Comms Simulink Tutorial

A Tutorial on Using Simulink™ and Xilinx™ System Generator to Build

Floating-point and Fixed-point Communication Systems

For EE225c, 2003

By Changchun Shi

Last Updated: March 10, 2003

Berkeley Wireless Research Center

EECS Department, University of California, Berkeley

1. Abstract

A simple communication system using root-raised-cosine filter on both transmitter and receiver will be built. The purpose is to show how to design a system in Simulink™ and Xilinx™ System Generator environment, starting from choosing the algorithm, to build the floating-point system, and to building the fixed-point system.

2. Motivation

Can we design a digital chip in a day? Research efforts in Berkeley Wireless Research Center (BWRC) and other places have indicated this is achievable. This tutorial will hopefully get you familiar with the design environment to reach this goal. Here Mathlab™, Simulink™ and Xilinx™ System Generator form the foundation, on top of which a number of Matlab™ scripts and Simulink™ libraries enable much of the design process automated. We will show how they can be utilized via designing a simplified transmitter-receiver system.

3. How to Start

You will need a computer with Matlab™, Simulink™, Xilinx™ System Generator installed in order to run through this tutorial. You will also need read/execute access to BWRC file server \\hitz.eecs.berkeley.edu\designs to use the Floating-point to Fixed-point Conversion (FFC) Tool. In addition if you want to learn how to map your design to FPGA, you need to refer to other tutorial such as System Generator Tutorial, Tutorial on BEE. To map to ASIC, you need to refer to Tutorial on Insecta.

A simple way to solve the problem is to login one of the MS Windows Remote Desktop Servers available in BWRC, namely intel2650-{1, 2, 3}.eecs.berkeley.edu, and a few others. You will need Remote Desktop Connection Client on your local PC to do that. If you are using Linux, you may use Rdesktop [Rdesktop]. Each of the Server has all the necessary tools installed correctly.

Once you have all the software ready, you need to map \\hitz.eecs.berkeley.edu\designs to your network drive, preferably H: disk.

The demo system in this tutorial can be found at H:\ffc\ffc_tutorial.mdl.

If you have never used one of the three tools before, you might want to ready Appendix A of this tutorial before you precede.

4. Algorithm Study

The system to be built here is a simplified version of that in your HW1[HW1]. Suppose we want to do base-band communication with 2-PAM modulation scheme at 1Mbits/sec. Under 2-PAM input symbols (1 symbol/ 1us), such as sequence choosing from binary integer {0,1}, are mapped into a data sequence choosing from {-A, A}. For convenience, we can let A = 1. The receiver needs a 2-PAM demodulator to map received signal into original integer. Suppose the channel impose additive white Gaussian random noise, but otherwise ideal.

Although we have assumed our channel is flat with no fading, in reality it could be band-limited (may also due to RF front-end filtering); thus rectangular base-band pulse in time domain (Sinc shape in frequency domain) through the channel will be clearly distorted. One technique to combating this is to have a low-pass pulse-shaping filter at the transmitter side [Proakis01]. For that one needs to first over-sample the data sequence at R MHz. This is usually done by an upsampler with integer R. A condition R>2 is necessary to satisfy Nyquist criteria.

However there are multiple reasons to make R even higher. One of them is to minimize

...