Recently, optimum signature sequence sets that maximize the capacity of sin
gle-cell synchronous code division multiple access (CDMA) systems have been
identified. Optimum signature sequences minimize the total squared correla
tion (TSC); they form a set of orthogonal sequences, if the number of users
is less than or equal to the processing gain, and a set of Welch bound equ
ality (WBE) sequences, otherwise. We present an algorithm where users updat
e their transmitter signature sequences sequentially, in a distributed fash
ion, by using available receiver measurements. We show that each update dec
reases the TSC of the set, and produces better signature sequence sets prog
ressively. We prove that the algorithm converges to a set of orthogonal sig
nature sequences when the number of users is less than or equal to the proc
essing gain. We observe and conjecture that the algorithm converges to a WE
E set when the number of users is greater than the processing gain. At each
step, the algorithm replaces one signature sequence from the set with the
normalized minimum mean squared error (MMSE) receiver corresponding to that
signature sequence. Since the MMSE filter can be obtained by a distributed
algorithm for each user, the proposed algorithm is amenable to distributed
implementation.