Treat wild ideas as testable hypotheses
Ternary-Six9 is an experimental communications concept. Here's how I keep unusual ideas honest — models, fixtures, baselines and logged measurements before any conclusions.
Update, October 2026. The method below produced real results. The biggest: the nine symbols were never the problem — the circle was. Placed on a sphere instead of a phase dial, the same nine symbols gain 4.55 dB, and a coil tilt of arccos(1/√3) makes the fixture isotropic. The full write-up, including what didn't survive, is in the Academic Theory paper.
Some of my projects are practical: a speedcubing device, an SEO platform. Some are pure curiosity. Ternary-Six9 is the second kind — an experimental communications and control concept built around a hierarchical 3/6/9/12 addressing scheme with ordered ternary identifiers, a nine-state synchronization cycle and defined frame structures.
It might lead somewhere. It might not. Either outcome is fine, as long as the method is sound. This post is about that method.
The problem with exciting ideas
Exciting ideas attract confirmation bias. You see what you hope to see, especially in noisy physical measurements. The defense isn't skepticism toward the idea; it's structure around the experiment.
Step 1: state the hypothesis in a falsifiable way
"Three-phase geometry is better" is not testable. "Signals driven at 120° spacing produce a measurably different field pattern at sensor position X than signals at 123° spacing, under identical drive conditions" is.
That's the kind of comparison I set up: synchronized microcontroller outputs drive a set of coils, and field sensors log what happens. The only thing that changes between runs is the phase relationship.
Step 2: build a fixture before you build a theory
Hand-placed coils and sensors give you hand-placed results. Move a sensor by a few millimeters and your data changes. So before collecting meaningful data, I designed OpenSCAD fixtures that hold the three coils and the sensors in fixed, repeatable geometry.
A fixture turns "I think it was about here" into "position 4, every time." It's the cheapest way to make physical experiments reproducible.
Step 3: control the variables you can
- The same microcontroller generates all phases, so timing comes from one clock.
- Drive levels are logged, not assumed.
- Ambient readings are recorded with outputs off, to know the noise floor.
- Runs are alternated (A, B, A, B) rather than batched, so slow drift — temperature, supply voltage — doesn't masquerade as an effect.
Step 4: log everything, conclude later
Every run produces a file with timestamps, settings and raw sensor values. Analysis happens afterwards, in Python, on the raw data. If the code that analyzes the data is written after seeing the data, I'm careful to note that — it's an easy way to fool yourself.
Step 5: earn the next stage
The roadmap only advances when the current stage produces something solid. Low-frequency coil experiments come before software-defined radio. SDR comes before any claim about RF propagation. Each stage has to justify the cost and complexity of the next.
Why bother?
Because the habits transfer. The same discipline — fixtures, baselines, controlled variables, raw logs — makes my mesh networking tests honest and my product prototypes reliable. Curiosity projects are where you practice the method with the lowest stakes.
And occasionally, a strange idea survives measurement. That's when it gets interesting.
Model it. Build it. Measure it. Compare it. Only then believe it.
Some links in these notes are affiliate links. If you buy through one, I may earn a commission at no extra cost to you. I only link to tools I use or would recommend.