Research project ยท Allen High School ยท Wi-Fi 7 & 6G

Wireless links fail
without warning.
WaveLynk predicts it.

We discovered a sharp instability boundary โ€” the Coherence Cliff โ€” where Zero-Forcing beamforming transitions from stable to catastrophically unreliable. WaveLynk uses a physics-derived metric to detect it before the link fails.

How It Works โ†’ View Results Read the Paper
โ†“0%
Outage rate reduction
vs. fixed ZF strategy ยท 100 trials
โ†“0%
Latency reduction
measured at weak signal conditions
100
Validated trials
hardware testbed + Monte Carlo simulation
The Problem

Zero-Forcing beamforming fails suddenly, not gradually

Modern Wi-Fi 7 and 6G routers use Zero-Forcing (ZF) precoding to serve multiple users simultaneously. ZF works by inverting the wireless channel matrix โ€” this gives peak performance when channel information is fresh.

The problem: as users move, the channel estimate becomes stale. When staleness crosses a critical threshold, matrix inversion amplifies errors exponentially โ€” producing sudden, catastrophic packet loss and latency spikes. There is no warning.

Learn the physics โ†’
The Coherence Cliff
STABLE REGION INSTABILITY ZONE CCI = ฮณ switch โ†— Fixed ZF (collapses) WaveLynk (stays stable) Performance CCI โ†’
The Core Metric

Conditioned Coherence Index (CCI)

A single physics-derived number that tells you how close the channel is to the cliff.

CCI(t) = ฮบ(H) ยท |Jโ‚€(2ฯ€ยทfDยทฯ„)| ยท SINR/(SINR + ฮฑ) ยท eโˆ’ฮฒยทฯ„/Tc
ฮบ(H)

Channel conditioning

How much ZF amplifies any CSI error โ€” the spatial instability factor.

Jโ‚€(ยท)

Doppler correlation

Bessel function measuring how fast the channel changes with user velocity.

SINR/ยท

Signal quality

Normalized signal-to-interference ratio โ€” weights contribution by link quality.

eโˆ’ฮฒยทฯ„/Tc

CSI aging

Exponential decay capturing how stale the channel estimate has become.

Switching rule: when CCI โ‰ฅ ฮณ = 0.6, WaveLynk switches from ZF to MRT beamforming. The threshold ฮณ = 0.6 is derived analytically from the Bessel function at a typical feedback delay ฯ„ โ‰ˆ 0.3ยทTc.
How It Works

Predictive, not reactive

Existing systems react to performance drops. WaveLynk computes CCI in real time and acts before the link collapses.

1

Monitor

Continuously estimate CSI age ฯ„, Doppler shift fD, and channel condition number ฮบ(H) from standard CSI feedback.

โ†’
2

Predict

Compute CCI in real time. When CCI approaches ฮณ = 0.6, the system is approaching the coherence cliff.

โ†’
3

Switch

Proactively transition from ZF to MRT before failure โ€” preserving throughput and preventing latency spikes.

Hardware Validated

Real Wi-Fi 7 testbed

100 independent measurement trials on a Netgear Wi-Fi 7 router with iperf3, confirming the theory against real-world wireless conditions.

See the data โ†’
Open Source

Full Python implementation

Complete source code: CCI computation, Jake's fading channel simulator, ZF/MRT precoders, and the adaptive controller โ€” all in reproducible notebooks.

Browse the code โ†’
Interactive

Live simulation models

Three interactive models let you tune channel parameters in real time and watch the CCI-based switching mechanism respond to approaching instability.

Try the models โ†’
Get started

Explore the work

Dig into the physics, run the simulations, or try the live interactive models.

How It Works โ†’ Results Live Models