When probes measure the same limiting distance
Divide a ring into successively smaller cells. One probe takes 1 tick on the first link of each cell and 3 on the second. Another takes 3 and 1. Both register the same time between cell boundaries and different times at interior ports. With one turn as the unit of time, the discrepancy disappears as the cells get smaller. Distinct angular positions remain a positive distance apart. The construction below starts with local pulses and clocks.
What must converge?
Attach the same endpoint labels across scales and specify the preparation, relay and detector. Use the first attainable delivery time for each ordered pair. For a unit-delay network where incoming pulses combine by Boolean union, local synchronous evolution supplies this minimum. For interfering signals, check which first arrivals the controls and detector can actually recover. A monitor that stops before any arrival records an incomplete observation. First arrival in a network · Interference controls
[T] For one ordered pair, the following three tests together are necessary and sufficient for two calibrated probes to have a common positive limiting time.
- At any requested precision all sufficiently late readings of the reference probe agree with each other within that tolerance. This is the Cauchy test over the whole tail of the measurements.
- The two normalized readings approach one another to every requested precision.
- Beyond some scale the reference reading stays above a fixed positive threshold.
Finite positive clock units may vary independently for each probe. The proof encodes their exact ratios by upper dyadic approximations with error strictly less than one place of the chosen precision. Upper rounding preserves the relay inequality, even for two legs each shorter than one such place. The limit is constructed as an equivalence class of Cauchy readings with its own order and addition. Idle preparation, relay composition and asymptotic agreement of outward and return times give the metric laws. One may identify labels at zero distance, or certify a positive separation for every distinct pair of original labels (with stable equality). Limit and uniqueness · Arbitrary calibrated clocks
A finite certificate from the dynamics
At resolution , let a macroscopic unit last fine ticks. To certify a comparison with the next resolution, realize every elementary passage by an actual itinerary in the other network, in each direction of comparison. A local inequality may credit time to the starting port and debit the final port. Credits at intermediate ports cancel, including on repeated loops. Let bound the remaining timing error for the whole transfer, expressed in the finer ticks. Write for the first arrival between two fixed labels. Then .
Supply a nonnegative error reserve satisfying
[T] The intervals with centre and radius nest. They determine a unique limiting time. It is positive if and only if some finite measurement has . A constant error bound permits . When the ticks came from rounding calibrated readings, include the rounding error in the comparison budget. An independent lower-bound certificate is a register whose increase along each link is at most that link's delay and whose rise between the two ports exceeds the reserve. Such a register exists exactly when the first time exceeds the reserve. The first-arrival clock from the source attains the maximum possible rise. The scale-wide error certificate remains a premise of this finite test. Local translations and error bounds · Common metric from these observations
What a return experiment establishes
Choose a destination, require a visit there, then measure the earliest return to the source. When the two minimal legs compose, this reads the sum of their first times. If, for every labelled pair, two probes agree on this sum up to vanishing normalized error, and each probe has vanishing outward/return bias, their one-way distances converge together whenever the reference converges. A common clock is required throughout.
On a strongly connected shared support, supply access and return paths to a root. Agreement of both probes on each access/return loop and the rooted loop for each actual link gives agreement on every directed closed trip. For an open trip any remaining timing difference then depends only on its endpoints. For the example with outward/return times and , repeating this difference in macroscopic units gives identical closed-trip times and different one-way readings at every resolution. A separate failure occurs when probes agree only on one complete turn: two rings with quarter times and have the same circumference and incompatible distances even though both are reciprocal. Cycle clocks · Ring counterexample
A continuum built from pulses
Use cells with the same total time per cell and permit different interior delays for each probe, cell and resolution. Between rational angular labels the normalized error from the shorter circular arc is at most , uniformly over labels. The arc formula follows from the two physically available directions and a lower bound on every walk including repeated turns. For four-tick cells the construction extends to all infinite binary addresses. It identifies the two expansions of a dyadic cut and the two sides of the seam. Antipodes are positively separated. The station specified by the first digits is at most away in the resulting metric. From links to the observed arc · Infinite addresses and dense stations
With two independent ring coordinates, allow one coordinate to move on each tick. Actual paths split into the two coordinate trips and either trip can be executed while the other coordinate stays still. This yields the sum of the two angular distances and a separated metric on their product. Adding a diagonal step changes travel times even while individual axis tests continue to agree. This sensitivity is relevant when inferring geometry from an instrument's permitted actions. The broader study of stable norms provides precedents for large-scale geometry of periodic metrics. Product dynamics · Burago, Ivanov and Kleiner
Ways to falsify the inference
- Slow drift. A reciprocal two-station cycle has normalized first arrivals between 1 and 2. Two identical probes agree at every scale and the difference between successive readings tends to zero. Nevertheless the readings return to 1 and 2 at arbitrarily late scales. The all-tail Cauchy test rejects this model. Pulse-level counterexample
- Missing stations. Agreement at named ports controls the whole network if the named ports cover it at sufficiently small cost in both directions. With uniform discrepancy on named pairs and coverage radii , there is a correspondence covering both networks with distortion at most in common ticks. To infer vanishing distortion, these errors must vanish after normalization. An unobserved remote branch can otherwise survive in the limit. Coverage theorem
- Shortcuts and direction bias. An independent first-arrival calculation on periodic grids measured 5,440 destination sites over four resolutions. A bounded endpoint correction gave decreasing error. A growing correction kept the one-way error at of a turn while preserving all constrained round trips. Adding a one-tick shortcut drove the normalized antipodal times from down to . These are exact results on the specified simulated networks. Reproducible experiment
The general construction yields a distance on the declared observations. A physical theory must independently establish the clock calibration, allowed transfers, resolvable modes, endpoint coverage and refinement law. Dimension, smoothness, signature and gravitational dynamics remain separate questions.