CD deemphasis

How to do this properly? I found an attenuation vs. frequency table on the internet, but I am not sure whether it's accurate. Also, which AA filter would be the most appropriate here? FFT filter?

Also, which AA filter would be the most appropriate here? FFT filter?

Since you are talking about two turnover points only, and a -6dB/oct slope, you should be able to implement this easily with a straightforward parametric eq implementation with just two bands. Since the rates are defined, and, like it says in one of the references, the time-constant, t, is related to the (-3 dB) cutoff frequency, f, by the relationship t = 1/(2*pi*f), you should be able to determine the settings quite easily. I haven't tried this, but I might give it a go later. I don't own any of the mythical CDs that are supposed to require this, so it would only be a theoretical exercise as far as I'm concerned - although it could be published as a preset, of course.

[*update*]

As for the given figures, a quick check reveals that they work out; if you rearrange the formula for the time constants, you'd get (for the first one - 50uS) a turnover frequency of 3.183kHz. This would be the point at which the response drops to -3dB - and that's pretty much what the table says. The rest follows...

*update* A quick look at the excel spreadsheet referenced gives an easier way to implement this, I think. If the Audition graphic EQ is used (it should sound fine for a first-order slope), then the thing to do is to modify the spreadsheet frequency values to the ISO 30-band ones that the EQ uses, and generate dB drops for those. You should easily be able to paste the formulae down and extend the table a bit, and after that, it's simply a matter of plugging the numbers into the EQ directly.

Thanks, I was not sure how to achieve the exact shape in the Parametric EQ. I'll probably use Graphic EQ. Is there any specific reason why you suggested Graphic EQ instead of FFT filter?

And yes, I have one of these CDs...

[*update*]

Well, I just rewrote the table for the ISO frequencies, extended the calculation cells and then tried to enter the values... and there is nowhere near enough resolution to do exactlywhat it says - although I really can't see it making that much difference.

And, I can't get sensible values into the parametric - that doesn't appear to give the result I was looking for at all. So far, the gentle roll-off in the graphic looks the best bet - just ignore the hundredths of a dB bits! And also, having an FIR filter implementation, it has a much better phase response. I'd post the revised spreadsheet, but first I would have to make a small modification to the website to allow .xls files, and I haven't checked that out yet.

*update* I've added the extension to the 'allowable' list, and it seems to work fine. Look at the lower half of the table for the ISO version.

Thanks a lot. My CD is going to be listenable. I could probably modify the xls file myself, but I needed input on filter choice.  And the FIR filter not only has better phase response but also entering values is much easier in AA than for other filters.  On the other hand, I had this thought that originally deemphasis was supposed to be done with a simple analogue filter built in the CD player. Thus, using an IIR filter could introduce phase distortion that was supposed and expected to happen in the analogue playback filter. But I seriously doubt if it was ever taken into account during pre-emphasis.

EDIT: one more thing: do you recommend any particular FFT size? I guess I should keep it fairly low, because high frequency resolution is not required, and I would avoid pre-ringing, is that correct?

[*yet another update...*]

EDIT: one more thing: do you recommend any particular FFT size? I guess I should keep it fairly low, because high frequency resolution is not required, and I would avoid pre-ringing, is that correct?

I believe that in practice, it's a compromise. Probably the best thing to do is to try it with an impulse source and see what happens. If I get a chance I'll try it later here and see what the difference, if any, is.

*yet another update...* I tried it using a low (800) and high (9000) setting, and I'm blowed if i can see any difference - there's nothing even vaguely like an overshoot visible at all, never mind ringing. According to the help file, there shouldn't be, anyway - the Q value of each filter is fixed low enough to prevent this. Which means that you can really do is to try it on the real thing and see what happens with a complex waveform. Unfortunately you can't use a spectrogram to determine what happens, because it uses essentially the same mechanism to generate the display as you are using to alter it - so it would be very difficult to work out what the results meant - if anything!

Bottom line - use your ears. If you like it, that's Good Enough!!!!

CD deemphasis for Audition from another forum:

http://www.hydrogenaudio.org/forums/index.php?s=&showtopic=38312&view=findpost&p=584438

Paul

Strikes me that it's not particularly accurate, that version...

Here's a script for CD de-emphasis (http://audiomastersforum.net/synforum/12/topic-12977.htm):

Collection: CD De-emphasis
Title: CD-Deemphasis
Description:
Mode: 4
Undo: 1

Selected: 0 to 126430 scaled 126430 SR 44100
Freq: Off
cmd: Channel Both

Selected: 0 to 126430 scaled 126430 SR 44100
Freq: Off
Comment: {BA6B165E-6CFE-4DDE-B77B-AF61E557D47A}
cmd: Filters\FFT Filter
1: 125
2: 0
3: 99
4: 40
5: 99
6: 73
7: 100
8: 105
9: 100
10: 136
11: 100
12: 192
13: 100
14: 243
15: 100
16: 291
17: 100
18: 334
19: 100
20: 375
21: 100
22: 413
23: 100
24: 449
25: 100
26: 483
27: 100
28: 515
29: 100
30: 559
31: 100
32: 601
33: 100
34: 640
35: 100
36: 677
37: 100
38: 711
39: 100
40: 744
41: 100
42: 784
43: 100
44: 822
45: 100
46: 858
47: 100
48: 892
49: 100
50: 924
51: 100
52: 961
53: 100
54: 997
55: 100
56: 1030
57: 100
58: 1062
59: 100
60: 1098
61: 100
62: 1131
63: 100
64: 1163
65: 100
66: 1198
67: 100
68: 1232
69: 100
70: 1263
71: 100
72: 1297
73: 100
74: 1329
75: 100
76: 1360
77: 100
78: 1392
79: 100
80: 1423
81: 100
82: 1455
83: 100
84: 1486
85: 100
86: 1518
87: 100
88: 1549
89: 100
90: 1580
91: 100
92: 1612
93: 100
94: 1643
95: 100
96: 1674
97: 99
98: 1706
99: 99
100: 1738
101: 99
102: 1769
103: 99
104: 1800
105: 99
106: 1832
107: 99
108: 1863
109: 99
110: 1894
111: 99
112: 1925
113: 99
114: 1956
115: 98
116: 1987
117: 98
118: 2018
119: 98
120: 2049
121: 98
122: 2080
123: 98
124: 2111
125: 97
126: 2142
127: 97
128: 2173
129: 97
130: 2204
131: 96
132: 2235
133: 96
134: 2266
135: 95
136: 2297
137: 95
138: 2328
139: 94
140: 2359
141: 94
142: 2390
143: 93
144: 2421
145: 93
146: 2452
147: 92
148: 2483
149: 91
150: 2514
151: 90
152: 2545
153: 89
154: 2576
155: 88
156: 2607
157: 87
158: 2638
159: 86
160: 2669
161: 85
162: 2700
163: 84
164: 2731
165: 82
166: 2762
167: 81
168: 2793
169: 80
170: 2824
171: 78
172: 2855
173: 76
174: 2886
175: 75
176: 2917
177: 73
178: 2948
179: 71
180: 2979
181: 69
182: 3010
183: 67
184: 3041
185: 65
186: 3072
187: 64
188: 3103
189: 62
190: 3134
191: 60
192: 3165
193: 58
194: 3196
195: 56
196: 3227
197: 54
198: 3258
199: 52
200: 3289
201: 50
202: 3320
203: 48
204: 3351
205: 46
206: 3382
207: 44
208: 3413
209: 42
210: 3444
211: 40
212: 3475
213: 38
214: 3506
215: 37
216: 3537
217: 35
218: 3568
219: 33
220: 3599
221: 31
222: 3630
223: 29
224: 3661
225: 28
226: 3692
227: 26
228: 3723
229: 24
230: 3754
231: 22
232: 3785
233: 20
234: 3816
235: 18
236: 3847
237: 16
238: 3878
239: 14
240: 3909
241: 11
242: 3940
243: 9
244: 3971
245: 6
246: 4002
247: 4
248: 4033
249: 1
250: 4096
251: 1
252: 125
253: 0
254: 99
255: 40
256: 99
257: 73
258: 100
259: 105
260: 100
261: 136
262: 100
263: 192
264: 100
265: 243
266: 100
267: 291
268: 100
269: 334
270: 100
271: 375
272: 100
273: 413
274: 100
275: 449
276: 100
277: 483
278: 100
279: 515
280: 100
281: 559
282: 100
283: 601
284: 100
285: 640
286: 100
287: 677
288: 100
289: 711
290: 100
291: 744
292: 100
293: 784
294: 100
295: 822
296: 100
297: 858
298: 100
299: 892
300: 100
301: 924
302: 100
303: 961
304: 100
305: 997
306: 100
307: 1030
308: 100
309: 1062
310: 100
311: 1098
312: 100
313: 1131
314: 100
315: 1163
316: 100
317: 1198
318: 100
319: 1232
320: 100
321: 1263
322: 100
323: 1297
324: 100
325: 1329
326: 100
327: 1360
328: 100
329: 1392
330: 100
331: 1423
332: 100
333: 1455
334: 100
335: 1486
336: 100
337: 1518
338: 100
339: 1549
340: 100
341: 1580
342: 100
343: 1612
344: 100
345: 1643
346: 100
347: 1674
348: 99
349: 1706
350: 99
351: 1738
352: 99
353: 1769
354: 99
355: 1800
356: 99
357: 1832
358: 99
359: 1863
360: 99
361: 1894
362: 99
363: 1925
364: 99
365: 1956
366: 98
367: 1987
368: 98
369: 2018
370: 98
371: 2049
372: 98
373: 2080
374: 98
375: 2111
376: 97
377: 2142
378: 97
379: 2173
380: 97
381: 2204
382: 96
383: 2235
384: 96
385: 2266
386: 95
387: 2297
388: 95
389: 2328
390: 94
391: 2359
392: 94
393: 2390
394: 93
395: 2421
396: 93
397: 2452
398: 92
399: 2483
400: 91
401: 2514
402: 90
403: 2545
404: 89
405: 2576
406: 88
407: 2607
408: 87
409: 2638
410: 86
411: 2669
412: 85
413: 2700
414: 84
415: 2731
416: 82
417: 2762
418: 81
419: 2793
420: 80
421: 2824
422: 78
423: 2855
424: 76
425: 2886
426: 75
427: 2917
428: 73
429: 2948
430: 71
431: 2979
432: 69
433: 3010
434: 67
435: 3041
436: 65
437: 3072
438: 64
439: 3103
440: 62
441: 3134
442: 60
443: 3165
444: 58
445: 3196
446: 56
447: 3227
448: 54
449: 3258
450: 52
451: 3289
452: 50
453: 3320
454: 48
455: 3351
456: 46
457: 3382
458: 44
459: 3413
460: 42
461: 3444
462: 40
463: 3475
464: 38
465: 3506
466: 37
467: 3537
468: 35
469: 3568
470: 33
471: 3599
472: 31
473: 3630
474: 29
475: 3661
476: 28
477: 3692
478: 26
479: 3723
480: 24
481: 3754
482: 22
483: 3785
484: 20
485: 3816
486: 18
487: 3847
488: 16
489: 3878
490: 14
491: 3909
492: 11
493: 3940
494: 9
495: 3971
496: 6
497: 4002
498: 4
499: 4033
500: 1
501: 4096
502: 1
503: 2
504: 0
505: 6000
506: 1
507: 0
508: 2
509: 0
510: 0
511: 1000
512: 100
513: 4
514: 0
515: 100
516: -10.0002
517: 0.069992
518: 24000
519: 37421056
520: 112329846
521: 187238637
522: 262081892
523: 336990683
524: 411899474
525: 486808265
526: 561717055
527: 636625846
528: 711534637
529: 786377892
530: 37421056
531: 112329846
532: 187238637
533: 262081892
534: 336990683
535: 411899474
536: 486808265
537: 561717055
538: 636625846
539: 711534637
540: 786377892
541: 0
542: 2
543: 1
544: 0
545: 1
546: 0
547: 0
548: 0
549: 1
550: 1
551: 44100
552: 0
553: 4096
554: 0
555: 0
556: 0
557: 0

Freq: Off
End:

Hehe! That'll  teach me not to look in the archive first!

If you run the script once (on pink noise, for instance), you can save the FFT settings as a preset, as well. Just open the FFT and save what's currently there, after you've run it, with a suitable name.

Hehe! That'll  teach me not to look in the archive first!

If you run the script once (on pink noise, for instance), you can save the FFT settings as a preset, as well. Just open the FFT and save what's currently there, after you've run it, with a suitable name.

Yeah, preset.  The curve corresponds pretty well with the data posted in the first post link.  This one is slightly non-curve-perfect, but it is what my Sony CD player delivers.  I may re-measure for mu Sony SACD player to see if it gives the same thing.  I'm going to have to look into the phase shift thing of the first post link.  If deemphasis assumes a phase-shift
reversal, it can be simulated digitally with the scientific filter "remove subsonic rumble" preset. 

BTW, is there a current method for sharing presets anymore with AA besides a script?

The curve corresponds pretty well with the data posted in the first post link.  This one is slightly non-curve-perfect, but it is what my Sony CD player delivers.  I may re-measure for mu Sony SACD player to see if it gives the same thing.  I'm going to have to look into the phase shift thing of the first post link.  If deemphasis assumes a phase-shift reversal, it can be simulated digitally with the scientific filter "remove subsonic rumble" preset. 

The values I created for the graphic EQ were generated from the same formula that resulted in the published list, but the more I think about it, the less happy I am that it's quite correct. The -3dB point isn't quite where it should be, as far as I can see. I'll investigate...

As for the phase issue - well, that might be moot. It would certainly be true that if the pre-emphasis was created with 'real' filters, complete with phase shift, you might want to replicate this - but in the absence of knowing exactly how the filter was created, you wouldn't necessarily get this correct anyway. And if, heaven forbid, somebody had used pre-emphasis generated using a digital filter, where would you be then? Probably one of two likely places, but which?

BTW, is there a current method for sharing presets anymore with AA besides a script?

I don't think that there is. The .xml file containing the data is a remarkably fragile thing, and you extract and modify stuff in it at your peril. Whenever I've tried adding data, it's said that the .xml file was corrupted, and promptly gone back to a default. So using scripts seems like an excellent method for doing this, one way or another.