Automobile cabin noise is generally categorized by the major contributing sources: powertrain, road, and wind. These three noise components have traditionally been evaluated at their dominant operating conditions, e.g., wide-open throttle for powertrain, which may not represent realistic customer driving conditions and can result in unnecessary vehicle cost/weight. A portable PC-based system using multichannel least-mean-square (LMS) adaptive filtering was developed and validated to provide the capability of separating powertrain, road, and wind noise components (time-domain) under realistic customer driving conditions. By judiciously selecting the locations of reference sensors on the engine and suspension with optimized filter length, the powertrain/road noise components can be successfully separated. Separated noise components from various vehicles can subsequently be used to synthesize total cabin noise for subjective sound quality evaluation. Results of vehicle noise separation/synthesis from various vehicles will be presented and accompanied by an audio demonstration. This newly developed adaptive noise separation/synthesis system using multichannel time-domain adaptive filtering was proven to be a useful tool for automobile noise separation/synthesis and could be extended to potential applications such as wind tunnel fan noise minimization, noise/vibration source/path identification, and benchmarking of vehicle system transfer functions under realistic driving conditions.