Loudspeakers for AEC: Measurement and Linearization

The performance of an AEC is hard-limited by the loudspeaker’s non-Linear-Time-Invariant-ness (non-LTI-ness), which is researched.

The performance of an AEC is hard limited by loudspeaker’s non-Linear-Time-Invariant-ness (non-LTI-ness). The assumption of “all loudspeakers are sufficiently LTI and essentially the same” had been been useful only for poor, essentially half-duplex AEC. Its validity is being verified on the sample of various 3"...5.25" loudspeaker drivers (plus about 10 more ommited as copycats):

- ADAM X-ART F5

- ADAM X-ART F5 AMT tweeter

- AIYIMA MJD4

- AIYIMA 3.5" silk film 12 Ohm 50W

- Bowers & Wilkins CWM-650 tweeter

- Focal PS130F

- Gooldwood Sound GW-205-W

- Gooldwood Sound GT-501

- HIVI Research DM-7500

- Infinity 4032

- LABO LB-PS1401D

- PYLE PRO PDBT19

- SB Acoustics SB12CACS25-4

- Sounderlink AMT-920 (aka NEO8)

- Vifa (Tymphany) NE95W-04

- JmLAB Spectral 908.1 as the reference.

LTI distortions vary wildly between loudspeakers. They can not be easily identified during a “regular” teleconferencing. Therefore, for an AEC striving to perform in the maximally transparent full-duplex mode, the aforementioned assumption is wrong and must be explicitly discarded. An adaptive filter can be more or less generic but a good AEC can not be generic. A better-performing AEC must be tied to a specific hardware.

Generally, there is not much point in improving AEC without improving loudspeakers first.