5-Axis Level Shifter — Board Map

The board

What to read off it before you touch anything:

• The four chip notches all face LEFT. That is the render agreeing with the Gerbers — the rectangular pin 1 pad sits at each socket’s bottom-left. Two independent methods, same answer.
• The terminals ARE labeled, even though they have no pin 1 pad: ENA− ENA+ DIR− DIR+ PUL− PUL+ reading left to right on J5 and J6. That silkscreen is the orientation cue for all five driver blocks.
• Eleven test points in two rows. Top: TP1 STEP_A, TP2 DIR_A, TP3 STEP_B, TP4 DIR_B, TP5 STEP_C, TP6 DIR_C. Bottom: TP7 STEP_D, TP8 DIR_D, TP9 ENA, TP10 STEP_E, TP11 DIR_E. The board says it plainly: TP9 = GLOBAL ENA GPIO21.
• Chip roles are on the silkscreen too — IC1 A/B and IC2 C/D for step and direction, IC3 ENA A–D, IC4 E/YAW. And 5 AXIS   PUL = STEP, which is the one piece of translation you need when reading a DM542T datasheet.
• J1 says ESP32 / USB LEFT and nothing else. It is the only part with no pad marker at all, so that text is doing the entire job. See the orientation checks below.

Connectors

ESP32 pin assignment

Buffers and passives

Pin 1 markers — what this board has and what it doesn’t

From the Gerber apertures, 2026-09-17. This is the copper, not the render.

Assembly order

1. Measure a terminal header pin first. Holes are 1.5 mm; the pluggable Phoenix-style part must clear that, and there must be nothing under the header but six pins in a line.
2. Underside passives before anything else — R1–R11 and C1–C4 are on the bottom and get awkward once tall parts are on the top.
3. DIP sockets, four of them. Orient by the board silkscreen notch.
4. J1 headers — two single-row 1×15 female strips, not a dual-row part. Push both onto the ESP32's pins first so the module holds them square, drop the assembly into the board, tack the corners, check it sits flat, then solder the rest.
5. Terminals J2 through J8.
6. Verify before the chips go in. Flash the pin walker, hold each signal, and meter its terminal — now far easier, since TP1–TP11 give you a pad per signal instead of a screw head.
7. Then one motor, lowest current, loose on the bench, nothing coupled to the frame.

★ The verified map — straight from the IPC-D-356 netlist

Read 2026-09-17 from esp32-stepper-level-shifter.d356, parsing 317 and 327 records. This is the manufacturing data, not the schematic and not the render. Use it as the answer key for the bare-board tracing below — trace first, then check.

Buffer channels — which channel carries which signal

On all four chips OE pins 1, 4, 10 and 13 go to GND (outputs permanently enabled — there is no tri-state on this board), pin 14 = 5V, pin 7 = GND. IC3’s four inputs are all on the one ENABLE_IN net, which is the design intent. IC4 channel 4 is a terminated spare — its input is tied to GND rather than left floating, and its output goes nowhere. That is better practice than leaving it open.

J1 — all thirty pads

⚠️ Counting convention, stated once so it cannot drift: the silkscreen says J1 ESP32 / USB LEFT, so position 1 is the LEFTMOST pad, at the USB end, counting left → right in each row. Upper and lower rows, not columns.

★ Verify the bare board first — before any part is soldered

Do this on a bare board, cold, with nothing fitted. Every pad is reachable from both sides, nothing is in the way, and continuity means what it says because there is no semiconductor anywhere in the path. You have three boards — keep one permanently bare as the tracer. That is what made the v1.0 verification possible without disturbing a working assembly.

The method, per channel. An SN74AHCT125 is four independent buffers. Socket input pad → J1 gives you which ESP32 pin drives it. Socket output pad → screw terminal gives you which signal it becomes. Counting the socket pins on THIS board — verified 2026-09-17 from the Gerbers, not the render: each socket has exactly one rectangular pad (aperture D16, 1.5 × 2.0 mm) and thirteen ovals; 4 squares and 52 ovals board-wide, which is 14 × 4 with no exceptions. That square pad is pin 1. All four sit at the bottom-left, and the chips run horizontally (pads span 15.24 mm in X, rows 7.62 mm apart). So: the notch faces LEFT, pin 1 is bottom-left, count 1→7 rightward along the bottom row, then 8→14 leftward along the top row. ⚠ An earlier draft of this page said “notch at the top, pin 1 top-left, count down the left” — that is the vertical orientation and is wrong for this board. It would have you counting the wrong way around the socket, which is exactly how v1.0 ended up 2-of-8 wrong. Confirm every beep by lifting a probe and hearing it stop — a probe tip bridging two adjacent pads gives a false hit, and that is what caused the pad 18-vs-19 mix-up on v1.0.

Fifteen live channels, two readings each. Which channel of a chip carries which signal is deliberately left blank — that is the answer the tracing produces, not an input to it. The design fixes only which chip handles which pair of axes; the router chose the channels, and the only place that is written down is the copper. Fill the last two columns in and they become the pin map. The four IC3 inputs should all land on the same J1 pad, GPIO 21 — that is correct and is the point of the design, one pin disabling every axis. IC4 channel 4 is unused.

⚠ Orientation — check before soldering J1

The socket takes the ESP32 either way round, and every pin is wrong if it is reversed. J1's thirty pads are identical ovals with no pin-1 marker — each chip got a square pad at pin 1, J2 and J7 got one, J1 did not. And the module covers that silkscreen once seated, so read it before the board disappears underneath.

Three checks, in this order. The module is never exposed until the first two pass.

1. Continuity — nothing powered, no module. Buzz from J7 to find J1's ground pad, and from J2's +5V to find its supply pad. On a 30-pin DevKit, GND and VIN are adjacent, at positions 14 and 15 of the left column, so finding the board's two power pads side by side confirms you have the right end.
   ⚠ Adjacency survives a flip — check the order too. VIN sits closer to the USB end than GND. So the 5V pad should be nearer the left edge, with ground one position further in. Reversed means the module would go in backwards.
   ⚠ GND is not unique: a 30-pin DevKit has GND at position 14 in both columns, plus 3V3 at the bottom right. A single ground beep can mislead if you lose track of which column you are on — which is exactly why you check 5V as well.
2. 5 V into J2 — still no module. J1's supply pad is on the same net, so it goes live. Walk a voltmeter across J1: one pad reads 5 V, one reads 0, the other twenty-eight read nothing. That finds both power pads with a meter rather than a beeper, and nothing is at risk with an empty socket.
3. Seat the module, then read 3V3. Its 3V3 pin should sit at 3.3 V. This is the good final confirmation precisely because you did not wire it — it comes from the module's own regulator, so seeing it means the supply reached the correct pin.

Verification record

Checked three independent ways before ordering, deliberately not trusting any single view of the design:

Order