← RFC

Hybrid TDMA Data Plane — Measured Result

Status: Stage 2 result. Date: 2026-06-08. Tag: sim-result-v3. Reproduce: cd sim && python3 -m experiments.tdma_vs_csma (writes tdma_vs_csma.png).

This is the headline Stage-2 result: the measured confirmation of the central design claim — TDMA eliminates the hidden-node loss that CSMA structurally cannot prevent. It is a relative comparison on identical topology and traffic, which the validation gate (north-georgia-validation.md) licensed.

Experiment

The same hidden-node hub as the baseline: K clients all heard by one hub, all mutually hidden from each other, each offering 1 msg/min (LongFast). Only the MAC changes:

• CSMA flood (Meshtastic-equivalent): clients can't sense each other (hidden), so they collide at the hub — ALOHA loss.
• Hybrid TDMA (Loomwave): the hub (INFRA) is the slot authority; each client gets its own slot referenced to the schedule beacon, so transmissions never overlap. Mutual hiding becomes irrelevant — clients never need to hear each other, only the beacon.

Result

K (hidden clients)	CSMA delivery	TDMA delivery	CSMA loss	TDMA loss	TDMA efficiency
10	0.885	0.998	11.5%	0.2%	92.5%
20	0.738	0.996	26.2%	0.4%	95.8%
50	0.499	0.992	50.1%	0.8%	97.9%
100	0.234	0.965	76.6%	3.5%	98.6%

• Collision-free by construction: the hub records < 5 collision decode failures under TDMA at K=50, versus thousands under CSMA — the slots make overlap impossible.
• Efficiency ≈ 98–99% at scale, matching the analysis (mac-layer-options.md §3). It is lower at small K (92.5% at K=10) because the per-frame beacon amortises over fewer slots — an honest overhead the analysis' "amortised beacon" assumption smooths over.
• The K=100 TDMA residual (3.5%) is latency, not collision: the frame period (~48 s for 100 LongFast slots) approaches the 60 s message interval, so Poisson bursts occasionally queue. It is a capacity/latency limit (the §3 ceiling), not a reliability failure — and exactly what adaptive modulation + spatial reuse (Q3, task #8) exists to relieve.
• Control overhead is just the beacon (< 15% of airtime, and falling with K), versus flooding's measured ~90% relay tax — the same lever seen from the MAC side.

Reading

The CSMA curve is the hidden-node problem the project brief names as the central challenge; the TDMA curve is the design answer. The gap widens with K — precisely the regime (an elevated relay gathering many mutually-hidden clients) where the analysis showed CSMA/CAD is helpless. The next results (INFRA-aggregated DV control budget, then adaptive modulation for the capacity ceiling the K=100 residual exposes) build on this same validated comparison.