Handouts

EE 163B

1. Sample eye diagrams for raised-cosine pulses: 0% and 100% excess bandwidth
2. Effect of channel dispersion on the eye diagram
3. The whitened matched filter channel model
4. Demo of the Viterbi algorithm

Matlab files and demos

1. DSB suppressed carrier demodulation using a Costas loop:
   • Download costasloop_init.m and costasloop_demo.mdl.
   • Start Matlab, type costasloop_init, hit return, and then run costasloop_demo at the prompt.

2. Intersymbol interference:
   • Download and run raisedcosine_demo.mdl.
   • Change the filter rolloff factor ("beta") in the R-C filter block and see the results.
   • Observe the eye diagram for different values of beta. Is the eye "more open" when beta = 1? Can you see much less sensitivity to errors in the sampling times when beta is close to 1?
   • Try increasing the noise level by changing sigma in the Rayl N block from 0.001 to 0.1. See how the scatter diagram smears, and also see the effect on the eye diagram.

3. Adaptive equalization:
   • Download and run zeroforcing.m - this is a zero-forcing equalizer with no noise. Change the channel time constant tau in the program. Notice how the equalizer "opens the eye" perfectly.
   • Download and run zeroforcing_noise.m - this is a zero-forcing equalizer with additive noise. Notice that the equalizer appears to amplify the noise.
   • Download and run wiener_equalizer.m - this is a Wiener filter equalizer with additive noise.
   • Download and run lms_equalizer.m - this is an adaptive equalizer using the LMS algorithm.

EE 163A

1. A little thought experiment
2. Notes on $\pi$/4-QPSK

Matlab files and demos

1. Quantization effects:
   • Download and run quantization_demo.m.
   • Change the number of bits to R = 7, 6, 5, 4, 3, 2, 1 and listen to the output. You will notice that at a certain quantization level and signal-to-quantization-noise ratio the noise due to quantization becomes audible.
   • You can also click below to hear the effects of quantization (to R bits or L = 2^R + 1 levels) without running the program above: Original, 7 bits, 6 bits, 5 bits, 4 bits, 3 bits, 2 bits, 1 bit.

2. Matched filtering:
   • Download and run matchedfilter.m.
   • Note that best performance is obtained with rectangular matched filtering of the rectangular pulse.
   • Run the program repeatedly until the point where matched filtering of the triangular pulse appears to be better than rectangular filtering.
   • Change the noise amplitude and observe the effects.
   • Download mc_matchedfilter.m - this simulates the detection performance of various filters over 10,000 runs.