EE-256 Semester Project Report

This report covers all the aspects of the project. It explains how things were executed and accomplished. The layout of the report is as follows:
1. 2. 3. 4. 5. 6.

MATLAB Filter Design and Analysis Filter Coefficients MATLAB Code C5416 DSK Program Results Using Simulated audio signal Conclusion and Comments

Filter Design and Analysis

FIR Low Pass Filter

Lowpass Filter with Cutoff Frequency 2400Hz 50 Magnitude (dB)

-50

-100

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

0 Phase (degrees)

-2000

-4000

-6000

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Group Delay of FIR Filter 49.5

49

48.5

48

47.5

100

200

300 samples

400

500

600

Pole-Zero Plot of FIR Filter 6

2 Imaginary Part 97

-2

-4

-6 -16 -14 -12 -10 -8 -6 Real Part -4 -2 0

IIR Notch Filter

Notch Filter with Notch Frequency 1200Hz 0 Magnitude (dB)

-20

-40

-60

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

100 Phase (degrees) 50 0 -50 -100

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Group Delay of IIR Filter 6

100

200

300 samples

400

500

600

Pole-Zero Plot of IIR Filter 1 0.8 0.6 0.4 Imaginary Part 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1 -0.5 0 Real Part 0.5 1

Filter Coefficients

The filter coefficients of our desired filters designed in MATLAB were converted to Q.15 using MATLAB. The function to convert them into Q15 Hex numbers, which were written down in the fir_coeff.dat file. These coefficients are listed below. FIR Coefficients 98 coefficients of FIR filter in Q.15 form
.word -13 .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word -75 -36 116 -50 -94 131 5 331 -57 -333 328 124 -476 234 .word .word .word .word -154 116 85 -200

-203 43 44 -76 8 82 60 181 -210 -40 271 -167 -178

.word .word .word .word .word .word .word

367 -564 21 650 -548 -338 942

.word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word

-356 -889 1207 148 -1795 1407 1508 -4109 1515 17251 17251 1515 -4109 1508 1407 -1795 148 1207 -889 -356 942

.word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word

-338 -548 650 21 -564 367 234 -476 124 328 -333 -57 331 -178 -167 271 -40 -210 181 60 -200

.word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word .word

85 116 -154 5 131 -94 -50 116 -36 -75 82 8 -76 44 43 -203 -13

IIR Filter Coefficient The IIR coefficients were converted to Q15 hex numbers and then converted to decimal format that was given to the main file. The decimal numbers were represented in signed 16-bit numbers. The coefficients are shown below in all the formats.
; IIR Filter coefficients (sequential locations) b0 .word 9598 ;257Eh ;b0 = 0.2929 b1 .word -11282 ;D3EEh ;b1 = -0.3343 b2 .word 9598 ;257Eh ;b2 = 0.2929 a1 .word -11282 ;D3EEh ;a1 = -0.3343 a2 .word -13572 ;CAFCh ;a2 = -0.4412

MATLAB Code
This is the code that I used to design the 2400Hz cutoff frequency low pass FIR Filter and 1200Hz Notch IIR filter. The magnitude and phase response, group delay and Pole-zero plot were seen for both the designed filters. An audio file named male.wav is passed to the system and two single frequency tones are superimposed on it. These frequencies are 3500Hz in order to test the Low Pass filter and 1200Hz in order to test the IIR filter. The frequency domain analysis was done for the system.
% **************************************************** % Design & Analysis of Filters % ****************************************************

% ******************* % FIR Filter % ******************* Fs = 8000; f = [0.3*Fs 0.32*Fs]; a = [1 0]; dev = [0.01 0.01]; % % % % Sampling frequency Cutoff frequencies Desired amplitudes deviations in the pass & stop bands

%Parks-McClellan optimal FIR filter order estimation [n,fo,ao] = firpmord(f,a,dev,Fs) b = firpm(n,fo,ao); % Filter Coeeficients % Ploting the Low Pass FIR Filter figure; freqz(b,1,512,Fs); title('Lowpass Filter with Cutoff Frequency 2400Hz'); % Group Delay

figure; plot(grpdelay(b,a)); title('Group Delay of FIR Filter');xlabel('samples'); % Pole-Zero Plot figure;zplane(b,1);title('Pole-Zero Plot of FIR Filter');

% ******************* % IIR Filter % ******************* wo=0.15*Fs/(0.5*Fs); % Notch Frequency bw=wo/.4; [bi,ai] = iirnotch(wo,bw); % Ploting the Low Pass FIR Filter figure; freqz(bi,ai,512,Fs); title('Notch Filter with Notch Frequency 1200Hz'); % Group Delay figure;plot(grpdelay(bi,ai)); title('Group Delay of IIR Filter');xlabel('samples'); % Pole-Zero Plot figure;zplane(bi,ai);title('Pole-Zero Plot of IIR Filter');

% ******************************************************* % Creating Input Signals % ******************************************************* w = [1:(3*Fs)]; [input,fs]=wavread('male.wav'); % sound file passed n=[1:length(input)]; % Add sinosoids with frequency f1=1200Hz and f2=2800Hz input=input'+0.04*sin(2*pi*.35*n)+0.05*sin(2*pi*n*.1507); % input=awgn(input,25); fft_input = abs((fft(input))); % Plot Input figure; plot((w(1:1.5*Fs)*(Fs/2))/(1.5*Fs),fft_input(1:1.5*Fs)); axis([1 4000 1 100]); title('Input Signal'); xlabel('Frequency (Hz)'); ylabel('Magnitude') % Play Input input = input(1:(3*Fs)); sound(input,Fs);

% *********************************************************** % Filtering of the input % *********************************************************** input = input(1:(3*Fs)); % just pick 3 seconds of the signal

% ******************* % LPF Output % ******************* % Generate Output fir_output = filter(b,1,input); fft_fir_output = abs((fft(fir_output))); % Plot Output figure; plot((w(1:1.5*Fs)*(Fs/2))/(1.5*Fs),fft_fir_output(1:1.5*Fs)); axis([1 4000 1 100]); title('Output of FIR Filter (Low-Pass)'); xlabel('Frequency (Hz)'); ylabel('Magnitude') % Play Output pause(3) sound(fir_output,Fs);

% **************************** % Notch Filter Output % **************************** % Generate Output iir_output = filter(bi,ai,fir_output); fft_iir_output = abs((fft(iir_output))); % Plot Output figure; plot((w(1:1.5*Fs)*(Fs/2))/(1.5*Fs),fft_iir_output(1:1.5*Fs)); axis([1 4000 1 100]); title('Output of IIR Filter (Notch)'); xlabel('Frequency (Hz)'); ylabel('Magnitude') % Play Output pause(3) sound(iir_output,Fs);

C5416.m File This code was used to record the sound coming out of the C5416 DSK board. The frequency response of the unprocessed input signal, the FIR output signal and the final output after the IIR filter were plotted using this code.
% ******************************************** % C5416 Sound Output Analysis % ******************************************** Fs=8000; input_board = wavrecord(3*Fs,Fs); fir_output_board = wavrecord(3*Fs,Fs);

iir_output_board = wavrecord(3*Fs,Fs);

% ******************************************************** % Frequency Response Using a Simulated Signal % ******************************************************** % Generate FFt of input_board fft_input_board = abs((fft(input_board))); % Plot input_board figure; plot((w(1:1.5*Fs)*(Fs/2))/(1.5*Fs),fft_input_board(1:1.5*Fs),'r'); axis([1 4000 1 100]); title('input board Signal'); xlabel('Frequency (Hz)'); ylabel('Magnitude') % Play Input sound(input_board,Fs)

% ******************* % LPF Output % ******************* % Generate FFT of Output fft_fir_output_board = abs((fft(fir_output_board))); % Plot Output figure; plot((w(1:1.5*Fs)*(Fs/2))/(1.5*Fs),fft_fir_output_board(1:1.5*Fs),'r'); axis([1 4000 1 100]); title('Output of FIR Filter (Low-Pass)'); xlabel('Frequency (Hz)'); ylabel('Magnitude') % Play Output pause(3) sound(fir_output_board,Fs);

% **************************** % Notch Filter Output % **************************** % Generate FFT of Output fft_iir_output_board = abs((fft(iir_output_board))); % Plot Output figure; plot((w(1:1.5*Fs)*(Fs/2))/(1.5*Fs),fft_iir_output_board(1:1.5*Fs),'r'); axis([1 4000 1 100]); title('Output of IIR Filter (Notch)'); xlabel('Frequency (Hz)'); ylabel('Magnitude')

% Play Output pause(3) sound(iir_output_board,Fs);

C5416 DSK Program

Following is the C5416 program used to implement the required system on the DSk board. It takes the audio input through Line In port on the board. It stores the signal in the form of sample because of Codec that is set to sample at 8KHz, using the following value on the Codec Clock Register in the initPCM3002.asm file:
; Codec clock Register (0000 1010) ; ; ; ; ; ; ; ; VAL_CLK_REG Bit7-4 = 0000: Reserved Bit3 = 1: Clock divisor selected 0: No divisor, 48 KHz sampling rate Bit2 = 0: Clock enabled Bit1-0 = 00: Clock divisor for 24 KHz sampling rate 01: Clock divisor for 12 KHz sampling rate 10: Clock divisor for 8 KHz sampling rate 11: Clock divisor for 6 KHz sampling rate 12h

.set

The samples received and transmitted are 16-bits each. This is achieved by changing the values of the transmit and receive control registers (RCR1, RCR2, XCR1, XCR2) in the file initMcBSP.asm:
; Receive Control Register 1 (0000 0000 0100 0000) ; ; ; ; VAL_RCR1 Bit15 = 0: Reserved Bit14-8 = 0000000: 1 word per frame Bit7-5 = 010: 16 bit receive word Bit4-0 = 00000: Reserved 0040h

.set

; Receive Control Register 2 (0000 0000 0100 0001) ; ; ; ; ; ; VAL_RCR2 Bit15 = 0: Single phase frame Bit14-8 = 00h: 1 word per frame Bit7-5 = 010: 16 bit receive word Bit4-3 = 00: No companding Bit2 = 0: Receive frame sync pulses not ignored Bit1-0 = 01: 1-bit data delay 0041h

.set

; Transmit Control Register 1 (0000 0000 0100 0000) ; Bit15 = 0: Reserved

; Bit14-8 = 00h: 1 word per frame ; Bit7-5 = 010: 16 bit transmit word ; Bit4-0 = 0h: Reserved VAL_XCR1 .set 0040h

; Transmit Control Register 2 (0000 0000 0100 0000) ; ; ; ; ; ; VAL_XCR2 Bit15 = 0: Single phase frame Bit14-8 = 00h: 1 word per frame Bit7-5 = 010: 16 bit transmit word Bit4-3 = 00: No companding Bit2 = 0: Transmit frame sync pulses not ignored Bit1-0 = 00: 0-bit data delay 0040h

.set

The main file of the project is as shown below. It waits for the receive interrupt and when a sample is received it calls an FIR filter whose coefficients are already loaded in a circular buffer. After the FIR filter the sample is in Register A. It then calls the IIR filter and the sample is processed upon.
************************************************************************ * * This program reads an input signal from the ADC , Passes it through a Lowpass FIR *filter and then that output through a IIR Notch Filter. It then writes the sample to * the DAC on the DSK5416 board. * ************************************************************************ .include .ref .ref .ref .def .def .def .def VAL_SP OutSamples .set .bss .bss .bss .sect .include .set "regs.asm" initC5416 initMcBSP2 initPCM3002 _c_int00 brint2_isr sample_receive sample_transmit 0x0500 y,1,1 CoefBuf, 99, 1 SampleBuf, 99, 1 "FIRcoeff" "coff_fir.dat" 98 ; initial stack address ; Allocate space for y(n)s ; Memory for coeff circular buffer ; Memory for sample circular buffer ; Filter coeff (seq locations) ; Number of filter taps N - 1

FIRcoeff Nm1

; Intermediate variables (sequential locations) wn .word 0 wnm1 .word 0 wnm2 .word 0 ; IIR Filter coefficients (sequential locations) b0 .word 9598

;initial w(n) ;initial w(n-1) =0 ;initial w(n-2)=0 ;b0 = 0.2929

b1 b2 a1 a2

.word .word .word .word .text

-11282 9598 -11282 -13572

;b1 ;b2 ;a1 ;a2

= = = =

-0.3343 0.2929 -0.3343 -0.4412

* The entry-point for the program _c_int00: stm ssbx call call call stm orm stm orm stm orm STM ST CALL ssbx rsbx wait_main: idle b nop nop nop #VAL_SP, SP INTM initC5416 initMcBSP2 initPCM3002 #SPCR1, MCBSP2_SPSA #0001h, MCBSP2_SPSD #SPCR2, MCBSP2_SPSA #0001h, MCBSP2_SPSD #0FFFFh, IFR #040h, IMR #OutSamples, AR6 #0, *AR6 fir_init SXM INTM 1 wait_main ; Define the stack ; Disable all interrupts ; Init the DSP processor ; Init the McBSP2 port ; Init the DSK CODEC ; Enable McBSP2 receiver ; Enable McBSP2 transmitter ; Clear pending interrupts ; Unmask McBSP2 RX int ; Clear output sample buffer ; AR6 points to OutSample buffer ; Init for filter calculations ; Select sign extension mode ; Enable all interrupts ; Wait for an RX interrupt

* Interrupt service routine for McBSP2 Receiver brint2_isr: call call (integer) call STH call rete STH sample_receive fir_filter A,1,*AR6 iir_filter A,1,*AR6 sample_transmit ; Receive the sample ; Call FIR filter routine, ; Store filtered sample ; Call IIR filter routine ; Store filtered sample (integer) ; Transmit the sample

* This procedure receives a 16-bit value from the ADC * Return with A (LSBs) = 16 bit received sample sample_receive: ld ldm ret * This procedure sends a 16-bit value in A (LSBs) to the DAC. sample_transmit: stlm ret * * * * * * * ---------------------------------------------------------------------------------FIR Filter Initialization Routine This routine sets AR2 as the pointer for the sample circular buffer, and AR3 as the pointer for coefficient circular buffer. BK = Number of filter taps - 1. AR0 = 1 = circular buffer pointer increment ---------------------------------------------------------------------------------ST ST STM RPT MVPD RPT ST STM ret nop nop nop * * * * * * * -------------------------------------------------------------------------------FIR Filter Routine Enter with A = the current sample x(n) - an integer, AR2 pointing to the location for the current sample x(n), and AR3 pointing to the q15 coefficient h(N-1). Exit with A = y(n) as q15 number. -------------------------------------------------------------------------------#CoefBuf,AR3 #SampleBuf,AR2 #Nm1,BK #Nm1 #FIRcoeff, *AR3+% #Nm1 - 1 #0h,*AR2+% #1,AR0 ; AR3 is the CB Coeff Pointer ; AR2 is the CB sample pointer ; BK = number of filter taps ; Place coeff in circular buffer ; Clear circular sample buffer ; AR0 = 1 = CB pointer increment A, MCBSP2_DXR1 ; Transmit 16 bits #0, A MCBSP2_DRR1, A ; Retrieve 16 bits

fir_init:

fir_filter: STL RPTZ MAC ret nop nop nop A, *AR2+0% A, #Nm1 - 1 *AR3+0%,*AR2+0%,A ; Place x(n)in the sample buffer ; A = 0 ; A = filtered sum (q15)

;--------------------------------------------------------------------------------; IIR Filter Subroutine ; Enter with A = x(n) as q15 number ; Exit with A = y(n) as q15 number

; Uses AR2 and AR3 ;--------------------------------------------------------------------------------iir_filter: pshm pshm AR2 AR3

;w(n)=x(n)+ a1.w(n-1)+ a2.w(n-2) STM #a2,AR2 STM #wnm2, AR3 MAC *AR2-,*AR3-,A MAC *AR2-,*AR3-,A STH A,1,*AR3 ; ; ; ; ; ; ; AR2 points to a2 AR3 points to w(n-2) A = x(n)+ a2.w(n-2) AR2 points to a1 & AR3 to w(n-1) A = x(n)+ a1.w(n-1)+ a2.w(n-2) AR2 points to b2 & AR3 to w(n) Save w(n)

;y(n)=b0.w(n)+ b1.w(n-1)+ b2.w(n-2) LD #0,A STM #wnm2,AR3 MAC *AR2-,*AR3-,A DELAY *AR3 MAC *AR2-,*AR3-,A DELAY *AR3 MAC *AR2,*AR3,A popm popm RET .end AR3 AR2 ; Return ; A = 0 ; AR3 points to w(n-2) ; A = b2.w(n-2) ; AR2 points to b1 & AR3 to w(n-1) ; w(n-1) -> w(n-2) ; A = b1.w(n-1)+ b2.w(n-2) ; AR2 points to b0 & AR3 to w(n) ; w(n) -> w(n-1) ; A = b0.w(n)+ b1.w(n-1)+ b2.w(n-2)

Frequency Response of a Simulated Signal

This section shows the results of the project. The input signal, low pass filter output and notch filter output are all shown in the frequency domain. The plots in red color are the ones obtained by processing upon the signal obtained from the DSK board. MATLAB Results

Input Signal 1000 900 800 700 600 Magnitude 500 400 300 200 100 0

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Output of FIR Filter (Low-Pass) 600

500

400 Magnitude

300

200

100

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Output of IIR Filter (Notch) 40 35 30 25 Magnitude 20 15 10 5 0

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

C5416 DSK Board Output

input board Signal 60

50

40 Magnitude

30

20

10

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Output of FIR Filter (Low-Pass) 60

50

40 Magnitude

30

20

10

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Output of IIR Filter (Notch) 80 70 60 50 Magnitude 40 30 20 10 0

500

1000

1500 2000 2500 Frequency (Hz)

3000

3500

4000

Comments and Conclusions

The project was successfully implemented and desired results were achieved. The FIR filter was a high order filter as expected, but since DSP processor are fast in implementing filter operations, this was not an issue. More important thing was to get the desired frequency response. The output in MATLAB using AWGN in the signal was also seen. The results were similar so the plots are not included. The outputs of MATLAB simulation and actual implementation were quite similar. This is evident from the correlation of the two plots and from the audio heard during the demonstration. The difference in amplitudes of the magnitude responses can be because different strength tones were used during implementation as compared to simulation. The simulated signals were created in MATLAB while NCH Tone Generator was used to generate tones for the input to board. Moreover, the output signal of the board was acquired for analysis by bringing the speaker to the laptop mic. Hence, there could be loss of amplitude. Overall, the results were satisfactory and the objective of the project was achieved.