Gpsspeed is an open-source printed circuit board by bocaj88, from the bocaj88/gpsspeed repository on GitHub, published under the MIT licence. It is a 2-layer board of 72 × 46 mm with 67 parts on 39 nets, routed with 506 tracks and 244 vias.
It is built around the ATGM336H-5N31, ESP32-C3-MINI-1-H4 and USBLC6-2SC6. Its parts include 18 resistors, 16 capacitors, 8 diodes, 4 ICs and 3 connectors. It belongs with the radio and wireless boards and the ESP32 and RISC-V boards of this library.
GPS → paddlewheel pulse emulator for PerfectPass, rebuilt as a JLCPCB-assembled
board. Replaces the Uno + perfboard hat.
Size
72 × 46 mm, 2 layers, 1.6 mm, 4× M3 holes
Power
12/24 V DC nominal, 32 V continuous max. 60 V polyfuse → 1.5 kW bidirectional TVS (±58 V clamp) → 100 V reverse-polarity Schottky → 22 µF/100 V bulk + 4 × 2.2 µF/100 V → LMR38010 buck (80 V operating, 85 V abs max). USB-C on the bench.
Brain
ESP32-C3-MINI-1-H4 (WiFi boot window, OTA updates; 105 °C part)
GPS
ATGM336H-5N31 (GPS+BeiDou, 10 Hz), u.FL for the existing active antenna
Output
ZXMS6004FF self-protected low-side switch (60 V, current limit, thermal shutdown, active clamp) behind a series Schottky (swapped +12 V/GND can't back-drive the dash), 33 V TVS on the drain; survives SIGNAL wired to +12 V; optional pull-up to the input rail via JP1 (ships open)
Connector
J1 screw terminal: 1 = +12 V, 2 = GND, 3 = SIGNAL (wires enter from the board edge)
LEDs
PWR (red), GPS (green), WIFI (green)
Rev 1.3 — third EE review (independent)
Review item
Change
Stop-ship: SIGNAL wired to +12 V shorts the supply through D5 and a bare 2N7002K the first time Q1 turns on
Q1 → ZXMS6004FF IntelliFET: current limit 0.7–1.7 A, thermal shutdown with auto-restart, short-circuit protected to VDS 36 V. Specified for 3.3 V logic (VIH 3 V, 0.6 Ω max at 3 V)
D4 (53 V max clamp) vs Q1 (60 V) margin thin
Q1's own active clamp is 60–70 V / 90 mJ, so a spike past D4 is absorbed by design instead of avalanching a bare MOSFET
(follows from the above)
SIGNAL_OUT / SIGNAL_DRAIN routed at 0.5 mm for the 1.7 A fault current
Bench items from that review (not schematic errors): output-low level with D5 in series against the real
PerfectPass input; USB-only startup at ~4.6 V into the buck. JP1 stays open on 24 V boats.
Q1 switching: ~15 µs on / ~60 µs off. Constant edge delays leave the output frequency unchanged; duty
shifts by ~1 % at 60 MPH.
Rev 1.2 — second EE review (reviewed the native KiCad files)
Review item
Change
C16 only 63 V against a 58 V clamp
22 µF / 100 V, same series and 6.3 × 7.7 can (RV100V22M6X8)
Battery-sense pin reaches 3.6 V at the clamp
R1 150k → 220k (×23): 32 V reads 1.4 V, the 58 V clamp 2.5 V. Firmware 4.0.2
Input ceramics borderline after DC-bias derating
4th 2.2 µF / 100 V (C17)
24 V pull-up would forward-bias D5 and make D4 clamp continuously
Q1 → 2N7002K 60 V, D4 → SMF33A (33 V standoff, 53 V clamp < 60 V)
Buck exposed pad: one full paste opening, no thermal path
Paste split into 4 windows (60 %); 2 tented GND vias just past the GND-pin end of the pad; pad tied solid into the pour
(found while re-checking)
Two u.FL stitching vias overlapped fanout vias (drill-to-drill). Fanout now avoids existing vias, and drill spacing fails the build
Not changed: L1 stays where it is (SW node ~8.7 mm total; moving it means relocating D2 in the
packed input corner — the reviewer rated it optional for prototypes). D2 stays an SS310: it has
to block up to ~53 V when USB and boat power are both connected, and USB power is bench-only.
D5 SS310 in series at J1 pin 3 (chosen over a resistor so a stiff head-unit pull-up can't break the low level; adds ~0.3 V)
Open via-in-pad under the ESP32
Removed; also stripped 4 open vias JLC's footprint put in the buck's exposed pad
Buck loop / u.FL grounding
Hand-routed C3–VIN–GND loop (~2 mm), SW straight into L1, FB on the quiet side; 8 stitching vias around the u.FL
Not changed (reviewed and accepted): USB diode-OR (verify startup during WiFi TX on the bench),
reboot on very deep crank dips (firmware ignores brownouts for rollback), SMF15A on the drain
(24.4 V clamp < Q1's 30 V).
Potting: keep resin off the ESP32 antenna end (top edge) — resin detunes the PCB antenna.
Leave an air pocket there, or pot only the lower two-thirds of the board.
What fixes the old problems
21 MPH cliff: gone. The square wave comes from the LEDC hardware peripheral off
the 40 MHz crystal; no software can stretch it. Measured on the real firmware code:
worst frequency error 0.02 % across 1–60 MPH (the Uno drifted up to 2.5 % fast
at 30 MPH from its 128 µs rounding).
One serial port: the GPS has its own hardware UART; USB-C serial is always free.
Tuning at the dock with a laptop: everything is edited from your phone.
Files
text
pcb/
gpsspeed.kicad_pro / .kicad_sch / .kicad_pcb open in KiCad 10
gpsspeed.kicad_dru two scoped DRC exceptions (explained inside)
lib/ exact JLC footprints/symbols/3D (easyeda2kicad)
gen/ the generators -- design.py is the netlist source of truth
out/jlcpcb/ gerbers.zip, bom_jlcpcb.csv, cpl_jlcpcb.csv <- upload these
out/ schematic PDF, assembly PDF, STEP, renders
firmware/gpsspeed_v4/ ESP32-C3 firmware; bin/ has prebuilt images
firmware/test/ host-side tests of the pulse + calibration code
Ordering at JLCPCB
PCB: upload out/jlcpcb/gerbers.zip. Defaults are right: 2 layers, 1.6 mm,
HASL lead-free, green. Qty 5 (the minimum).
PCB Assembly: on, Standard, Top side, qty 5. (Economic won't take
the ESP32 module -- JLC flags every ESP32-C3-MINI-1 "Standard PCBA only".)
Upload bom_jlcpcb.csv and cpl_jlcpcb.csv.
In the parts review, every line should match an LCSC number. J1 (screw
terminal) is through-hole; if Economic won't place it, untick it and solder its
3 pins by hand.
In the placement preview, check polarity/pin-1 on: D1–D5, C16 (+), LED1–3, U1–U4, Q1,
J2, J3. Footprints came from JLC's own library, so rotations should already
match, but this is the one step a human should eyeball.
Rev 1.1 quote (2026-09-26), for reference: 5 assembled boards $162.84, + UPS ~$37 + US
customs duty ~$61 (JLC collects it up front with DDP shipping) − $10 coupon ≈ $251.
Bring-up
Before power: meter +12 V to GND on J1 — must not beep.
(or arduino-cli upload — build command is in gpsspeed_v4.ino). If it won't
connect, hold BOOT, tap RST, release BOOT.
Serial monitor at 115200: expect chars: climbing and ant:OK once the antenna
is on. ant:OPEN = antenna not connected / passive; ant:SHORT = cable fault.
Plug in the antenna, go outside, wait for the GPS LED to go solid.
Then wire the boat: +12 V, GND, SIGNAL to the paddlewheel signal line.
Using it on the boat
Power on → join WiFi GPSSpeed-XXXX (open, no password; XXXX is fixed per board) → the page pops up, tap Done → later just open http://gpsspeed.local
(Safari; a captive-portal sheet may open on its own).
The access point lives 2 minutes after power-on, extends while your phone is
connected, then turns the radio off. To get back in: cycle power.
Calibrate once: tap Self-test, and at each step type what the dash shows into
Dash shows → Save point. One pass gets the dash within ~0.01 MPH at the
calibrated speeds (tested against a simulated nonlinear dash).
While GPS is searching the output holds a steady 5 MPH (never sweeps).
Firmware updates:Firmware → Upload the -ota.bin. A bad build that
crash-loops rolls back to the previous one automatically; brownouts from engine
cranking do not count as crashes.
Things I could not verify without hardware
CASIC config commands ($PCAS01/02/03) follow ZhongKe's documented format. If
the module ignores one, the firmware still works at its defaults (9600 baud, 1 Hz)
— the web UI shows the actual baud/rate it achieved.
Default Hz/MPH is 4.42, the Uno's effective constant at 15–26 MPH. Expect to
run the self-test once.
Active antenna assumed (GT-U7 kits ship with one). For a passive antenna,
remove L2.
Wiring polarity: the board survives +12 V/GND swapped (D5 keeps the swap from
back-driving the dash input) and SIGNAL wired to +12 V (Q1 current-limits and thermal-cycles).
Still meter it before powering.
JP1 on 24 V boats: leave it open. Bridged at 28.8 V, R15 (0603) runs near its 0.1 W
rating; the dash's own pull-up is the normal case anyway.
It measures 72 × 46 mm and has 2 copper layers, 1.6 mm thick.
What parts does the Gpsspeed use?
67 parts, built around the ATGM336H-5N31, ESP32-C3-MINI-1-H4 and USBLC6-2SC6. 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 Gpsspeed KiCad files?
From the bocaj88/gpsspeed repository on GitHub, at the commit this page was made from: the board is pcb/gpsspeed.kicad_pcb and its schematic pcb/gpsspeed.kicad_sch. PCBKit opens the board in the browser, with nothing to install.
Can I use the Gpsspeed 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).
Plant Monitor is a way to view data from a remote humidity sensor on web interface, which then tells you when you need to water your plants plus the air humidity, soil humidity, air temp, and air pressure.
Warum habe ich mich dagegen entschieden, das Sim Modul selber zu implementieren? Ich habe kein Chip gefunden, der günstiger als das fertige Modul war, und ein Datenblatt in Englischer Sprache hat.
bocaj88/gpsspeed
