wip

Makefile

@@ -34,12 +34,27 @@ never-gonna-give-you-up.wav : never-gonna-give-you-up.mp4
 	  -ar 15704 \
 	  -y $@
 
+rickroll-17s.wav.s : rickroll-17s.wav pack-vocoder.js
+	node pack-vocoder.js $< $@
+
+rickroll-17s.wav : rickroll-17s.mp4
+	ffmpeg -i "rickroll-17s.mp4" \
+	  -vn \
+	  -t 2.0 \
+	  -acodec pcm_u8 \
+	  -ac 1 \
+	  -ar 15000 \
+	  -y $@
+
 %.o : %.s
 	ca65 -v -t atari -o $@ $<
 
-%.xex : %.o dither4.o never-gonna-give-you-up.wav.o atari-asm-xex.cfg
+rickroll.xex : rickroll.o dither4.o never-gonna-give-you-up.wav.o atari-asm-xex.cfg
 	ld65 -v -C ./atari-asm-xex.cfg -o $@ dither4.o never-gonna-give-you-up.wav.o $<
 
+vocoder.xex : rickroll.o vocoder.o rickroll-17s.wav.o atari-asm-xex.cfg
+	ld65 -v -C ./atari-asm-xex.cfg -o $@ vocoder.o rickroll-17s.wav.o $<
+
 clean :
 	rm -f *.o
 	rm -f *.s.png

dither-image.js

@@ -79,8 +79,7 @@ class RGB {
         let lm = val & 15;
         let crlv = cr ? 50 : 0;
 
-        /*
-        let phase = ((cr - 1) * 25 - 58) * (2 * Math.PI / 360);
+        let phase = ((cr - 1) * 25 - 33) * (2 * Math.PI / 360);
 
         let y = 255 * (lm + 1) / 16;
         let i = crlv * Math.cos(phase);
@@ -89,8 +88,8 @@ class RGB {
         let r = y + 0.956 * i + 0.621 * q;
         let g = y - 0.272 * i - 0.647 * q;
         let b = y - 1.107 * i + 1.704 * q;
-        */
 
+        /*
         // PAL
         let phase = ((cr - 1) * 25.7 - 15) * (2 * Math.PI / 360);
 
@@ -101,8 +100,9 @@ class RGB {
         let r = y + 0.956 * i + 0.621 * q;
         let g = y - 0.272 * i - 0.647 * q;
         let b = y - 1.107 * i + 1.704 * q;
+        */
 
-        return new RGB(r, g, b).clamp().fromSRGB();
+        return new RGB(r, g, b).clamp().fromNTSC();
     }
 
     map(callback) {
@@ -1004,11 +1004,9 @@ ${byte2byte(odd(frame.palette3))}
 .align 1024
 displaylist:
     ; 24 lines overscan
-    .repeat 2
+    .repeat 3
         .byte $70 ; 8 blank lines
     .endrep
-    ; include a DLI to mark us as frame 0
-    .byte $f0 ; 8 blank lines
 
     ; ${height} lines graphics
     ; ANTIC mode e (160px 2bpp, 1 scan line per line)

dither4.s

@@ -27,7 +27,6 @@ sample_ptrh = $85
 sample_ptr = sample_ptrl
 scanline = $86
 audiotemp = $87
-frame_counter = $89
 
 ;height = 160
 height = 192
@@ -66,6 +65,7 @@ audio_high_byte:
         byteseq $3
         byteseq $4
         byteseq $5
+        byteseq $6
         byteseq $7
         byteseq $8
         byteseq $9
@@ -102,12 +102,6 @@ audio_high_byte:
     lda #.hibyte(displaylist)
     sta DLISTH
 
-    ; Set up the DLI handler
-    lda #.lobyte(dli_handler)
-    sta VDSLSTL
-    lda #.hibyte(dli_handler)
-    sta VDSLSTH
-
     ; Disable VBI but allow Reset and DLI
     lda #$a0
     sta NMIEN
@@ -158,25 +152,24 @@ wait_loop:
         sta AUDC1          ; 4 cyc
     .endmacro
 
-    .macro audio_prep
+    .macro audio_prep ; 8-9 cycles
         ; Y is VCOUNT at entry
         lda (sample_ptr),y ; 5/6 cyc
         sta audiotemp      ; 3 cyc
     .endmacro
 
-    ; call with A pre-loaded to audiotemp
-    .macro audio_play_lo
-        ;lda audiotemp ; 3 cyc
-        and #$0f ; 2 cyc
-        ora #$10 ; 2 cyc
-        sta AUDC1 ; 4 cyc
+    .macro audio_play_lo ; 8 cycles
+        ; A is loaded with packed audio byte at entry
+        and #$0f       ; 2 cyc
+        ora #$10       ; 2 cyc
+        sta AUDC1      ; 4 cyc
     .endmacro
 
     ; clobbers Y
     .macro audio_play_hi ; 12 cycles
-        ldy audiotemp ; 3 cyc
+        ldy audiotemp         ; 3 cyc
         lda audio_high_byte,y ; 5 cyc
-        sta AUDC1 ; 4 cyc
+        sta AUDC1             ; 4 cyc
     .endmacro
 
     .macro audio_inc
@@ -191,17 +184,12 @@ wait_loop:
         cmp #.hibyte(audio_samples_end) ; 2 cyc
         bmi audio_cont                  ; 2 cyc
 
-        sta WSYNC
-
         ; 10 cycles, optional
         lda #.lobyte(audio_samples)     ; 2
         sta sample_ptrl                 ; 3
         lda #.hibyte(audio_samples)     ; 2
         sta sample_ptrh                 ; 3
 
-        sta WSYNC
-        ldy VCOUNT             ; 4 cycles
-
         audio_cont:
     .endmacro
 
@@ -215,26 +203,22 @@ wait_loop:
         sty scanline ; 3 cycles
         inner_scanline frame_offset, 0 ; 23-26 cycles before break, 12 cycles after
 
-        ldy scanline ; 3 cycles
-        audio_prep
-        audio_play_lo
+        ldy scanline  ; 3 cycles
+        audio_prep    ; 8-9 cycles
+        audio_play_lo ; 8 cycles
 
         ldy scanline   ; 3 cycles
         inner_scanline frame_offset, 128 ; 23-26 cycles before break, 12 cycles after
 
-        audio_play_hi
+        audio_play_hi  ; 12 cycles
 
-        ; pair cleanup: 6 cycles
         ldy VCOUNT             ; 4 cycles
         bne each_scanline_pair ; 2 cycles
 
+        ; Do bookkeeping during vblank!
         audio_inc      ; 22-32 cycles
+        ;ldy VCOUNT             ; 4 cycles
 
-        ; frame cleanup: 11 cycles
-        lda frame_counter ; 3 cycles
-        eor #1            ; 2 cycles
-        sta frame_counter ; 3 cycles
-        ;jmp wait_start    ; 3 cycles
         jmp each_frame    ; 3 cycles
     .endscope
     .endmacro
@@ -244,10 +228,3 @@ run_frame1:
     run_frame 0
 
 .endproc
-
-
-.proc dli_handler
-    lda #0
-    sta frame_counter
-    rti
-.endproc

pack-vocoder.js

@@ -0,0 +1,102 @@
+import wavefile from 'wavefile';
+let WaveFile = wavefile.WaveFile;
+
+import {default as dct} from 'dct';
+
+const sampleRate = 15000;
+let frameRate = 60;
+let samplesPerFrame = sampleRate / frameRate;
+
+import {
+    readFileSync,
+    writeFileSync
+} from 'fs';
+
+function audio2voices(samples) {
+    let voices = [];
+
+    let floatSamples = samples.map((byte) => ((byte / 256) - 0.5) * 2);
+    let transformed = dct(floatSamples);
+    console.log('audio2voices');
+    console.log(floatSamples);
+    console.log(transformed);
+    //throw new Error('xxxxx');
+
+    let freqs = transformed.map((_f, i) => i * frameRate);
+
+    let bands = new Float64Array(256);
+    for (let i = 0; i < transformed.length; i++) {
+        let amplitude = transformed[i];
+        let freq = freqs[i];
+        if (freq == 0) {
+            continue;
+        }
+        let divisor = Math.floor(sampleRate / freq) - 1;
+        if (divisor > 255) {
+            continue;
+        }
+        bands[divisor] += amplitude;
+    }
+    console.log(bands);
+
+    for (let i = 0; i < 4; i++) {
+        let divisor = 0;
+        let max = 0;
+        for (let j = 0; j < bands.length; j++) {
+            if (bands[j] > max) {
+                divisor = j;
+                max = bands[j];
+            }
+        }
+        let amplitude16 = Math.floor(max * 7) + 8;
+        voices.push(divisor, amplitude16);
+    }
+    return voices;
+}
+
+function byte2byte(arr) {
+    let lines = [];
+    for (let i=0; i < arr.length; i++) {
+        lines.push(`.byte ${arr[i]}`);
+    }
+    return lines.join('\n');
+}
+
+function output2assembly(output) {
+    return `
+.segment "AUDIO"
+
+    .export audio_samples
+    .export audio_samples_end
+
+audio_samples:
+    ${byte2byte(output)}
+audio_samples_end:
+    .byte 24
+
+`;
+}
+
+function wav2assembly(buffer) {
+    let wav = new WaveFile(buffer);
+    let samples = wav.getSamples();
+    let seconds = samples.length / sampleRate;
+    let frames = Math.floor(seconds * frameRate);
+
+    let output = [];
+    for (let i = 0; i < frames * samplesPerFrame; i += samplesPerFrame) {
+        let voices = audio2voices(samples.slice(i, i + samplesPerFrame));
+        output.push(...voices);
+    }
+
+    console.log(output);
+    return output2assembly(output);
+}
+
+let infile = process.argv[2];
+let outfile = process.argv[3];
+
+let buffer = readFileSync(infile);
+let asm = wav2assembly(buffer);
+writeFileSync(outfile, asm, 'utf-8');
+

pack-wav.js

@@ -20,7 +20,7 @@ class Dither {
     }
 
     to4bit(val8) {
-        let val = (val8 / 255) - this.err;
+        let val = (val8 / 255) + this.err;
         if (val < 0) {
             val = 0;
         }
@@ -29,7 +29,7 @@ class Dither {
         }
         let val4 = Math.round(val * 15);
         let dithered = (val4 / 15);
-        this.err = (dithered - val);
+        this.err = (val - dithered);
         return val4;
     }
 }

package-lock.json

@@ -10,6 +10,7 @@
         "wavefile": "^11.0.0"
       },
       "devDependencies": {
+        "dct": "^0.1.0",
         "eslint": "^8.36.0"
       }
     },
@@ -814,6 +815,16 @@
         "node": ">= 8"
       }
     },
+    "node_modules/dct": {
+      "version": "0.1.0",
+      "resolved": "https://registry.npmjs.org/dct/-/dct-0.1.0.tgz",
+      "integrity": "sha512-/uUtEniuMq1aUxvLAoDtAduyl12oM1zhA/le2f83UFN/9+4KDHXFB6znEfoj5SDDLiTpUTr26NpxC7t8IFOYhQ==",
+      "dev": true,
+      "license": "MIT",
+      "engines": {
+        "node": ">=0.12.0"
+      }
+    },
     "node_modules/debug": {
       "version": "4.3.4",
       "resolved": "https://registry.npmjs.org/debug/-/debug-4.3.4.tgz",
@@ -2603,6 +2614,12 @@
         "which": "^2.0.1"
       }
     },
+    "dct": {
+      "version": "0.1.0",
+      "resolved": "https://registry.npmjs.org/dct/-/dct-0.1.0.tgz",
+      "integrity": "sha512-/uUtEniuMq1aUxvLAoDtAduyl12oM1zhA/le2f83UFN/9+4KDHXFB6znEfoj5SDDLiTpUTr26NpxC7t8IFOYhQ==",
+      "dev": true
+    },
     "debug": {
       "version": "4.3.4",
       "resolved": "https://registry.npmjs.org/debug/-/debug-4.3.4.tgz",

package.json

@@ -6,6 +6,7 @@
     "wavefile": "^11.0.0"
   },
   "devDependencies": {
+    "dct": "^0.1.0",
     "eslint": "^8.36.0"
   }
 }

vocoder.s

@@ -0,0 +1,230 @@
+SAVMSC = $58
+VDSLST = $200
+VDSLSTL = $200
+VDSLSTH = $201
+COLPF0 = $D016
+COLPF1 = $D017
+COLPF2 = $D018
+COLPF3 = $D019
+COLBK = $D01A
+
+AUDC1 = $D201
+DMACTL = $D400
+DLISTL = $D402
+DLISTH = $D403
+WSYNC = $D40A
+VCOUNT = $D40B
+NMIEN = $D40E
+
+temp1l = $80
+temp1h = $81
+temp1 = temp1l
+temp2l = $82
+temp2h = $83
+temp2 = temp2l
+sample_ptrl = $84
+sample_ptrh = $85
+sample_ptr = sample_ptrl
+scanline = $86
+audiotemp = $87
+
+;height = 160
+height = 192
+bytes_per_line = 40
+pages_per_frame = 32
+lines_per_frame = 262
+;scanline_offset = 31 + (40 - 24) / 2
+scanline_offset = 30
+scanline_max = (lines_per_frame - scanline_offset) / 2
+
+.data
+
+.import audio_samples
+.import audio_samples_end
+
+.import frame1_top
+.import frame1_bottom
+.import frame1_palette1_even
+.import frame1_palette1_odd
+.import frame1_palette2_even
+.import frame1_palette2_odd
+.import frame1_palette3_even
+.import frame1_palette3_odd
+.import displaylist
+
+audio_high_byte:
+    .scope
+        .macro byteseq val
+            .repeat 16
+                .byte val | $10
+            .endrep
+        .endmacro
+        byteseq $0
+        byteseq $1
+        byteseq $2
+        byteseq $3
+        byteseq $4
+        byteseq $5
+        byteseq $6
+        byteseq $7
+        byteseq $8
+        byteseq $9
+        byteseq $a
+        byteseq $b
+        byteseq $c
+        byteseq $d
+        byteseq $e
+        byteseq $f
+    .endscope
+
+.code
+
+.export start
+
+.proc start
+    ; Set up the audio sample buffer
+    lda #.lobyte(audio_samples)
+    sta sample_ptrl
+    lda #.hibyte(audio_samples)
+    sta sample_ptrh
+
+    ; Disable display DMA
+    lda #$00
+    sta DMACTL
+
+    ; Disable VBI and DLI but allow Reset
+    lda #$20
+    sta NMIEN
+
+    ; Set up the display list
+    lda #.lobyte(displaylist)
+    sta DLISTL
+    lda #.hibyte(displaylist)
+    sta DLISTH
+
+    ; Disable VBI but allow Reset and DLI
+    lda #$a0
+    sta NMIEN
+
+    ; Manually wait for first scan line
+wait_vblank:
+    sta WSYNC
+    lda VCOUNT
+    bne wait_vblank
+
+    ; Re-enable display DMA
+    lda #$22
+    sta DMACTL
+
+wait_start:
+    ; Wait for the vblank
+    ; Resynchronize the scanline counter
+wait_loop:
+    ldy VCOUNT    ; 4 cycles
+    bne wait_loop ; 2 cycles
+
+    .macro inner_scanline frame_offset, line_offset
+        ; Y should be VCOUNT at entry
+        ; it'll fire on unused lines, but harmlessly
+
+        ; 23-26 cycles before break
+        ; Leisurely memory fetches
+        lda frame1_palette1_even + frame_offset + line_offset - scanline_offset / 2,y ; 4/5 @FIXME alternate
+        pha                                                                           ; 3
+        ldx frame1_palette2_even + frame_offset + line_offset - scanline_offset / 2,y ; 4/5
+        lda frame1_palette3_even + frame_offset + line_offset - scanline_offset / 2,y ; 4/5
+        tay                                                                           ; 2
+        pla                                                                           ; 3
+
+        ; Wait for horizontal blank
+        sta WSYNC                                                                     ; 4
+
+        ; 12 cycles after break
+        ; Update color registers as fast as possible
+        sta COLPF0 ; 4
+        stx COLPF1 ; 4
+        sty COLPF2 ; 4
+    .endmacro
+
+    .macro audio_play_raw
+        ;ldy VCOUNT ; set on entry
+        lda (sample_ptr),y ; 5/6 cyc
+        sta AUDC1          ; 4 cyc
+    .endmacro
+
+    .macro audio_prep ; 8-9 cycles
+        ; Y is VCOUNT at entry
+        lda (sample_ptr),y ; 5/6 cyc
+        sta audiotemp      ; 3 cyc
+    .endmacro
+
+    .macro audio_play_lo ; 8 cycles
+        ; A is loaded with packed audio byte at entry
+        and #$0f       ; 2 cyc
+        ora #$10       ; 2 cyc
+        sta AUDC1      ; 4 cyc
+    .endmacro
+
+    ; clobbers Y
+    .macro audio_play_hi ; 12 cycles
+        ldy audiotemp         ; 3 cyc
+        lda audio_high_byte,y ; 5 cyc
+        sta AUDC1             ; 4 cyc
+    .endmacro
+
+    .macro audio_inc
+        ; 22 cycles
+        lda sample_ptrl ; 3 cyc
+        clc             ; 2 cyc
+        adc #131        ; 2 cyc
+        sta sample_ptrl ; 3 cyc
+        lda sample_ptrh ; 3 cyc
+        adc #0          ; 2 cyc
+        sta sample_ptrh ; 3 cyc
+        cmp #.hibyte(audio_samples_end) ; 2 cyc
+        bmi audio_cont                  ; 2 cyc
+
+        ; 10 cycles, optional
+        lda #.lobyte(audio_samples)     ; 2
+        sta sample_ptrl                 ; 3
+        lda #.hibyte(audio_samples)     ; 2
+        sta sample_ptrh                 ; 3
+
+        audio_cont:
+    .endmacro
+
+    .macro run_frame frame_offset
+    .scope
+        ; each scanline is 228 color clocks
+        ; that's 114 CPU cycles
+        ; minus 41-43 for DMA leaves 71-73 clock cycles per line
+
+    each_scanline_pair:
+        sty scanline ; 3 cycles
+        inner_scanline frame_offset, 0 ; 23-26 cycles before break, 12 cycles after
+
+        ldy scanline  ; 3 cycles
+        audio_prep    ; 8-9 cycles
+        audio_play_lo ; 8 cycles
+
+        ldy scanline   ; 3 cycles
+        inner_scanline frame_offset, 128 ; 23-26 cycles before break, 12 cycles after
+
+        audio_play_hi  ; 12 cycles
+
+        ldy VCOUNT             ; 4 cycles
+        bne each_scanline_pair ; 2 cycles
+
+        ; Do bookkeeping during vblank!
+        audio_inc      ; 22-32 cycles
+        ;ldy VCOUNT             ; 4 cycles
+
+        jmp each_frame    ; 3 cycles
+    .endscope
+    .endmacro
+
+each_frame:
+run_frame1:
+    run_frame 0
+
+.endproc