Spatial Filtering
Beamforming and spatial windowing across array elements to isolate signals by direction and reject off-axis energy.
Deep-Tech Research · Array Signal Processing
Advancing the math behind
Advancing the math behind
adaptive receive arrays
Kanad Laboratory is a deep-tech research effort into the core algorithms for multi-element arrays — spatial filtering, direction finding, adaptive nulling, and on-array neuromorphic classification. We're at the deep research point of this evolution right now.
Capabilities
Core research focus areas
The open problems we're actively researching, designed to compose into a single real-time array-processing chain.
Angle of Arrival Vectorization
Vectorized AoA estimation across multiple sources and snapshots, mapping steering vectors to bearing solutions.
Space-Time Adaptive Processing
Joint spatial and temporal (STAP) filtering to suppress clutter and jamming while preserving slow-moving targets.
Sample Covariance Matrix Inversion
Numerically stable estimation and inversion of the sample covariance matrix, with diagonal loading and rank control.
Carrier-Phase Distortion Compensation
Per-channel phase alignment and distortion correction to keep coherent processing accurate across the aperture.
Real-Time Telemetry
Live streaming of array health, weight vectors, and interference state for monitoring and closed-loop control.
Independent Nulling Matrix Scales
Per-null gain scaling within the adaptive weight matrix, allowing independent depth control for each interferer.
Edge-Computed Neuromorphic Spectrum Classification
Spiking neural inference on-array to classify emitters and modulation directly from the spectrum, at low power on the edge.
The Platform
One adaptive chain, from capture to weights
The focus areas compose into a single processing chain: estimate the covariance, invert it, form spatial and space-time weights, apply independent nulls, then classify the residual spectrum — all within the array's coherent processing interval.
- Deterministic timing — updates bounded to the coherent processing interval.
- Portable kernels — the same math across CPU, DSP, and FPGA targets.
- Numerical stability — loaded, rank-controlled covariance inversion.
- Full observability — telemetry taps at every stage of the chain.
import kanad as sf
chain = sf.ArrayChain(elements=64)
R = chain.covariance(loading=1e-3)
Rinv = chain.invert(R, rank=48)
w = chain.stap_weights(Rinv,
aoa=chain.estimate_aoa(),
null_scales="independent")
# classify on-array, real time
chain.run(weights=w,
classifier=sf.Neuromorphic(),
telemetry=sf.RealTime())
Our Research
The deep research behind each focus area
Each capability is an active line of inquiry — here is the harder question we're working on underneath it.
Spatial Filtering
Deriving beamformer weights that stay robust when steering vectors and array geometry are only partially known.
Angle of Arrival Vectorization
Vectorizing multi-source AoA so bearing estimates hold up under low snapshot counts and correlated arrivals.
Space-Time Adaptive Processing
Reduced-dimension STAP formulations that converge with limited training data and non-stationary clutter.
Sample Covariance Matrix Inversion
Loading and rank-selection strategies that keep the inverse well-conditioned when samples are scarce.
Carrier-Phase Distortion Compensation
Estimating and correcting per-channel phase drift without disturbing the coherent aperture.
Real-Time Telemetry
Exposing the internal adaptive state at rate without perturbing the processing timeline.
Independent Nulling Matrix Scales
Controlling null depth per interferer independently while preserving the mainlobe response.
Edge Neuromorphic Classification
Mapping spectrum classification onto spiking models that infer on-array within a tight power budget.
About us
A deep-tech research team
Kanad Laboratory is a small group of scientists and engineers working on adaptive array processing. We're at the deep research point of this evolution — exploring covariance estimation, adaptive weighting, AoA, phase compensation, and on-array neuromorphic classification, and turning that research into methods that can run on real-time hardware.
Have an array-processing problem?
Send us a note about your system and one of our scientists or engineers will follow up. You can also reach us directly at contact@kanadlabs.in.