Sprinter/mame
2
0
Fork 0
mirror of https://github.com/holub/mame synced 2025-05-30 17:41:47 +03:00
Code Issues Actions Packages Projects Releases Wiki Activity

Added proper interpolation sub-cycles ('PC' cycles and A/B subcycles of each) and updated comments

Browse Source
This commit is contained in:
Jonathan Gevaryahu 2008-05-20 01:53:25 +00:00
parent 3c6278aef8
commit ae6bf98511
1 changed files with 84 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

109
src/emu/sound/tms5220.c
View File

@ -10,6 +10,7 @@
Chirp/excitation table fixes by Lord Nightmare
Various fixes by Lord Nightmare
Modularization by Lord Nightmare
Sub-interpolation-cycle parameter updating added by Lord Nightmare
Much information regarding these lpc encoding comes from US patent 4,209,844
US patent 4,331,836 describes the complete 51xx chip
@ -93,9 +94,9 @@ struct tms5220
UINT16 previous_energy; /* needed for lattice filter to match patent */
UINT8 interp_count; /* number of interp periods (0-7) */
UINT8 sample_count; /* sample number within interp (0-24) */
UINT16 pitch_count;
UINT8 interp_count; /* number of samples within each sub-interpolation period, ranges from 0-24 */
UINT8 sample_count; /* number of samples within the ENTIRE interpolation period, ranges from 0-199 */
UINT16 pitch_count; /* pitch counter; provides chirp rom address */
INT32 u[11];
INT32 x[10];
@ -595,28 +596,83 @@ tryagain:
}
}
}
else if (tms->interp_count == 0)
else
{
/* Update values based on step values */
/*mame_printf_debug("\n");*/
interp_period = tms->sample_count / 25;
tms->current_energy += ((tms->target_energy - tms->current_energy) >> interp_coeff[interp_period]);
if (tms->old_pitch != 0)
tms->current_pitch += ((tms->target_pitch - tms->current_pitch) >> interp_coeff[interp_period]);
/*mame_printf_debug("*** Energy = %d\n",tms->current_energy);*/
for (i = 0; i < 10; i++)
{
tms->current_k[i] += ((tms->target_k[i] - tms->current_k[i]) >> interp_coeff[interp_period]);
}
switch(tms->interp_count)
{
/* PC=X X cycle, rendering change (change for next cycle which chip is actually doing) */
case 0: /* PC=0, A cycle, nothing happens (calc energy) */
break;
case 1: /* PC=0, B cycle, nothing happens (update energy) */
break;
case 2: /* PC=1, A cycle, update energy (calc pitch) */
tms->current_energy += ((tms->target_energy - tms->current_energy) >> interp_coeff[interp_period]);
break;
case 3: /* PC=1, B cycle, nothing happens (update pitch) */
break;
case 4: /* PC=2, A cycle, update pitch (calc K1) */
if (tms->old_pitch != 0)
tms->current_pitch += ((tms->target_pitch - tms->current_pitch) >> interp_coeff[interp_period]);
break;
case 5: /* PC=2, B cycle, nothing happens (update K1) */
break;
case 6: /* PC=3, A cycle, update K1 (calc K2) */
tms->current_k[0] += ((tms->target_k[0] - tms->current_k[0]) >> interp_coeff[interp_period]);
break;
case 7: /* PC=3, B cycle, nothing happens (update K2) */
break;
case 8: /* PC=4, A cycle, update K2 (calc K3) */
tms->current_k[1] += ((tms->target_k[1] - tms->current_k[1]) >> interp_coeff[interp_period]);
break;
case 9: /* PC=4, B cycle, nothing happens (update K3) */
break;
case 10: /* PC=5, A cycle, update K3 (calc K4) */
tms->current_k[2] += ((tms->target_k[2] - tms->current_k[2]) >> interp_coeff[interp_period]);
break;
case 11: /* PC=5, B cycle, nothing happens (update K4) */
break;
case 12: /* PC=6, A cycle, update K4 (calc K5) */
tms->current_k[3] += ((tms->target_k[3] - tms->current_k[3]) >> interp_coeff[interp_period]);
break;
case 13: /* PC=6, B cycle, nothing happens (update K5) */
break;
case 14: /* PC=7, A cycle, update K5 (calc K6) */
tms->current_k[4] += ((tms->target_k[4] - tms->current_k[4]) >> interp_coeff[interp_period]);
break;
case 15: /* PC=7, B cycle, nothing happens (update K6) */
break;
case 16: /* PC=8, A cycle, update K6 (calc K7) */
tms->current_k[5] += ((tms->target_k[5] - tms->current_k[5]) >> interp_coeff[interp_period]);
break;
case 17: /* PC=8, B cycle, nothing happens (update K7) */
break;
case 18: /* PC=9, A cycle, update K7 (calc K8) */
tms->current_k[6] += ((tms->target_k[6] - tms->current_k[6]) >> interp_coeff[interp_period]);
break;
case 19: /* PC=9, B cycle, nothing happens (update K8) */
break;
case 20: /* PC=10, A cycle, update K8 (calc K9) */
tms->current_k[7] += ((tms->target_k[7] - tms->current_k[7]) >> interp_coeff[interp_period]);
break;
case 21: /* PC=10, B cycle, nothing happens (update K9) */
break;
case 22: /* PC=11, A cycle, update K9 (calc K10) */
tms->current_k[8] += ((tms->target_k[8] - tms->current_k[8]) >> interp_coeff[interp_period]);
break;
case 23: /* PC=11, B cycle, nothing happens (update K10) */
break;
case 24: /* PC=12, A cycle, update K10 (do nothing) */
tms->current_k[9] += ((tms->target_k[9] - tms->current_k[9]) >> interp_coeff[interp_period]);
break;
}
}
if (tms->old_energy == 0)
{
/* generate silent samples here */
tms->excitation_data = 0x00; /* I'm not sure if this is actually RIGHT, the current_energy may be forced to zero when we've just passed a zero energy frame. However, this does work too. May be removed later. */
tms->excitation_data = 0x00; /* This is NOT correct, the current_energy is forced to zero when we just passed a zero energy frame because thats what the tables hold for that value.
However, this code does no harm. Will be removed later. */
}
else if (tms->old_pitch == 0)
{
@ -642,13 +698,16 @@ tryagain:
tms->excitation_data = chirptable[tms->pitch_count];
}
/* Update LFSR every clock, like patent shows */
bitout = ((tms->RNG >> 12) & 1) ^
((tms->RNG >> 10) & 1) ^
((tms->RNG >> 9) & 1) ^
((tms->RNG >> 0) & 1);
tms->RNG >>= 1;
tms->RNG |= bitout << 12;
/* Update LFSR *20* times every sample, like patent shows */
for (i=0; i<20; i++)
{
bitout = ((tms->RNG >> 12) & 1) ^
((tms->RNG >> 10) & 1) ^
((tms->RNG >> 9) & 1) ^
((tms->RNG >> 0) & 1);
tms->RNG >>= 1;
tms->RNG |= bitout << 12;
}
buffer[buf_count] = clip_and_wrap(lattice_filter(tms)); /* execute lattice filter and clipping/wrapping */