SI Hub-and-Spoke Reversibility Benchmark
Direct unit converters often accumulate compounding rounding errors when chaining reciprocal factors because they rely on uncoordinated decimal multipliers. This interactive benchmark tests reciprocal numerical stability across 40,000 conversion hops (10,000 cycles across four length units) comparing uncalibrated direct pairs against an SI canonical base hub.
Live In-Browser 40,000-Hop Stress Test
10,000 cycles: Fathoms → Inches → Meters → Nautical Miles → Fathoms
Why Most Unit Converters Drift
In traditional unit converters, every unit pair is stored as an arbitrary floating-point conversion factor (e.g. 1 fathom = 1.8288 m, 1 m = 0.546806649... fathoms). Because binary computers cannot precisely represent arbitrary base-10 decimals, small rounding errors accumulate every time a value is converted back and forth.
The Moonglade Hub-and-Spoke Solution
Instead of a fragile N-by-N conversion grid, Moonglade implements a two-step canonical routing architecture:
- Spoke to Hub: Every unit converts to its exact SI base dimension (e.g., Meter, Kilogram, Kelvin, Joule) using exact integer rational fractions (N / D).
- Hub to Spoke: The canonical SI base value converts to the target unit by multiplying by the exact reciprocal fraction (D / N).
While rational definitions are algebraically exact, client-side browser execution relies on standard IEEE-754 double-precision arithmetic (Float64). Conversions involving affine offsets (such as Celsius to Fahrenheit) or non-terminating binary fractions can exhibit machine-epsilon variations (e.g., round-tripping 0.1 °C can produce 0.10000000000000142). The benefit of SI canonical routing is eliminating uncontrolled compound drift between disparate unit pairs by anchoring every spoke to a single verified reference definition.