Novel blind channel estimators based on the finite alphabet property of inf
ormation symbols are derived in this paper for OFDM and related multicarrie
r code-division multiple access (MC-CDMA) systems. The resulting algorithms
are applicable not only to standard OFDM transmitters with cyclic prefix,
but also to the recently proposed zero padded OFDM transmissions that impro
ve symbol recovery at the expense of altering the transmitter and complicat
ing the equalizer. Based on FFT-processed received data, channel identifiab
ility is guaranteed regardless of channel zero locations and various channe
l estimation algorithms become available by trading off complexity for perf
ormance. Unlike existing blind channel estimators, the proposed alternative
s require short data records especially for PSK transmissions. The inherent
scalar ambiguity is easily resolved because it has unit amplitude and phas
e values drawn from a finite set. Decoupling channel from symbol estimation
enables a phase-directed operation that improves upon decision-directed sc
hemes that are known to suffer from error propagation. Practical issues are
also addressed including the presence of frequency guard intervals, conste
llation and power loading, various frame designs, coded transmissions as we
ll as semi-blind and online implementations for systems with training seque
nces. The algorithms are tested with simulations and also compared with exi
sting alternatives in a realistic HIPERLAN/2 setting.