1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
//! Game Boy DMG APU (Audio Processing Unit).
//!
//! Implements all four DMG sound channels, the Frame Sequencer, master volume
//! and panning (NR50/NR51/NR52), wave RAM, and a mono sample output pipeline
//! compatible with the existing `sample_ready()` / `get_sample()` interface.
//!
//! ## Register map
//!
//! | Address | Register | Description |
//! |---|---|---|
//! | $FF10–$FF14 | NR10–NR14 | CH1 Pulse + sweep |
//! | $FF16–$FF19 | NR21–NR24 | CH2 Pulse |
//! | $FF1A–$FF1E | NR30–NR34 | CH3 Wave |
//! | $FF20–$FF23 | NR41–NR44 | CH4 Noise |
//! | $FF24 | NR50 | Master volume / VIN routing |
//! | $FF25 | NR51 | Channel terminal panning |
//! | $FF26 | NR52 | APU power / channel status |
//! | $FF30–$FF3F | — | Wave RAM (32 × 4-bit samples) |
use crate::trace_apu;
use serde::{Deserialize, Serialize};
use std::collections::VecDeque;
use crate::gb::model::CgbModel;
use super::channel1::Channel1;
use super::channel2::Channel2;
use super::channel3::Channel3;
use super::channel4::Channel4;
/// DMG clock rate in M-cycles per second (4 194 304 Hz / 4).
const DMG_MCYCLES_PER_SEC: f32 = 1_048_576.0;
/// Upper bound for queued audio samples awaiting frontend retrieval.
const MAX_PENDING_SAMPLES: usize = 16_384;
/// Cutoff frequency for the GB APU AC-coupling high-pass filter (~7 Hz removes DC bias).
const HP_CUTOFF_HZ: f32 = 7.0;
// Note: The Frame Sequencer is clocked by DIV-APU falling edges (512 Hz = every 2048 M-cycles)
// from the Timer, not by an internal countdown timer. See `clock_div_apu()`.
/// Frame Sequencer 8-step table.
///
/// Each entry is a bitmask:
/// - bit 0 = clock length counters
/// - bit 1 = clock frequency sweep (CH1 only)
/// - bit 2 = clock volume envelopes
#[rustfmt::skip]
const FS_TABLE: [u8; 8] = [
0b001, // step 0: length
0b000, // step 1: —
0b011, // step 2: length + sweep
0b000, // step 3: —
0b001, // step 4: length
0b000, // step 5: —
0b011, // step 6: length + sweep
0b100, // step 7: envelope
];
/// 8-step duty-cycle waveforms shared by CH1 and CH2 (bit 7 = first step output).
pub(super) const DUTY_TABLE: [[u8; 8]; 4] = [
[0, 0, 0, 0, 0, 0, 0, 1], // 12.5%
[1, 0, 0, 0, 0, 0, 0, 1], // 25%
[1, 0, 0, 0, 0, 1, 1, 1], // 50%
[0, 1, 1, 1, 1, 1, 1, 0], // 75%
];
/// Game Boy DMG APU.
#[derive(Debug, Clone, Serialize, Deserialize)]
pub struct Apu {
ch1: Channel1,
ch2: Channel2,
ch3: Channel3,
ch4: Channel4,
/// NR50 ($FF24): master volume and VIN routing.
nr50: u8,
/// NR51 ($FF25): channel terminal panning.
nr51: u8,
/// APU power flag (NR52 bit 7). When `false` most registers are frozen.
powered: bool,
/// Current Frame Sequencer step (0–7).
/// Clocked by DIV-APU falling edges via `clock_div_apu()`.
fs_step: u8,
/// Low-frequency divider flag. Toggled each M-cycle to track 2 MHz alignment.
/// Used to calculate the startup delay for pulse channels on trigger.
#[serde(default)]
lf_div: bool,
/// FS-step counter (`div_divider`) advanced once per DIV-APU event (i.e.,
/// per Frame Sequencer step at 512 Hz). Sub-M-cycle APU machinery gates
/// on its low bits:
/// * `& 1 == 1` — length / NR14 retrigger "extra length tick" glitch
/// * `& 3 == 3` — sweep tick (drives `Channel1::sweep_countdown`)
/// * `& 7 == 7` — envelope volume countdown decrement
#[serde(default)]
div_divider: u8,
/// When `true`, the next DIV-APU falling edge is skipped.
/// Set when the APU is powered on while the DIV-APU bit is already HIGH.
/// Per SameSuite div_write_trigger_10: "Starting the APU while bit 4 of
/// the DIV register is set causes the APU to skip the first DIV-APU event."
#[serde(default)]
skip_next_div_apu_event: bool,
/// When `true`, the next DIV-APU event is serviced without advancing
/// `div_divider`. The
/// first event after a power-on-high skip still runs APU edge side effects,
/// but keeps the seeded divider phase for length/envelope/sweep timing.
#[serde(default)]
div_apu_event_without_div_increment: bool,
/// Fractional M-cycle accumulator for sample generation.
sample_acc: f32,
/// Number of M-cycles between output samples.
cycles_per_sample: f32,
/// Pending output samples queued until the frontend consumes them.
#[serde(default)]
pending_samples: VecDeque<f32>,
/// `true` when running a CGB-compatible ROM (header byte 0x0143 is 0x80 or 0xC0).
/// Gates CGB-specific APU differences (length counter behavior on power off/on).
is_cgb: bool,
/// CGB hardware revision for model-specific APU quirks.
#[serde(default)]
cgb_model: CgbModel,
/// High-pass filter state: previous mix input (before filter).
#[serde(default)]
hp_prev_in: f32,
/// High-pass filter state: previous filter output.
#[serde(default)]
hp_prev_out: f32,
/// High-pass filter coefficient (α = f_s / (f_s + 2π·f_c)), derived from
/// the output sample rate and a fixed ~7 Hz cutoff. Updated whenever
/// `set_sample_rate()` is called. Not serialized — recomputed on load.
#[serde(skip, default = "Apu::default_hp_rc")]
hp_rc: f32,
}
impl Apu {
/// Create a new APU in the power-off state.
pub fn new(is_cgb: bool) -> Self {
let mut ch1 = Channel1::new();
// Default CGB revision = CgbE (latest); CgbBus overrides via
// `set_cgb_model` once it knows the configured revision.
ch1.set_model(is_cgb, CgbModel::CgbE);
Self {
ch1,
ch2: Channel2::new(),
ch3: Channel3::new_with_mode(is_cgb),
ch4: Channel4::new(),
nr50: 0x00,
nr51: 0x00,
powered: false,
fs_step: 0,
lf_div: true, // Initialize to 1
div_divider: 0,
skip_next_div_apu_event: false,
div_apu_event_without_div_increment: false,
sample_acc: 0.0,
cycles_per_sample: DMG_MCYCLES_PER_SEC / 44_100.0,
pending_samples: VecDeque::with_capacity(MAX_PENDING_SAMPLES),
is_cgb,
cgb_model: CgbModel::CgbE,
hp_prev_in: 0.0,
hp_prev_out: 0.0,
hp_rc: Self::compute_hp_rc(44_100.0),
}
}
/// Compute the high-pass filter coefficient for a given output sample rate.
///
/// Uses a target cutoff of ~7 Hz (well below audible range) to remove DC
/// bias while leaving all audible content intact.
/// Formula: α = f_s / (f_s + 2π·f_c)
fn compute_hp_rc(sample_rate: f32) -> f32 {
sample_rate / (sample_rate + 2.0 * std::f32::consts::PI * HP_CUTOFF_HZ)
}
/// Default HP filter coefficient (44.1 kHz baseline), used by serde when
/// deserializing old save states that predate this field.
fn default_hp_rc() -> f32 {
Self::compute_hp_rc(44_100.0)
}
/// Set the CGB hardware revision and propagate it to subchannels that
/// consume it (currently CH1 sweep timing). Should be called once at
/// bus construction; default model in `Apu::new` is correct for DMG.
pub fn set_cgb_model(&mut self, cgb_model: CgbModel) {
self.cgb_model = cgb_model;
self.ch1.set_model(self.is_cgb, cgb_model);
}
/// CGB-0/A/B ignore the current NRx4 length-enable bit for the extra length
/// clock; only the previously disabled state matters.
fn cgb_early_extra_length_clock(&self) -> bool {
self.is_cgb
&& matches!(
self.cgb_model,
CgbModel::Cgb0 | CgbModel::CgbA | CgbModel::CgbB
)
}
/// CGB-B CH3 keeps its active flag set until the next non-trigger NR34
/// write when this quirk clocks length from 1 to 0 with length-enable clear.
fn cgb_b_delayed_ch3_length_disable(&self) -> bool {
self.is_cgb && self.cgb_model == CgbModel::CgbB
}
/// Set the output sample rate in Hz (default: 44 100).
pub fn set_sample_rate(&mut self, rate: f32) {
self.cycles_per_sample = DMG_MCYCLES_PER_SEC / rate;
self.hp_rc = Self::compute_hp_rc(rate);
self.pending_samples.clear();
}
/// Returns the current output sample rate in Hz.
pub fn sample_rate(&self) -> f32 {
DMG_MCYCLES_PER_SEC / self.cycles_per_sample
}
/// Returns `true` when an audio sample is ready to be collected.
pub fn sample_ready(&self) -> bool {
!self.pending_samples.is_empty()
}
/// Consume and return the pending audio sample, or `None` if not ready.
pub fn take_sample(&mut self) -> Option<f32> {
self.pending_samples.pop_front()
}
fn push_pending_sample(&mut self, sample: f32) {
if self.pending_samples.len() == MAX_PENDING_SAMPLES {
self.pending_samples.pop_front();
}
self.pending_samples.push_back(sample);
}
/// Advance the APU by `m_cycles` M-cycles.
///
/// Clocks all active channels and accumulates the fractional sample counter.
/// The Frame Sequencer is NOT clocked here — it is driven by DIV-APU
/// falling edges via `clock_div_apu()`.
pub fn tick(&mut self, m_cycles: u8) {
for _ in 0..m_cycles {
self.tick_one();
}
}
/// Advance CH3 by exactly one APU cycle (used in CGB double-speed mode).
///
/// In double-speed mode CH3 is ticked once per CPU M-cycle (1 APU cycle)
/// rather than twice every two M-cycles (2 APU cycles), giving the
/// 1-cycle granularity required for accurate wave-position timing.
pub(crate) fn tick_ch3_one_apu_cycle(&mut self) {
if self.powered {
self.ch3.tick_apu_cycles(1);
}
}
/// Advance all channels **except CH3** by `m_cycles` M-cycles.
///
/// Used in CGB double-speed mode where CH3 is ticked separately via
/// `tick_ch3_one_apu_cycle` for finer timing granularity.
pub(crate) fn tick_except_ch3(&mut self, m_cycles: u8) {
for _ in 0..m_cycles {
self.tick_one_except_ch3();
}
}
/// Advance CH1, CH2, CH4, sweep, and the sample mixer by one M-cycle.
/// CH3 is intentionally excluded — see `tick_except_ch3`.
fn tick_one_except_ch3(&mut self) {
self.lf_div = !self.lf_div;
self.ch1.tick();
self.ch2.tick();
self.ch4.tick();
self.ch1.sweep_tick();
self.sample_acc += 1.0;
if self.sample_acc >= self.cycles_per_sample {
self.sample_acc -= self.cycles_per_sample;
let sample = self.mix();
self.push_pending_sample(sample);
}
}
/// Clock the APU envelope units due to a DIV-APU **rising** edge.
///
/// The GB APU envelope has a two-phase mechanism driven by the DIV-APU bit:
/// - **Rising edge** (this function): if a channel's envelope countdown has
/// reached zero, arm the clock flag and reload the countdown so that the
/// volume tick fires at the very next falling edge. This is the "phantom
/// tick" mechanism needed for zombie-mode / NRx2-glitch accuracy.
/// - **Falling edge** (`clock_div_apu`): fire the volume tick for any channel
/// whose clock flag was armed by the most recent secondary event.
pub fn clock_div_apu_secondary(&mut self) {
if !self.powered {
return;
}
self.ch1.clock_envelope_secondary();
self.ch2.clock_envelope_secondary();
self.ch4.clock_envelope_secondary();
}
/// Clock the APU frame sequencer due to a DIV-APU falling edge.
///
/// The GB APU frame sequencer is clocked by the falling edge of DIV bit 4
/// (bit 12 of the 16-bit internal counter, or bit 13 on CGB double-speed).
/// This occurs:
/// - Naturally every 8192 T-cycles (2048 M-cycles) as the counter wraps
/// - Artificially when DIV is written while the DIV-APU bit was HIGH
///
/// Each call steps the frame sequencer by one step (0-7), clocking the
/// appropriate length counters, sweep, and/or envelope units.
///
/// If the skip_next_div_apu_event flag is set (APU was powered on while
/// the DIV-APU bit was HIGH), this event is skipped and the flag is cleared.
pub fn clock_div_apu(&mut self) {
// Do nothing if APU is powered off.
if !self.powered {
return;
}
// Skip the first event if APU was powered on while DIV-APU bit was high.
if self.skip_next_div_apu_event {
self.skip_next_div_apu_event = false;
self.div_apu_event_without_div_increment = true;
trace_apu!(3; "GB APU skipping DIV-APU event (power-on with DIV-APU bit high)");
return;
}
let increment_divider = !self.div_apu_event_without_div_increment;
if increment_divider {
// Increment div_divider at the start of serviced DIV-APU events,
// except for the first post-skip serviced event, which preserves the
// seeded power-on-high phase. Low bits drive length/sweep/envelope
// sub-events.
self.div_divider = self.div_divider.wrapping_add(1);
} else {
self.div_apu_event_without_div_increment = false;
}
trace_apu!(3; "GB APU FS step={} length={} sweep={} envelope={}",
self.fs_step,
(FS_TABLE[self.fs_step as usize] & 0x01) != 0,
(FS_TABLE[self.fs_step as usize] & 0x02) != 0,
(FS_TABLE[self.fs_step as usize] & 0x04) != 0);
let flags = FS_TABLE[self.fs_step as usize];
// Fire envelope tick for any channel whose clock flag was armed by the
// most recent secondary (rising-edge) event. Must run at *every* FS step
// so that phantom ticks land on the correct falling edge, not just step 7.
self.ch1.clock_envelope_primary();
self.ch2.clock_envelope_primary();
self.ch4.clock_envelope_primary();
if flags & 0x01 != 0 {
trace_apu!(4; "GB APU clock length counters");
self.ch1.clock_length();
self.ch2.clock_length();
self.ch3.clock_length();
self.ch4.clock_length();
}
if flags & 0x02 != 0 {
trace_apu!(4; "GB APU clock CH1 sweep");
self.ch1.clock_sweep(self.lf_div);
}
// Decrement envelope countdowns every 8th FS event (64 Hz).
// Uses div_divider bit pattern: fires when
// div_divider & 7 == 7, i.e. on the 7th, 15th, 23rd … falling edges.
if self.div_divider & 7 == 7 {
trace_apu!(4; "GB APU decrement envelope countdowns");
self.ch1.clock_envelope_decrement();
self.ch2.clock_envelope_decrement();
self.ch4.clock_envelope_decrement();
}
self.fs_step = (self.fs_step + 1) & 7;
}
/// Advance by exactly one M-cycle.
///
/// Clocks channel frequency timers and generates samples.
/// Frame Sequencer is NOT clocked here — see `clock_div_apu()`.
fn tick_one(&mut self) {
// Toggle the low-frequency divider (tracks 2 MHz alignment).
self.lf_div = !self.lf_div;
// ── Channel frequency timers ───────────────────────────────────────
self.ch1.tick();
self.ch2.tick();
self.ch3.tick();
self.ch4.tick();
// ── CH1 sweep machinery (1 MHz cadence; gated internally) ─────────
// Hardware behavior `GB_apu_run` advances sweep state at half the
// M-cycle rate; `Channel1::sweep_tick` toggles its own phase flag and
// performs work every other call.
self.ch1.sweep_tick();
// ── Sample output ──────────────────────────────────────────────────
self.sample_acc += 1.0;
if self.sample_acc >= self.cycles_per_sample {
self.sample_acc -= self.cycles_per_sample;
let sample = self.mix();
self.push_pending_sample(sample);
}
}
/// Mix all four channels through NR50/NR51 into a mono f32 in `[-1.0, 1.0]`.
///
/// Applies a first-order high-pass filter (simulating the DMG output
/// capacitor / AC coupling) to remove DC bias and centre the signal
/// around 0. The filter coefficient α = f_s / (f_s + 2π·f_c) gives a
/// cutoff of roughly 7 Hz at any configured sample rate, removing DC
/// while leaving all audible content unaffected.
fn mix(&mut self) -> f32 {
if !self.powered {
return 0.0;
}
let samples = [
self.ch1.output(),
self.ch2.output(),
self.ch3.output(),
self.ch4.output(),
];
// NR51 bits 7-4 = left enable (CH4/3/2/1), bits 3-0 = right enable.
let left_mix = Self::mix_terminal(samples, self.nr51 >> 4);
let right_mix = Self::mix_terminal(samples, self.nr51 & 0x0F);
// NR50 bits 6-4 = left master volume (0-7), bits 2-0 = right (0-7).
let left_volume = ((self.nr50 >> 4) & 0x07) as f32 / 7.0;
let right_volume = (self.nr50 & 0x07) as f32 / 7.0;
let raw = (left_mix * left_volume + right_mix * right_volume) / 2.0;
// High-pass filter: out[n] = α × (out[n-1] + in[n] − in[n-1]).
// α = f_s / (f_s + 2π·f_c), with f_c ≈ 7 Hz, updated by set_sample_rate().
// Removes DC bias and produces bipolar output centred around 0.
let hp_out = self.hp_rc * (self.hp_prev_out + raw - self.hp_prev_in);
self.hp_prev_in = raw;
self.hp_prev_out = hp_out;
let final_out = hp_out.clamp(-1.0, 1.0);
trace_apu!(5; "GB APU mix ch1={:.3} ch2={:.3} ch3={:.3} ch4={:.3} left={:.3} right={:.3} raw={:.3} hp={:.3} out={:.3}",
samples[0], samples[1], samples[2], samples[3],
left_mix * left_volume, right_mix * right_volume,
raw, hp_out, final_out);
final_out
}
/// Sum channel samples gated by a 4-bit enable mask (bit i enables samples[i]).
fn mix_terminal(samples: [f32; 4], enable_mask: u8) -> f32 {
samples
.iter()
.enumerate()
.map(|(i, &s)| if enable_mask & (1 << i) != 0 { s } else { 0.0 })
.sum::<f32>()
/ 4.0
}
// ── Register read ──────────────────────────────────────────────────────
/// Read an APU register.
///
/// Returns $FF for write-only bits / unused registers.
/// When powered off, registers NR10–NR51 return their cleared values
/// (with unused bits set per the normal masking), consistent with DMG hardware.
/// NR52 and wave RAM are always readable.
pub fn read_register(&self, addr: u16) -> u8 {
match addr {
// CH1
0xFF10 => self.ch1.read_nr10(),
0xFF11 => self.ch1.read_nr11(),
0xFF12 => self.ch1.read_nr12(),
0xFF13 => 0xFF, // write-only
0xFF14 => self.ch1.read_nr14(),
// CH2
0xFF15 => 0xFF, // unused
0xFF16 => self.ch2.read_nr21(),
0xFF17 => self.ch2.read_nr22(),
0xFF18 => 0xFF, // write-only
0xFF19 => self.ch2.read_nr24(),
// CH3
0xFF1A => self.ch3.read_nr30(),
0xFF1B => 0xFF, // write-only
0xFF1C => self.ch3.read_nr32(),
0xFF1D => 0xFF, // write-only
0xFF1E => self.ch3.read_nr34(),
// CH4
0xFF1F => 0xFF, // unused
0xFF20 => 0xFF, // write-only
0xFF21 => self.ch4.read_nr42(),
0xFF22 => self.ch4.read_nr43(),
0xFF23 => self.ch4.read_nr44(),
// Master
0xFF24 => self.nr50,
0xFF25 => self.nr51,
0xFF26 => self.read_nr52(),
// Wave RAM
0xFF30..=0xFF3F => self.ch3.read_wave_ram(addr),
_ => 0xFF,
}
}
fn read_nr52(&self) -> u8 {
// Bits 6-4 are unused and read as 1 on DMG hardware.
let mut nr52 = 0x70;
nr52 |= u8::from(self.powered) << 7;
nr52 |= u8::from(self.ch1.is_active_for_nr52());
nr52 |= u8::from(self.ch2.is_active()) << 1;
nr52 |= u8::from(self.ch3.is_active()) << 2;
nr52 |= u8::from(self.ch4.is_active()) << 3;
nr52
}
/// Read PCM12 ($FF76, CGB only): digital outputs 1 & 2 [read-only].
///
/// Low nibble = CH1 digital output (0-15), high nibble = CH2 (0-15).
pub fn read_pcm12(&self) -> u8 {
let ch1 = self.ch1.digital_output();
let ch2 = self.ch2.digital_output();
ch1 | (ch2 << 4)
}
/// Read PCM34 ($FF77, CGB only): digital outputs 3 & 4 [read-only].
///
/// Low nibble = CH3 digital output (0-15), high nibble = CH4 (0-15).
pub fn read_pcm34(&self) -> u8 {
let ch3 = self.ch3.digital_output();
let ch4 = self.ch4.digital_output();
ch3 | (ch4 << 4)
}
// ── Register write ─────────────────────────────────────────────────────
/// Write NR52 ($FF26) with knowledge of the current DIV-APU bit state.
///
/// When the APU is powered on while the DIV-APU bit (bit 12 or bit 13 in
/// double-speed) is HIGH, the first DIV-APU event should be skipped.
/// The bus should call this method for NR52 writes instead of write_register.
///
/// `div_apu_bit_high` should be true if the DIV-APU bit is currently HIGH.
pub fn write_nr52_with_div_state(&mut self, val: u8, div_apu_bit_high: bool) {
let power_on = self.write_nr52(val);
if power_on && div_apu_bit_high {
self.arm_skip_next_div_apu_event();
}
}
/// Write an APU register.
///
/// When powered off, writes to NR10–NR51 are ignored (length counters are
/// writable on DMG even when powered off, but we keep it simple here).
/// Writes to NR52 are always honoured to allow power-on.
///
/// Note: For NR52 ($FF26) writes, prefer `write_nr52_with_div_state()` to
/// correctly handle the DIV-APU skip behavior on power-on.
pub fn write_register(&mut self, addr: u16, val: u8) {
self.write_register_with_apu_phase(addr, val, None);
}
/// Write an APU register, optionally supplying bit-packed CGB double-speed
/// phase bits for trigger startup timing.
pub fn write_register_with_apu_phase(
&mut self,
addr: u16,
val: u8,
double_speed_phase_bits: Option<u8>,
) {
self.write_register_with_apu_phase_and_div_counter(
addr,
val,
double_speed_phase_bits,
None,
);
}
pub fn write_register_with_apu_phase_and_div_counter(
&mut self,
addr: u16,
val: u8,
double_speed_phase_bits: Option<u8>,
div_counter: Option<u16>,
) {
// NR52 is always writable.
if addr == 0xFF26 {
self.write_nr52(val);
return;
}
// Wave RAM is accessible regardless of APU power state.
if (0xFF30..=0xFF3F).contains(&addr) {
self.ch3.write_wave_ram(addr, val);
return;
}
if !self.powered {
// On DMG, length counters remain writable when APU is off.
// On CGB, all writes (including length) are rejected when off.
if !self.is_cgb {
match addr {
0xFF11 => self.ch1.write_nr11_length_only(val),
0xFF16 => self.ch2.write_nr21_length_only(val),
0xFF1B => self.ch3.write_nr31_length_only(val),
0xFF20 => self.ch4.write_nr41_length_only(val),
_ => {}
}
}
return;
}
// Extra length clocking is keyed directly off the DIV-APU divider LSB,
// not `fs_step`: the power-on-high skip path seeds `div_divider` before
// the first serviced FS step so the trigger-time length glitch observes
// the same phase as hardware.
let extra_clk = self.div_divider & 1 == 1;
let cgb_early_extra_length_clock = self.cgb_early_extra_length_clock();
let cgb_b_delayed_ch3_length_disable = self.cgb_b_delayed_ch3_length_disable();
match addr {
0xFF10 => self.ch1.write_nr10(val, self.lf_div),
0xFF11 => self.ch1.write_nr11(val),
0xFF12 => self.ch1.write_nr12(val),
0xFF13 => self
.ch1
.write_nr13_with_apu_phase(val, double_speed_phase_bits),
0xFF14 => self.ch1.write_nr14_with_apu_phase_and_length_quirk(
val,
extra_clk,
self.lf_div,
double_speed_phase_bits,
cgb_early_extra_length_clock,
div_counter,
),
0xFF15 => {}
0xFF16 => self.ch2.write_nr21(val),
0xFF17 => self.ch2.write_nr22(val),
0xFF18 => self.ch2.write_nr23(val),
0xFF19 => self.ch2.write_nr24_with_apu_phase_and_length_quirk(
val,
extra_clk,
self.lf_div,
double_speed_phase_bits,
cgb_early_extra_length_clock,
),
0xFF1A => self.ch3.write_nr30(val),
0xFF1B => self.ch3.write_nr31(val),
0xFF1C => self.ch3.write_nr32(val),
0xFF1D => self.ch3.write_nr33(val),
0xFF1E => self.ch3.write_nr34_with_length_quirk(
val,
extra_clk,
cgb_early_extra_length_clock,
cgb_b_delayed_ch3_length_disable,
),
0xFF1F => {}
0xFF20 => self.ch4.write_nr41(val),
0xFF21 => self.ch4.write_nr42(val),
0xFF22 => self.ch4.write_nr43(val),
0xFF23 => self.ch4.write_nr44_with_apu_phase_and_length_quirk(
val,
extra_clk,
double_speed_phase_bits,
cgb_early_extra_length_clock,
),
0xFF24 => {
trace_apu!(2; "GB APU write NR50=0x{:02X} left_vol={} right_vol={}", val, (val >> 4) & 0x07, val & 0x07);
self.nr50 = val;
}
0xFF25 => {
trace_apu!(2; "GB APU write NR51=0x{:02X} left_en=0x{:X} right_en=0x{:X}", val, val >> 4, val & 0x0F);
self.nr51 = val;
}
_ => {}
}
}
fn write_nr52(&mut self, val: u8) -> bool {
let was_powered = self.powered;
self.powered = val & 0x80 != 0;
let power_on_transition = !was_powered && self.powered;
if was_powered && !self.powered {
trace_apu!(1; "GB APU power off");
// Power off: clear all NR10–NR51 registers and HP filter state.
self.ch1.power_off();
self.ch2.power_off();
self.ch3.power_off();
self.ch4.power_off();
self.nr50 = 0x00;
self.nr51 = 0x00;
self.hp_prev_in = 0.0;
self.hp_prev_out = 0.0;
// Clear skip flag on power-off to prevent leaking into a later power-on.
self.skip_next_div_apu_event = false;
self.div_apu_event_without_div_increment = false;
// Reset Sub-M-cycle FS-step counter alongside `fs_step` so
// that sub-events keyed off `div_divider` low bits stay aligned
// across power cycles.
self.div_divider = 0;
} else if power_on_transition {
trace_apu!(1; "GB APU power on");
// Power on: reset frame sequencer step and skip flag.
// The frame sequencer is clocked by DIV-APU events (not an internal timer),
// so we only reset the step counter here. The skip flag is cleared first,
// then write_nr52_with_div_state() will re-arm it if DIV-APU bit is HIGH.
self.fs_step = 0;
self.div_divider = 0;
self.skip_next_div_apu_event = false;
self.div_apu_event_without_div_increment = false;
// Reset CH4 counter model state (alignment, counter) on APU init,
self.ch4.apu_init();
// On CGB, powering on resets all length counters to 0.
if self.is_cgb {
self.ch1.length_counter = 0;
self.ch2.length_counter = 0;
self.ch3.length_counter = 0;
self.ch4.length_counter = 0;
}
}
power_on_transition
}
/// Arm the skip flag for the next DIV-APU event.
///
/// Call this when the APU is powered on while the DIV-APU bit (bit 12 or
/// bit 13 in double-speed) is currently HIGH. Per SameSuite div_write_trigger_10:
/// "Starting the APU while bit 4 of the DIV register is set causes the APU
/// to skip the first DIV-APU event."
pub fn arm_skip_next_div_apu_event(&mut self) {
// Initializes this skipped-start phase to 1 so trigger-time
// length glitches see `div_divider & 1 == 1`, even before the first
// non-skipped DIV-APU event advances the divider. `extra_clk` uses this
// LSB directly, so the following NR14 trigger can clock/reload length
// in the same phase as hardware.
self.div_divider = 1;
self.skip_next_div_apu_event = true;
trace_apu!(2; "GB APU armed skip_next_div_apu_event (power-on with DIV-APU bit high)");
}
/// Expose ch3 wave RAM read for DmgBus (used during CH3 playback quirk).
pub fn read_wave_ram(&self, addr: u16) -> u8 {
self.ch3.read_wave_ram(addr)
}
/// Sub-M-cycle FS-step counter used by sub-M-cycle APU machinery.
///
/// Advanced once per DIV-APU event in `clock_div_apu`. Phase 3 consumes
/// the low two bits to drive the per-128 Hz sweep tick.
#[allow(dead_code)] // Direct readers land in Phase 6 (NR10 glitch).
pub(crate) fn div_divider(&self) -> u8 {
self.div_divider
}
}
impl Default for Apu {
fn default() -> Self {
Self::new(false)
}
}
#[cfg(test)]
mod tests {
use super::*;
fn powered_apu() -> Apu {
let mut apu = Apu::new(false);
apu.write_register(0xFF26, 0x80);
apu
}
/// Creates a powered CGB APU configured for model-specific quirk tests.
fn powered_cgb_apu_for_model(model: CgbModel) -> Apu {
let mut apu = Apu::new(true);
apu.set_cgb_model(model);
apu.write_register(0xFF26, 0x80);
apu
}
/// `div_divider` is a Frame-Sequencer-step counter exposed for the
/// Sub-M-cycle sub-M-cycle sweep machinery. It must advance by
/// exactly one per serviced DIV-APU event so the low bits drive
/// * `& 1 == 1` — extra-length-tick glitch (256 Hz tick, half of 512 Hz)
/// * `& 3 == 3` — sweep tick (128 Hz, every 4 FS steps)
/// * `& 7 == 7` — envelope volume countdown (64 Hz)
///
///
#[test]
fn test_div_divider_advances_per_fs_step() {
let mut apu = powered_apu();
let start = apu.div_divider();
apu.clock_div_apu();
assert_eq!(apu.div_divider(), start.wrapping_add(1));
apu.clock_div_apu();
apu.clock_div_apu();
assert_eq!(apu.div_divider(), start.wrapping_add(3));
}
#[test]
fn test_div_divider_does_not_advance_on_skipped_event() {
// Power-on quirk: skip_next_div_apu_event suppresses both the FS
// dispatch AND the div_divider increment. The following DIV-APU event is
// serviced without incrementing the divider.
let mut apu = powered_apu();
apu.write_register(0xFF11, 0x3E);
apu.write_register(0xFF12, 0x80);
apu.write_register(0xFF14, 0xC0);
assert_eq!(apu.ch1.length_counter, 2);
apu.arm_skip_next_div_apu_event();
assert_eq!(apu.div_divider(), 1);
apu.clock_div_apu();
assert_eq!(apu.div_divider(), 1);
assert_eq!(apu.ch1.length_counter, 2);
apu.clock_div_apu();
assert_eq!(apu.div_divider(), 1);
assert_eq!(apu.ch1.length_counter, 1);
apu.clock_div_apu();
assert_eq!(apu.div_divider(), 2);
assert_eq!(apu.ch1.length_counter, 1);
}
#[test]
fn test_div_divider_low_two_bits_align_every_4_fs_steps() {
let mut apu = powered_apu();
// Advance until the low 2 bits are zero, then verify the
// every-4-FS-step boundary lands on `& 3 == 3` exactly once
// per group of 4.
while apu.div_divider() & 3 != 0 {
apu.clock_div_apu();
}
let mut hits = 0u32;
for _ in 0..16 {
apu.clock_div_apu();
if apu.div_divider() & 3 == 3 {
hits += 1;
}
}
assert_eq!(hits, 4, "expected 4 boundary hits over 16 FS steps");
}
#[test]
fn test_div_divider_resets_on_nr52_power_cycle() {
// Sub-events keyed off `div_divider` low bits must stay aligned
// with `fs_step` after an NR52 power-off / power-on cycle.
let mut apu = powered_apu();
for _ in 0..5 {
apu.clock_div_apu();
}
assert_ne!(apu.div_divider(), 0);
// Power off via NR52.
apu.write_register(0xFF26, 0x00);
assert_eq!(apu.div_divider(), 0);
// Power on again — should still be 0 (and fs_step is also 0).
apu.write_register(0xFF26, 0x80);
assert_eq!(apu.div_divider(), 0);
}
#[test]
fn test_cgb_b_extra_length_clock_ignores_current_length_enable() {
let mut apu = powered_cgb_apu_for_model(CgbModel::CgbB);
apu.write_register(0xFF11, 0x3E);
assert_eq!(apu.ch1.length_counter, 2);
apu.clock_div_apu();
apu.write_register(0xFF14, 0x00);
assert!(!apu.ch1.length_en());
assert_eq!(apu.ch1.length_counter, 1);
}
#[test]
fn test_cgb_c_extra_length_clock_requires_current_length_enable() {
let mut apu = powered_cgb_apu_for_model(CgbModel::CgbC);
apu.write_register(0xFF11, 0x3E);
assert_eq!(apu.ch1.length_counter, 2);
apu.clock_div_apu();
apu.write_register(0xFF14, 0x00);
assert!(!apu.ch1.length_en());
assert_eq!(apu.ch1.length_counter, 2);
}
#[test]
fn test_nr52_power_on_bit_readable() {
let apu = powered_apu();
assert_eq!(apu.read_register(0xFF26) & 0x80, 0x80);
}
#[test]
fn test_nr52_power_off_bit_readable() {
let apu = Apu::new(false);
assert_eq!(apu.read_register(0xFF26) & 0x80, 0x00);
}
#[test]
fn test_nr52_unused_bits_read_as_1() {
let apu = Apu::new(false);
assert_eq!(apu.read_register(0xFF26) & 0x70, 0x70);
}
#[test]
fn test_power_off_clears_nr50() {
let mut apu = powered_apu();
apu.write_register(0xFF24, 0x77);
assert_eq!(apu.read_register(0xFF24), 0x77);
apu.write_register(0xFF26, 0x00);
assert_eq!(apu.nr50, 0x00);
}
#[test]
fn test_power_off_clears_nr51() {
let mut apu = powered_apu();
apu.write_register(0xFF25, 0xF3);
apu.write_register(0xFF26, 0x00);
assert_eq!(apu.nr51, 0x00);
}
#[test]
fn test_power_off_ignores_nr50_write() {
let mut apu = Apu::new(false);
apu.write_register(0xFF24, 0x77);
assert_eq!(apu.nr50, 0x00, "NR50 write must be ignored when APU is off");
}
#[test]
fn test_power_on_allows_nr50_write() {
let mut apu = powered_apu();
apu.write_register(0xFF24, 0x77);
assert_eq!(apu.nr50, 0x77);
}
#[test]
fn test_nr51_write_read_roundtrip() {
let mut apu = powered_apu();
apu.write_register(0xFF25, 0xAB);
assert_eq!(apu.read_register(0xFF25), 0xAB);
}
#[test]
fn test_sample_not_ready_initially() {
let apu = Apu::new(false);
assert!(!apu.sample_ready());
}
#[test]
fn test_sample_ready_after_enough_ticks() {
let mut apu = Apu::new(false);
apu.set_sample_rate(44_100.0);
apu.tick(30);
assert!(apu.sample_ready(), "expected a sample after 30 M-cycles");
}
#[test]
fn test_multiple_samples_accumulate_until_consumed() {
let mut apu = Apu::new(false);
apu.set_sample_rate(44_100.0);
// A web video frame advances far more than one output sample period
// before JS drains audio. Generated samples must queue instead of
// overwriting/dropping everything after the first pending sample.
apu.tick(72);
let mut sample_count = 0;
while apu.take_sample().is_some() {
sample_count += 1;
}
assert!(
sample_count >= 3,
"expected at least 3 queued samples, got {sample_count}"
);
}
#[test]
fn test_take_sample_clears_ready_flag() {
let mut apu = Apu::new(false);
apu.tick(30);
apu.take_sample();
assert!(!apu.sample_ready());
}
#[test]
fn test_sample_is_silent_when_powered_off() {
let mut apu = Apu::new(false);
apu.tick(30);
let s = apu.take_sample().unwrap_or(0.0);
assert_eq!(s, 0.0);
}
#[test]
fn test_frame_sequencer_advances_on_clock_div_apu() {
let mut apu = powered_apu();
assert_eq!(apu.fs_step, 0);
// Frame sequencer is now clocked by explicit DIV-APU events,
// not by the internal fs_timer. One call to clock_div_apu()
// advances by one step.
apu.clock_div_apu();
assert_eq!(
apu.fs_step, 1,
"fs_step must be 1 after one clock_div_apu()"
);
}
#[test]
fn test_frame_sequencer_wraps_at_8() {
let mut apu = powered_apu();
// Clock frame sequencer 8 times to verify wrap
for _ in 0..8 {
apu.clock_div_apu();
}
assert_eq!(apu.fs_step, 0, "fs_step must wrap back to 0 after 8 steps");
}
#[test]
fn test_registers_return_cleared_values_when_powered_off() {
let mut apu = powered_apu();
apu.write_register(0xFF26, 0x00);
assert_eq!(apu.read_register(0xFF10), 0x80);
assert_eq!(apu.read_register(0xFF12), 0x00);
}
#[test]
fn test_nr52_ch1_active_bit_reflects_channel1_state() {
let apu = Apu::new(false);
assert_eq!(apu.read_register(0xFF26) & 0x01, 0x00);
}
fn powered_cgb_apu() -> Apu {
let mut apu = Apu::new(true);
apu.write_register(0xFF26, 0x80);
apu
}
#[test]
fn test_cgb_power_off_rejects_nr41_length_write() {
let mut apu = powered_cgb_apu();
apu.write_register(0xFF20, 0x3F);
apu.write_register(0xFF26, 0x00);
apu.write_register(0xFF20, 0x3F);
assert_eq!(
apu.ch4.length_counter, 0,
"CGB must reject length writes when APU is off"
);
}
#[test]
fn test_cgb_power_off_rejects_nr11_length_write() {
let mut apu = Apu::new(true);
apu.write_register(0xFF11, 0x3F);
assert_eq!(
apu.ch1.length_counter, 0,
"CGB must reject CH1 length writes when APU is off"
);
}
#[test]
fn test_cgb_power_off_rejects_nr21_length_write() {
let mut apu = Apu::new(true);
apu.write_register(0xFF16, 0x3F);
assert_eq!(
apu.ch2.length_counter, 0,
"CGB must reject CH2 length writes when APU is off"
);
}
#[test]
fn test_cgb_power_off_rejects_nr31_length_write() {
let mut apu = Apu::new(true);
apu.write_register(0xFF1B, 0xFF);
assert_eq!(
apu.ch3.length_counter, 0,
"CGB must reject CH3 length writes when APU is off"
);
}
#[test]
fn test_dmg_power_off_allows_nr41_length_write() {
let mut apu = Apu::new(false);
apu.write_register(0xFF20, 0x3F);
assert_eq!(
apu.ch4.length_counter, 1,
"DMG must allow length writes when APU is off"
);
}
#[test]
fn test_cgb_power_on_resets_length_counters() {
let mut apu = powered_cgb_apu();
apu.write_register(0xFF11, 0x00);
apu.write_register(0xFF16, 0x00);
apu.write_register(0xFF1B, 0x00);
apu.write_register(0xFF20, 0x00);
assert_eq!(apu.ch1.length_counter, 64);
assert_eq!(apu.ch4.length_counter, 64);
apu.write_register(0xFF26, 0x00);
apu.write_register(0xFF26, 0x80);
assert_eq!(
apu.ch1.length_counter, 0,
"CGB power-on must reset CH1 length"
);
assert_eq!(
apu.ch2.length_counter, 0,
"CGB power-on must reset CH2 length"
);
assert_eq!(
apu.ch3.length_counter, 0,
"CGB power-on must reset CH3 length"
);
assert_eq!(
apu.ch4.length_counter, 0,
"CGB power-on must reset CH4 length"
);
}
#[test]
fn test_mix_hp_filter_produces_negative_transient_when_input_drops() {
let mut apu = powered_apu();
apu.hp_prev_in = 1.0;
apu.hp_prev_out = 0.5;
let sample = apu.mix();
assert!(
sample < 0.0,
"HP filter must produce a negative transient when input drops from high to zero, got {sample}"
);
}
#[test]
fn test_mix_hp_filter_output_stays_in_bipolar_range() {
let mut apu = powered_apu();
apu.write_register(0xFF11, 0x80);
apu.write_register(0xFF12, 0xF0);
apu.write_register(0xFF14, 0x80);
apu.write_register(0xFF16, 0x80);
apu.write_register(0xFF17, 0xF0);
apu.write_register(0xFF19, 0x80);
apu.write_register(0xFF1A, 0x80);
apu.write_register(0xFF1C, 0x20);
apu.write_register(0xFF1E, 0x80);
apu.write_register(0xFF21, 0xF0);
apu.write_register(0xFF23, 0x80);
apu.write_register(0xFF25, 0xFF);
apu.write_register(0xFF24, 0x77);
let mut violations = 0u32;
for _ in 0..60_000 {
apu.tick(1);
if let Some(s) = apu.take_sample()
&& !(-1.0..=1.0).contains(&s)
{
violations += 1;
}
}
assert_eq!(
violations, 0,
"mix() must never exceed [-1.0, 1.0]; found {violations} violation(s)"
);
}
#[test]
fn test_mix_active_channel_produces_negative_samples_after_filter_converges() {
let mut apu = powered_apu();
apu.write_register(0xFF11, 0x80);
apu.write_register(0xFF12, 0xF0);
apu.write_register(0xFF13, 0x80);
apu.write_register(0xFF14, 0x84);
apu.write_register(0xFF25, 0x01);
apu.write_register(0xFF24, 0x07);
let mut found_negative = false;
for _ in 0..71_340 {
apu.tick(1);
if let Some(s) = apu.take_sample()
&& s < 0.0
{
found_negative = true;
break;
}
}
assert!(
found_negative,
"HP filter must produce negative samples once DC is removed from an oscillating channel"
);
}
#[test]
fn test_pcm12_returns_zero_when_channels_off() {
let apu = powered_cgb_apu();
// CH1 and CH2 inactive → PCM12 should be $00
assert_eq!(apu.read_pcm12(), 0x00);
}
#[test]
fn test_pcm34_returns_zero_when_channels_off() {
let apu = powered_cgb_apu();
// CH3 and CH4 inactive → PCM34 should be $00
assert_eq!(apu.read_pcm34(), 0x00);
}
#[test]
fn test_pcm12_nibble_packing() {
let mut apu = powered_cgb_apu();
// Trigger CH1 with volume 7 and 50% duty (high state)
apu.write_register(0xFF12, 0x70); // vol=7, no envelope
apu.write_register(0xFF11, 0x80); // 50% duty
apu.write_register(0xFF13, 0xFF); // freq low = 0xFF for fast period
apu.write_register(0xFF14, 0x87); // trigger + length disable, freq high=7
// Trigger CH2 with volume 5
apu.write_register(0xFF17, 0x50); // vol=5, no envelope
apu.write_register(0xFF16, 0x80); // 50% duty
apu.write_register(0xFF18, 0xFF); // freq low = 0xFF for fast period
apu.write_register(0xFF19, 0x87); // trigger + length disable, freq high=7
// Tick to advance duty_pos to position 5 (high in 50% duty).
// freq=0x7FF -> period=4 T-cycles = 1 M-cycle per advance.
// startup delay ~8 T-cycles = 2 M-cycles.
// After ~2 ticks, first_sample_zero is cleared on first advance.
// duty_pos at tick N ≈ (N - 2) mod 8.
// For duty_pos=5, need N=7 (7-2=5).
apu.tick(7);
let pcm12 = apu.read_pcm12();
// Low nibble = CH1, high nibble = CH2
// At duty_pos=5 (50% duty), output should be volume (7 for CH1, 5 for CH2)
let ch1_out = pcm12 & 0x0F;
let ch2_out = (pcm12 >> 4) & 0x0F;
// At least one of them should output (depends on duty cycle position)
assert!(
ch1_out > 0 || ch2_out > 0,
"At least one pulse channel should output non-zero in PCM12"
);
}
#[test]
fn test_pcm_returns_zero_when_apu_powered_off() {
let mut apu = powered_cgb_apu();
// Trigger some channels
apu.write_register(0xFF12, 0xF0);
apu.write_register(0xFF14, 0x80);
apu.write_register(0xFF17, 0xF0);
apu.write_register(0xFF19, 0x80);
// Power off APU
apu.write_register(0xFF26, 0x00);
// PCM registers should return 0 when powered off
assert_eq!(apu.read_pcm12(), 0x00);
assert_eq!(apu.read_pcm34(), 0x00);
}
#[test]
fn test_ch1_length_counter_clocked_by_div_apu_event() {
// Simulates the scenario in SameSuite channel_1_stop_div:
// Setup CH1 with length_counter = 1, then clock the frame sequencer
// via DIV-APU event, which should stop the channel.
let mut apu = powered_apu();
// Setup CH1: duty 50%, length_load = 63 → counter = 64 - 63 = 1
apu.write_register(0xFF11, 0x80 | 0x3F);
assert_eq!(apu.ch1.length_counter, 1, "CH1 length counter should be 1");
// Enable DAC and set volume
apu.write_register(0xFF12, 0x80); // volume = 8, no envelope
// Trigger with length enabled
apu.write_register(0xFF14, 0xC0);
// Channel should be active after trigger
assert!(
apu.ch1.is_active(),
"CH1 should be active after trigger with DAC on"
);
assert!(
apu.ch1.length_en(),
"CH1 length should be enabled after NR14 write"
);
// Clock the frame sequencer via DIV-APU event.
// Since fs_step starts at 0 after power-on, FS_TABLE[0] = 0x01 clocks length.
apu.clock_div_apu();
// Length counter should have decremented from 1 to 0, stopping the channel
assert_eq!(
apu.ch1.length_counter, 0,
"CH1 length counter should be 0 after clocking"
);
assert!(
!apu.ch1.is_active(),
"CH1 should be inactive after length counter reaches 0"
);
// PCM12 should show 0 for CH1 now
assert_eq!(
apu.read_pcm12() & 0x0F,
0,
"CH1 digital output should be 0 after channel stops"
);
}
#[test]
fn test_channel_1_delay_high_freq() {
// Test mimics SameSuite channel_1_delay test:
// freq = $7FF, duty = 75% ($C0), volume = 8 ($80)
// Expected: after 2 M-cycles, output should be 0
// after 3 M-cycles, output should be 8
let mut apu = powered_cgb_apu();
// Power off and on APU to reset
apu.write_register(0xFF26, 0x00); // APU off
apu.write_register(0xFF26, 0xFF); // APU on
// Set up channel 1 like the test ROM
apu.write_register(0xFF13, 0xFF); // NR13: freq low = $FF
apu.write_register(0xFF11, 0xC0); // NR11: duty = 75%
apu.write_register(0xFF12, 0x80); // NR12: volume = 8, no envelope
apu.write_register(0xFF14, 0x87); // NR14: trigger + freq high = 7 → freq = $7FF
// At this point, trigger happened. duty_pos should be 0.
// 75% duty = [0,1,1,1,1,1,1,0], so output at pos 0 = 0.
// After 2 M-cycles (0 NOPs in SameSuite), should be at position 0 (output = 0)
apu.tick(2);
let pcm12_at_2 = apu.read_pcm12() & 0x0F;
assert_eq!(
pcm12_at_2, 0x00,
"Output should be 0 at 2 M-cycles (0 NOPs)"
);
// After 1 more M-cycle (1 NOP in SameSuite), should be at position 1 (output = 8)
apu.tick(1);
let pcm12_at_3 = apu.read_pcm12() & 0x0F;
assert_eq!(pcm12_at_3, 0x08, "Output should be 8 at 3 M-cycles (1 NOP)");
assert!(apu.ch1.is_active(), "CH1 should be active after trigger");
}
// Note: Restart timing test removed - the retrigger delay behavior is complex
// and depends on delay_between value in non-trivial ways. The SameSuite restart
// tests verify this behavior, but getting all edge cases right requires more
// investigation. See channel_1_restart.asm comment:
// "It appears that after restarting, the start delay from the 'delay' test
// is actually 1 tick shorter. The countdown for the next sample is reset,
// but the new pulse's first sample will be the next sample the old pulse
// would have played (i.e. the current sample index or phase does not reset)"
}