CMA ON MICROWAVE CHANNEL

The file simulate_cma.m is Matlab code which generates noisy received signals from a linear channel model driven by a QAM source constellation and adapts a FSE (T/2) using CMA. User selectable variables are in the Input Variables section, and the channel is selected in the Channel section of the code.

The channel used in the simulation may be obtained from the SPIB database at Rice University.

simulate_cma.m

The following subroutines are needed to run the simulation.
isi.m
residual.m

The following example demonstrates CMAs ability to adapt a FSE blindly from a received sequence synthesized from a shortened version (length-16) of chan2.mat from the SPIB database at Rice University. The CMA-FSE successfuly reduces the cluster variance so that transfer to a Decision Directed mode is possible should further error rate reduction be desired.

The source is 16 QAM, white and equiprobable. The equalizer is length-16 and initialized with a unity center spike and all other taps zero, and the stepsize is 0.001. The SNR at the equalizer input is 35dB. The received signal is normalized to (near) unity power.

The plots show

The magnitude of the T/2-spaced channel impulse response
The magnitude of the T-spaced
combined channel-equalizer impulse response after 30,000 iterations
The noisy received data before equalization
The equalized data.