A bench tool for pH-electrode boards (electrometer front ends). A battery-powered simulator feeds a known voltage E to the board through a high series resistor, the way a glass electrode would. An ESP32-C3 with an ADS1115 measures E and the board output S, works out the board's gain and zero from two calibration points, and from then on shows the board's input leakage current - the fault that makes a board ignore a perfectly good electrode while still reacting to a low-impedance test source. Readings go to a 128x64 OLED and to a web page (EN / UA) with live charts. The ESP32-C3 also measures its USB 5 V supply through a 15 kΩ / 10 kΩ divider on GPIO4, so supply changes are recorded alongside the high-impedance measurements.
Why it exists: a board that ignored a working electrode. How the leakage is worked out: docs/leak-measurement.md.
The same circuit as a schematic: docs/schematic.pdf (KiCad 10 project in hardware/).
- A: the simulator voltage in mV, and the pH it stands for (
pH = 7 - A/59.16, 25 °C). - S: the board output in V, and the pH the board reports after calibration.
- Leakage current in nA, how many times it exceeds your limit, and a chart over time.
- The board's gain and zero (two-point line E → S), the noise on S and the board's supply current.
- USB supply voltage, measured by the ESP32-C3 ADC on GPIO4 and calibratable against a multimeter.
- Electrode equilibrium: how fast a real electrode settles after it is dipped (time to equilibrium, t63, drift).
- History in the browser (1 point/s, ~14 h), downloadable as CSV.
| Controller | ESP32-C3 SuperMini |
| ADC | ADS1115 module (I2C 0x48), powered from 3.3 V |
| Display | SSD1306 128x64 OLED, I2C 0x3C |
| Simulator | LR44 cell, DPDT ON-OFF-ON switch (polarity), 6.8 kΩ, 10 kΩ trimmer |
| Series resistor | 50 MΩ (five 10 MΩ SMD resistors in series), shorted by a push button for calibration |
| Board connector | BNC female; JST 3-pin (S, +5 V, GND) |
| S divider | R3 = R4 = 10 kΩ (the ADS gets S/2) |
| Current shunt | Rsh (R5) = 10 Ω in the board's ground |
| USB supply divider | R7 = 15 kΩ from USB +5 V to GPIO4; R6 = 10 kΩ from GPIO4 to GND (USB = 2.5 × pin voltage) |
Wire by the pin names printed on the modules. Never connect the BNC centre (the high-impedance node) to the ADS - its input would itself become the leakage.
Connect the JST to the board under test by function - +5 V to its supply, GND to its ground, S to its output - and check the pin order at both ends before you power it up. The JST here has +5 V in the middle; the pH board of this project has its ground there (P2: 1 VCC, 2 com, 3 S), so the cable between them is crossed.
PlatformIO Core 6.2.0 or newer (pioarduino needs it). In the
folder with platformio.ini:
pio run -t upload -t monitorEvery version is pinned in platformio.ini - the platform
(pioarduino 55.03.312-1 = Arduino core 3.3.12, ESP-IDF 5.5.5, GCC 14.2) and each library
(Adafruit ADS1X15 2.6.2, Adafruit BusIO 1.17.4, U8g2 2.36.18, AsyncTCP 3.5.0,
ESPAsyncWebServer 3.12.1). The first build downloads them. Do not loosen the pins, or
the project will stop building in a year.
platformio.ini names no port: with one board plugged in, PlatformIO finds it on its own.
With several (another ESP on the same USB hub, say), tell it which one. Either once:
pio run -t upload --upload-port /dev/cu.usbmodem1101or for good, in platformio_local.ini next to platformio.ini (ignored by git, picked up
by extra_configs):
[env:esp32-c3-devkitm-1]
upload_port = /dev/cu.usbmodem1101
monitor_port = /dev/cu.usbmodem1101Find the name with pio device list. The ESP32-C3 SuperMini uses native USB, so it shows
up as /dev/cu.usbmodem… on macOS, /dev/ttyACM0 on Linux and COMx on Windows.
No network is compiled in. On first start the calibrator cannot join anything, so it brings up its own access point:
| Network | pH-Calibrator |
| Password | calib1234 |
| Page | http://192.168.4.1/ (a phone usually opens it by itself) |
Open the Wi-Fi block at the bottom of the page, press Find networks, pick yours,
type the password and press Save and restart. The calibrator joins it and from then on
answers at http://calibrator.local/ (or the IP it shows on the OLED for a few seconds
after start). Only 2.4 GHz networks work.
If the saved network is out of reach, the calibrator falls back to the same access point after 20 s; Forget the network sends it there straight away. A network that drops after the calibrator has joined it is reconnected automatically.
For your own builds you can also put a default network in src/wifi_secret.h (ignored by
git). It is used only until a network is saved on the page:
#pragma once
#define WIFI_SSID "my-network"
#define WIFI_PASS "my-password"The page has no password: use it on a network you trust.
Channel A (the simulator voltage, ADS1115 AIN0-AIN1) has a sign that depends on which way round the two wires are. Check it once after assembly:
- Press and hold the button (the series resistor is shorted).
- Put a multimeter on DC mV across the BNC: red to the centre, black to the shield.
- Compare its sign with A on the OLED or the page.
Same sign - nothing to do. Opposite sign - in src/main.cpp change
A_SIGN (-1.0f to 1.0f or back) and flash again, or swap the AIN0 and AIN1 wires. The
size should match the meter to within a few tenths of a mV; Autozero A on the page
removes a constant offset (with the trimmer at its zero end).
The sign matters: pH on the page is 7 - A/59.16 (positive mV = acid), and the leakage
is calculated from the difference between A and the board's input, so a wrong sign gives
a wrong pH and a meaningless leakage figure.
- Button pressed: set some E with the trimmer, Take point 1. Set a clearly different E (flip the polarity switch), Take point 2. The line is ready.
- Enter your series resistor in Resistor R, MΩ (50 by default).
- Button released: read the leakage. Watch the chart while a washed board dries.
Details, formulas and pitfalls: docs/leak-measurement.md.
The page's «Download CSV» keeps only time, pH and S. To keep every reading - E, S, the
board input, the leakage and the calibration line - log what the page itself polls:
python3 tools/calibrator_log.py -o run.csv (type a note and Enter to mark a phase).
USB CDC, 115200 baud. One CSV line millis,A_mV,S_V,pH per measurement cycle (about
0.4 s: each channel averages 25 samples). Commands:
z |
autozero A |
Z |
clear the autozero |
r |
print the autozero |
v |
wiring check: each ADS input AIN0..AIN3 against GND, in volts |
- The pH shown for A assumes the ideal Nernst slope at 25 °C; there is no temperature compensation - it is a simulator, not a meter.
- Leakage resolution is set by R and noise, not by the arithmetic. With 50 MΩ one ADC count is under 1 pA; mains hum and drift on the high-impedance node are larger. A bigger R (1 GΩ) needs the node in air or on PTFE, inside a metal box.
- The board under test is powered from USB 5 V. A board whose zero follows its supply will shift a little if the USB voltage changes between calibration and measurement.
- Channel A needs a board. The simulator's reference - the trimmer leg, the BNC shield and A1 - has no divider of its own. With a board connected the readings are valid; without one A1 floats and A means nothing. A 10 kΩ / 10 kΩ divider from 3.3 V to GND with 100 nF, its midpoint on A1, would make A valid on its own; the author's unit works without it.
Open hardware, in three parts (see LICENSE):
- hardware (wiring diagram, schematics, boards): CERN-OHL-S v2;
- firmware and scripts: GPL-3.0-or-later;
- documentation and images: CC BY-SA 4.0.
You may build, modify, share and sell the device. A modified version you distribute must stay under the same licences, with its sources published.