

Thermo
Companion KiCad project for "Thermo", a specialized automotive-grade board: sibling to manifold-pcb and ecu-pcb, same 12V/AEC-Q/engine-bay-rated constraints as the rest of the family, but purpose-built for two specific…
- Size
- 85.3 × 51.9 mm
- Copper layers
- 6
- Thickness
- 1.6 mm
- Parts
- 62
- Nets
- 44
- Pads
- 248
- Tracks, vias
- 983, 96
- Pours
- 2
- Routed
- Yes
- KiCad
- KiCad 10.0 (file 20260206)
- Created
- 14 Aug 2026
- Updated
- 12 Oct 2026
Overview
Thermo is an open-source printed circuit board by jessiepullaro414, from the jessiepullaro414/Thermo repository on GitHub, published under the MIT licence. It is a 6-layer board of 85.3 × 51.9 mm with 62 parts on 44 nets, routed with 983 tracks and 96 vias.
Its parts include 27 capacitors, 13 resistors, 6 ICs, 5 connectors and 2 diodes.
README
README.mdThermo — KiCad schematic + PCB
Companion KiCad project for "Thermo", a specialized automotive-grade board:
sibling to manifold-pcb and
ecu-pcb, same 12V/AEC-Q/engine-bay-rated constraints as the rest of the
family, but purpose-built for two specific jobs instead of Manifold's
general-purpose Nano-pinout I/O:
- Read a real automotive resistive temperature sending unit (a GM-style
closed-element NTC sensor, 2-wire isolated, -40°F to 210°F) via a
TI ADS1118-Q1 ADC. This board originally targeted a K-type thermocouple
instead; the thermocouple front-end (differential input, open-lead
detection, cold-junction compensation) was fully removed and replaced
with a simple resistive-divider circuit once the sensor choice changed -
see
build_schematic.py's own comments at the U5/J4 section for the real reasoning and the sensor's real published resistance curve. - Drive one on-board NEMA14-class stepper motor (real chosen motor:
StepperOnline 14C19S1504GF5-025RS, 1.5A/phase) via a TI DRV8434A-Q1
driver - current-limited in hardware to match (see
R4/R5in the schematic).
Status
Schematic is fully wired and ERC-clean (only the same 4 expected
tool-limitation findings manifold-pcb's own schematic has). The PCB is
placed, real-footprint, and routed except for 1 net that needs a quick
hand-finish in the KiCad GUI - route_board.py tries several automatic
passes and reports exactly which connection is left; everything else is
DRC-clean. A real BOM and connector pinout reference (with 3D renders) is
published at jessiescars.com/thermo-spec.html.
Why this project exists
I need a thermostatic actuator replacement for my 1972 Alfa Romeo GTV. Thermo reads real engine temperature from an off-board sending unit and drives a small stepper motor in its place - electronic control standing in for a mechanical/wax-based actuator that's original-1972 hard to find good replacement parts for.
How it's built
Same script-driven, kiutils/kicad-cli-verified workflow as manifold-pcb
and ecu-pcb: the generator scripts (build_schematic.py, build_pcb.py,
route_board.py, run_drc.py) are the source of truth - always regenerate
from them, never hand-edit Thermo.kicad_sch/Thermo.kicad_pcb directly.
Every real part, pin number, and package on this board was verified against
a real manufacturer datasheet, not guessed - see the inline comments in
build_schematic.py for the citations (NXP's S32K1xx Reference Manual, TI's
ADS1118-Q1/DRV8434A-Q1/LMR33630-Q1/LM74700-Q1/TLV733P-Q1 datasheets, TE/
Deutsch's real DTM06 connector data, Molex's KK-254 datasheet, Phoenix
Contact's MKDS datasheet).
Real hardware summary
- MCU: NXP S32K144 (Arm Cortex-M4F, AEC-Q100), LQFP-64. Only the pins this board actually uses are exposed (LPSPI0 + 4 stepper GPIO + the fixed crystal/SWD/RESET/power set) - not a general-purpose breakout.
- Temp sensor ADC: TI ADS1118-Q1 (AEC-Q100 Grade 1), VSSOP-10 (DGS0010A,
hand-built footprint - see
build_u5_footprint.py). - Stepper driver: TI DRV8434A-Q1 (AEC-Q100), VQFN-24 (RGE0024N, hand-
built footprint with the real 2.45x2.45mm exposed pad - see
build_u6_footprint.py; neither of KiCad's two bundled RGE0024 variants had the right EP size for this exact part). - Power stage: reused verbatim from
manifold-pcb's own proven 12V input-protection + 5V buck + 3.3V LDO circuit (TI LM74700-Q1 ideal-diode reverse-battery protection, LMR33630-Q1 buck, TLV733P-Q1 LDO, Nexperia PMV37ENEA MOSFET). - Connectors: a two-part pigtail architecture, not a single connector mounted on the board. Each of J4 (temp sensor)/J5 (stepper)/J6 (power) is a small real board-mount connector (Molex KK-254 / Phoenix Contact MKDS) wired to a short pigtail, with a sealed TE/Deutsch DTM06 connector crimped on the free/outboard end - DTM06 itself turned out to be a wire-to-wire-only part family with no PCB-mount variant in this pin-count range, confirmed directly against TE's own product data before this architecture was finalized.
- Board: 6-layer (F.Cu / GND plane / signal / signal / +3V3 plane / B.Cu),
rounded corners, 4 corner mounting holes (M3). Each IC is placed as a
cluster with its own decoupling/boot/charge-pump/crystal parts at its
pins (
CLUSTERSinbuild_pcb.py), not scattered by the area packer.
Every assembled part carries an LCSC field in the schematic - the exact
part JLCPCB places, with its voltage/current rating in the Value - and
export_fab.py builds the JLCPCB BOM/CPL from those fields. The BOM in
Thermo_pinout.html predates the 2026-10-02 review
below and is out of date.
Regenerating the board
python build_schematic.py # writes Thermo.kicad_sch (+ Thermo.kicad_sym, sym-lib-table)
python build_pcb.py # writes Thermo.kicad_pcb, unrouted
python route_board.py # routes it with FreeRouting, pours GND/+3V3 zones
python run_drc.py # real kicad-cli DRC, human-readable report
python export_fab.py # Thermo_gerbers.zip + fab_output/Thermo-JLC-BOM.csv / -CPL.csvOr run the whole pipeline in order with build.py:
python build.py # full pipeline
python build.py --skip-route # skip the slow FreeRouting passroute_board.py's own routing pass is stochastic - it retries automatically
several times per run and keeps the best result, but IMPORTANT: only ever
run it against a fresh, unrouted board from build_pcb.py. Re-running it
against a board that already has zones filled has reliably hung FreeRouting
indefinitely in testing (confirmed: the java process sits at flat memory
usage with zero output, a real deadlock, not "just needs more time") - if
you need to re-route, regenerate from build_pcb.py first.
Setup (one-time, same as manifold-pcb): install a JRE, and download
freerouting-2.2.4.jar
into tools/. The jar, not the installer - FreeRouting's own docs
specify the jar for headless/CLI use. Not committed to this repo (57.8MB,
a generic third-party tool, not project source).
Known open items
- 1 net needs manual routing -
route_board.py's output names exactly which one on the run that produced the currentThermo.kicad_pcb. - Prototype-grade parts: JLCPCB doesn't stock the -Q1 versions of the ADS1118 or DRV8434A, so the prototype uses the commercial versions (same die, pinout, package). Production should use the -Q1s.
- No firmware yet: temperature conversion (ratiometric, see below), homing against J7's switch, stall-threshold learning, sensor-fault handling, brownout recovery and holding-current strategy are all unwritten and untested. The board can't replace the wax actuator until homing and recovery are proven on the bench.
- Thermal budget unmeasured: driver/motor/enclosure temperature at the 1.5A current limit, and whether the lead screw holds position with the driver asleep (no holding current), need bench testing.
- Order options: choose "Via Covering: Epoxy Filled & Capped" (the 9 thermal vias under U6 are in-pad) and "Confirm Parts Placement: Yes".
- U1 (S32K144) isn't stocked at JLCPCB: hand-solder it (LQFP-64, 0.5mm pitch) or consign it. Order assembly with "Confirm Parts Placement: Yes" - no per-package rotation corrections are applied to the CPL.
Fixed in the 2026-10-02 pre-fab review
Every IC's pins and support parts were re-checked against the real TI datasheets. Real bugs found and fixed:
- Crystal shorted: Y1 had OSC_IN on both crystal terminals (a 3225 crystal is pins 1-3 diagonal, 2/4 are the lid). It could never have oscillated.
- Capacitor voltage ratings: values had no ratings, so JLCPCB matched a 10V cap onto the 12V rail. All parts now carry explicit LCSC numbers.
- Buck power stage: L1 was matched to a 150mA signal inductor; now a 2.2uH/6A AEC-Q200 part, with TI's specified 2x22uF output and 10uF + 2x 100nF input caps for the 1.4MHz variant.
- TVS clamp too high: SMCJ33A clamped up to 53V against the buck's 38V abs max; now SMCJ24A.
- DRV8434A straps: now 1/8 microstep (M0=M1=DVDD) and STL_REP grounded (TI's way to disable stall reporting; it was floating, which isn't allowed). Footprint is now TI's exact land pattern - nothing interpolated.
- ADC input: 10k series resistor added so a sender wire shorted to 12V can't destroy the ADS1118.
- Placement: decoupling/boot/charge-pump caps had landed 10-56mm from their ICs; now 2-7mm.
Fixed in the 2026-10-04 follow-up review
- Real surge margin: buck is now the 60V LMR36015AQRNXRQ1 (genuine AEC-Q100, same footprint), so the weakest part on the 12V rail is the stepper driver at 50V abs max - above SMCJ24A's 38.9V max clamp. TVS is now the AEC-Q101 SMCJ24AQ-13-F.
- Sender short-to-battery: D2 blocks the R12 back-feed path into +3V3; U5's AIN1 measures the pull-up node, and firmware computes R_sender = 1k x V(AIN0) / (V(AIN1) - V(AIN0)) (ratiometric).
- Position recovery: J7 home/limit switch input (U1 pin 23), and the driver's STL_MODE/STL_REP now go to U1 (pins 24/39) for stall learning and jam detection.
- Solid planes: 6-layer stackup; the router may no longer put traces on the GND (In1) or +3V3 (In4) planes.
- Release gates:
route_board.pyandrun_drc.pyexit non-zero on an incomplete or failing board,build.pystops at the first failure, andexport_fab.pyrefuses to write order files unless DRC passes. - Firmware scope: the ADS1118-Q1 reads a raw resistive-divider voltage;
firmware needs a real NTC resistance-to-temperature lookup/interpolation
against the sensor's published curve (see
build_schematic.py's J4/U5 comments for the real reference points) - this is new scope, not shared with either sibling project's firmware.
Fixed in the 2026-10-11 BOM check against LCSC
Every LCSC number in the schematic was checked against LCSC's own listing. 27 of 29 matched; two did not:
- Q1 ordered the wrong MOSFET: C3647059 is a Leiditech part that reuses
the name "PMV37ENEA" (100 mOhm RDSon, 3A, 715pF, -55..150C, not AEC-Q101),
not the Nexperia part this design was verified against. Now C553095, the
genuine Nexperia PMV37ENEAR (49 mOhm, 3.5A, 450pF, -55..175C, AEC-Q101).
Stock warning: LCSC and JLCPCB each show only 1 unit of C553095 as of
this check. Fine for one prototype; for more boards buy it from
DigiKey/Mouser/Farnell and hand-fit it (remove its
lcsc=argument soexport_fab.pyleaves it out of the JLC BOM/CPL), or re-qualify a stocked AEC-Q101 60V SOT-23 N-FET. - Y1 was mislabelled, not mis-ordered: C367183 is a Taitien XXDCELNANF-8MHZ (8MHz, 12pF load, 3225 4-pin), exactly what the circuit needs, but the schematic called it a Yangxing X322508MSB4SI. Renamed. Its rated range is only -40..+85C, below the 125C the rest of the family targets, so check it before any engine-bay use.
Parts
62| Qty | Value | Footprint | References |
|---|---|---|---|
| 4 | MountingHole_3.2mm_M3 | MountingHole:MountingHole_3.2mm_M3 | MH1, MH2, MH3, MH4 |
| 3 | 22uF X7R 25V 1210 (buck COUT) | Capacitor_SMD:C_1210_3225Metric | C2, C27, C28 |
| 2 | 100nF X7R 50V | Capacitor_SMD:C_0603_1608Metric | C6, C7 |
| 2 | 18pF C0G 50V | Capacitor_SMD:C_0603_1608Metric | C4, C5 |
| 1 | 1.00k 1% VREF upper | Resistor_SMD:R_0603_1608Metric | R4 |
| 1 | 1.00k 1% sender pull-up | Resistor_SMD:R_0603_1608Metric | R12 |
| 1 | 1.50k 1% VREF lower (IFS=1.5A) | Resistor_SMD:R_0603_1608Metric | R5 |
| 1 | 100k 1% FB top | Resistor_SMD:R_0603_1608Metric | R2 |
| 1 | 100nF X7R 50V (BOOT) | Capacitor_SMD:C_0603_1608Metric | C11 |
| 1 | 100nF X7R 50V (LM74700 VCAP) | Capacitor_SMD:C_0603_1608Metric | C10 |
| 1 | 100nF X7R 50V (U5 VDD) | Capacitor_SMD:C_0603_1608Metric | C13 |
| 1 | 100nF X7R 50V ADC filter | Capacitor_SMD:C_0603_1608Metric | C24 |
| 1 | 100nF X7R 50V HOME_SW debounce | Capacitor_SMD:C_0603_1608Metric | C31 |
| 1 | 100nF X7R 50V VPULL sense filter | Capacitor_SMD:C_0603_1608Metric | C30 |
| 1 | 10k 1% ADC input protection | Resistor_SMD:R_0603_1608Metric | R13 |
| 1 | 10k 1% HOME_SW pull-up | Resistor_SMD:R_0603_1608Metric | R15 |
| 1 | 10k 1% RESET pull-up | Resistor_SMD:R_0603_1608Metric | R1 |
| 1 | 10k 1% SPI_CS pull-up | Resistor_SMD:R_0603_1608Metric | R11 |
| 1 | 10k 1% STL_REP pull-up | Resistor_SMD:R_0603_1608Metric | R17 |
| 1 | 10k 1% VPULL sense | Resistor_SMD:R_0603_1608Metric | R14 |
| 1 | 10k 1% nFAULT pull-up | Resistor_SMD:R_0603_1608Metric | R6 |
| 1 | 10nF X7R 50V (U6 VM pin 1) | Capacitor_SMD:C_0603_1608Metric | C17 |
| 1 | 10nF X7R 50V (U6 VM pin 8) | Capacitor_SMD:C_0603_1608Metric | C18 |
| 1 | 10uF X7R 50V 1210 (U6 VM bulk) | Capacitor_SMD:C_1210_3225Metric | C19 |
| 1 | 10uF X7R 50V 1210 (buck CIN) | Capacitor_SMD:C_1210_3225Metric | C1 |
| 1 | 15uH 3.8Asat (Bourns SRN6045TA-150M, AEC-Q200) | Inductor_SMD:L_Bourns_SRN6045TA | L1 |
| 1 | 1N4148W (VPULL back-feed block) | Diode_SMD:D_SOD-123 | D2 |
| 1 | 1nF X7R 50V (TRQ_CNT/STL_TH) | Capacitor_SMD:C_0603_1608Metric | C23 |
| 1 | 1uF X5R 50V | Capacitor_SMD:C_0603_1608Metric | C8 |
| 1 | 1uF X5R 50V (LDO out) | Capacitor_SMD:C_0603_1608Metric | C3 |
| 1 | 1uF X5R 50V (VCC) | Capacitor_SMD:C_0603_1608Metric | C12 |
| 1 | 20pF C0G 50V (FB feed-forward CFF) | Capacitor_SMD:C_0603_1608Metric | C29 |
| 1 | 220nF X7R 25V (VCP-VM) | Capacitor_SMD:C_0603_1608Metric | C21 |
| 1 | 220nF X7R 50V (buck CHF, VIN pin 10) | Capacitor_SMD:C_0603_1608Metric | C26 |
| 1 | 220nF X7R 50V (buck CHF, VIN pin 2) | Capacitor_SMD:C_0603_1608Metric | C25 |
| 1 | 22nF X7R 50V (CPH-CPL) | Capacitor_SMD:C_0603_1608Metric | C20 |
| 1 | 24.9k 1% FB bottom | Resistor_SMD:R_0603_1608Metric | R3 |
| 1 | 470nF X7R 25V (DVDD) | Capacitor_SMD:C_0603_1608Metric | C22 |
| 1 | 47k 1% HOME_SW series | Resistor_SMD:R_0603_1608Metric | R16 |
| 1 | 8MHz 12pF (Taitien XXDCELNANF-8MHZ) | Crystal:Crystal_SMD_3225-4Pin_3.2x2.5mm | Y1 |
| 1 | ADS1118IDGSR (proto; ADS1118-Q1 for production) | TI_DGS0010A_VSSOP-10:TI_DGS0010A_VSSOP-10_3x3mm_P0.5mm | U5 |
| 1 | DRV8434ARGER (proto; DRV8434A-Q1 for production) | TI_RGE0024N_VQFN-24:TI_RGE0024N_VQFN-24-1EP_4x4mm_P0.5mm_EP2.45x2.45mm | U6 |
| 1 | DTM06-2S nickel, 12V input (sealed) | TerminalBlock_Phoenix:TerminalBlock_Phoenix_MKDS-1,5-2-5.08_1x02_P5.08mm_Horizontal | J6 |
| 1 | DTM06-2S nickel, GM-style resistive temp sender input (sealed) | Connector_Molex:Molex_KK-254_AE-6410-02A_1x02_P2.54mm_Vertical | J4 |
| 1 | DTM06-2S nickel, home/limit switch input (sealed) | Connector_Molex:Molex_KK-254_AE-6410-02A_1x02_P2.54mm_Vertical | J7 |
| 1 | DTM06-4S nickel, NEMA14 stepper output (sealed) | Connector_Molex:Molex_KK-254_AE-6410-04A_1x04_P2.54mm_Vertical | J5 |
| 1 | LM74700-Q1 (AEC-Q100 G1) | Package_TO_SOT_SMD:SOT-23-6 | U3 |
| 1 | LMR36015AQRNXRQ1 60V 400kHz (AEC-Q100) | TI_RNX0012C_VQFN-HR:TI_RNX0012C_VQFN-HR-12_2x3mm_P0.5mm | U2 |
| 1 | Mini blade holder, 2A Littelfuse 297 fuse (SAE J2077/ISO 8820-3) | Fuse:Fuseholder_Blade_Mini_Keystone_3568 | F1 |
| 1 | NXP S32K144 automotive (AEC-Q100) | Package_QFP:LQFP-64_10x10mm_P0.5mm | U1 |
| 1 | PMV37ENEA automotive (AEC-Q101) | Package_TO_SOT_SMD:SOT-23 | Q1 |
| 1 | SMCJ24AQ-13-F TVS (AEC-Q101) | Diode_SMD:D_SMC | D1 |
| 1 | SWD Tag-Connect | Connector:Tag-Connect_TC2030-IDC-NL_2x03_P1.27mm_Vertical | J2 |
| 1 | TLV73333PQDBVRQ1 3.3V (AEC-Q100 G1) | Package_TO_SOT_SMD:SOT-23-5 | U4 |
| 1 | ferrite 1k@100MHz 2.5A (FBMH3225HM102NT) | Inductor_SMD:L_1210_3225Metric | L2 |
Similar boards
- 6 layers32 × 32 mm

- 6 layers40 × 40 mm

- 6 layers10 × 17.3 mm

OpenRX Mono
OpenDrone-hw/OpenRX-Mono
Open-source dual-band ExpressLRS receiver with one LR1121 radio and one U.FL antenna.
62 parts6CERN-OHL-S-2.0 - 6 layers72.1 × 65.5 mm

Frankfurt eth pic
grodansparadis/can4vscp-frankfurt-eth
Frankfurt Ethernet to CAN4VSCP bridge
62 parts2 - 6 layers120 × 40 mm

CM5CarrierV2
AsadImranRafique/CM5-Carrier-Board-Rev2
Open-source six-layer Raspberry Pi CM4/CM5 carrier board with USB-C, 25 W PoE, M.2 NVMe, microSD, I2S, and expansion-board connectors.
62 parts1Apache-2.0 - 6 layers143.5 × 83.8 mm

- 6 layers72 × 44 mm

- 6 layers50.8 × 50.8 mm

Questions
How big is the Thermo PCB?
It measures 85.3 × 51.9 mm and has 6 copper layers, 1.6 mm thick.
What parts does the Thermo use?
62 parts. The full list of its parts is on this page, with links to each part's symbol and footprint where PCBKit's library has them.
Where can I download the Thermo KiCad files?
From the jessiepullaro414/Thermo repository on GitHub, at the commit this page was made from: the board is Thermo.kicad_pcb and its schematic Thermo.kicad_sch. PCBKit opens the board in the browser, with nothing to install.
Can I use the Thermo design in my own project?
Its repository publishes it under the MIT licence: you may reuse it on that licence's terms (keep the author's notice, and for a copyleft licence share your changes alike).