#voron #voroncascade #cnc #electronics #commissioning #checklist # Overview What the bench still has to answer for Rev O of [[Voron Cascade Electronics Wiring]] and [[Voron Cascade Electrical Configuration]]. It comes from the independent review of Rev L on 2026-11-12; the codes in brackets are that review's finding numbers. N01 to N06 are the new findings of the second review, of Rev N, on 2026-11-14, and R3-01 to R3-08 those of the third, on 2026-11-15. Part 1 needs a meter and a camera and nothing powered. Part 2 is powered checks, in the order they become possible. Add each result as a sub-bullet under its item. The decisions the review asked for (S2, S3, S5, S6, HW-06) were made on 2026-11-12, and the branch-fuse question (EL11) on 2026-11-13: none are fitted. All are recorded in the wiring note's Decisions. The second review came back on 2026-11-14 and did not pass Rev N. Three HAL changes followed the same day, the first of them corrected that evening (the SPI link stops 0.5 s after any trip of the E-stop latch; KM1 dropping out trips the latch; the latch cannot be reset while KM1 reads "in") and the simulator was rerun with corrected scripts ([[Voron Cascade Electrical Configuration#Simulator run (2026-11-13)]]); those changes have not been reviewed. The third review came back on 2026-11-15 and did not pass it either. Erik's decisions that day (wiring note, Decisions): the fault its blockers sit under, a FlexiHAL that answers the Pi but ignores its commands, is an accepted limit with the E-stop button as the stop; the door now acts through LinuxCNC only, and only while the spindle is commanded on; and the bench is next, not a fourth paper review. **So this note is now the plan.** The order: the Aux rework, then part A on the FlexiHAL alone, starting with the input-resistor check, because the bare-board procedure holds `Bend` closed with a wire link and that is the load the check is about. The two open items at the top of part 1 still gate wiring the bay: the Aux rework and the nameplates. Two more checks come before the steps that depend on them: the input resistors (HW-04, part A) before any input is left closed for long, and the step timing (SW-07, part D) before the drivers are run. The door change has not been reviewed. Part C repeats the simulator's checks on the real board. That review also pointed out what some checks here cannot show: ten closures or twenty presses say that something works today, not how long a contact lasts. Those items now say so. # TODO ## Part 1: information to collect (nothing powered) The two items marked as gates hold up wiring: do not wire the bay until they are done. Two checks in part 2 gate what follows them: the input resistors (part A, first item) and the step timing (part D, first item). The Aux pull-up question and the PE strip are answered. - [x] **FlexiHAL Aux pull-ups (HW-01).** Board unpowered, Aux plugs empty. Ohms from `Aux0 +` (red lead) to `Aux0 −` (black lead), then Aux1, Aux2, Aux3. About 1 kΩ means the schematic's pull-up resistors are fitted, and an Aux output held on at 24 V would overheat them; open circuit means they are not. Record the board revision printed on the silkscreen too. - 2026-11-14: fitted. PCB revision A7, printed on the back. `Aux +` to `Aux −` reads 1.00 kΩ on Aux0, Aux1, Aux2 and Aux3; Mist reads open. R81 and R82 are marked `01B` (1.00 kΩ, 1 %). P17 is on MAIN (24 V). Decision: remove R56, R77, R81 and R82 (wiring note, Decisions). - [ ] ==Gates wiring== **Aux rework (HW-01).** Remove R56, R77, R81 and R82. R81 and R82 are the pair by the `CYC/ST` label; R56 and R77 are the two in the column nearest the AUX1 terminal. A resistor to remove has one pad on an `Aux +` screw and the other on an `Aux −` screw. Leave R70, R66, R76 and R73 beside them: gate pull-downs, same marking. Afterwards: `Aux +` to `Aux −` (red lead on `+`) reads open on all four, and each of the four that stay still reads about 1 kΩ. No Aux output on at 24 V before this is done. - [ ] ==Gates wiring== **VFD and spindle nameplates (EL01, V1, V2).** Photograph both. VFD: model, input volts / phases / amps, output amps and kW. Spindle: volts, amps, Hz, rpm, poles. Also the title page of the manual that came in the box, and its pages for F004, F153, F155, F164, F165 and F169. - 2026-11-14, VFD: G-Penny `H100-1.5C2-1B`. Power 1.5 kW; input 1PH 110 V 50/60 Hz; output 3PH 0–110 V, 12.5 A, 0–1000 Hz. No input current is printed. It is the 110 V unit. Its output tops out at 110 V, so the spindle's plate has to say 110 V; note its amps against the 12.5 A. Still to come: the spindle's plate, the manual's title page and its parameter pages. - [x] **PE strip (EL02).** Does the 7641K51 jumper bar seat and tighten in the 7641K81 grounding block next to the green blocks? - 2026-11-12: no. The bar does not fit the 7641K81; it has two screw terminals, so block 5 is linked to block 6 by a 12 AWG wire, and the 2020 frame lug moves to the plate stud (wiring note, Grounding). - 2026-11-13: the rear rail is 119 mm and the end stops are 8 mm (McMaster 5575N22), so a second 7641K81 goes in as block 6. The link now runs from block 6 to block 7 and the frame lug is back on the rail group. - 2026-11-15: back to one grounding block. Erik replaced one 7641K81 with a plain green block and moved the other to the right-hand end of the group: blocks 5–10 green on one jumper bar, block 11 the 7641K81, and a 12 AWG wire from block 10 to block 11. The incoming earth lands on a green block. ==Check that a jumper bar spanning six blocks is in stock.== - [ ] **E-stop contact blocks (EL13).** On the GCX1131 in stock: how many contact blocks are fitted, and the terminal numbers printed on each. It needs two NC; the operator ships with one. - [ ] **Terminal markings as printed** on Q1, the Z1 filter and the VFD (mains in, motor out, RS485, PE). The wire labels use functional names for these until the real ones are known. - [ ] **Mains inlet and cord (EL04).** Part numbers on the inlet and the cable connector in hand, the cord's plug (5-15P or 5-20P) and its gauge. The drawings' part number, NAC3MPXXA, is a powerCON XX, not the TRUE1 the notes used to name; the review gives the XX as 20 A / 127 V in North America and 16 A / 250 V in Europe, so read the rating off the part too. - [x] **FlexiHAL commons (W24).** Continuity between the `G` pins of Start, Pause, Aend, Door, Halt and Bend, and between `Aux1 +` and `Aux2 +`. Common means the top-box cable is six cores, not eight. - 2026-11-14: all common. The 24 V `−` screw beeps to `Halt G`; the `G` pins of Halt, Door, Bend, Aend, Pause and Start beep to each other; `Aux1 +` beeps to `Aux2 +`. So the top-box cable is six cores, and `Door G` on XT3 0 V (rung 7) is the board's own 0 V. The pin order comes from the board itself: its back prints `12V SIG GND` at every input block, which from the top, with the inputs along the upper edge, is G, S, V left to right. - [ ] **Endstop sensors.** Label photo: NPN or PNP, NO or NC, supply range. - [ ] **Tool setter and probe (HW-05).** Meter across the two wires, at rest and pressed: NO or NC. Same for the 3D probe if there is one. - [ ] **Door switches.** Part number, and which tabs are COM / NO / NC by meter. - [ ] **Touchscreen (D4).** Native resolution, size, and orientation as mounted. QtDragon HD is a fixed 1920 × 1056 layout. - [ ] **Software versions.** LinuxCNC (`linuxcnc_var LINUXCNCVERSION`, or the title bar), the Flexi-Pi image name and date, and the Remora firmware file flashed to the FlexiHAL. - [ ] **Pi UART overlay (HW-03).** On the Pi: `grep -A6 uart2-pi5 /boot/firmware/overlays/README` (or `/boot/overlays/README`). Confirms the overlay name and that it puts UART 2 on GPIO 4 / 5. - [ ] **Spindle connector.** Continuity from GX16 pin 4 to the spindle body. - [ ] **24 V loads (EL11).** Current drawn by the Daylight bar and by the bay fan at 24 V, from the label or a meter, for the 24 V budget. (Branch fuses were decided against on 2026-11-13.) What that leaves open is whether the thin leads survive a short: note the gauge of the two USB-C pigtails and of the fan and light leads, if it is printed. - [ ] **Twin ferrule at K5.** Do two 18 AWG wires in a 2 × 1.0 mm² twin ferrule seat in a 52003 relay terminal? The 0 V link at K5 11 needs it, and since 2026-11-15 so does K5 A2, where KM1's coil return joins the A2 → 11 link. - [ ] **F1, F2 and the 52003 (EL08).** The part number on the 1 A fuses, for their clearing curve, and any short-time or overload figure published for the 52003's contacts. The SDR-120 can push 7.5 A for about 3 s; nothing yet shows whether the fuse or the relay contact gives way first. - [ ] **12 AWG bonds (EL03).** Does a 12 AWG ferrule or lug go into the FG terminals of the NDR-480, SDR-120 and MDR-60, onto the inlet's earth tab and into the 7641K81? Does a high-strand 12 AWG fit the drag chain with the other cables? The 48 V −V → PE bond is the exception: Erik accepted it on 2026-11-14 as 18 AWG in a twin ferrule with U4's −V feed. Check that a 2 × 1.0 mm² twin ferrule seats in the NDR's +V and −V terminals; three of the four carry one. - [ ] **DIN rail stock (Layout J).** Slot pitch and slot length of the rail in hand: the lower rail's hole spacing (100 mm) and its marking jig assume 25 mm pitch. Also what is left of the piece the 119 mm rear rail was cut from: the lower rail needs 330.5 mm now that it runs on under the VFD. And the width of a 7641K51 end cover, which Layout J takes as 2.8 mm. - [ ] **H100 mounting pattern and size.** Hole spacing across and up, hole diameter, where the upper-right hole sits from the body's top and right edges, and the body's width, height and depth. Layout J and the rail-mount reference (`media/cascade-vfd-rail-mount-ref`) assume holes 77 × 137 mm apart on a 90 × 150 × 110 mm body. Wider than about 98 mm and the body overlaps the wire-guard nut. What the back of the case is, plastic or a bare heatsink, and how warm it runs: the printed rail mount sits against it. Also which way the power and motor terminals face: there is 17 mm under the VFD to the filter. - [ ] **When they arrive:** KM1 coil resistance A1–A2 (about 64 Ω is the 9 W coil the design assumes, about 96 Ω a 6 W one, which is what the third review reads in the HMC data sheet's coil table); whether an 18 AWG ferrule clamps firmly in KM1's A1 / A2, 13 / 14 and 21 / 22; the rating printed for KM1's built-in aux contacts (the valve now runs through 13–14); the valve plug's terminal range and the circuit printed on it (D5: its sheet says 16–12 AWG and a 0.24–0.32 in cable, and the review says its suppressor is a Zener, not a plain diode). ## Part 2: bench checks (powered) ### A. FlexiHAL and Pi alone, on bench 24 V Since 2026-11-14 the HAL needs KM1's status contact to behave: the screen's reset is refused while `Bend` reads "KM1 in", which is what an unconnected input reads, and machine-on is refused while it reads "KM1 out". On the bare board a wire link across `Bend S` and `Bend G` stands in for the contact: fitted is KM1 out, removed is KM1 in. Fit it, reset the E-stop on the screen, then take it out before machine-on. Putting it back while the machine is on trips the E-stop latch, as KM1 dropping out would, and so does a ragged pull that remakes the link for more than 5 ms: fit it and reset again. A small toggle switch on the pair is easier. Do the first item below before anything else: this procedure holds `Bend` closed, and that is the load it asks about. - [ ] ==Do this first.== **Input current and resistor (HW-04).** mA through one closed contact on Bend or Door. The schematic implies about 11 mA, which is about 117 mW in a series resistor published as 1 kΩ, 0.1 W: reading 11 mA confirms the overload, it does not clear it. So also find that resistor on the board, note its size and marking, and take its temperature after ten minutes with the input closed. Bend is closed whenever KM1 is out, which on the bare board means whenever the wire link above is fitted. Door is closed whenever a door is open, and since 2026-11-15 that can be for as long as a job is being set up. HALT is closed the whole time the E-stop is out, but the schematic gives it a circuit of its own (R45 2 kΩ, R57 and R91 4.7 kΩ; A7 schematic sheets 9 and 13), so the 117 mW figure is not its: measure HALT's current and find its parts as well (R3-07). - [ ] **Input polarity in `halshow`.** The HAL pin value at rest and actuated for HALT (jumper in the NC position; the HALT LED should be off at rest), DOOR, A_LIMIT, B_LIMIT, CYCLE_START, FEED_HOLD, and one axis limit with a real sensor. The inputs only update while the SPI link is up, and the link now stops 0.5 s after LinuxCNC goes into E-stop. For this check hold it up by hand: `halcmd unlinkp flexi.SPI-enable`, `halcmd setp flexi.SPI-enable 1`, then one click on the screen's E-stop reset, without pressing POWER ON (KM1 then stays out). The link stays up whether or not the reset is accepted. Restart LinuxCNC afterwards. `power-present` must read false with KM1 out and true with it in, and `power-lost`, the other pin of the same input, the opposite. DOOR needs particular care since 2026-11-15: a wrong polarity shows nowhere until the spindle is started, and it would stop the spindle with the door shut and let it run with the door open. `door-open` must read false with the door shut and true with it open; on the bare board a link across `Door S` and `Door G` is "door open". - [ ] **PROBE input (HW-05).** Value with the tool setter at rest, pressed and unplugged; the probe at rest and triggered; both triggered together. Do this before any probing move. - [ ] **Lamps.** Each button LED on 24 V through 4.7 kΩ, then 2.2 kΩ: bright enough in room light? On an Aux output: fully dark when the output is off? Do the Aux part only after the Aux rework. - [ ] **RS485 port.** With `dtoverlay=uart2-pi5`, which `/dev/ttyAMA*` appears. - [ ] **5 V.** Volts at the Pi header and at the display with everything running; `vcgencmd get_throttled` reads `0x0`. - [ ] **5 V isolation.** Ohms from the 5 V rail's 0 V to the 24 V rail's 0 V, and to PE, with the display's USB and HDMI plugged in. Expect open; note any path. - [ ] **Time to drop K2 (SW-02, N01).** Three timings; a scope on the output, or just "instant / about a second / never". (a) On-screen E-stop to Aux0 turning off: at once. (b) The same E-stop to `flexi.SPI-status` going false: about 0.5 s (`halscope`, or watch the inputs freeze in `halshow`). Then open the HALT loop instead: the link has to stop 0.5 s after that trip too, and after a reset that is refused. (c) The firmware's timeout: `kill -9` of LinuxCNC with K2 and Mist on, to Aux0 and Mist turning off; the same with the Pi's power pulled. By the firmware source that is 0.7 to 0.8 s after the last packet. The simulator's board model was written from the same source, so it says nothing about the real board; if the outputs stay on much longer than a second, the stop in (b) has no backstop behind it and the HAL needs another look. The fault N01 describes, a board that answers reads but takes no commands, cannot be made on the bench; (c) is the part of it that can be measured. Since 2026-11-15 that fault is an accepted limit with the E-stop button as the stop, and (c) says how good the backstop behind it is. - [ ] **The firmware's separate reset input (parked option, information only).** With K2 and Mist on: `halcmd setp flexi.PRU-reset 1` (Pi GPIO 25). Do the outputs go off, how fast, and does the board come back by itself? By the source the firmware stops its main loop and its own watchdog restarts it after about 2 s. Nothing in the HAL uses this pin; it might one day give a way to switch the board off that does not depend on the board taking a command. - [ ] **On-screen E-stop during a move (simulator finding 2).** No drives needed: scope a step output, or watch it on a logic probe. With a program moving one axis at 3000 mm/min, press the on-screen E-stop, twenty or thirty times. Do the step pulses ever carry on after the press, and for how long? With the HAL as first written 1 of 40 did in the simulator, for 0.55 s. With the HAL as it stands (run of 2026-11-15), 40 simulated E-stops all ended the same way: the step frequency was already zero when the E-stop reached the HAL in 38, went to zero in the same cycle in 1 and 2 ms later in 1; K2's command went off 1 ms after it in all 40; and the axis moved 0.05 mm at most after that point. Counted from the script's call it moved 1.1 to 1.6 mm, by step count, most of it before LinuxCNC had acted on the request. The earlier figure, "none of 80", counted 40 runs of a HAL that has since been replaced and was taken after the stop call had returned (N03, N05). The pulses must stop every time. ### B. Control ladder on the bench Relays, fuses, E-stop, POWER ON and door switches wired as rungs 1–4, with KM1 or a 24 V relay in its place. Since 2026-11-15 KM1's A2 goes to K5 A2 (0 V), not to K5 14: the door relay is no longer in rung 4. No mains. Where LinuxCNC takes part in a check, wire an NC contact of KM1 or of the stand-in relay to `Bend` as in rung 6: since 2026-11-14 the screen's reset is refused without it. - [ ] **Latch.** POWER ON latches K3; the E-stop drops it; releasing the E-stop does not bring it back; neither does cycling 24 V. - [ ] **POWER ON held (S5).** Hold it and release the E-stop; hold it while 24 V returns. Does KM1 pick up with no fresh press? - [ ] **Door (rungs 3 and 7; changed 2026-11-15).** Latch set, LinuxCNC out of E-stop, KM1 in. Open a door switch, then close it. K5 must drop and pull in again, `door-open` must follow it in `halshow`, and KM1 must stay in throughout: the door is no longer in KM1's coil circuit. With the spindle not commanded nothing else may happen, no E-stop and no machine-off. (Until 2026-11-15 this item was review S1: the door dropped KM1 and it had to stay out until the screen reset.) - [ ] **Order of operations.** Screen reset before POWER ON, and after. Same result? - [ ] **Supply faults.** Pull and restore F1, then F2, then the FlexiHAL's 24 V feed. What comes back by itself? - [ ] **Blown-fuse indicator (EL08).** With the latch set and LinuxCNC in E-stop (K2 out), make F1 open with a blown or dummy fuse in the closed holder; opening the lever may disconnect the indicator as well. K3 must drop and the white lamp go dark. The KN-F10L24DC's indicator LED passes about 2 mA around an open fuse; record what the lamp and K3 do. - [ ] **24 V dip (S5, N02).** Turn the supply down slowly. Volts at which KM1 drops and at which K3 drops; anything that re-arms on the way back up. With LinuxCNC connected and out of E-stop: when KM1 drops and K3 does not, LinuxCNC must go into E-stop and K2 must open, so that KM1 stays out when the voltage comes back. Then the short case: break KM1's coil wire for an instant, ten times. LinuxCNC in E-stop and the axes un-homed every time? In the simulator 30 dips of 32 to 300 ms all ended that way, and KM1 never came back by itself. That was with the board taking its commands. With a board that ignores them the simulator showed the exception Erik has accepted: KM1 back in 0.12 s after the dip began, and in for about a second (R3-02). That fault cannot be made on the bench. - [ ] **POWER ON contact (EL09).** Twenty presses with K2 already closed, some of them short stabs. Chatter, sticking, a KM1 that fails to seal? A short stab can make the button itself break the coil current before K3 seals. This shows whether it works today, not how long an unrated contact lasts. - [ ] **KM1 coil.** Holding current, pick-up and drop-out volts, and release time with the HMX1 suppressor fitted (scope). - [ ] **KM1's NC aux on Bend (EL07).** Twenty operations: does `power-present` read false every time KM1 is out? The contact's rated minimum is 17 V and the input gives it 12 V, and a contact that fails to conduct reads as power present. Then the reset check added on 2026-11-14: with LinuxCNC in E-stop and KM1 out, pull one wire of the Bend pair and click the screen's reset. It must be refused, with `km1-stuck` true in `halshow`; reconnect, and the reset goes through. Twenty good operations do not show that a contact run below its rated minimum stays reliable; the reset check is what catches it when it stops. What it catches is a contact that reads "in" when KM1 should be out. It is not a weld detector: a KM1 with a welded main contact may still read "out" (R3-03). - [ ] **24 V off, Pi alive (HW-02).** What DOOR and HALT read. ### C. Rev O HAL LinuxCNC's simulator ran these with a model in place of the FlexiHAL: on 2026-11-13, twice on 2026-11-14 with the three HAL changes and corrected scripts, and on 2026-11-15 after the door change and the third review ([[Voron Cascade Electrical Configuration#Simulator run (2026-11-13)]]). The second review found that the first run's "53 checks" included three rows that asserted nothing and left out outputs it claimed to have checked. The present scripted run has 82 assertions, the valve with K4 forced on and all ten board outputs among them, and the door test has 23; all pass. That is evidence about the HAL's logic and LinuxCNC's behaviour, no more: the board model was written from the firmware source by the same hand as the HAL, and nothing has run on the board. The last item is hardware (the valve's feed through KM1's contact), which the simulator only modelled. Tick them here when the real board has done the same. - [ ] Loads with no errors or warnings. - [ ] FEED HOLD only pauses; CYCLE START only resumes, and does nothing when idle. (Known and left: a second resume asked for within about a quarter of a second of the first can be lost, N06. Two quick AIR presses both act, R3-06.) - [ ] CYCLE START held, then FEED HOLD pressed: it pauses and stays paused (SW-06). - [ ] AIR: one change per press; held while the machine comes on, it does nothing; E-stop and M9 turn it off; the lamp follows M7 / M9. - [ ] Lamps: yellow steady when paused; green steady when running, flashing when paused. - [ ] Machine-on is refused while KM1 is out, with "can't enable motion, enable input is false" (SW-03). - [ ] The axes are un-homed after an on-screen E-stop and after an E-stop-button trip (SW-05). Try the button five times, some during a move. Before the `power-lost` line the simulator left the axes homed in 8 of 20 hardware trips; with it and the latch trip of 2026-11-14, in none of 20 (run of 2026-11-15). A door stop is not an E-stop any more and must leave the axes homed (door items below). - [ ] KM1 dropped by itself with the machine on (pull the KM+ wire): since 2026-11-14 LinuxCNC goes into E-stop, K2 opens, the spindle and air commands stop and the axes are un-homed. Put the wire back: KM1 stays out until the screen reset (S5, N02). - [ ] Screen reset with the E-stop button still in: K2 does not click, and LinuxCNC stays in E-stop (SW-11). Screen reset with the door open: accepted, since 2026-11-15. - [ ] On-screen E-stop: K2 drops at once, not after a pause. About 0.5 s later the link stops: `flexi.SPI-status` goes false and `halshow` shows the inputs' last values until the next reset. K2, Mist and both lamp outputs are off in E-stop. - [ ] E-stop-button trip: the link stops about 0.5 s later, as after the on-screen E-stop. A reset refused with the button still in starts the link, and it stops again. (The link follows the E-stop latch since the evening of 2026-11-14; before that it stopped only after the on-screen E-stop.) - [ ] **Door, spindle off (2026-11-15).** Machine on and homed, spindle off. Open the door: nothing happens. With it open, jog an axis, run an MDI move, press AIR and home an axis: all work. Then an on-screen E-stop, the reset, and machine-on, still with the door open: all work. - [ ] **Door, spindle on.** Before the VFD is connected, watch the signal `spindle-forward` in `halshow`: it is the run command. Run a program that starts the spindle and moves an axis, and open the door. The run command goes false at once, the axis stops, the screen shows "Process Aborted", the spindle command is cleared and the air goes off. The machine stays on, KM1 stays in and the axes stay homed. Shut the door: nothing restarts. Five times. - [ ] **Spindle start with the door open.** M3 from MDI, the screen's spindle button, and a program, each with the door open: the run command never goes true, and the screen shows "Process Aborted". - [ ] **A short "door open" reading.** With the spindle on, flick a door switch open and shut as fast as possible: a full stop, and nothing restarts. - [ ] **The door's backup.** In a terminal: `halcmd unlinkp halui.abort`. Start the spindle program and open the door: the run command still goes off at once, and 1 s later LinuxCNC goes into E-stop, K2 and KM1 drop and the axes are un-homed. `halcmd net door-abort halui.abort` puts the pin back, or restart LinuxCNC. Then, with the abort connected, the same program with `M51 P0` as its first line, which makes LinuxCNC ignore the two inhibit pins: the abort alone has to stop the spindle and the program, with the machine left on. - [ ] **QtDragon and the door's pins.** At the first start with the real screen: no "pin is already linked" error for `halui.abort`, `spindle.0.inhibit` or `motion.feed-inhibit`. The simulator did not load QtDragon; the sample's post-GUI files do not use these pins. - [ ] Screen reset with one wire of the Bend pair pulled: refused, `km1-stuck` true (EL07; the same check as in part B, here on the real KM1). - [ ] E-stop pressed with K4 forced on: the valve still closes (S4). ### D. One axis on 48 V - [ ] **Step timing (SW-07).** Scope PU and DR at the driver through reversals: DR settled at least 125 µs before the next pulse; PU low for at least 2.5 µs at top speed. - [ ] **Bus voltage (EL06).** PSU trimmed to 48.0 V measured; then the peak on the 48 V bus in the hardest stop from top speed. It has to stay under the CL57Y's 50 V. The NDR's own over-voltage trip is at 56 to 65 V, so the supply staying up proves nothing here. - [ ] **Driver unpowered (SW-03).** Command a move with 48 V off. What LinuxCNC shows, and what happens when 48 V returns. - [ ] **Z with 48 V off (S3).** Does the head drop, and how far? - [ ] **Driver fault (HW-06).** Stall the axis or pull the encoder plug. Does the driver alarm, and does LinuxCNC notice? As designed today it will not, and there is no following error either: the position LinuxCNC reads back is the board's own step count. ### E. Mains, VFD and spindle - [ ] **Bonding, before first mains (EL02, EL03).** Inlet PE pin to the plate, frame, gantry, each DIN rail, the PSU cases, filter, VFD chassis, spindle body and panel skins: 0.1 Ω or less each. - [ ] **Inrush (EL05).** Ten KM1 closures at the intervals the machine will really see (an E-stop and a reset are a few seconds apart; since 2026-11-15 the door no longer cycles KM1). Does Q1 hold? That is all this shows. Contact life and how often the H100 may be powered up need the data: KM1's making or capacitive-load rating, and any limit on power-ups per hour in the H100 manual. - [ ] **Modbus (V1).** Which serial settings answer; a register dump; whether a frequency write is in 0.1 Hz or 0.01 Hz; the run and stop words. Spindle unloaded, low speed. Compare with [[G-Penny H100 VFD Modbus Integration]]. - [ ] **Base frequency (V2).** F004 and the rated-voltage setting match the spindle nameplate. - [ ] **Comms loss.** Unplug RS485 while running. Does it stop, and after how long? - [ ] **Mains return (V2).** With a run command standing, drop KM1 and close it again. The spindle must not restart by itself; record F153 and F155. - [ ] **Door stop with the real spindle (2026-11-15).** At 24 000 rpm in air, open the door. Seconds from the door opening to the spindle starting to slow (the run command goes to the VFD over Modbus at vfdmod's polling rate, so expect a fraction of a second), and to standstill. The VFD's stop mode has to be ramp-to-stop, not coast: record the parameter, the deceleration time, and whether the VFD trips on overvoltage at that setting. Compare with the coast-down below: the door stop should be the quicker of the two. - [ ] **Coast-down (S3).** Seconds from KM1 dropping at 24 000 rpm to standstill. - [ ] **Inlet current (EL01).** True-RMS clamp at the inlet: idle, spindle at 24 000 rpm in air, a heavy cut with the axes moving, and the peak while the spindle runs up (max-hold). The H100's plate gives no input current, and by arithmetic even the ≈7–8 A motor-current setting could mean 13–15 A here (wiring note, mains budget), so start with light cuts and read the clamp before trusting the setting. It has to stay under Q1's 15 A with margin. # References - [[Voron Cascade Electronics Wiring]]: Rev O, on hold - [[Voron Cascade Electronics Layout]]: Layout J, where everything sits in the bay - [[Voron Cascade Electrical Configuration]] - [[G-Penny H100 VFD Modbus Integration]] - [[Electronics]] # Changelog ## 2026-11-12 - Created from the independent review of Rev L. - Added two wire-fit checks (twin ferrule at K5, 18 AWG at KM1) after the panel wire was re-specified to the stock on hand. - Open decisions in the Overview brought up to date: S3 decided (written rule); S6 and HW-06 added to the list. - Brought in line with Rev M: the decisions are marked as made, the first three items now gate wiring, and checks are added for the NC status contact, the valve interlock, the 12 AWG bonds and the machine-on gate. - PE strip check answered (the jumper bar does not fit the 7641K81); two items now gate wiring. ## 2026-11-13 - PE strip item updated: second grounding block on the rear rail, link from block 6 to block 7. - Added the DIN rail stock check for Layout I (slot pitch, remaining length, end cover width); the wiring note's revision letter in References corrected to M. - Branch fuses (EL11) decided against; the 24 V loads item reworded. Added the blown-fuse indicator check for F1 (EL08). - Added the H100 mounting pattern and size check for Layout I and the VFD jig. - Simulator run of the Rev M HAL recorded: part C says what the simulator already passed, un-homing is a check of its own, and part A has a new check for an on-screen E-stop made during a move. - `power-lost` line approved and in the HAL: the un-homing checks in part C and the polarity check in part A follow. - SPI link kept up through an on-screen E-stop, approved and in the HAL: the two on-screen E-stop checks in part A updated, and one added to part C. ## 2026-11-14 - Aux pull-ups answered: fitted, 1.00 kΩ on all four, board revision A7. The gate is now the rework that removes them, with its own check. Input pin order recorded from the board's silkscreen. - FlexiHAL commons answered: all common, so W24 is six cores. - Revision letter moved to N with the wiring note; no check changes with it. - H100 moved to a printed mount on the lower rail: the rail-stock check asks for 330.5 mm, and the H100 check names the mount reference and asks what the back of the case is. - Second independent review (of Rev N) and the three HAL changes Erik approved from it. Part A: a wire link on `Bend` stands in for KM1 on the bare board, the polarity check holds the link up by hand, the input-resistor check asks for the part and its temperature (HW-04), K2's timing is three measurements (SW-02, N01), and the E-stop-during-a-move figures are the rerun's (N03, N05). Part B: the dip tests with LinuxCNC connected (S5, N02) and the refused reset (EL07). Part C: what the rerun covered, and the items follow the new HAL. Parts D and E and part 1 say what a check can and cannot show, and ask for the data the review wants (EL01, EL03, EL04, EL05, EL06, EL08, EL09, EL11, HW-06). - VFD nameplate recorded under the nameplate gate (`H100-1.5C2-1B`, 110 V in, 0–110 V 12.5 A out); the spindle's plate and the manual pages are still open, so the gate stays. The inlet-current check says why the motor-current setting is not yet a mains limit. - 48 V −V → PE bond accepted by Erik as 18 AWG in a twin ferrule; the bond check now only asks whether the twin ferrules seat in the NDR's terminals. - The SPI link follows the E-stop latch (approved by Erik): part A's timing check and part C ask for the link to stop after a HALT or door trip and after a refused reset, not only after the on-screen E-stop. The simulator's assertion count is 72. ## 2026-11-15 - Third independent review (of the Rev N package) and Erik's decisions: the fault its blockers sit under is an accepted limit, the door acts through LinuxCNC only, and the bench is next. The overview says so and gives the order: Aux rework, then part A with the input-resistor check first (it is now the first item of part A), because the bare-board procedure holds `Bend` closed. The input-resistor and step-timing checks are named as coming before the steps that depend on them. - Door change. Part B's door item now asks that KM1 stays in when the door opens; the POWER ON item no longer needs K5 closed; the twin-ferrule item covers K5 A2 as well as K5 11. Part C has six new door items (spindle off, spindle on, a spindle start with the door open, a short reading, the backup, QtDragon's pins) and its reset, un-homing and link items no longer treat the door as a trip. Part A's polarity check says why DOOR needs care. Part E has the door stop with the real spindle, and the inrush item no longer counts door openings as KM1 closures. - Third review's corrections: HALT has its own input circuit (R3-07); the reset check is not a weld detector (R3-03); quick double presses described as they are (R3-06); the dip item names the accepted exception (R3-02); the E-stop-during-a-move figures are those of the run of 2026-11-15, by step count. Part C's intro gives the present counts: 82 assertions in the scripted run and 23 in the door test. - Added, as information only: what the firmware's separate reset input (`flexi.PRU-reset`) does on the real board. - Revision letter moved to O with the wiring note (Erik's decision); no check changes with it. Part C's heading follows. - Rear-rail PE group back to one grounding block, at block 11 (Erik's decision): result added under the PE strip item, with the jumper bar to check. The layout is at J and has its own note, [[Voron Cascade Electronics Layout]]: the two items that name the layout say J, and References lists the note.