Signal Processing, Redone
A high-level overview of our approach. We can't disclose specifics — that's the IP — but we can explain the problem space, our philosophy, and why our results differ from conventional methods.
Where GNSS Positioning Hits a Wall
Modern GNSS receivers collect an enormous amount of data. Multiple constellations, multiple frequencies, carrier phase measurements with millimeter-level resolution. The raw measurements are far more precise than the final position solution.
The gap between measurement precision and positioning accuracy is created in the processing pipeline. Conventional architectures — developed over decades — handle ambiguity resolution, atmospheric correction, and noise filtering in a sequence that was designed when the data was sparser and the constraints were different.
Today's receivers are using yesterday's processing architecture to handle today's data. The result is a system limited by algorithm, not hardware.
A Unified Processing Framework
Rather than improving individual components of the GNSS processing chain, we redesigned the framework that connects them.
Observation-Level Integration
Instead of processing observations through a fixed sequence of filters and estimators, our framework considers the full observation vector jointly — preserving correlations that conventional pipelines discard.
Realtime Phase Calibration
Methods for removing antenna and hardware phase bias from GNSS signals in real time — improving positioning accuracy without requiring pre-calibrated reference equipment.
Noise-Tolerant Design
Every stage of our processing chain is designed to handle noise sources that typically degrade conventional solutions — from multipath to hardware bias to atmospheric residuals — without special-case logic.
Multi-Frequency Exploitation
All available frequency bands are exploited simultaneously within a single mathematical framework — not combined post-hoc. This preserves the full information content of multi-frequency observations.
Rapid Convergence
The unified framework converges to fixed-integer solutions more quickly because it doesn't accumulate error through sequential processing stages. Fewer stages, fewer opportunities for information loss.
Constellation-Agnostic Core
The core processing logic doesn't depend on any particular constellation's signal structure. New constellations and signal types integrate naturally — they become additional observations in the same framework.
Backed by Data
We validate our approach against established reference stations and published benchmarks. Every performance claim is measured, not modeled.
Our validation methodology is thorough: controlled environments with known truth, extended observation campaigns across diverse geometries, and comparison against multiple reference implementations.
The results speak for themselves — and are documented in our filed IP and academic publications where disclosure is appropriate.
Independent Verification
Our results have been independently verified against high-precision reference data. Performance improvements are quantifiable and reproducible.
For specific validation results, see our IP & Publications page — or contact us for a technical deep-dive under NDA.