A four electrode earth resistivity meter (Wenner array) built from scratch in about a month, for roughly $150 in parts. Commercial instruments that make the same measurement start above $20,000.
This repo is written as an engineering log. Four ADS1115 ADCs died during development, all from the same root cause, and the failure log is probably the most useful file here. Read it before powering anything up.
The custom PCB, designed after the breadboard build below produced its results. KiCad sources, gerbers, renders and review notes are in hardware/.
| Test | Result |
|---|---|
| Bench, known 100 Ω resistor | measured 98.6 Ω (1.4% error) |
| Field, moist Georgia clay, a = 0.5 m | 71-72 Ω·m apparent resistivity, 1.24 mA injected |
No 2D image yet. What exists today is a validated single point measurement and one good field reading, both from the breadboard build. Full status below.
Contact resistance is the whole problem. Where a steel rod meets soil this build measured 40-80 kΩ, and a two electrode measurement puts that directly in series with the tens of ohms of ground you actually want. So the roles are split across four electrodes: current goes in through the outer pair (C1, C2), and the voltage it creates in the ground is read across a separate inner pair (P1, P2). The potential pair feeds an instrumentation amplifier and draws essentially no current, so its contact resistance drops no voltage and falls out of the result. On the current side, contact resistance only limits how much current flows, and the current is measured across a shunt rather than assumed.
From there it is Ohm's law: R = V/I, and apparent resistivity ρ = 2πa·R for electrode spacing a. Details in docs/theory.md.
flowchart LR
mega[Arduino]
bridge[L298N H-bridge]
muxa[MUX A]
muxb[MUX B]
muxcd[MUX C and D]
amp[AD620]
shunt["1 kΩ shunt"]
adc[ADS1115]
earth((earth))
mega -->|IN1, IN2| bridge
bridge --> muxa -->|C1| earth
earth -->|C2| muxb --> shunt --> bridge
earth -->|P1, P2| muxcd --> amp -->|A0| adc
shunt -.->|A2, A3| adc
adc -.->|I2C| mega
The H-bridge flips the injection polarity every ~150 ms. Differencing the two half cycles cancels electrode self-potential and amplifier offset, and stacking 20 cycles averages the noise down. The AD620 stage gain is 76 on the breadboard build, measured with a multimeter during calibration rather than taken from the trimmer setting. Four 16-channel muxes let any of 16 electrodes play any of the four roles.
- Bench validation against a known resistor
- One field reading
- 16 channel electrode switching built on the breadboard, 12 of 16 channels working (contact faults, see the failure log)
- Automated Wenner survey firmware, outputs pyGIMLi format
- Custom PCB designed: Arduino Nano plus shift registers, 2 layers, DRC clean, gerbers generated, firmware written and compiled
- PCB fabricated, assembled and brought up (one design issue to settle first, see the review notes in hardware/)
- Non-polarising potential electrodes (steel rods drift within minutes)
- 2D survey of real ground
In reading order:
- Theory: Ohm's law through the ground, Wenner geometry, why the polarity alternates, why stacking works
- Hardware: the signal chain, wiring reference, and five design constraints learned the hard way
- Calibration: the offset/gain procedure in the order that converges, with the numbers from this build
- Failure log: symptom, root cause, and fix for every failure, including the four dead ADCs
- Field procedure: how the field reading was taken and what limits survey duration
- Build guide: power, physical layout, every connection pin by pin, assembly order, pre-power checklist
- Software guide: toolchain, uploading, the bring-up walkthrough, capturing a survey, running the inversion
- Firmware: seven bring-up sketches for the breadboard and one for the PCB. Run them in order.
- PCB: the board, what changed from the breadboard, shift register wiring, BOM, fabrication, review notes
diy-ert/
├── docs/ theory, hardware, calibration, failure log, field
│ procedure, build guide, software guide
├── firmware/ bring-up sketches 00-06 (Mega breadboard), 07 (Nano PCB)
├── hardware/ PCB: KiCad sources, gerbers, renders, schematic PDF,
│ BOM, DRC/ERC reports
├── analysis/ pyGIMLi inversion script, capture cleaner, format notes
└── data/ data format spec and the one validated field reading
Roughly $150 in parts including spares. The PCB bill of materials is in hardware/README.md; the parts that are not on the board (rods, supply, wire, the breadboard-only trimmers) are listed in the build guide.
MIT, see LICENSE.