Skip to Main Content
The constant modulus algorithm (CMA) for blind multiuser detection is studied in this paper. A general expression for the solutions is given, which shows there are many undesired stationary points. Therefore, methods that guarantee that CMA converges to the desired point are needed. We present approximate conditions that set initialization in the lock region. Simulations show that CMA with variable step size has quicker convergence speed than MOE-RLS, while obtaining better performance.