Texecom power reset

Texecom Remote Power Reset

Project summary: project-summary.html: one page with the reset timing, fail-safe behaviour, schematic, PCB layers, parts and bench tests. View it rendered (GitHub shows .html files as source), or download the file and open it in a browser.

A single board with an ESP32-C3 (Tasmota) and two on-board relays. It sits inside a Texecom Premier Elite 64-W housing and can remotely "hard power reset" the panel by briefly cutting both its battery and 16VAC feeds.

Status: Design rev A with on-board relays. Schematic done (ERC clean, netlist verified). PCB rev A routed: DRC clean (0 violations, 0 unconnected), GND pour on the bottom only, with an engraved top legend (Mill_Legend). Gerbers + drill + legend in fab/. Not yet milled or tested.

Background

The panel occasionally raises an "AUX 12V output failed" fault. The only way to clear it is to remove all power (battery and AC) and let the panel restart. This board lets you do that remotely from Home Assistant.

The root cause (probably an overloaded or tripping AUX output) is being investigated separately. This board is a remote recovery tool, not a fix.

Requirements

Panel

Function

Fail-safe

Power

Hardware

Parts on hand

Wiring concept

                  ┌─────────────────────────── Main PCB ───────────────────────────┐
Battery +  ──► J1 BAT IN + ──┬──────────────────────────► K1 COM ─┐  NC            │
                             │                                    └──────► J2 BAT OUT + ──► Panel BATT +
                             └─ F1 → D1 → Pololu 5V → ESP32-C3, 74HCT4538, relay coils │
Battery −  ──► J1 BAT IN − ───── GND ─────────────────────────────────► J2 BAT OUT − ──► Panel BATT −
                                                                                       │
16VAC ~A   ──► J3 AC IN ~A ───────────────────────────► K2 COM ─┐  NC                  │
                                                                └──────► J4 AC OUT ~A ──► Panel AC ~A
16VAC ~B   ──► J3 AC IN ~B ───────────────────────────────────────────► J4 AC OUT ~B ──► Panel AC ~B
                  └────────────────────────────────────────────────────────────────┘

Each panel power path passes through only two screw terminals (in and out) and one relay contact.

Panel facts (Texecom INS176-15 installation manual)

From the Premier Elite 12-W, 24-W & 48/64-W sections:

Relay rating check: each relay path carries at most about 1.6A (the battery PTC limit) or the panel's AC input current. The SRD-05VDC-SL-C contacts are rated 10A. OK.

Kick-start consequence: the panel only restarts after a reset if AC is present. A reset during a real AC outage would leave the panel off until someone presses the kick-start button. The Home Assistant script therefore checks that AC is present before it resets (see below).

Design: rev A (on-board relays)

Block diagram

 Battery + ─J1─┬─ F1 PTC ─ D1 ─┬─ C1/C2/D2 ─ U1 Pololu ─► +5V ─┬─ D3 ─► SuperMini 5V
               │               │  bulk/TVS                   ├─► U2 74HCT4538 VCC
               │               │                             └─► relay coils, LEDs
               └───────────────┼──────────────► K1 COM ─ NC ─► J2 BAT OUT +
                               │
 GPIO4 ─ R9 1k ─┬─ reset ─┐                                          +5V
                └─ RC ────┴─► U2 ch1 (7 s max) ─ Q ─ R7 1k ─► Q1 2N2222A ─► K1 coil ┤ D4 flyback, D6 LED
 GPIO3 ─ R10 1k ─┬─ reset ─┐                                         +5V
                 └─ RC ────┴─► U2 ch2 (7 s max) ─ Q ─ R8 1k ─► Q2 2N2222A ─► K2 coil ┤ D5 flyback, D7 LED

 J3 AC IN ~A ─► K2 COM ─ NC ─► J4 AC OUT ~A        J3 ~B ─► J4 ~B (straight through)

1. Power input and protection

Current budget (estimated):

2. Controller: ESP32-C3 SuperMini

• Plugs into two 1×8 female headers, so it can be removed for flashing or replacement.
• Relay control pins: GPIO4 → channel 1 (K1, battery) and GPIO3 → channel 2 (K2, AC). (Swapped from the first draft when the SuperMini footprint was corrected; this pairing keeps the tracks from crossing.)
  • Neither is a strapping pin (unlike GPIO2, 8 and 9), nor a USB pin (18, 19) or the UART0 pin that prints the boot log (GPIO21).
  • Both are plain inputs with no internal pull-up at reset (ESP32-C3 datasheet, Table 2-1), so they can't pulse a relay while the ESP boots. GPIO6 (MTCK) and GPIO9 do have a weak pull-up at reset and are avoided.
• R1 and R2: 100k pull-downs on the 74HCT4538 side of R9/R10. They hold the resets low (relays off) while the ESP is booting, unpowered or removed.
• R9 and R10: 1k series resistors between each GPIO and the 74HCT4538. They protect the ESP when the 5V rail is off. See the note in §3.
• The V2 SuperMini's onboard LED is a WS2812 RGB LED on GPIO8 (a strapping pin). It isn't used as a status LED; leave GPIO8 unassigned in Tasmota.
• Orientation: the module mounts component (button) side up with its USB-C at the right-hand board edge (engraved "USB →" on the legend), so it can be flashed in place. The antenna end therefore points into the board, about 8.5mm above the copper on the header sockets. If WiFi is weak inside the panel housing (especially a metal one), use the V2's IPEX external-antenna option (move the small antenna-select link as described on the module's info page) and fit the optional U.FL antenna from the BOM (Micro Robotics BAT-P60-WIFI-2, a stick-on flexible PCB antenna on a 100mm lead).

3. Hardware timeout: 74HCT4538 dual monostable (at 5V)

One chip (TI CD74HCT4538E, DIP-16) covers both channels. Each channel behaves as:

relay on = GPIO high AND less than T seconds since the GPIO went high

Why a 74HCT at 5V (not a 74HC at 3.3V): it runs from the same 5V rail as the relay drivers, and it gives a full 5V drive to the transistor bases. HCT inputs use TTL thresholds (high = 2.0V or above, low = 0.8V or below), so the ESP's 3.3V GPIO drives them reliably. (It was chosen originally when bench test 1b showed the relay module's input needed 4.77mA; with on-board relays the choice still stands.)

Wiring for each channel. TI pin names are used here; Nexperia's datasheet swaps the letters A and B, but the pin numbers and functions are identical.

Timing: T = 0.7 × REXT × CEXT = 0.7 × 1M × 10µF ≈ 7 s nominal (TI datasheet: τ = 0.7·RX·CX, RX min 5k; noise susceptibility "may occur for RX > 1M", so 1M is the upper limit). It needs to stay above 3.5 s (the longest PulseTime) and, ideally, below 30 s. Bench test 3 confirms the real value. Tune it by lowering REXT or changing CEXT. Keep the RXCX traces short.

• CEXT must be low-leakage. With REXT at 1M the timing current is only about 5µA, so an aluminium electrolytic (leakage in the µA range) would upset the timing badly. Use an X7R/X5R ceramic. Avoid Y5V ceramics, which can lose 50–80% of their capacitance.
• Power-down protection: TI requires protection when CX is 0.5µF or more. The 100Ω in series with CX limits the discharge current into pins 2/14 if 5V collapses quickly.
• Use 74HCT4538 (TI CD74HCT4538E from Communica). A 74HC4538 won't reliably read 3.3V inputs at 5V, and a CD4538 has too little output drive.

ESP protection (does 5V reach the ESP?): no. The ESP pins only drive the 74HCT4538's inputs, which are high-impedance CMOS gates. No 5V output connects to any ESP pin, so the ESP never sees more than its own 3.3V.

The one corner case is the 5V rail being off while the ESP is powered, for example on USB for flashing (D3 blocks USB from the 5V rail). If the ESP then drives a GPIO high, current could flow through the chip's input clamp diode into the unpowered 5V rail. R9/R10 (1k) limit that to under 3mA, which is harmless for both parts. They cost nothing in normal operation, because the HCT inputs draw no current.

3a. Relay drivers and relays

Relay pin-out (KiCad Relay_SPDT_SANYOU_SRD_Series_Form_C, matches the Songle SRD): 1 = COM, 4 = NC, 3 = NO, 2 and 5 = coil.

2N2222A pin-out warning: TO-92 2N2222A parts are usually E-B-C, but some makers use other orders. The footprint is TO-92 inline with pin 1 = E, 2 = B, 3 = C. Check the pin-out of the actual transistors before soldering.

Fail-safe chain: GPIO low / ESP dead → U2 reset low → Q low → R13/R14 hold the base low → coil off → NC closed → panel powered. U2 unpowered or pulled from its socket gives the same result.

4. Connectors

Screw terminals: all KF301-2P (5.0mm pitch, 16A / 250V, 22–14 AWG).

5. Field wiring

Wire: 0.5–0.75mm² (20–18 AWG) for the battery and AC runs, with ferrules in the screw terminals.

Recommended: an in-line fuse (about 3A) on the battery + lead, close to the battery. The battery current now runs through the PCB, and a 7Ah SLA battery can deliver very high currents into a short before the panel's own F6 PTC (which sits downstream) can do anything.

6. PCB

• 2-layer, about 70 × 50mm (estimated): two relays (19 × 15.5mm each) in, and the relay module, J5, J6, J7 and the control lead out. The housing dimensions are still open.
• Placement: J1–J4 along one edge with K1/K2 right behind them, so the battery and AC paths are short. The SuperMini sits at the opposite edge, as far from the relays as possible, with its USB-C at the board edge.
• Heavy-current traces (J1 → K1 → J2, J3 → K2 → J4, and the BAT− and AC ~B straight-throughs): at least 2mm wide on 1oz copper (about 3A at 10°C rise), or copper pours.
• Clearance: the 16VAC lines swing about ±23V relative to GND (the panel's bridge rectifier references them to 0V). Keep at least 1mm from logic traces. This is low voltage, so no mains-level creepage is needed.
• Footprint pitch: the KF301 is 5.0mm. Use a 5.0mm footprint (not 5.08mm) so that interlocked blocks line up.
• Through-hole wherever possible: axial resistors and diodes, radial capacitors, TO-92, DIP-16 in a socket, 2.54mm headers. 1206 SMD is acceptable (hand-solderable) and is used for F1, C5/C6 and several resistors.
• GND pour on the bottom layer only. The top layer carries only tracks, pads and the engraved legend (§6a).
• 2 × M3 mounting holes.

6a. CNC-milling rules (H8)

Process: KiCad → Gerbers + Excellon drill → FlatCAM (isolation and drill G-code) → CNC. Double-sided, with FlatCAM flip alignment using alignment pins. Tool: 30° V-bit, 0.1mm tip. The machine has milled 0.7mm tracks reliably (see the example board), and all common drill sizes are available.

Milled boards have no plated through-holes, no solder mask and no silkscreen, so the layout follows these rules:

Routing: I place the parts and then try routing. If that's poor, use the user's preferred autorouter, the KiCad routing tools (Go implementation), with the rules above as design rules.

6b. KiCad project

Layout: power path across the top. Channel 2 (GPIO3, K2 AC relay) is in the middle band and channel 1 (GPIO4, K1 battery relay) in the lower band. This order avoids wire crossings from the SuperMini. J3/J4 (AC) sit next to K2, and J2 (BAT OUT) next to K1. The BAT_IN+ wire runs from J1 over the top and down the right side to K1 COM.

Reference designators match BOM.md: U1 Pololu, U2 74HCT4538 (units A/B = channels 1/2, unit C = power), U3 SuperMini, J1–J4 KF301 terminals, K1/K2 relays, Q1/Q2 2N2222A, F1 PTC, D1/D3 1N5822, D2 P6KE18A, D4/D5 1N4007, D6/D7 LEDs, C1–C8, R1–R16, MH1/MH2 M3 holes.

Verification: ERC 0 errors, 0 warnings. The KiCad netlist was checked net by net against the intended circuit (32 nets). The only unconnected pins are the intentional ones (relay NO, Q̅ outputs, Pololu EN, unused SuperMini pins).

6c. PCB layout (rev A)

Drill sizes:

How the "no top joint on covered pads" rule is enforced (so the router can't create a connection that can't be soldered):

• KiCad forces plated pads onto both copper layers, so the covered footprints (terminals, relays, DIP socket, headers, TO-92, electrolytics, LEDs) carry a per-pad F.Cu keepout (no tracks, no vias, no pour) inside the footprint. These are in the *_Mill footprints in texecom-power-reset.pretty/.
• There is no top pour at all. The .kicad_dru rule "no top pour on covered pads" (zone_connection none on F.Cu) is kept as a guard in case a top pour is ever added again. DRC connectivity is therefore honest: every connection it counts is solderable.

Assembly notes for the milled board:

Two assembly drawings, both printable at 2:1 on A4 landscape (print at 100%). The vias carry the same numbers on both.

• Top side (PDF): every top-side part with its reference and value, its outline with polarity and pin-1 marks, and the 20 pads that need a top solder joint ringed in orange (the same pads as the rings on the engraved legend).
• Bottom side (PDF): there's no silkscreen on the bottom, so this shows the board as seen from below (mirrored), with the eleven 1206 parts labelled and all 16 vias numbered. The through-hole parts appear in grey for orientation.

1. Vias: fit a wire through each of the 16 vias and solder both sides first.

2. SMD (1206) parts are on the bottom (F1, C5, C6, R7–R12, R15, R16): solder them first, on the bottom.

3. Axial parts and disc capacitors (R1–R6, R13, R14, D1–D5, C2, C4, C7, C8): mount them 1–2mm off the board so the iron can reach the top pads, and solder every pad on both sides. The holes aren't plated, so the top and bottom pads are only joined by solder on the lead. Twenty of these pads carry a top track, and for those the top joint is essential (some have no bottom track at all). They're marked on the board with a small engraved ring (○ = SOLDER TOP) next to the pad. Soldering the other top pads is harmless and keeps the rule simple.

   Part	Type	Pads with a top track	Net
   R1	Axial resistor	1, 2	U2A reset, GND
   R2	Axial resistor	1	U2B reset
   R3	Axial resistor	1	+5V
   R4	Axial resistor	1	+5V
   R5	Axial resistor	1, 2	U2A reset, U2A trigger
   R6	Axial resistor	1, 2	U2B reset, U2B trigger
   R13	Axial resistor	none
   R14	Axial resistor	2	GND
   D1	Axial diode (1N5822)	1	D1 cathode
   D2	Axial diode (P6KE18A)	1	D1 cathode
   D3	Axial diode (1N5822)	none
   D4	Axial diode (1N4007)	1, 2	+5V, D4 anode
   D5	Axial diode (1N4007)	1	+5V
   C2	Disc capacitor	1	D1 cathode
   C4	Disc capacitor	1, 2	+5V, GND
   C7	Disc capacitor	1	U2A trigger
   C8	Disc capacitor	1	U2B trigger

   Solder the top joint first, then the bottom. Afterwards, check continuity from each ringed pad to the far end of its top track.

4. Everything else (terminals, relays, socket, headers, TO-92s, electrolytics, LEDs) is soldered on the bottom only; their top pads are isolated rings.

   • SuperMini (U3): solder male pins to the module and two 1×8 female headers to the board (bottom side). Plug the module in button side up, USB-C towards the right-hand board edge, following the engraved "USB →" arrow. Before plugging it in, check that the square pad (pin 1, 5V) is at the USB end of the row nearest the relays.

5. After testing, coat the board (no solder mask).

Milling order (double-sided, alignment pins, height map):

1. Alignment-pin holes first, outside the board outline, so the pins are ready for the flip.
2. Bottom side: probe a height map (no holes yet, so the probe can't drop into one), then mill the B.Cu isolation, mirrored around the pin axis (4 passes, see Isolation passes below).
3. Flip onto the alignment pins and probe a new height map on the top (the top face isn't flat in the same way as the bottom).
4. Top isolation (F.Cu, not mirrored, 4 passes).
5. Legend (Mill_Legend, not mirrored) with the same V-bit, Z zero and height map; see below.
6. Drill from the top (drill file not mirrored), with a firm sacrificial backing so exit burrs on the bottom pads stay small. Re-zero Z after each drill change (the height map isn't needed for drilling through). Mill the 3.2mm M3 holes as circles if you don't have a 3.2mm drill.
7. Board outline last (2.4mm tool, 4 passes to −2.0mm, no tabs). The board is held on double-sided tape, so it stays put after the final pass; tape near the edges as well as the middle. The drills go through the tape, so clean adhesive off the small bits if holes start to come out ragged.
8. Lightly deburr the bottom pads before soldering.

Isolation passes (copper layers): use 4, not 1

The first board was milled with one pass of the V-bit (about 0.14mm wide at −0.07mm). Isolation routing cuts a line around the outline of each copper shape; it doesn't clear the space between shapes. With one pass:

• Top: there's no pour, so every pad and track sits in unused, floating copper only about 0.14mm away. Solder can bridge a pad to that copper, and if the same piece touches another pad, you get a short.
• Bottom: the Gerber has 0.8mm gaps between tracks, pads and the GND pour. One pass traces both edges of each gap: narrow gaps get cleared completely (the cuts merge), wider or irregular ones (for example around thermal-relief spokes) keep a sliver of copper between the two cuts. That's why the separation looked wide in some places and thin in others.

Settings for both copper layers (FlatCAM Beta, Isolation tool):

That gives about 0.45–0.5mm of isolation from every copper edge: the 0.8mm bottom gaps are cleared completely (no slivers, even look), and the top pads get about 0.5mm to the floating copper. The legend stays a single Follow pass. FlatCAM's Non-Copper Clearing would remove all unused copper, but with a 0.1mm V-bit it's very slow; the extra passes give the clearance that matters.

Checking a milled board: under magnification, scrape out any copper slivers with a sharp hobby knife and brush the dust away. Then check with a continuity meter for shorts between neighbouring pads, and from each pad to the copper around it, especially the terminal, relay and 5V pads, before soldering.

Alignment holes with a 2.4mm fishtail end mill (at 12,000rpm):

• Pin fit: a flat-bottomed end-mill hole comes out very close to the bit size. If the pins are also 2.4mm, test on scrap first. If they're too tight, use a slightly smaller pin, or mill the holes (FlatCAM "mill holes") for a snug sliding fit.
• First hole: if the bit squeals or the resin turns brown, increase the feed a little rather than slowing down.
• These are general carbide-in-FR4 figures for hobby spindles; if the bit's supplier lists feeds, use those.

Hole drills at 12,000rpm (carbide PCB drills in FR4):

• PCB drills are designed for 30,000–100,000rpm. At 12,000rpm each turn must take a smaller bite, or thin bits break, so the feed scales with the diameter (about 0.005–0.015mm per revolution).
• A single plunge to −2.0mm (no pecking) is fine for a 1.6mm board at these sizes. Set Feedrate Z per tool in FlatCAM's drilling job and regenerate the drill G-code.
• The 0.8mm bit is the fragile one: keep it at the low end, make sure it runs true, and clean tape adhesive off it. On the larger bits, if a bit squeals or the hole edges go brown, raise the feed a step rather than slowing down.
• These are general figures for hobby spindles; if the drill supplier lists feeds, use those.

Engraving the legend (FlatCAM):

The legend Gerber is made of 0.15mm-wide lines. Normal isolation routing would cut around each line (a double outline); Follow geometry runs the tool once down the centre of each line, so the V-bit cuts a single thin groove.

1. Load fab/texecom-power-reset-Mill_Legend.gbr together with the F.Cu Gerber. It shares their origin, so it lines up with the top isolation and the alignment pins. Apply exactly the same offset or double-sided move/mirror to the legend as to F.Cu (select both and apply it once).
2. Make the Follow geometry:
   • FlatCAM Beta (8.99x): select the legend Gerber, open the Isolation Routing tool and tick Follow. Tool diameter and passes don't matter with Follow. Generate: this creates a …_follow geometry.
   • FlatCAM 8.5: select the legend Gerber; in the Selected tab under Isolation Routing tick Follow, then Generate Geometry.
   • Check the result: text, outlines and rings should be single centre lines, not pairs of outlines.
3. CNC job on the follow geometry: tool type V-shape, 0.1mm tip, 30°, Cut Z about −0.05 to −0.07mm (just through 35µm copper, giving a line about 0.12–0.14mm wide), multi-depth off. Use the feeds, speed and travel Z from the top isolation job. A deeper cut widens the lines and blurs the small text (1.2mm references, 1.5mm terminal labels), so try a short test on scrap first.
4. The cut is very shallow, so use the top height map (auto-levelling); without it, lines fade out on high spots and cut wide on low ones.
5. The legend was generated at least 0.35mm clear of everything on the top copper, so even a slightly deep cut only marks floating (unconnected) copper. If you change the layout, regenerate it with tools/make_legend.py (run with KiCad's Python; see the script header) rather than editing it by hand.

FlatCAM menu names differ a little between versions.

7. Tasmota configuration

Build: tasmota32c3.bin (flash through the SuperMini's USB-C, with the board unpowered from the battery; D3 isolates USB from the relay rail).

Configuration → Configure Module (Generic ESP32-C3):

Console:

Backlog PowerOnState 0; PulseTime1 30; PulseTime2 35; Sleep 100; FriendlyName1 Panel Battery Cut; FriendlyName2 Panel AC Cut

If the SuperMini has trouble holding a WiFi connection (a known issue with some SuperMini antenna layouts), try WifiPower 8.5. On this board the antenna end points inward over the PCB, so if the signal inside the panel housing is poor, switch the V2 to its IPEX external antenna.

8. Home Assistant script (example)

Entity IDs are placeholders. Adjust them, and the AC-OK condition, to match the Tasmota and Texecom integrations.

script:
  texecom_hard_power_reset:
    alias: "Texecom hard power reset"
    mode: single
    sequence:
      # Never reset without AC: the panel would not restart without the kick-start button.
      - condition: state
        entity_id: binary_sensor.texecom_ac_ok        # placeholder
        state: "on"
      - action: switch.turn_on
        target:
          entity_id:
            - switch.texecom_reset_panel_battery_cut   # placeholder
            - switch.texecom_reset_panel_ac_cut        # placeholder

9. Bench tests before installing

Do these with a bench supply of about 12–13.8V on J1, before connecting the panel.

1. Power-up: the +5V rail reads 4.9–5.2V, both LEDs stay off, and both relays stay released. The NC contacts show continuity: J1+ to J2+, and J3 ~A to J4 ~A.
2. Normal pulse: turn on Power1. K1 clicks, D6 lights, and J1+ to J2+ opens for 3.0 s, then everything releases. Repeat with Power2 (K2, D7, 3.5 s).
3. Hardware timeout: set PulseTime1 0 (no software timeout) and turn on Power1, so the GPIO stays high. K1 should release after about 5–9 s (7 s nominal) and stay off. Afterwards, restore PulseTime1 30. Repeat for channel 2.
4. Fail-safe: unplug the SuperMini, and separately pull U2 from its socket. Both relays must stay released (NC closed) and the LEDs off.
5. USB isolation: with J1 disconnected, plug in USB. The relays, LEDs and +5V rail must not power up.
6. Idle current: measure the J1 current with WiFi connected. The target is about 15mA or less.

Bill of materials

The full BOM and sourcing tracker (on hand / Micro Robotics / Communica) is in BOM.md.

Commercial alternative: ESP8266 2 × 30A relay board

If milling the custom board isn't worth it, Micro Robotics' ESP8266-2CH-30A (R236, 83 × 68mm) can do the same job with the same wiring (checked against its schematic):

• Wiring: battery + through relay 1 (COM → NC), one AC line through relay 2 (COM → NC); battery − and the other AC line straight to the panel. Power the board (7–28V input, LM2596 5V buck) from the battery side of relay 1's COM, through the inline blade fuse from the BOM, so it's never switched off with the panel.
• Fail-safe: each relay has an NPN driver with a 1k base resistor and a 10k pull-down, plus a flyback diode and LED. If the ESP crashes, reboots or is removed, the relays drop out and the panel keeps its power.
• Move relay 1 off GPIO16. The relays reach the ESP through links on header P5 (Relay 1 ↔ GPIO16, Relay 2 ↔ GPIO14). GPIO16 goes high briefly at boot, which could blip the battery relay on every ESP restart, so wire Relay 1 to GPIO12 or GPIO13 instead (both brought out on P5, no boot glitch).
• Tasmota (tasmota.bin, Generic module): Relay1 = GPIO12 (or 13), Relay2 = GPIO14, Button1 = GPIO0, LedLink_i = GPIO5 (blue LED; check the polarity). Use the same PowerOnState 0; PulseTime1 30; PulseTime2 35. Flash through the UART header with a 3.3V USB-TTL adapter (IO0 to GND while flashing); later updates go over the air.

What it gives up compared with the custom board:

Verdict: a sound substitute, especially with relay 1 moved to GPIO12/13. The custom board is kept for the hardware timeout, lower idle current, USB-C flashing and the external-antenna option.

Open items

• Free space inside the 64-W housing (dimensions still to come), including height for the relays (about 15.5mm)
• Panel fuse ratings: the battery fuse is 1.6A PTC, which is well within the 10A relay rating
• Relay module trigger polarity and 3.3V drive (tests 1a/1b): done 2026-09-25 (high-level trigger, 4.77mA at 3.3V). No longer relevant, because the module was replaced by on-board relays.
• KiCad schematic: done with on-board relays, drawn with wires, ERC clean, netlist verified (2026-09-25)
• Check the pin-out of the 2N2222A transistors actually bought (E-B-C expected)
• PCB layout: rev A placed and routed, DRC clean, Gerbers and drill in fab/ (2026-09-25)
• Mill, assemble and bench-test rev A (section 9)

Decisions log