Tips and tricks fo fixed point operation

Tips and tricks for fixed point operations either for DSP processor or FPGA

Tips

Using Graphedit to get elementary streams

Open Graphedit
-File->Render Media File, Select the container format or ‘ctrl + R’
Select the file which is to be rendered

Remove all except the file for example here spin4.mpa by selecting each and press delete button

-graph->Insert Filters or ctrl + F

Select DirectShow Filters as shown in figure, expand it

Select Dump in DirectShow Filters, click on Insert Filter

Select the files to which audio and video should be dumped

Connect the output pin of video of spin4.mpa to VID_f15.m2v by left clicking a dragging the pointer to input of VID_f15.m2v

Click on the play button, you might have got the required files

Commonly used C code DSP applications

Finding Endianess:

enum byte_order { unknown, bigEndian, littleEndian };

static enum byte_order determine_endianess (void)

{

char s[sizeof (long) + 1];

union {

long longval;

char charval[sizeof (long)];

} probe;

probe.longval = 0x41424344L; /* ABCD in ASCII */

strncpy (s, probe.charval, sizeof (long));

s[sizeof (long)] = '\0';

/* fprintf( stderr, "byte order is %s\n", s ); */

if (strcmp (s, "ABCD") == 0)

return bigEndian;

else if (strcmp (s, "DCBA") == 0)

return littleEndian;

else

return unknown;

}

Endianess conversion: Big Endian to Little Endian

C code:

#define BE2LE_32(_x) \
((_x) <<>> 8 & 0x0000FF00 \
| (_x) >> 24 & 0x000000FF)

#define BE2LE_16(_x) ((_x) <<>> 8 & 0x00FF)

Saturation

#define SATURATE(x, max, min) \
( ((x) > (max)) ? (max) : ((x) < (min)) ? (min) : (x) ) #define TRUNCATE(x, Y, Z, DATA_SIZE) \ (((((x) >> ((Y##_##F) - (Z##_##F)))) << ((DATA_SIZE) - (Z##_##W) - (Z##_##F))) >> ((DATA_SIZE) - (Z##_##W) - (Z##_##F)))
/* <--Frac part trunc--->*/
/* <---------------------Truncating the MSB---------------------------------->*/
/* <----------------------------------Sign Extending to do the arithmetic operations in C------------------------------> */

#define RND_OFF(x, Y, Z) \
( ((x) + (1 << (((Y##_##F) - (Z##_##F)) - 1))) >> ((Y##_##F) - (Z##_##F)) )

USAGE:
definition Q1:
#define Q_FMT1_W 3 // whole part
#define Q_FMT1_F 14 // fractional part

definition Q2:
#define Q_FMT2_W 3 // whole part
#define Q_FMT2_F 7 // fractional part

Rounding off:
RND_OFF(val, Q_FMT1, Q_FMT2);

Saturate:
SATURATE(val, 255, -256);

Truncate:
TRUNCATE(val, Q_FMT1, Q_FMT2, 31);