The original MixChecker contains only the simulation of the linear behavior of the devices. We not only added the modeling of non-linear behavior, we have also completely rewritten the original processing to achieve phase linearity, lower latency and better resolution in all parts of the frequency spectrum. Read more about the science behind the algorithms...
Parallel Warped Filters
The linear part of the model uses transfer function modeling by fixed-pole parallel filters based on the dual-band warped filter design.
More info here: Balázs Bank and Germán Ramos, "Improved Pole Positioning for Parallel Filters Based on Spectral Smoothing and Multiband Warping," IEEE Signal Processing Letters, vol. 18, no. 5.
Thanks to the filter design on a warped frequency scale, the frequency resolution of the model takes into account the frequency resolution of the auditory system. The dual-band design brings even more precise modeling at low frequencies without reducing precision at high frequencies..
The parallel structure of 2nd-order minimum-phase filters brings extremely low latency, minimal phase distortion, and high stability.
Extended Wiener Model
The non-linear part of the model uses proprietary model based on the Wiener model.
The nonlinear model extends the standard Wiener model and introduces dynamic nonlinearity and frequency-dependent nonlinearities while keeps the computing demands low.
The extended Wiener model contains technology that keeps constant energy of the output signal for all distortion levels. It allows to simulate distortion produced by the device at higher SPL without increasing SPL produced by your studio monitors. This also protects your studio monitors against mechanical damage.
For the seamless switching among devices, the overall gain of all devices is normalized to the same objective loudness using algorithm derived from Loudness Unit (LU) meter (ITU-R BS.1770-3)
The scale of the distortion level is calibrated in real SPL measured in 1m distance in free field.
All characteristics of the devices used for identification of model parameters were measured using APx525 Audio Analyzer with acoustic option in anechoic chamber. Headphones were measured using B&K Type 4128-C.