5.2 Fir filter lab

Introduction

This module is a lab assignment for implementing an FIR filter using a circular buffer.

Reading

• For a description of FIR Direct-Form structures see the module FIR Filter Structures .
• For a description of circular buffers see Circular Buffers .
• FIR Filter Implementation .

Filter implementation

In this lab you will be writing a function fir_filt that will take as its input one new sample and produce one output sample. The fir_filt function will have the following basic structure:

Store the input value in a circular buffer

Compute the new output value

Return the new output value

Store the coefficients in in a header file

Here are some tips for writing the function.

• Store the coefficients as float values.
• Store the input as Int16 values (use a global variable or a local static variable). This is not very good programming practice but we will do it to simplify the processing.
• The intermediate sum in the filter calculation should be stored in a float variable initialized to 0.0.
• Be sure to cast the output float variable to an Int16 before output. e.g. return (Int16) var;
• When calling the function use a statement like outbuf[2*i] = fir_filt(inbuf[2*i]);

Pre-laboratory

• Design a Bandpass FIR filter with the following requirements:

1. Bandpass
2. Stopband attenuations: 40 dB
3. Passband: 1300Hz to 2000 Hz
4. Transition widths: 600 Hz
5. Sample rate: 8000 samples/sec.
6. DO NOT USE BUILT IN FUNCTIONS IN MATLAB for filter designs. You can use functions like window and sinc .

• For the filter design, your prelab report should include a plot of the filter coefficients and the magnitude and phase plots of the filter spectrum with non-normalized Hz as the x-axis.
• Save the coefficients to an ASCII data file to be used in your C function below. It would be advantageous to have this completely automated so that you could simply change the bandpass region and output all of the above for a new frequency range (in MATLAB see fopen , fprintf , fclose ). Your header file should have the coefficients in an array, be saved in a format like this: 3.007445391758509e-003 and be data type float .
• Write a C function called fir_filt that takes as its input one input time point and outputs one filtered time point. Use the coefficients designed above. Use a circular buffer to store the saved values of the input.
• Write a program that filters one channel of the CODEC input and zeros the other channel. The data should be filtered with the function fir_filt .
• Your function fir_filt should be located in a separate file from you main program. Also, create a header file to contain the declaration of your function fir_filt .
• Your files should be organized something like this:

1. DSK6713_audio.c - contains your main code for sending and receiving data and filtering the signal.
2. bpf.h - contains your filter coefficients and a macro definition for the length of your filter (e.g. #define N 70 ). This file will be included in fir_filt.c .
3. fir_filt.c - contains your FIR filter function called fir_filt .
4. fir_filt.h - contains the declaration of your FIR filter fir_filt . This will be included in the DSK6713_audio.c file.

Laboratory

• Run the program on the DSK and demonstrate the filtered channel. When you build the project you may need to have the -o3 option selected. This will help optimize the code. If you are debugging the code you will need to set this back to None since the optimized code does not work well in debugging.
• Use an input sinusoid and measure the amplitude of the output at different frequencies to test the frequency response. You could use a spectrum analyzer to get a power spectral density of your filter or you could use your laptop to analyze the system.
• You should verify your spectrum that you determined in your pre-lab work. Use a good experimental write-up for this lab .