grounding block is fine. It gives the rail and the −V bond a screw terminal that can be seen and metered, for 2 mm of driver pitch.
## 2026-11-13: VFD 5 mm lower (bolt head behind its upper mounting hole)
Drawing the VFD jig turned up a fault in Layouts H and I: the upper-right VFD hole at a 340, h 204.05 sat on the rim of an M8 bolt head inside the gantry upright's T-slot, 1 mm behind the plate, so the drill and tap would have run into it. Both layouts had described that spot only as "the T-slot of the gantry upright". Erik approved lowering the VFD 5 mm: body h 55.6–205.6, holes at (263, 62.05) and (340, 199.05). The upper screw then clears the bolt head by 2.35 mm. The cost is 5 mm of room under the VFD (17 mm to the filter, was 22); XT3 and KM1 gain the same 5 mm above it (28.4 and 13.6 mm), and the fit test still shows no collisions. The alternatives were worse: the other T-slot (a 310) puts the VFD into the lower rail, the VFD's other diagonal puts its lower-right hole on the T-nut of the M6 bolt at (340, 75), and raising it instead needs more than 15 mm. Two earlier statements are corrected at the same time, both from reading the CAD mesh instead of bounding boxes: the printed front-brace cover is hollow behind the right-hand rail hole and the VFD's lower-left hole, with its wall about 27 mm back, so nothing has to come off for drilling; and the VFD's clearance to the wire guard is 7.9 mm to the nut, not 12 mm to the fitting.
## 2026-11-13: No DC branch fuses; F1 and F2 stay as they are (review EL11, EL08)
The review asked for short-circuit protection on the small DC branches (EL11): the two USB-C leads on the 5 V supply, the fan and light on the 24 V, and the three 18 AWG driver feeds on the 48 V. Erik decided to fit none of them: not necessary for a hobby machine. Each branch relies on its supply's own limit, as in the UEB design. What that leaves is set out under [[Voron Cascade Electronics Wiring#What stays on, what gets switched]]: the 48 V and 24 V supplies shut down within about 3 s of a short and the 5 V supply does not, so its USB-C pigtails are the least protected wires in the bay. The option that was turned down was three inline blade-fuse holders (Pi, display, fan + light) with none on the 48 V; a third fuse terminal on the upper rail would also have needed KM1 moved back 6 mm.
F1 and F2 stay 1 A time-delay fuses from stock (EL08). The review is right that such a fuse does not hold a fault to 1 A, so the claim that F1 keeps the K3 contact inside its 1 A DC-13 rating is withdrawn: the fuses protect the branch wiring, and the 6 A relay contacts carry the fault for the moment it takes the fuse to clear. One check goes to [[Voron Cascade Bench Checks]]: with the latch set, an open F1 must drop K3, because the fuse terminal's blown-fuse LED passes about 2 mA around an open fuse.
## 2026-11-13: Layout I, and the lower-rail holes re-spaced
Layout I replaces Layout H as the bay placement this note defers to. It is Layout H with what has been decided, measured and cut since: the 119 mm rear rail with 15 blocks, the 8 mm end stops (which move F1, F2 and XT3 1 mm forward and leave 6.4 mm in front of KM1), 12 AWG bonding, and the Rev M wiring behind the controls. K2 and K3 swap places, so the relays run K2 K3 K4 K5 as drawing 3 shows them. Nothing else in the bay moved, and the fit test still shows no collisions. The lower-rail screws go from a 40 / 135 / 230 to a 40 / 140 / 240: Layout H's 95 mm spacing does not land on a rail slotted at the usual 25 mm pitch, and 100 mm does. Erik is printing a marking jig for them that registers in the two bottom corners of the frame opening ([[Voron Cascade Electronics Layout#Marking jig for the lower-rail holes]]). The alternative was to keep the Layout H positions and drill the rail to suit. Layout I also raises two things to check: the 119 mm rear rail and the 233.5 mm lower rail no longer both come out of the BOM's 350 mm piece, and with the 12 AWG gantry jumper a 4 mm air line fits the 10×20 drag chain only on top of the encoder cable.
## 2026-11-13: Second grounding block on the rear rail
The rear rail has more room than Layout H allowed for: Erik cut it to 119 mm, and the end stops in hand (McMaster 5575N22) are 8 mm, so a fifteenth block fits with about 7 mm to spare. The fourth 7641K81 from stock goes in as block 6, beside block 5. The two grounding blocks are joined through the rail they clamp to. The incoming earth and the wire link to the green blocks share block 6, so the path from the inlet to every device earth stays in copper; the plate stud and the 2020 frame lug land on block 5. That puts the frame lug back on the rail group (it had moved to the plate stud the day before) and leaves one screw spare. XT2 moves up one block: L-SAFE is 12–13 and N-SAFE 14–15. The other choice was a sixth green block, which would have needed a jumper bar that spans six. (Reversed on 2026-11-15: Erik took that other choice, with the one grounding block at block 11. Entry above.)
## 2026-11-12: PE group linked by wire (review EL02)
Erik tried the part: the 7641K51 jumper bar does not seat in the 7641K81 grounding block, as the review suspected. Block 5 is therefore linked to the jumpered green blocks 6–10 by a 12 AWG wire from its second screw to block 6. The link takes two of the group's twelve screws, and eleven conductors already landed there, so one had to move: the 2020 frame lug is now fed from a second ring lug on the plate stud. The alternative is a seventh PE block, which needs one more green block and another 6 mm on a rear rail that is already tight.
## 2026-11-12: Rev M — review fixes and accepted limits
Erik approved these after the independent review of Rev L. **Fixed:** K2 follows the E-stop latch, so a door or HALT trip keeps drive power off until the screen reset (S1). The valve's feed runs through KM1's NO aux contact, so every hardware trip closes the valve (S4); the status signal moves to the NC aux and gates machine-on through `motion.enable` (SW-03). CYCLE START and AIR act once per fresh, debounced press; the HAL line that would not load is gone; the E-stop latch runs after its inputs are read (SW-06, SW-04, SW-11). Bonding conductors go from 14 to 12 AWG, sized for the 20 A branch breaker (EL03).
**Accepted as they are, for a personal desktop mill with one operator:** the single-relay door channel, described as a convenience interlock (S2); a held or welded POWER ON button re-arming the hardware latch (S5); no hardware OFF other than the E-stop button and the breaker (S6); driver alarms left unwired until after the first cuts (HW-06). Access after a stop is a written rule (S3, next entry).
**Alternatives for the valve interlock:** KM1's spare main pole (its terminals start at 16 AWG, and it would put a 24 V circuit between the line and neutral poles), or an add-on auxiliary block (not checked against Layout H's clearances). Using the built-in NO pair is free, but it moves the status signal to the NC pair, where a broken wire reads as power present.
## 2026-11-12: Access after a stop is a written rule (review S3)
The review's S3 is that dropping the contactor does not make the enclosure safe to reach into: the spindle coasts, Z can sink with no holding torque, and nothing locks the door. This is a personal desktop mill with one operator, so a written rule covers it, and no door lock, spindle brake or Z brake is added. The rule sits with the other consequences under the E-stop chain; the coast-down time and the Z sink still have to be measured and written into it.
## 2026-11-12: Panel wire from the shelf, no single wire thinner than 20 AWG
The single wires that were 24 AWG (the FlexiHAL's relay-coil and input pairs, HALT, DOOR, the door switches, the front-panel lamps and AIR, the tool setter) are now 20 AWG blue for every + and S conductor and 18 AWG white/blue for every 0 V, G and −. That is the MTW on hand, along with 18 AWG blue and 14 AWG black, white and green/yellow. 24 AWG had been picked for the small terminals, but nothing needed it: the plugs for the FlexiHAL's 3.5 mm headers are listed for 28–16 AWG, and the review (EL10) pointed out that several of those wires land on XT3 blocks whose range starts at 22 AWG. The wires that land on KM1 are 18 AWG blue. The jacketed cables (sensor leads, RS485, the top-box multicore) are unchanged, and 18 AWG black for the 48 V + runs was ordered the same day, together with 18 AWG white.
## 2026-11-11: Rev L — no POWER OFF, 16 mm buttons, cycle start and AIR
Layout H's button problem (the 66 mm GCX stacks don't fit behind the front panel) turned into a review of what each control is for. The red POWER OFF button had NC contacts in both the latch chain and the HALT loop, which makes it a second E-stop that doesn't latch; with the E-stop for emergencies and the on-screen E-stop for the routine case (it drops K2, and K2 drops KM1) it had no job left, and neither did the red "power off" lamp. The Millennium UEB, which this design ports, has only a rocker and an E-stop and does its whole power latch in software; keeping a hardware **on** and letting software do the **off** keeps the property that matters — nothing software does can restore power past an open E-stop or door. Dropping the button also removed the only NC contact the small buttons couldn't provide, so the controls could become the Adafruit 16 mm illuminated momentaries already in hand: POWER ON (white) and AIR (blue) in the front panel's outer holes, the middle hole plugged until AUX has a function, and FEED HOLD (yellow) with a new CYCLE START (green) in a printed two-button box on the top panel — the two controls worth a tactile button mid-cut. CYCLE START is resume-only to begin with and its lamp flashes while paused; FEED HOLD only pauses; AIR toggles the mist output through LinuxCNC so the button and M7 / M9 share one state. White rather than green for POWER ON, so green means only "running / press to resume". Considered and dropped: IDEC LB-series 16 mm buttons (round, 24 V LED, 2 CO — they would have supplied the NC contact, but stock was thin and the button they were for went away), a three-button top box with POWER ON on top, and an AIR switch wired straight to the valve. The E-stop stays a GCX1131 in the top panel; a printed top plate that carries the box and moves the E-stop about 10 mm toward the door is being drawn.
## 2026-11-10: Rev K — McMaster 7641K51 blocks everywhere, colours, Layout H
Konnect-It blocks lasted one revision: there is a stock of McMaster 7641K51 (TE, 6 mm, choose-a-colour) on hand, so XT1, XT2, PE and XT3 are all 7641K51, and only F1/F2 stay Konnect-It fuse terminals. Colour scheme chosen by conductor class rather than circuit: black = un-switched line, red = KM1-switched line, white = every grounded conductor (N, N-SAFE, 24 V 0 V), green = PE, blue = +24 V, gray = labelled 24 V control nets; beige / orange / yellow unused. Placement now follows the main Cascade agent's Layout H: XT1 / PE / XT2 on a rear-panel rail (13 blocks), and K2–K5 + F1 F2 + XT3 in the ≈121 mm gap on the upper rail between the relays and KM1 — which is why XT3 is trimmed to 12 blocks with a few doubled conductors at the buttons and relays. Drawing 3 is no longer a bay layout (that is Layout H's job) but a terminal-strip drawing, `media/cascade-terminals.svg`.
## 2026-11-10: Rev J — Konnect-It terminal strip for XT3, and a door relay
Asked how signals were distributed, the honest answer was "they aren't": power went through 221-415 lever nuts and the control-ladder joins were implied. Counting the conductors showed XT3's two lever nuts short by one +24 port and seven 0 V ports, with the latch, door and lamp joins unaccounted for. XT3 is now a Konnect-It KN-T12 screw-terminal strip (gray 24 V nets, blue 0 V, jumper bars, the two KN-F10L24DC fuse terminals in the same strip) on a 150 mm rail on the bottom 2020; XT1 / XT2 stay on the Cascade's Wago mounts. Mains distribution could move to the same strip later if the lever nuts annoy. Konnect-It rather than the laser cell's WAGO TOPJOB S stock, to match the fuse terminals already in use.
Checking the joins also exposed a Rev C–I fault: the KW10 door switches were drawn with their NO contact in the 24 V contactor rung *and* their NC contact across the FlexiHAL `Door` input, but a KW10 has one common, so the NC side would have carried node L's 24 V into the input (or floated). A fourth 52003, **K5**, is now the door relay: both switches in series drive its coil from the fused 24 V, its NO contact sits in KM1's coil return, and its NC contact (common at 0 V) feeds `Door`. The ladder is ten rungs; rung numbers shifted by one from rung 3 on.
## 2026-11-10: Rev I — 22 mm illuminated GCX buttons
The BOM's M12/M16 flat buttons (the front-panel DXF actually has Ø10 holes at 40 mm pitch) are replaced by 22 mm AutomationDirect GCX illuminated operators — the same family as the laser cell's panel and the GCX1131 E-stop already in stock. The 30 mm bezels fit the existing pitch; the holes get drilled to Ø22.3. With 2-NC heads on both the E-stop and the red button, the HALT path becomes a fail-safe NC series chain (XOR jumper to the non-default position). The LEDs become status: green = latched, red = motor/VFD power off (from KM1's spare NC aux), yellow = feed hold (from FlexiHAL Aux1 via `halui.program.is-paused`). Rung 9 added. An ECX1040-2 NO pack was considered for the red button before the all-NC arrangement made it unnecessary.
## 2026-11-10: Rev H — air-blast valve on the Mist output
A third 52003 (K4) driven by the FlexiHAL `Mist` output switches a 24 VDC 2/2 NC air valve (Y1) through F2, so M7/M9 in G-code control a chip-clearing air blast. Chosen over driving the valve directly from the FlexiHAL because a 10 W valve coil exceeds the output's 250 mA limit. `Flood` stays free for a coolant pump and the Lovato contactor idea (vac auto-start) would hang off a fourth 52003 if ever wanted.
## 2026-11-10: Rev G — slimline interface relays
K2 and K3 are Murrelektronik 52003s from the laser-cell stock instead of the AutomationDirect ice-cube relays: 6.2 mm each on the rail, 1 CO rated 1 A DC-13 at 24 V (the KM1 coil is 0.38 A), 24 V AC/DC rectified input (no diode, and the FlexiHAL output sees 17 mA). Because the 52003 has a single contact, rung 3 is fed from the latch node (node L, between the seal-in and the K3 coil) rather than through a second K3 contact — same logic, one contact per relay. The two spare poles the ice cubes offered are gone; a future vac auto-start gets its own 52003.
## 2026-11-09: Rev E/F — 5 V supply
The 5 V rail only feeds the Pi 5 and the display, and the FlexiHAL cannot source it (its isolated MCU domain is fed *by* the Pi). Rev E picked a DDR-60G-5 DC-DC off the 24 V (isolated, no third mains device); Rev F the same evening went to the **MDR-60-5** from McMaster (3718N152) because the DDR was a week out and the MDR ships next day: 10 A, 40 mm wide, DC-OK relay, 78 % efficient. Also weighed: HDR-60-5 (cheapest, 6.5 A), Phoenix UNO-PS/1AC/5DC/40W (8 A, 35 mm, pricey), TDK DRB50-5-1 (6 A), Omron S8VK-G03005 (5 A, premium), RHINO PSRS-05-30 (5 A, 73 %). The WDR-60-5, which matches the SDR/NDR housing, is a 180–550 VAC unit and cannot run on 120 V.
## 2026-11-09: Rev D — drawn to the hardware in hand
The first drafts assumed the BOM's open-loop LDO motors with DM542T drivers, an LRS-350-48 and IRM supplies. The vault's [[Electronics]] note already records the real kit: StepperOnline 3-CLYS30-V20 closed-loop drivers and 5 A encoder motors, NDR-480-48, Delta 20DRGS5 line filter, HDR-30-5, uFlexiNET. Drawings and this note now match it; the DM542T ENA/ALM workarounds are gone (CL57Y MF stays open, ALM polarity already suits the FlexiHAL).
## 2026-11-08: Rev C — hardware START/STOP latch and door in the contactor chain
Adopted from a friend's working Cascade: a latching relay K3 (START NO + seal-in, STOP NC and E-stop NC in the chain) in series with the LinuxCNC-enable relay K2 and both door switches ahead of the KM1 coil. Power needs a physical START after any E-stop/STOP, can never come on with the door open regardless of software, and LinuxCNC can still depower on alarm. The green button becomes POWER ON (cycle start from the GUI); the red one needs an NC contact.
## 2026-11-08: Rev B — IronHorse 24 VDC contactor and SDR-120-24
Swap the BOM's 120 V-coil WEG contactor (already ordered; now spare) for an IronHorse HMC-18B30-11-DS (24 VDC coil, built-in 1NO/1NC aux, HMX1-SSVRC-DC suppressor) and the owned HDR-30-24 for a Mean Well SDR-120-24. The whole control chain is 24 V, the power-present status comes off the contactor's own aux contact, and the 24 V rail has headroom. F1's holder and fuse come from the laser-cell Konnect-It stock.
## 2026-11-08: FlexiHAL + Remora/LinuxCNC instead of BTT Scylla
Controller is the Expatria FlexiHAL, not the Scylla. Stack is LinuxCNC via Remora with the Pi on the FlexiHAL GPIO header. The planned USB-RS485 adapter is dropped: Expatria's sample config drives the board's own RS485 from the Pi UART, and an H100 vfdmod config already exists.
## 2026-11-08: UEB architecture, Cascade parts, electronics inside the bay
Port the UEB's un-switched / contactor-switched split onto the Cascade hardware rather than building a separate box.
## 2026-11-08: E-stop scope
E-stop drops the 48 V motor rail and the VFD mains through KM1; 24 V, 5 V, FlexiHAL and Pi stay up.
## 2026-11-08: Mains
110 V H100 kit on a 120 V / 20 A circuit with a NEMA 5-20P cord. Breaker (15 A) and inlet (16 A) are the limiting elements, so the VFD current limit caps spindle power rather than relying on Q1.
# Completed TODO items
Struck through in the wiring note's TODO on 2026-11-16. Where a line ends in an arrow and a highlight, that remainder is still open and stays in the wiring note's TODO.
- ~~Confirm the K4 → KM1 gap~~ (Layout I: 119.4 mm; the end stops in hand are McMaster 5575N22, 8 mm, so the group is 112 to 113 mm) → ==measure the end cover; KM1 can still slide back about 6 mm.==
- ~~Decide the lower-rail bond~~ (2026-11-13: the 7641K81 at a 248–254 with the drivers at 31 mm pitch, as Layout I draws it)
- ~~Decide the two HAL changes the simulator run proposed~~ (both approved 2026-11-13 and in the HAL: un-homing on loss of drive power, and the SPI link kept up through an on-screen E-stop; since 2026-11-14 the link stops 0.5 s after it, and KM1 dropping out trips the latch) → ==prove all of it on the bench ([[Voron Cascade Bench Checks]], parts A, B and C).==
- ~~Decide where the three GCX buttons go~~ (Rev L: no GCX buttons — 16 mm Adafruit buttons, AIR and POWER ON in the front panel, CYCLE START and FEED HOLD in a printed box on the top panel; see [[Voron Cascade Electronics Wiring#Operator controls]])
- ~~CAD check before drilling: bezels vs display, depth behind the panel~~ (Layout H: bezels fit on the front, 66 mm stacks don't — see the button decision above)
- ~~Confirm the H100's input voltage on its nameplate~~ (2026-11-14: `H100-1.5C2-1B`, 110 V single-phase in, 0–110 V 12.5 A out) → ==the spindle's nameplate and the manual's parameter pages are still to come.==
- ~~Order the 5 V supply~~ (MDR-60-5 ordered from McMaster 2026-11-09, 3718N152)
- ~~Confirm the E-stop head is 1NO + 1NC~~ (Rev I: GCX1131 with 2 NC; rung 5 is an NC chain and the HALT jumper moves)
- ~~Rail budgets and depths~~ (Layout H, fit-tested 2026-11-10: rear rail → 115 mm piece on the stock holes (since cut to 119 mm, 2026-11-13); NDR-480 ends 18.5 mm from the door, MDR-60 54.5 mm; upper rail 375 mm as CAD, lower 233.5 mm; the upper needs its own 375 mm piece) → ==rail stock: the lower rail is 330.5 mm since 2026-11-14 (it runs on under the VFD), so with the 119 mm rear rail it cannot come from the BOM's 350 mm piece; the upper rail needs its own 375 mm piece as well, so a 1 m length covers both (Layout J).==
# Wiring note changelog
The `# Changelog` section of the wiring note as it stood on 2026-11-16, before its entries were cut to one line each.
## 2026-11-08
- Created from a research pass over the Cascade repo/manual/BOM, the UEB wiring reference, the Flexi-HAL README/schematic and the remora-flexi-hal sample config. Three scripted SVG drawings added under `media/`.
- Rev B the same day: IronHorse HMC-18B30-11-DS (24 VDC coil) and SDR-120-24 replace the WEG contactor and HDR-30-24; rungs 2 and 4 redrawn; parts table, cable schedule and TODO updated. K2 part number corrected to 782-2C-24D / 782-2C-SKT (781 is the SPDT family).
- Rev C the same day: hardware START/STOP latch K3 and the door switches added to the KM1 coil chain (per a fellow builder's wiring); ladder now six rungs; button roles, parts, cable schedule, HAL table and TODO updated.
## 2026-11-09
- Rev D: reconciled with [[Electronics]] — CL57Y-V20 closed-loop drivers and 23HS40-5004D-E1000 motors replace the assumed DM542T/LDO set, NDR-480-48 replaces LRS-350-48, HDR-30-5 replaces IRM-30-5, Delta 20DRGS5 is a fitted part. Encoder cables added to the schedule; DIN-rail budget and bay depth checks added; H100 Modbus discrepancy between the vault note and Expatria's ini flagged. Note moved into the vault alongside [[Electronics]].
- Rev E: PSU3 is now a DDR-60G-5 DC-DC off the 24 V rail instead of a mains 5 V supply; rails table, budgets, cable schedule (W21), parts table, configuration and TODO updated; 5 V supply comparison recorded under Decisions.
- Rev F: PSU3 changed to the MDR-60-5 (McMaster 3718N152, ordered) for lead time; back on the mains side with FG, W21 removed, budgets and rail fit updated, decision and comparison extended.
## 2026-11-10
- Rev G: K2/K3 are Murrelektronik 52003 slimline relays (laser-cell stock); rung 3 is now fed from the latch node so each relay needs one contact; parts table, TODO and order list updated, 782-2C relays dropped.
- Rev H: rungs 7–8 add K4 (third 52003) on the FlexiHAL Mist output driving the Y1 air-blast valve through F2; cable W21, parts, budgets, an Air blast configuration section, TODO and the layout's cable list updated.
- Rev I: 22 mm GCX illuminated buttons (GCX1202/1201/1203-24L) and the GCX1131 E-stop replace the flat buttons and McMaster head; HALT is now an NC series chain (jumper non-default); rung 9 lamps (green = latched, red = power off, yellow = feed hold); W22, parts, HAL table, TODO updated. Configuration settings consolidated in the new [[Voron Cascade Electrical Configuration]] note.
- Rev J: XT3 becomes a Konnect-It KN-T12 terminal strip (18 blocks + the two fuse terminals) on a 150 mm rail on the bottom 2020, with a block-by-block table in the new Terminals section and on the bay drawing; K5 (fourth 52003) added as the door relay after the shared-common fault in the KW10 wiring was found — the ladder is ten rungs and rungs 3–10 renumbered; architecture, rails, cable schedule (W13, W14, W17, W20, W22, W23), parts, TODO and all three drawings updated. KN-F10 corrected to the 8 mm KN-F10L24DC that is actually in stock.
- Rev K: all terminals are McMaster 7641K51 blocks (XT1 / PE / XT2 on the rear-panel rail per Layout H, XT3 trimmed to 12 blocks in the upper-rail gap); colour scheme added; `cascade-bay-layout` retired in favour of `cascade-terminals` (placement is the main agent's Layout H); Terminals section, cable schedule, parts, TODO updated.
- Rev K (Layout H response): the main agent's fit test folded in — upper-rail order and 119.4 mm gap, rear rail → 115 mm piece, plate stud / frame lug / VFD bolt positions, two rail footprints to mask, lower-rail 7641K81 vs washer left to decide, drag-chain contents corrected to one motor + encoder + endstop + PE (4 mm air line), W1 / W2–W4 / W5–W7 / W21 routing fixed (Y1 outside the cabinet), GCX depth 66 mm → the front-panel buttons don't fit and the location is an open decision (top-panel row recommended).
- Rev K (later): NDR-480 confirmed to have two output pairs — U2 + U3 twin-ferruled into one, U4 in the other (W16, rails table, NDR section, dwg 1 and 2). PE block 5 becomes a 7641K81 grounding block (in stock) jumpered to the green blocks; bonding plan added under Grounding (plate stud, 7641K81 on each plate rail carrying the 0 V / −V bonds, frame, bed, gantry, carriages, spindle; ≤ 0.1 Ω acceptance); rear rail is 14 blocks.
## 2026-11-11
- Rev L: operator controls redone after the 2026-11-11 control review. SB3 POWER OFF and the H3 red lamp are gone (rung 1 = E-stop NC1 → [POWER ON ∥ seal-in]; rung 5 HALT = E-stop NC2 alone; KM1 21–22 unused); SB1 POWER ON is a white 16 mm Adafruit button in the front panel; SB2 FEED HOLD moves to a printed two-button box on the top panel beside a new SB4 CYCLE START (resume-only, lamp H4 on Aux2); new SB5 AIR on the `Aend` input toggles mist; the bare button LEDs get 4.7 kΩ series resistors (R1 R2 R4 R5). New [[Voron Cascade Electronics Wiring#Operator controls]] section and decision; overview, architecture, rungs 1 / 2 / 5 / 6 / 10, consequences, XT3 table, W14 / W22 / new W24, parts, FlexiHAL jumper notes, HAL table, TODO and all three drawings updated. K2 is now described as what it is — in while LinuxCNC is out of E-stop, not "machine on" — and two HAL lines that would not have loaded are corrected.
## 2026-11-12
- Rev L correction, found while preparing the review bundle and checked against the LinuxCNC 2.9 source: the AIR button's HAL needs a `toggle` ahead of `toggle2nist` — on 2.9 `toggle2nist` acts on every change of its input, so a momentary button wired straight in would have blown air only while held — and a machine-on gate, because LinuxCNC refuses coolant commands unless the machine is on (AIR therefore needs F2, not just F1). FEED HOLD's behaviour when nothing is running noted. HAL text is in [[Voron Cascade Electrical Configuration]]; drawing 2's HAL note updated.
- Added [[Voron Cascade Electronics Wiring#Wire numbers and labels]]: a wire-numbering rule (number = net, label = number + the terminal that end lands on), wire colours per [[Industrial Control Panel Wire Size and Color Standards]], and a full wire list with the label for each end of every conductor and cable. It also pins down two things the note left open: which block in each jumpered group takes which conductor, and that K2 11 picks up node L by a link from K3 A1.
- Independent review of Rev L received (11 blockers, 28 majors): hold notice added at the top. Documentation errors it found are corrected here and on the drawings — the operator sequence uses the screen's buttons (on QtDragon HD the F1 / F2 keys only switch off); `SCALE`, not `STEP_SCALE`; re-homing is not enforced by the sample INI; the kit's motor cables are unshielded; RS485 is the Pi's UART 2 on GPIO 4 / 5; DOOR cannot assert when the whole 24 V rail is down; drawing 3 no longer suggests dropping XT3:12, which carries the 0 V → PE bond. The design changes the review calls for wait for Rev M.
- Panel wire re-specified to the MTW on hand: 20 AWG blue for every + and S conductor, 18 AWG white/blue for every 0 V, G and −, 18 AWG blue at KM1; no single wire is 24 AWG any more. The 48 V + runs stay 18 AWG black, which is on order. Cable schedule, wire table, label tables, parts and the order list updated, with a decision entry; drawings 1 and 3 updated to match.
- Review S3 decided: access after a stop is a written rule, with no door lock or brake. Rule added under the E-stop chain, with a decision entry.
- **Rev M**, approved by Erik after the review. K2 follows the E-stop latch instead of `user-enable-out`, so a door or HALT trip keeps drive power off until the screen reset (S1). The valve's feed runs through KM1's NO aux 13–14 (S4); the status signal moves to the NC aux 21–22 and gates machine-on through `motion.enable` (SW-03). CYCLE START and AIR act once per fresh press (SW-06); the HAL is in [[Voron Cascade Electrical Configuration]]. Bonding conductors are 12 AWG (EL03). Accepted and written down: the single-relay door channel (S2), a held POWER ON re-arming the latch (S5), E-stop button to disarm and Q1 to isolate (S6), driver alarms unwired until after the first cuts (HW-06). Rungs 2, 6 and 9, the consequences list, operator controls, grounding, cable schedule, wire labels (`24F2` re-routed, new `24K`, `BEND` on KM1 21 / 22), parts, TODO and all three drawings updated. Still on hold for the Aux pull-up, PE jumper and nameplate checks; branch fuses (EL11) are not decided.
- PE group (review EL02, answered on the part): the jumper bar does not fit the 7641K81, so block 5 links to block 6 by a 12 AWG wire and the 2020 frame lug moves to the plate stud. Grounding, terminal tables, wire labels, parts, drawing 3 and the hold notice updated; two checks now gate wiring.
## 2026-11-13
- Rear rail, from Erik's parts: the rail is cut to 119 mm and the end stops are McMaster 5575N22 (8 mm), so a fifteenth block fits. The fourth 7641K81 goes in as block 6. The incoming earth and the link to the green blocks share it, the plate stud and frame lug land on block 5, the frame lug is back on the rail group, and one PE screw is spare. XT2 is renumbered 12–15. Grounding, terminal tables, wire labels, parts, TODO and drawing 3 updated, with a decision entry.
- Layout I: this note now defers to [Cascade Layout I](https://claude.ai/artifact/SMifPvbaUPLKZVGp382NJC) instead of Layout H. Upper-rail positions updated for the 8 mm end stops (F1 F2 at a 234.8–250.8, XT3 at a 250.8–325.6, 6.4 mm in front of KM1). Lower-rail holes re-spaced to a 40 / 140 / 240, with a new [[Voron Cascade Electronics Layout#Marking jig for the lower-rail holes]] section, the drawing and DXF in `media/cascade-lower-rail-jig.*`, and a decision entry. Grounding row 9 and the drag-chain TODO carry Layout I's chain fill for the 12 AWG jumper; rail-stock and lower-rail TODOs added; Related notes and drawing 3 point to Layout I, and drawing 3 gives the K4 → KM1 gap as the fit-tested 119.4 mm (it said ≈ 121).
- Layout I, relay order: the upper-rail relays run K2 K3 K4 K5 at a 202–226.8 (Layout H had K3 K2 K4 K5), so the layout and drawing 3 agree.
- Review EL11 and EL08 decided. No branch fuses on the 5 V, the 24 V fan / light or the 48 V driver feeds: Erik accepted the omission. F1 and F2 stay 1 A time-delay. The claim that F1 keeps the K3 contact inside its DC-13 rating is withdrawn and the F1 / F2 parts rows reworded; a paragraph under the rails table says what each supply does into a short; the hold notice is updated and a decision entry added. One check added to [[Voron Cascade Bench Checks]].
- VFD lowered 5 mm (Layout I): the old upper-right mounting hole ran into an M8 bolt head in the upright's T-slot, 1 mm behind the plate. New "VFD mounting holes and marking jig" section (replaced on 2026-11-14 by [[Voron Cascade Electronics Layout#The H100 on the lower rail]]) with `media/cascade-vfd-jig.*`, and a decision entry; the VFD bolt position, the masking and KM1 TODOs and the XT3 clearance follow the new height. Corrections from the CAD: the front-brace cover is hollow behind the right-hand rail hole (the rail-jig sheet is Rev B, which also names the part on the bottom extrusion as the Cascade's 20 mm cable mount), and the VFD is 7.9 mm from the wire-guard nut, not 12 mm. The cable-entry paragraph now carries Layout I's drag-chain figures.
- Review fixes approved by Erik (D2, D3, HW-07, HW-08, SW-10, SW-12). The KM1 status pair is cable W25 in the schedule and on its two label rows, and W15's route names the un-switched branches to the SDR and MDR (D2). N and N-SAFE are described where they are (blocks 3–4 and 14–15), the driver ribbons list both alarm pins, and the HAL table says that HALT now goes through the OR with DOOR (D3). A bullet under FlexiHAL jumpers gives both sets of terminal names and warns about the pin order of the input blocks (HW-07). The 27 V Zener on the FlexiHAL's input is no longer called a clamp (HW-08). The air-blast note no longer asks a `timedelay` for a spindle lead (SW-10), and the HAL table says that the sample still drives `STEP_EN` (SW-12). Drawings 1 to 3 follow: Z1's earth lead and KM1's rail on drawing 1, the Bend terminal with KM1 21–22 and both alarm pins on drawing 2, the neutral wording on drawing 3.
- Lower-rail bond decided: the 7641K81 at a 248–254 with the drivers at 31 mm pitch. Grounding row 3, the `PE XPE:11` label row and the TODO updated; decision entry added.
- Rev M HAL run in LinuxCNC's simulator ([[Voron Cascade Electrical Configuration#Simulator run (2026-11-13)]]). It loads and behaves as this note describes, with two exceptions that wait for a decision: LinuxCNC often leaves the axes marked homed after a door, E-stop-button or KM1 trip, and an on-screen E-stop made during a move can leave the board stepping for about 0.6 s. The hold notice, rung 2, the operator sequence and the TODO say so; the claim that the INI enforces the re-home after any E-stop is withdrawn.
- Un-homing on loss of drive power approved by Erik and added to the HAL (`power-lost`). Rung 6, the operator sequence, the HAL table, the hold notice and the TODO updated; decision entry added; drawing 2's HAL box has the line.
- SPI link kept up through an on-screen E-stop, approved by Erik and added to the HAL: `user-enable-out` feeds the latch instead of `flexi.SPI-enable`. Rung 2, the HAL table's footnote, the hold notice and the TODO updated; decision entry added; drawing 2's HAL box has the line.
## 2026-11-14
- FlexiHAL checked on the bench (PCB revision A7). The Aux pull-ups the review warned about are fitted: 1.00 kΩ on all four Aux outputs, open on Mist. Erik is removing R56, R77, R81 and R82; hold notice, the FlexiHAL jumpers section and the TODO updated, decision entry added, and drawing 2's settings box has the line. Terminal names and input pin order confirmed from the board's silkscreen (HW-07): the two flags on that bullet are gone.
- FlexiHAL commons checked: the input `G` pins are common with each other and with the 24 V `−` terminal, and `Aux1 +` is common with `Aux2 +`. W24 is six cores; its schedule row, the label note and the cable-label row say so, and rung 7 notes that `Door G` on XT3 0 V is the board's own 0 V.
- **Rev N.** The revision letter moves to N in this note, the Configuration note, the Bench Checks note, the three drawings and the Layout I page, for the second independent review. Nothing in the design changes with the letter: the decision entry of the same date lists what Rev N contains. The hold notice is rewritten for it.
- H100 moved to the lower rail on a printed DIN mount: Erik's decision, made after the Rev N review bundle was cut, so the bundle still shows the bolted mounting. The lower rail is 330.5 mm, a 21.5–352, with a fourth screw H4 at a 340. The [[Voron Cascade Electronics Layout#The H100 on the lower rail]] section replaces "VFD mounting holes and marking jig" and embeds the new reference drawing, `media/cascade-vfd-rail-mount-ref.*`; the lower-rail jig drawing is at Rev C with a fourth bore. Grounding row 4 (the VFD's plate bond is gone), the parts table, the masking, lower-rail and VFD TODOs and the rail-stock item follow; decision entry added. Drawings 1 to 3 do not change.
- Second independent review (of Rev N) and what followed. Three HAL changes approved by Erik: the SPI link stops 0.5 s after LinuxCNC's own E-stop (N01), KM1 dropping out trips the latch (S5, N02), and the latch cannot be reset while KM1 reads "in" (EL07). Rungs 2 and 6, the anti-restart bullet, the operator sequence, the HAL table and the hold notice follow; decision entry added. Also recorded: the input series resistors (HW-04) under the FlexiHAL jumpers, presses under 0.25 s apart counting as one (N06), and that the ≈7–8 A motor-current setting is not a proven mains limit (EL01). Corrected statements: KM1 could come back by itself after a 24 V dip, with no stuck button (S5); POWER ON's contact can break the coil current on a short press (EL09); the 52003 is not shown to survive the 24 V supply's fault current (EL08); the 48 V supply's OVP does not protect the drivers (EL06); the inlet is not established to be a TRUE1 (EL04); the 48 V −V bond's gauge and termination are flagged as unsettled (EL03, D1). Drawing 1 no longer says an unwired driver alarm shows as a following error (HW-06) or names the inlet as a TRUE1; drawing 2's HAL box has the new lines; drawing 3 no longer offers twin-ferrule links (D1).
- 48 V −V → PE bond: Erik accepted it as it is in the wire labels, 18 AWG in a twin ferrule with U4's −V feed, as a deviation from the 12 AWG rule. Grounding intro and row 3, the label row and drawings 1 and 3 say so; decision entry added.
- VFD nameplate read: `H100-1.5C2-1B`, 110 V single-phase in, 0–110 V 12.5 A out, no input current printed. Mains budget, hold notice, H100 section and TODO updated. Drawing 1's VFD box has the nameplate and no longer says the VFD is bolted to the side plate, which the rail-mount decision had made wrong.
- The SPI link follows the E-stop latch, approved by Erik: it stops 0.5 s after any trip, not only after the on-screen E-stop as the afternoon's fix had it. Rung 2, the HAL table's footnote, the hold notice and the TODO follow; decision entry added, with the case it does not cover; drawing 2's HAL box has the line.
## 2026-11-15
- Third independent review received (of the Rev N package of 2026-11-14): it does not pass Rev N. Its two new blockers were reproduced in the simulator. Erik's decisions are in two new decision entries: the "board answers but ignores commands" fault is an accepted limit, with the E-stop button as the stop; the corrections are made; the bench comes next, not a fourth paper review. The hold notice is rewritten.
- **The door acts through LinuxCNC only, and only while the spindle is commanded on** (Erik's decision). One wire moves: KM1 A2 goes to K5 A2 (0 V) instead of K5 14, so K5's contact is out of KM1's coil circuit; the wire is now `24N`, 18 AWG white-blue. The overview, the architecture block, the E-stop chain section (retitled: the door is no longer in it), rungs 2, 3, 4, 7 and 9, the consequences list, the operator sequence, the XT3 table, cable W20, the wire-number and label rows, the HAL table, the TODO and drawings 1 to 3 follow.
- Accepted limit written into rung 2 and the anti-restart bullet. Withdrawn: "KM1 stayed out, 15 of 15" in the decision entry of 2026-11-14, and "can no longer bring drive power back by itself" without its condition in rung 6 (R3-01, R3-02).
- Corrections from the third review. Rung 6 and the second review's decision entry say what the reset check detects and that it is not a weld detector (R3-03). CYCLE START and AIR: two quick AIR presses both act (R3-06). HALT's own input circuit is named under the FlexiHAL jumpers, so the 117 mW figure applies to Bend and Door only (R3-07). W19 no longer routes the frame bond through the plate stud (R3-08). The valve plug's suppressor is no longer called a diode. The mains budget no longer offers a 240 V feed as a fix. KM1's coil power is flagged: 9 W assumed here, 6 W in the data sheet's table, to be settled by measuring the coil. The hold notice names the input-resistor and step-timing checks as coming before the steps that depend on them.
- **Rev O**, Erik's decision. The revision letter moves to O in this note, the Configuration note, the Bench Checks note, the three drawings and the Layout I page. It covers the entries above for this date: the door, the accepted limit and the corrections. Nothing else changes with the letter. Decision entry added; the hold notice and the line under the Overview heading say Rev O.
- **Rear-rail PE group: one grounding block, at block 11** (Erik's decision, later the same day). Blocks 5–10 are plain green and jumpered, block 11 is the 7641K81, and a 12 AWG wire links 10 to 11. The incoming earth and every device earth land on the green blocks; the lower-rail feed moves to block 6 and the gantry jumper to block 7. Grounding intro and row 2, the colour table, the rear-rail table, the label rows, the parts row (26 blocks, six of them green; three 7641K81 in use), the TODO and drawing 3 follow; decision entry added, and the entry of 2026-11-13 it reverses is marked. The wiring stays at Rev O.
- **Layout J**, in its own note. Bay placement now lives in [[Voron Cascade Electronics Layout]]; this note's references to the layout say Layout J, and Related notes lists the new note in place of the published Layout I page.
- Door change, one line that was missed: the panel-wire paragraph under Wire numbers and labels still listed `KM-` among the 18 AWG blue wires at KM1. The wire on KM1 A2 is `24N`, 18 AWG white/blue, and the paragraph says so.
## 2026-11-16
- The last three entries of 2026-11-15 (the rear-rail PE group, Layout J, the missed panel-wire line) reached the vault today. Erik had edited this note by hand on the evening of 2026-11-15, after the Rev O version was written. Claude's session could not read that edit, and with Erik's approval this version was written over it, so whatever he changed then is not in this text.
- Decisions sorted newest to oldest, at Erik's request. The six entries of 2026-11-08 moved from the top of the section to the bottom, with Rev C and Rev B ahead of the four original decisions of that day. No entry is reworded.
# References
- [[Voron Cascade Electronics Wiring]]: the current design, the decisions in force and the open TODO items
- [[Voron Cascade Electrical Configuration]], [[Voron Cascade Electronics Layout]], [[Voron Cascade Bench Checks]]: the companion notes, each with its own changelog