[Urgent] G.726 ADPCM implementation

[bythesea]

I'm trying to implement G.726 ADPCM using the source codes released by Sun Microsystems to the public domain. I subsequently tested my program using the test vectors released by ITU but there were many errors (ie discrepancies between the ADPCM codewords generated by my prog and the ITU-given codewords).

Can anyone who is experienced in the above implementation advise me on how can I go about debugging my program. I had spent 4 days trying to solve the problem but to no avail!!

ps. the codes released by Sun is for the the implementation for G.721 but it is similar to what i want (ie 32kbps) so i believe it's safe to use it directly. or am i wrong?

static short qtab_721[7] = {-124, 80, 178, 246, 300, 349, 400};

static short _dqlntab[16] =
{
-2048, 4, 135, 213, 273, 323, 373, 425,
425, 373, 323, 273, 213, 135, 4, -2048
};

static short _witab[16] =
{
-12, 18, 41, 64, 112, 198, 355, 1122,
1122, 355, 198, 112, 64, 41, 18, -12
};

static short _fitab[16] =
{
0, 0, 0, 0x200, 0x200, 0x200, 0x600, 0xE00,
0xE00, 0x600, 0x200, 0x200, 0x200, 0, 0, 0
};

// static short _fitab[8] = {0, 0, 0, 1, 1, 1, 3, 7};

/*
* Encodes the input value of linear PCM, A-law or u-law data sl and returns
* the resulting code. -1 is returned for unknown input coding value.
*/
int ADPCM_encoder(int sl, int in_coding, struct g72x_state *state_ptr)
{
short sezi, se, sez; // se is the estimated signal

short d; // SUBTA => Compute diff signal by subtracting signal estimate
// from input signal (or quantized reconstructed signal in decoder)
short sr;// ADDB => Addition of quantized diff signal and signal estimate
// to form reconstructed signal
short y; // MIX => Form linear combination of fast and slow
// quantizer scale factors
short dqsez; // ADDC => Obtain sign of addition of quantized diff signal
// and partial signal estimate
short dq, i;

static int count = 0;
FILE *debugPtr;
debugPtr = fopen("sad.txt", "a");

switch (in_coding)
{ // linearize input sample to 14-bit PCM
case AUDIO_ENCODING_ULAW:
sl = ulaw2linear((unsigned char)sl) >> 2; // sl is complement of orig codeword?
break;
case AUDIO_ENCODING_ALAW:
sl = alaw2linear((unsigned char)sl) >> 2;
break;
case AUDIO_ENCODING_LINEAR:
sl >>= 2; // 14-bit dynamic range
break;
default:
return (-1);
}
// **********************
// * ADAPTIVE PREDICTOR *
// **********************
sezi = predictor_zero(state_ptr); // computes estimated signal
// from 6-zero predictor

sez = sezi >> 1; // right-shift by 1-bit
se = (sezi + predictor_pole(state_ptr)) >> 1;// computes estimated
// signal from 2-pole predictor
// *********************************
// * DIFFERENCE SIGNAL COMPUTATION *
// *********************************
d = sl - se; // difference between actual and estimated signal

// **********************
// * ADAPTIVE QUANTIZER *
// **********************
/* quantize the prediction difference */
y = step_size(state_ptr); // computes quantization step size
i = quantize(d, y, qtab_721, 7); // quantizes the estimated difference
// and returns the ADPCM code
// ******************************
// * INVERSE ADAPTIVE QUANTIZER *
// ******************************
dq = inv_adapt_quan(i & 8, _dqlntab, y); // quantized est diff

// ***********************************
// * RECONSTRUCTED SIGNAL CALCULATOR *
// ***********************************
sr = (dq < 0) ? se - (dq & 0x3FFF) : se + dq;

dqsez = sr + sez - se; /* pole prediction diff. */

fprintf(debugPtr, "%6d %6d %6d %6d %6d %6d\n", sl, se, d, y, i, count++);

update(4, y, _witab << 5, _fitab, dq, sr, dqsez, state_ptr);

fclose(debugPtr);

return (i); // ADPCM codeword
}