Phase Resolved near Field Dipole Reconstruction for Multilayer Board Far Field Escape Prediction
Phase-resolved near-field dipole reconstruction predicts far-field escape emissions within 2 dB, eliminating un-phased scan errors before SAC testing.
Singular Value Decomposition
Matrix Factorization
Mathematical procedures decompose a complex data matrix into orthogonal components to extract principal structures and filter noise. Singular value decomposition separates a multi-port circuit measurement matrix or a field-scanning matrix into a set of singular values and singular vectors. This mathematical tool helps engineers identify the dominant modes of electromagnetic coupling and reduce the dimensionality of complex data.
The boundary of this decomposition lies in its assumption of linearity, and it cannot directly separate non-linear combinations of signals.
Noise Reduction
Signal processing for multi-channel board testing utilizes this matrix factorization to separate signal from measurement noise. When a high-speed tester collects return loss data from dozens of board channels, singular value decomposition isolates the primary coupling mechanisms from random measurement variation. This isolation is achieved by keeping only the largest singular values and setting the smaller, noise-dominated values to zero.
The reconstructed matrix has a much higher signal-to-noise ratio, which improves the accuracy of subsequent defect localization algorithms. This process assumes that the noise is randomly distributed and that the true signal is represented by the dominant singular values.
Diagnostic Verification
Post-processing verification checks the fraction of total variance captured by the selected singular vectors. If the chosen vectors represent less than the required threshold of the original data, more singular values must be retained in the model. This check ensures that critical signal features are not discarded during the noise reduction process.