IPB

Welcome Guest ( Log In | Register )

Why 24bit/48kHz/96kHz/, If 16bit/44.1kHz is good enough?
William
post Dec 29 2005, 13:45
Post #1





Group: Members
Posts: 288
Joined: 8-January 02
Member No.: 959



Yes, I have searched the forum.
Yes, maybe I am dumb.

But it seems I cannot find the answer.

Why do we need 24bit/48kHz/96kHz/192kHz if 16bit/44.1kHz is good enough? Are there any situations that 16bit/44.1kHz simply cannot satisfy? In other words, is there any real need for the higher bit depth and sampling rate?

Thanks for answering.

This post has been edited by William: Dec 29 2005, 13:47
Go to the top of the page
+Quote Post
 
Start new topic
Replies
AndyH-ha
post Jan 9 2006, 08:28
Post #2





Group: Members
Posts: 2209
Joined: 31-August 05
Member No.: 24222



The Sampling Theorem says that signals containing frequencies up to 1/2 the sampling frequency (S/2) can be exactly represented by S samples per second. Restated, the highest frequency that can be recorded, and played back, is equal to 1/2 the sample rate. It isn't approximately, or "in the neighborhood of", or "a degraded representation of", it is exactly the same waveform, right up to S/2 (e.g. 22, 050 Hz at 44,100 Hz). No information is lost.

The input must be bandlimited to no more than S/2 (the Nyquist frequency) or there will be distortion. If higher frequencies exist in the input they will be partially sampled, creating false information that portrays them as lower frequencies in the audio band. These are called alias images or foldover distortion. If S is the sample rate and F is some frequency between S/2 and S, then its image is created at S - F. Thus as F approaches S, the images are created at lower and lower audio frequencies.

Since it isn't feasible to build analogue filters that truly act as "brick walls," a stop band filter is employed that begins attenuating the input several thousand Hz below the Nyquist frequency. By S/2 the signal level is down sufficiently that aliasing isn't a significant problem. It is real world filter insufficiency, not any sampling theory constraints, that may limit actual performance to something less than the Nyquist frequency.

It is easier to do a better job by sampling at a much higher frequency and handling the details in the digital domain, either in hardware, or especially in software. In actual practice many ADCs, probably those in most of our soundcards, sample at several million times per second, use some form of sigma-delta conversion, then filter down to the desired sample rate and bit depth.

Sampling means that the continuous analogue waveform voltage can be measured only in terms of a limited number of discrete digital values. Although frequency information can be captured exactly (up to S/2), amplitude accuracy is limited by the size of the intervals into which the amplitude (from zero to maximum) must be assigned (e.g .01V, .02V, .03V, ...) The input is always recorded as exactly one of those values although the actual sampled input voltage is usually not exactly one of those values. The error can be as great as 1/2 interval.

Increasing the bit depth increases the number of possible values, increasing accuracy. Going from 16 bits to 24 bits increase the accuracy by a factor of 256, or decreases the average error amplitude by 1/256.

As signal level decreases the errors become larger relative to the signal. Also as signal level decreases, the number of bits, or levels, between it and zero decreases. This provides relatively fewer choices to recorded its value. Thus the increased number of quantization levels available with greater bit depth becomes more important at lower input levels.

At high signal levels the quantization error is experienced as white noise but at very low signal level it is experienced as distortion. For every bit of depth added, the error, or quantization noise, is decreased by 6dB. Distortion of low level signals (e.g. -90dB to -96dB) is much greater when using 16 bit rather than 24 bit.

There are of course many other considerations to getting good digital audio reproduction . Whether or not some of the factors discussed here are always audible, they are real and measurable.
Go to the top of the page
+Quote Post
bug80
post Jan 9 2006, 10:19
Post #3





Group: Members
Posts: 403
Joined: 23-January 05
From: The Netherlands
Member No.: 19254



QUOTE (AndyH-ha @ Jan 9 2006, 09:28 AM)
The Sampling Theorem says that signals containing frequencies up to 1/2 the sampling frequency (S/2) can be exactly represented by S samples per second. Restated, the highest frequency that can be recorded, and played back, is equal to 1/2 the sample rate. It isn't approximately, or "in the neighborhood of", or "a degraded representation of", it is exactly the same waveform, right up to S/2  (e.g. 22, 050 Hz at 44,100 Hz). No information is lost.
*

I thought that only frequencies under half the sampling rate can be represented exactly (so, Fs/2 is the first frequency that cannot be represented anymore). See also wikipedia.

QUOTE (Qjimbo @ Jan 9 2006, 01:20 AM)
There's also the more debatable issue that higher frequencies add to the warmth of the sound, even though they're inaudible on a concious level. Hence the term analog warmth, where valves would add random high frequencies to the audio.
*

Well, as fas as I know these added harmonics are well within audible range. smile.gif
Go to the top of the page
+Quote Post

Posts in this topic
- William   Why 24bit/48kHz/96kHz/   Dec 29 2005, 13:45
- - Lyx   QUOTE (William @ Dec 29 2005, 01:45 PM)Why do...   Dec 29 2005, 13:53
|- - krabapple   QUOTE (Lyx @ Dec 29 2005, 07:53 AM)QUOTE (Wil...   Jan 6 2006, 20:04
- - William   I heard someone saying that increasing the samplin...   Dec 29 2005, 14:09
|- - Garf   QUOTE (William @ Dec 29 2005, 03:09 PM)I hear...   Dec 29 2005, 14:13
|- - William   QUOTE (Garf @ Dec 29 2005, 01:13 PM)Normally,...   Dec 29 2005, 14:22
|- - Garf   QUOTE (William @ Dec 29 2005, 03:22 PM)Would ...   Dec 29 2005, 14:45
- - bizangoin   I totally agree with you all. Increasing sample ra...   Dec 29 2005, 14:32
- - William   I am sorry. Thanks.   Dec 29 2005, 14:59
- - singaiya   QUOTE (William @ Dec 29 2005, 04:45 AM)Why do...   Dec 29 2005, 18:44
|- - bug80   QUOTE (singaiya @ Dec 29 2005, 07:44 PM)I hav...   Dec 29 2005, 22:16
- - rosshmusic   I agree that it makes no difference to end users.....   Dec 29 2005, 21:32
|- - Garf   QUOTE (rosshmusic @ Dec 29 2005, 10:32 PM)I a...   Dec 29 2005, 21:46
|- - listen   I had a thread here: http://www.hydrogenaudio.org/...   Dec 29 2005, 22:24
|- - rosshmusic   QUOTE (Garf @ Dec 29 2005, 04:46 PM)QUOTE (ro...   Dec 29 2005, 22:55
||- - Garf   QUOTE (rosshmusic @ Dec 29 2005, 11:55 PM)htt...   Dec 29 2005, 23:58
||- - rosshmusic   QUOTE (Garf @ Dec 29 2005, 06:58 PM)QUOTE (ro...   Dec 30 2005, 02:59
||- - bug80   QUOTE (rosshmusic @ Dec 30 2005, 03:59 AM)whe...   Dec 30 2005, 13:54
|- - krabapple   QUOTE (Garf @ Dec 29 2005, 03:46 PM)QUOTE (ro...   Jan 6 2006, 20:10
- - Tool462   ok, i'm a newb so don't give me some slack...   Dec 29 2005, 22:50
|- - Garf   QUOTE (Tool462 @ Dec 29 2005, 11:50 PM)ok, i...   Dec 30 2005, 00:01
- - Revliskciuq   I have to step out and disagree that there is no a...   Dec 30 2005, 03:11
|- - Garf   QUOTE (Revliskciuq @ Dec 30 2005, 04:11 AM)I ...   Dec 30 2005, 13:59
||- - jmvalin   QUOTE (Garf @ Dec 30 2005, 09:59 PM)Practise ...   Dec 31 2005, 16:22
||- - Garf   QUOTE (jmvalin @ Dec 31 2005, 05:22 PM)Actual...   Dec 31 2005, 16:31
||- - WmAx   QUOTE (jmvalin @ Dec 31 2005, 11:22 AM)but I...   Dec 31 2005, 16:33
||- - Shade[ST]   QUOTE (WmAx @ Dec 31 2005, 09:33 AM)Just to p...   Dec 31 2005, 19:25
||- - WmAx   QUOTE (Shade[ST),Dec 31 2005, 02:25 PM]QUOTE ...   Dec 31 2005, 20:52
||- - enry2k   First: What about ensuring extremely controlled li...   Jan 6 2006, 01:08
||- - Pio2001   QUOTE (William @ Dec 30 2005, 05:03 AM)And wh...   Jan 6 2006, 06:48
||- - krabapple   QUOTE (enry2k @ Jan 5 2006, 07:08 PM)First: W...   Jan 6 2006, 20:27
||- - enry2k   Thank you for quoting the thread, it is really i...   Jan 8 2006, 20:36
|- - crimsontide   QUOTE (Revliskciuq @ Dec 29 2005, 08:11 PM)I ...   Mar 13 2006, 14:36
|- - Cyaneyes   QUOTE (crimsontide @ Mar 13 2006, 08:36 AM)I ...   Mar 13 2006, 15:55
||- - crimsontide   QUOTE (Cyaneyes @ Mar 13 2006, 08:55 AM)QUOTE...   Mar 13 2006, 16:49
|- - Societal Eclipse   QUOTE (crimsontide @ Mar 13 2006, 09:36 AM)An...   Mar 13 2006, 17:40
- - William   OK, so here are some more technical questions I fo...   Dec 30 2005, 04:03
|- - probedb   QUOTE (William @ Dec 30 2005, 04:03 AM)1) The...   Jan 5 2006, 14:28
- - boombaard   so, on a slightly different note, would the 24/96 ...   Dec 30 2005, 14:24
|- - Garf   QUOTE (boombaard @ Dec 30 2005, 03:24 PM)so, ...   Dec 30 2005, 14:25
|- - boombaard   QUOTE (Garf @ Dec 30 2005, 02:25 PM)QUOTE (bo...   Dec 30 2005, 14:37
|- - Garf   QUOTE (boombaard @ Dec 30 2005, 03:37 PM)how ...   Dec 30 2005, 15:13
||- - boombaard   QUOTE (Garf @ Dec 30 2005, 03:13 PM)QUOTE (bo...   Dec 30 2005, 15:46
||- - jussi   well, and what about the HDCD (20bit) technology? ...   Dec 30 2005, 15:52
|||- - Garf   QUOTE (jussi @ Dec 30 2005, 04:52 PM)well, an...   Dec 30 2005, 16:00
||- - Garf   QUOTE (boombaard @ Dec 30 2005, 04:46 PM)QUOT...   Dec 30 2005, 15:56
|- - Crissaegrim   Definitive answer... Artists record at as high as ...   Dec 30 2005, 17:02
|- - bug80   QUOTE (Crissaegrim @ Dec 30 2005, 06:02 PM)De...   Dec 30 2005, 17:09
- - William   Well, it sounds like these 24bits / whatever kHz a...   Dec 30 2005, 15:54
- - Wintershade   Sorry for popping in like that, but here's my ...   Dec 30 2005, 16:04
|- - bug80   QUOTE (Wintershade @ Dec 30 2005, 05:04 PM)So...   Dec 30 2005, 16:25
||- - Wintershade   QUOTE (bug80 @ Dec 30 2005, 05:25 PM)Do you h...   Jan 2 2006, 12:45
|- - krabapple   QUOTE (Wintershade @ Dec 30 2005, 10:04 AM)So...   Jan 6 2006, 20:16
- - Axon   Some DSP algorithms may benefit from a high sampli...   Dec 30 2005, 19:39
- - Lyx   Regarding clipping and samplerate: according to a ...   Dec 30 2005, 19:58
|- - bug80   QUOTE (Lyx @ Dec 30 2005, 08:58 PM)Regarding ...   Dec 31 2005, 12:23
|- - Garf   QUOTE (Lyx @ Dec 30 2005, 08:58 PM)As you kno...   Dec 31 2005, 14:26
- - Raptus   >16bit would be necessary if someone wanted to ...   Dec 30 2005, 20:07
- - NoXFeR   On a sidenote; recorders may want higher frequenci...   Dec 31 2005, 00:05
|- - bug80   QUOTE (NoXFeR @ Dec 31 2005, 01:05 AM)On a si...   Dec 31 2005, 12:16
|- - NoXFeR   QUOTE (bug80 @ Dec 31 2005, 12:16 PM)QUOTE (N...   Dec 31 2005, 15:18
|- - WmAx   QUOTE (NoXFeR @ Dec 31 2005, 10:18 AM)The poi...   Dec 31 2005, 15:29
- - Lyx   One more reason to instead simply use proper level...   Dec 31 2005, 12:33
|- - bug80   QUOTE (Lyx @ Dec 31 2005, 01:33 PM)One more r...   Dec 31 2005, 13:33
|- - Garf   QUOTE (bug80 @ Dec 31 2005, 02:33 PM)QUOTE (L...   Dec 31 2005, 14:20
|- - bug80   QUOTE (Garf @ Dec 31 2005, 03:20 PM)QUOTE (bu...   Dec 31 2005, 14:23
|- - Garf   QUOTE (bug80 @ Dec 31 2005, 03:23 PM)QUOTE (G...   Dec 31 2005, 14:31
- - kwwong   The main reason why sampling rates > 44.1Khz is...   Dec 31 2005, 14:30
|- - Garf   QUOTE (kwwong @ Dec 31 2005, 03:30 PM)The mai...   Dec 31 2005, 14:41
||- - kwwong   QUOTE (Garf @ Dec 31 2005, 07:41 AM)QUOTE (kw...   Jan 1 2006, 04:30
||- - kwwong   QUOTE (kwwong @ Dec 31 2005, 09:30 PM)QUOTE (...   Jan 1 2006, 04:39
|- - WmAx   QUOTE (kwwong @ Dec 31 2005, 09:30 AM)As for ...   Dec 31 2005, 15:39
- - Garf   Another thing to consider is that with proper dith...   Dec 31 2005, 14:51
|- - marcan   QUOTE (Garf @ Dec 31 2005, 05:51 AM)Another t...   Jan 5 2006, 15:57
|- - enry2k   QUOTE (marcan @ Jan 5 2006, 06:57 AM)QUOTE (G...   Jan 8 2006, 21:28
|- - Garf   QUOTE (marcan @ Jan 5 2006, 04:57 PM)QUOTE (G...   Jan 10 2006, 12:33
- - Garf   I think what he's trying to say is this: If y...   Dec 31 2005, 15:31
- - Lyx   I think he means playing 192khz sample at 44khz WI...   Dec 31 2005, 15:31
- - ivalladt   QUOTE (Garf @ Dec 29 2005, 03:13 PM)The quest...   Dec 31 2005, 19:13
- - dekkersj   QUOTE (Garf @ Dec 31 2005, 02:51 PM)Another t...   Jan 2 2006, 15:41
- - Qjimbo   Ok, back to the original question of the thread. T...   Jan 9 2006, 00:20
|- - Lyx   QUOTE (Qjimbo @ Jan 9 2006, 12:20 AM)Firstly ...   Jan 9 2006, 16:17
|- - marcan   QUOTE (Lyx @ Jan 9 2006, 07:17 AM)QUOTE (Qjim...   Jan 9 2006, 17:59
|- - Woodinville   QUOTE (Lyx @ Jan 9 2006, 07:17 AM)This thread...   Jan 10 2006, 00:38
|- - Garf   QUOTE (Woodinville @ Jan 10 2006, 01:38 AM)We...   Jan 10 2006, 12:42
|- - Woodinville   QUOTE (Garf @ Jan 10 2006, 03:42 AM)QUOTE (Wo...   Jan 10 2006, 23:41
- - CSMR   Somehow this question brings out all the know-alls...   Jan 9 2006, 01:52
- - AndyH-ha   The Sampling Theorem says that signals containing ...   Jan 9 2006, 08:28
|- - bug80   QUOTE (AndyH-ha @ Jan 9 2006, 09:28 AM)T...   Jan 9 2006, 10:19
|- - Woodinville   QUOTE (AndyH-ha @ Jan 8 2006, 11:28 PM)T...   Jan 10 2006, 00:41
- - AndyH-ha   My immediately available source is Principles of D...   Jan 9 2006, 18:02
|- - SebastianG   QUOTE (AndyH-ha @ Jan 9 2006, 06:02 PM)o...   Jan 9 2006, 18:30
- - stephanV   @crimsontide: You can never prove a negative. The...   Mar 13 2006, 17:49
|- - crimsontide   QUOTE (stephanV @ Mar 13 2006, 10:49 AM)@crim...   Mar 13 2006, 18:32
- - harlekeyn   16bit means 2^16 = 65536 possible values for each ...   Mar 19 2006, 06:58
|- - Pio2001   QUOTE (harlekeyn @ Mar 19 2006, 07:58 AM)If y...   Mar 19 2006, 22:07
|- - crimsontide   Well I agree with what Tristan said For example ...   Mar 20 2006, 11:03
|- - Woodinville   QUOTE (crimsontide @ Mar 20 2006, 02:03 AM)I ...   Mar 20 2006, 21:19
- - harlekeyn   WHEN COMPARING digital audio to digital photograph...   Apr 4 2006, 16:48
- - benski   With somewhat expensive studio equipment (RME Hamm...   Apr 4 2006, 16:56
5 Pages V   1 2 3 > » 


Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 30th September 2014 - 16:30