When designing and using an ANVC system it is of interest how the system will be affected by different disturbances in the error microphones/accelerometers. With the use of a reference signal adaptive feedforward systems can attenuate a primary field substantially given that the noise field is tightly correlated with the reference signal. Questions that often arise in these applications are therefore how disturbances affect the performance: Will random noise in the error microphones affect the algorithm? Will error microphone tonal components affect the algorithm? Will speech in the error microphones affect the algorithm? How is convergence speed reduced by disturbances? In this paper, a filtered-X LMS feedforward controller has been investigated. Different types of noise sources have been injected and convergence speed as well as global reduction have been analyzed, given different contamination levels.