Bare Metal Digest — Kardella Park Field Notes

I started the conversation with a question I'd been turning over for a while: what's the next thing to attack in terms of fossil fuel dependency?

We'd done the obvious work. Thirteen kilowatts of panels on the roof. Sixteen kilowatt-hours of battery in the garage. A Kia EV3 in the driveway. Pulse Solar — the home energy management system I've been building — now handles Aurora Power Hours events autonomously, switching the inverter and managing EV charging based on tariffs and battery state. The energy stack is largely sorted.

But energy is only part of it. Fuel moved the food on my plate. Fertiliser from natural gas. Cold chain on diesel. Real dependencies, just less visible than the power bill.

So: what's next?

The conversation went somewhere I didn't expect.

What Was Already in the Shed

Two hours in, we were talking about soil moisture sensors.

Not because I asked about them. The conversation had traced a path from fossil fuel dependency through food production to the specific problem of keeping two raised vegetable beds watered without someone remembering to do it.

And here's the thing: I already had most of what I needed.

Sitting in the shed at Kardella Park: a 20W solar panel. A Powertech PWM charge controller. A 12V 12Ah SLA battery. A DC voltage regulator. A weatherproof hard case with moulded cable glands already built in.

I'd bought most of this for various reasons over the years. None of it had been assembled into a system. The conversation looked at what existed and connected it — meaningfully different from generating a parts list from nothing.

The system that emerged: solar to battery through controller, 12V stepped down to power a Raspberry Pi Zero 2W, Pi reads capacitive soil moisture sensors and fires a 12V solenoid when a bed drops below threshold. Drip irrigation to both beds. Rain forecast via Open-Meteo. Telegram alert every morning.

Entirely off-grid. Entirely autonomous. Self-contained in a case I already owned.

The Pressure Problem

The detail that could have bitten me on installation day came up in passing.

The solenoid valve I'd found — a DFRobot 12V normally closed 1/2" unit, $17.50 from Core Electronics — requires a minimum working pressure of 0.02 MPa to open fully. That's roughly 2 metres of water head in a gravity-feed system.

We were planning to run it off a tank mounted on the back of a new shed, gravity-feeding down to the beds. The shed is upslope of the beds. But "upslope" and "2 metres of elevation" are not the same thing, and the difference matters.

The conversation flagged it. I'll measure it before the tank goes in. If the head isn't sufficient, I need a different valve or a small pump. This is the texture of useful planning — not the grand architecture, but the constraints that sit quietly inside it until they become problems.

The Wiring Diagram

At some point I asked for a wiring diagram and one appeared — an interactive SVG, every component labelled, colour-coded: amber for power, green for water, coral for control, dashed lines for data.

It's pattern matching at scale — Pi GPIO to MCP3008 over SPI, relay modules switching 12V solenoids, all documented and well-trodden. What was useful was having it assembled for my components, my property, my constraints, in one conversation rather than across a dozen separate searches.

The diagram fed a handover document for Claude Code — a markdown file with full hardware inventory, GPIO pin assignments, Python architecture, systemd service configuration, calibration notes, and a checklist of what wasn't done yet. The kind of document I'd normally write after a project, written before I've touched a single component.

The Parts List

By the end of the conversation I had a Core Electronics cart open.

  • Capacitive Soil Moisture Sensor v2.0 × 2 — $5.90
  • MCP3008 8-Channel ADC — $7.90
  • 5V Single Channel Relay Module — $3.95
  • Raspberry Pi Zero 2W with Soldered Headers — $32.75
  • 12V Solenoid Valve 1/2" — $17.50
  • 100 Pack Diodes (includes 1N5408 for solenoid flyback protection) — $7.70
  • Professional Female to Female Jumper Wires 40 × 20cm — $7.30

Total: $83.00 before shipping. Bunnings covers the drip lines, poly fittings, and inline filter. Jaycar covers the fuse holders.

The solar panel, charge controller, battery, regulator, and enclosure were already owned. The system costs under $200 in new parts to build.

I placed the order.

The Code

The Python side is about 200 lines, give or take. Most of it is obvious: read the ADC over SPI, poll soil moisture every ten minutes, fire the valve if a bed drops below its threshold, log the reading, sleep.

The non-obvious parts are the interlocks. The valve has a maximum safe open time — if the Pi crashes while the valve is open, the tank empties into a single bed. So the open duration is bounded twice: a cycle length set by the moisture loop, and a hard cap set by a watchdog that doesn't trust the main loop to behave. Rain check before any cycle — Open-Meteo call for the next 12 hours, precipitation above 3mm cancels the cycle. Telegram at 7am with the previous day's readings and anything that got skipped.

The code doesn't exist yet. The handover document does. When the parts arrive I'll point Claude Code at the document — probably two hours of generation, bench testing against a Pi 4, then deploying to the Zero 2W once that arrives.

Which is the second thing the conversation bought me: not just the parts list, but a code spec precise enough that the implementation becomes the easy part.

What This Is Not

It's not a story about AI replacing engineering judgment. The 2-metre pressure requirement came from a product datasheet. The call to use capacitive sensors rather than resistive — because resistive ones corrode in soil within months — is the kind of thing any experienced maker would know, and I'd have arrived at it eventually.

What the conversation did was compress the distance between "what's the next thing to attack" and "parts ordered." It held the context of a full afternoon of planning — property layout, spring timelines, a recovering partner who'll step into a working system, chicken relocation logistics, beehive positioning, gravity-feed hydraulics — and kept connecting the threads.

The hard case in the shed is real. The cable glands are real. The garlic goes in the ground this week regardless of whether the irrigation system is working.

That's the point. The conversation didn't build the system. It made it possible to start.

David Thomas is building Hyperdella from a rural property in northern Tasmania. Pulse Solar, the home energy management system referenced in this piece, is one of approximately twenty products in active development. Bare Metal Digest covers the real working stories behind the stack.