<< Chapter < Page Chapter >> Page >

Does the output match your expectations based on the theory? Does this application illustrate any limitations of the FFT implementation?(Hint: note that most of the values in the FFT input are zero.) The previously-given C implementation uses a similar algorithm as theTI FFT; take a look at the C code for help. What are the limitation(s) of the FFT that show up in this application?

In lab4b.h M sets the number of autocorrelation points that are calculated. What is the maximum valueof M that allows the reference routines to run in real time? In determining this value you may find it useful to connect awave-function generator to the input and copy inputs into display_inputs . You may limit M to powers of 2 minus 1 .

Quiz information

From your prelab experiments, you should be able to describe the effect of windowing and zero-padding on FFT spectralanalysis. In your DSP system, experiment with different inputs, changing N and the type of window. Can you explain what happens as the input frequency is increased beyond the Nyquist rate? Does the X k 2 coincide with what you expect from Matlab? What is the relationship between the observed spectrum and the DTFT?What would happen if the FFT calculation takes longer than it takes to fill inputs with N samples? How long does it take to compute each FFT? What are the tradeoffs between writing code in C versus assembly?

Appendix a:

lab4main.c

1 /* v:/ece420/54x/dspclib/lab4main.c */ 2 /* dgs - 9/14/2001 */ 3 /* mdk - 2/10/2004 C FFT update */ 4 5 #include "v:/ece420/54x/dspclib/core.h" 6 7 /* comment the next line to use assembly fft */ 8 #define C_FFT 9 10 #ifdef C_FFT /* Use C FFT */ 11 12 #include "window.h" 13 #include "lab4.h" /* Number of C FFT points defined here */ 14 15 /* function defined in lab4fft.c */ 16 void fft(void); 17 18 /* FFT data buffers */ 19 int real[N]; /* Real part of data */ 20 int imag[N]; /* Imaginary part of data */ 21 22 #else /* Use assembly FFT */ 23 24 #define N 1024 /* Number of assembly FFT points */ 25 26 /* Function defined by c_fft_given.asm */ 27 void bit_rev_fft(void); 28 29 /* FFT data buffers (declared in c_fft_given.asm) */ 30 extern int bit_rev_data[N*2]; /* Data input for bit-reverse function */ 31 extern int fft_data[N*2]; /* In-place FFT & Output array */ 32 extern int window[N]; /* The Hamming window */ 33 34 #endif /* C_FFT */ 35 36 37 /* Our input/output buffers */ 38 int inputs[N]; 39 int outputs[N]; 40 41 volatile int input_full = 0; /* volatile means interrupt changes it */ 42 int count = 0; 43 44 45 interrupt void irq(void) 46 { 47 int *Xmitptr,*Rcvptr; /* pointers to Xmit & Rcv Bufs */ 48 int i; 49 50 static int in_irq = 0; /* Flag to prevent reentrance */ 51 52 /* Make sure we're not in the interrupt (should never happen) */ 53 if( in_irq ) 54 return; 55 56 /* Mark we're processing, and enable interrupts */ 57 in_irq = 1; 58 enable_irq(); 59 60 /* The following waitaudio call is guaranteed not to 61 actually wait; it will simply return the pointers. */ 62 WaitAudio(&Rcvptr,&Xmitptr); 63 64 /* input_full should never be true... */ 65 if( !input_full ) 66 { 67 for (i=0; i<BlockLen; i++) 68 { 69 /* Save input, and echo to channel 1 */ 70 inputs[count] = Xmitptr[6*i] = Rcvptr[4*i]; 71 72 /* Send FFT output to channel 2 */ 73 Xmitptr[6*i+1] = outputs[count]; 74 75 count++; 76 } 77 } 78 79 /* Have we collected enough data yet? */ 80 if( count >= N ) 81 input_full = 1; 82 83 /* We're not in the interrupt anymore... */ 84 disable_irq(); 85 in_irq = 0; 86 } 87 88 89 main() 90 { 91 /* Initialize IRQ stuff */ 92 count = 0; 93 input_full = 0; 94 SetAudioInterrupt(irq); /* Set up interrupts */ 95 96 while (1) 97 { 98 while( !input_full ); /* Wait for a data buffer to collect */ 99 100 /* From here until we clear input_full can only take * 101 * BlockLen sample times, so don't do too much here. */ 102 103 /* First, transfer inputs and outputs */ 104 105 #ifdef C_FFT /* Use C FFT */ 106 /* I n s e r t c o d e t o f i l l */ 107 /* C F F T b u f f e r s */ 108 109 #else /* Use assembly FFT */ 110 /* I n s e r t c o d e t o f i l l */ 111 /* a s s e m b l y F F T b u f f e r s */ 112 113 #endif /* C_FFT */ 114 115 /* Done with that... ready for new data collection */ 116 count = 0; /* Need to reset the count */ 117 input_full = 0; /* Mark we're ready to collect more data */ 118 119 /**********************************************************/ 120 /* Now that we've gotten the data moved, we can do the */ 121 /* more lengthy processing. */ 122 123 #ifdef C_FFT /* Use C FFT */ 124 125 /* Multiply the input signal by the Hamming window. */ 126 /* . . . i n s e r t C / a s m code . . . */ 127 128 /* Bit-reverse and compute FFT in C */ 129 fft(); 130 131 /* Now, take absolute value squared of FFT */ 132 /* . . . i n s e r t C / a s m code . . . */ 133 134 /* Last, set the DC coefficient to -1 for a trigger pulse */ 135 /* . . . i n s e r t C / a s m code . . . */ 136 137 /* done, wait for next time around! */ 138 139 140 #else /* Use assembly FFT */ 141 142 /* Multiply the input signal by the Hamming window. */ 143 /* . . . i n s e r t C / a s m code . . . */ 144 145 /* Bit-reverse and compute FFT in assembly */ 146 bit_rev_fft(); 147 148 /* Now, take absolute value squared of FFT */ 149 /* . . . i n s e r t C / a s m code . . . */ 150 151 /* Last, set the DC coefficient to -1 for a trigger pulse */ 152 /* . . . i n s e r t C / a s m code . . . */ 153 154 /* done, wait for next time around! */ 155 156 157 #endif /* C_FFT */ 158 159 } 160 }

Get Jobilize Job Search Mobile App in your pocket Now!

Get it on Google Play Download on the App Store Now




Source:  OpenStax, Digital signal processing laboratory (ece 420). OpenStax CNX. Sep 27, 2006 Download for free at http://cnx.org/content/col10236/1.14
Google Play and the Google Play logo are trademarks of Google Inc.

Notification Switch

Would you like to follow the 'Digital signal processing laboratory (ece 420)' conversation and receive update notifications?

Ask