Came across an interesting article debating/debunking the merits of sampling rates higher than 44.1Khz/16bit as a delivery mechanism for audio, and thought I would share it:
24/192 Music Downloads are Very Silly Indeed
To be honest I'm not sure which side of the debate I fall on. For a long time I thought 44.1/16 was plenty, in recent years I'm not so sure.
Not because I think we can hear sounds above 22Khz (I certainly can't, my hearing only goes to 17.5Khz, and has done since my 20's) but more because of imperfections in practical (read - low cost / commonly available) digital to analog chips and reconstruction filter designs/algorithms.
Not all DAC chips are created equal especially in cheap consumer hardware which surrounds us everywhere, and if your useful signal range goes quite close to the nyquist frequency then the differences between different chips and reconstruction techniques may be noticeable and start to diverge from a theoretically ideal response. (A perfect brick wall filter is not realisable, even in the digital domain)
Although it would seem to be a bit wasteful on number of bits, (one of the article authors main points) it does make some sense to me not to cut things too fine. In any other field it would be considered engineering safety margin. 192Khz does seem completely over the top to me but if 96Khz/24bit is realisable, why not ?
Is there a shortage of bits in the world, a shortage of hard drive space or internet bandwidth ? In the early days of the internet mp3's became popular because they could be downloaded over dialup and stored on small hard drives of the day.
Today both hard drives and internet speeds have reached the point where downloading and storing losslessly compressed 44.1/16 audio is perfectly practical, but very few sources exist. In a few more years downloading and storing losslessly compressed 96Khz/24bit will be perfectly practical.
If bits are cheap, why not ? Am I crazy ?
One comment of the article which I simply can't agree with is "It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate."
(With "moderate bitrate" being weasel words to me...)
MP3/AAC encoding has improved in leaps and bounds in the last 10 years from the point where I could pretty readily tell an mp3 version of something from the CD version regardless of bitrate (swishy sounding treble being one obvious cue) to the point where I find well encoded 256Kbit AAC difficult to distinguish from the original, but I can still hear a difference on certain songs with critical listening on good speakers/headphones.
If I'm honest a lot of my music is in 256Kbit AAC and it is "good enough" in most cases, but anything I've ripped from physical CD's has now been re-ripped in lossless formats, and I can notice the difference, even though its still only 44.1/16. (Expectation bias ? Who knows...) If I could download music in 44.1/16 lossless from iTunes instead of 256Kbit AAC, I would.
What are others thoughts on the article, and sample rates higher than 44.1/16 and/or lossy encoding ? (two separate issues really, kind of rolled in together in the original article)
24/192 Music Downloads are Very Silly Indeed
To be honest I'm not sure which side of the debate I fall on. For a long time I thought 44.1/16 was plenty, in recent years I'm not so sure.
Not because I think we can hear sounds above 22Khz (I certainly can't, my hearing only goes to 17.5Khz, and has done since my 20's) but more because of imperfections in practical (read - low cost / commonly available) digital to analog chips and reconstruction filter designs/algorithms.
Not all DAC chips are created equal especially in cheap consumer hardware which surrounds us everywhere, and if your useful signal range goes quite close to the nyquist frequency then the differences between different chips and reconstruction techniques may be noticeable and start to diverge from a theoretically ideal response. (A perfect brick wall filter is not realisable, even in the digital domain)
Although it would seem to be a bit wasteful on number of bits, (one of the article authors main points) it does make some sense to me not to cut things too fine. In any other field it would be considered engineering safety margin. 192Khz does seem completely over the top to me but if 96Khz/24bit is realisable, why not ?
Is there a shortage of bits in the world, a shortage of hard drive space or internet bandwidth ? In the early days of the internet mp3's became popular because they could be downloaded over dialup and stored on small hard drives of the day.
Today both hard drives and internet speeds have reached the point where downloading and storing losslessly compressed 44.1/16 audio is perfectly practical, but very few sources exist. In a few more years downloading and storing losslessly compressed 96Khz/24bit will be perfectly practical.
If bits are cheap, why not ? Am I crazy ?
One comment of the article which I simply can't agree with is "It's true enough that a properly encoded Ogg file (or MP3, or AAC file) will be indistinguishable from the original at a moderate bitrate."
(With "moderate bitrate" being weasel words to me...)
MP3/AAC encoding has improved in leaps and bounds in the last 10 years from the point where I could pretty readily tell an mp3 version of something from the CD version regardless of bitrate (swishy sounding treble being one obvious cue) to the point where I find well encoded 256Kbit AAC difficult to distinguish from the original, but I can still hear a difference on certain songs with critical listening on good speakers/headphones.
If I'm honest a lot of my music is in 256Kbit AAC and it is "good enough" in most cases, but anything I've ripped from physical CD's has now been re-ripped in lossless formats, and I can notice the difference, even though its still only 44.1/16. (Expectation bias ? Who knows...) If I could download music in 44.1/16 lossless from iTunes instead of 256Kbit AAC, I would.
What are others thoughts on the article, and sample rates higher than 44.1/16 and/or lossy encoding ? (two separate issues really, kind of rolled in together in the original article)
Last edited:
Interesting article. He makes a good point about audible intermodulation from inaudible ultrasonic signals, although how real an issue this is in practice is unclear as ultrasonic levels would be low even when present? I have a theory: some people prefer their sound to be accompanied by low-level intermodulation and interference - evidence for this is the popularity of things like 'tube buffers' (often badly designed cathode followers, which add noise and distortion), cheap (quality, design) but expensive (price) interconnects which can be poorly screened or even unscreened so allowing in RFI, NOS DACs with no attempt at reconstruction filtering so pushing lots of ultrasonics into the amplifier.
Misunderstandings of the sampling theorem are widespread and persistent, as he says. Many people have little idea what effect the anti-aliasing and reconstruction filters have, and often get them mixed up with each other.
So is he right? The hard evidence says yes, but anecdotes to the contrary persist. How much of this is due to real issues, and how much is due to ignorance and postmodernism, I don't know. Let the war begin!!
PS many modern recordings are so poorly engineered that the issue is purely academic anyway!
Misunderstandings of the sampling theorem are widespread and persistent, as he says. Many people have little idea what effect the anti-aliasing and reconstruction filters have, and often get them mixed up with each other.
So is he right? The hard evidence says yes, but anecdotes to the contrary persist. How much of this is due to real issues, and how much is due to ignorance and postmodernism, I don't know. Let the war begin!!
PS many modern recordings are so poorly engineered that the issue is purely academic anyway!
For SPECIFIC records, I can hear the difference between 16/44 and 24/96 (or DSD - that is at 24/88 level). Past the 96kHz is just waste of bandwidth and degrading DAC performances.
But for other recordings, there is nothing to gain from going Hi-Res. If original is garbage, garbage remains.
PS: Any DAC unfiltered output will output "garbage" past 20kHz mark, including DAC's that are 16/44 only. Lots of DYI eliminate the filters (because it is easy and takes no skills to demolish something) and claim that they hear "improvements". Personally I prefer to hear the music the way was created, not with extra harmonics and aliasing.
But for other recordings, there is nothing to gain from going Hi-Res. If original is garbage, garbage remains.
PS: Any DAC unfiltered output will output "garbage" past 20kHz mark, including DAC's that are 16/44 only. Lots of DYI eliminate the filters (because it is easy and takes no skills to demolish something) and claim that they hear "improvements". Personally I prefer to hear the music the way was created, not with extra harmonics and aliasing.
Last edited:
I have a couple of Denons that upsample to 192kHz. DAC's upsample that futher (PCM1794 goes up to 8x). That is good for a CD recording because it relaxes the demands for the output analog filters and can provide a better upsampling than the one integrated in DAC cips. 384kHz will be probably used so the signal is not OS in the DAC cip (AD1955 for example allows by-passing of the internal OS filter).
But from purelly reproduction, the 24/96 is more than sufficient.
Now, in studio work, where are lots of audio files manipulation in digital domain... I can see the reason for higher sample rates. For the sake of complete compatibility with future delivery in PCM (at any consumer sample rate) or DSD, they use 24bit/352.8kHz.
But from purelly reproduction, the 24/96 is more than sufficient.
Now, in studio work, where are lots of audio files manipulation in digital domain... I can see the reason for higher sample rates. For the sake of complete compatibility with future delivery in PCM (at any consumer sample rate) or DSD, they use 24bit/352.8kHz.
Last edited:
Must be pure marketing as I can't think of one engineering reason to do this. Bits not on the recording cannot be recovered and no ADC gets close to 24bit performance. No DAC that I've seen maintains the same measured performance figures as the sampling rate goes up. I'm with Dan Lavry - slower is better. The optimum sample rate is somewhere between 48 and 96k I reckon.
I do agree that the optimal is between 48 and 96... I have many HD recording from HD tracks, Linn, Analekta, etc...
For testing purposes I downloaded the 44.1 and MP3 to compare and there is a noticeable difference in sound... Let's forget the MP3 which is horrid...
Seems less "compressed" but a good recorded 16/44.1 CD sounds really good too!
For testing purposes I downloaded the 44.1 and MP3 to compare and there is a noticeable difference in sound... Let's forget the MP3 which is horrid...
Seems less "compressed" but a good recorded 16/44.1 CD sounds really good too!
There's a saying that is as true now as it was when it was first said (before the introduction of CD):
The wider you open the window, the more muck flies in.
Monsieur Fourier originally.
IMHO 24x96 is preferable for a solid state jukebox 192 is overkill.
SACD or LP is superior for a reference playback.
A square wave is nothing more than a large number of sine waves combined together.
Same goes for a triangle wave and all music.
The article shows the Fletcher Munson Curves, even these show the range of human hearing is from - 8 db to a threshold of pain above 140 db at the HF limit. Even if you use 130 db then the range required is 23 bits. What is presented is that at frequencies below half the sample rate there are more samples available to reproduce the desired tone, so that with the proper manipulation you can save 7 bits of information. That really limits you to full resolution only on bass frequencies.
The second assumption is that if you have 192K sampling rate you will be encoding information above the range of human hearing (30! Khz.) that is not particularly correct. What you usually want to do is maintain phase linearity to your desired cutoff frequency. Most folks understand you don't want to put 90 Khz signals into an audio system. It will do nasty things to the equipment.
The second assumption is that if you have 192K sampling rate you will be encoding information above the range of human hearing (30! Khz.) that is not particularly correct. What you usually want to do is maintain phase linearity to your desired cutoff frequency. Most folks understand you don't want to put 90 Khz signals into an audio system. It will do nasty things to the equipment.
Whoops, I managed to cross-post in the Dgitial Source section:
http://www.diyaudio.com/forums/digital-source/208239-xiph-org-24-192-releases.html
I guess we can merge those, I've no idea which should be the right place though.
http://www.diyaudio.com/forums/digital-source/208239-xiph-org-24-192-releases.html
I guess we can merge those, I've no idea which should be the right place though.
Exactly. People who dispute this just don't understand sampling. It is very instructive to play with DSP and input frequencies near the nyquist limit without an anti-aliasing filter.
I have crunched 24/96 down to 16/44 and heard no difference at all. Does that mean I am deaf, no because some guys (Meyer and Moran) did the same thing in 2007 and published a JAES paper about it and under double blind conditions, nobody could tell the difference with music signals.
Last edited:
Datz 'cause your amps are too big!
Put me in the "192Khz is a waste of bandwidth" camp. 24/96 seems nice. I was happy with 48KHz DAT when that was alive.
I've been downloading HD-audio in 24/96 and would be happy to buy $0.99 iTunes in this format. I think it will take a company like Apple to offer 24/96 at the same price as 16/44 to make HD-audio come to life.
The Nyquist–Shannon sampling theorem states that perfect reconstruction of a signal is possible when the sampling frequency is greater than twice the maximum frequency of the signal being sampled, or equivalently, when the Nyquist frequency (half the sample rate) exceeds the highest frequency of the signal being sampled.
It is desirable to have a sampling frequency greater than twice the desired system bandwidth so that a steep digital filter and a less steep analog anti-aliasing filter can be used in exchange for a steep analog anti-aliasing filter. The reason for wanting a less steep analog anti-aliasing filter is that the digital filter is not subject to any component variations thus always giving the filter response the designer have chosen.
Analog long playing records and cassette tapes have a SNR of approximately 50 to 60 dB, with a theoretical max of 70db.
The SNR of the CD-DA is 96 dB. The Dynamic Range is 96 db.
The SNR of the DVD-A is 108 db. The Dynamic Range is 108 db.
The SNR of the SACD-DA is 120 dB. The Dynamic Range is 108 db.
Only the best front ends and amplifiers can achieve SNR = 120db at even half rated power.
The Nyquist–Shannon sampling theorem states that perfect reconstruction of a signal is possible when the sampling frequency is greater than twice the maximum frequency of the signal being sampled, or equivalently, when the Nyquist frequency (half the sample rate) exceeds the highest frequency of the signal being sampled.
It is desirable to have a sampling frequency greater than twice the desired system bandwidth so that a steep digital filter and a less steep analog anti-aliasing filter can be used in exchange for a steep analog anti-aliasing filter. The reason for wanting a less steep analog anti-aliasing filter is that the digital filter is not subject to any component variations thus always giving the filter response the designer have chosen.
Analog long playing records and cassette tapes have a SNR of approximately 50 to 60 dB, with a theoretical max of 70db.
The SNR of the CD-DA is 96 dB. The Dynamic Range is 96 db.
The SNR of the DVD-A is 108 db. The Dynamic Range is 108 db.
The SNR of the SACD-DA is 120 dB. The Dynamic Range is 108 db.
Only the best front ends and amplifiers can achieve SNR = 120db at even half rated power.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.