The architecture, in four parts
Every Excité loudspeaker is built on the same architecture: direct measurement of every driver non-linearity that meaningfully contributes to distortion, fed into a real-time, feed-forward correction loop running on FPGA, ahead of per-driver amplification. The architecture is the same in every line; what scales is the driver complement and the cabinet that holds it.
This page is the architecture, broken out into the four areas that make it work: correction, metrology, measurement methodology, and signal path.
Correction grounded in direct measurement
A loudspeaker's distortion is not random. Every meaningful contribution — motor non-linearity, suspension stiffening at excursion, voice-coil inductance dependence on position and current, surround resonance, thermal modulation of drive strength — is governed by parameters that can, in principle, be measured. What the parameters are doing right now, in this driver, at this signal level, can be known.
Excité's entire architecture is built on that premise. We measure those parameters directly, both during driver characterization and continuously during operation, and we use the measurements to compute a feed-forward correction in real time, per-driver, on FPGA. Nothing in the loop is inferred from terminal signals. Nothing is estimated from a generic model. The correction is built from what we know.
That requires correction architecture, metrology, measurement methodology, and a signal path that all hold together. The four cards below are how we've broken the architecture out for reading.
One architecture. Four places to look
Each area is its own engineering problem; together they form the loop that runs through every Excité loudspeaker.
What the parameters are doing right now, in this driver, at this signal level, can be known — and corrected for