Abstract
Random numbers are used in several applications, including simulations, digital signal processing, machine learning, and others. For high-speed implementations, FPGAs are among the most suitable technologies. In this context, the Linear Feedback Shift Register (LFSR) is a widely used approach due to its parallel digital structure. However, the free-running sequence of the LFSR presents correlation among neighboring samples. This autocorrelated behavior makes the LFSR unsuitable for certain applications. This work proposes a multi-core parallel implementation with different seeds, in which the resulting autocorrelation sequence becomes impulsive. By combining the outputs of multiple LFSR instances, it is possible to overcome this limitation while maintaining circuit simplicity. The FPGA implementation uses low hardware resources and achieves a high clock frequency, demonstrating its efficiency for high-speed applications.