[bigshot]

Thanks hciman77.

Leporello, I have heard digital audio that sounds "clicky and hitchy" when it's sampled incorrectly, or when the processor or hard drive is overworked. It's entirely possible that he actually is hearing choppy sound. It's probably his computer.

See ya
Steve

 

[IPodPJ]

Quote:

Originally Posted by chesebert /img/forum/go_quote.gif Definitive XRCD vs 24bit/96Khz Review of Erick Friedman Violin Show Piece I have the XRCD version and I think its one of those most well mastered/recorded CDs I have. It's got the XRCD2, and the 20bitk2 processing. I downloaded the 24/96 tape transfer version from HighDefTapeTransfers WOW what a difference 24/96 tape transfer makes; the high def tape transfer made the XRCD version sounds like mp3!!! .............

That mimicks what I was saying about how good the sound quality is on those recordings. They are certainly amazing. If you have a soundcard that will output at 24/96 via ASIO and a DAC that will play the file, you ought to listen to them. You will immediately hear how bad they make the CD format sound.

 

[bigshot]

Now that you're back, perhaps you could let us know what software and dithering algorithm you used to resample the tracks. I'd be happy to knock a sample down without choppiness if you'd like.

See ya
Steve

 

[chesebert]

I asked a friend with Pro Tools HD (he is in the music making business) to downsample couple of my 24bit files (I had to burn several on a DVD for him). It took a couple days for him to get back to me but I am glad to know that even with industry-strength processing and professional TLC, the 16/44 is still inferior to 24/96 and it sounded no different than my own 16/44 conversion that I made with one of many pro-sumer audio editing softwares available on the market today

 

[IPodPJ]

I didn't use anything special. One way I tried it was using the Secret Rabbit Code resampler for Foobar. Another way was to convert the files in Audacity for Mac. I don't use dithering in Foobar. When the option is turned on, you can hear the noise it generates.

The difference between native audio recorded in 24/96 and redbook CD isn't hard to hear. What's harder to hear is the difference between 96 kHz upsampled redbook audio and non-upsampled redbook audio. With a good transport and DAC, it shouldn't sound very different at all. But sometimes upsampling helps to eliminate a lot of the fatigue associated with 44.1 kHz audio; other times it kills the clarity.

 

[bigshot]

If you downsample from 96 to 44.1 without dither, it's going to sound bad. I bet that's the problem. I don't know if Peak will open a flac file. If it does, I'd be happy to downsample a couple of minutes properly for you and send it back so you can hear what it should sound like.

It sounds like your experience with upsampling might be caused by different types of dithering too. There shouldn't be any difference at all between redbook and upsampled redbook.

A good way to test the quality of dithering is to up and downsample the same track over and over to see how many generations you can go without hearing a change. With a pro grade audio program it should be able to go more times than you have the patience for with no real degradation.

See ya
Steve

 

[Seidhepriest]

Some relevant information:

Digital Sampling & Fidelity Issues

Some points: CD Audio's top frequency is 22050 Hz (nominal), at which frequency there's no difference between saw, sine, square waves - it's all "sine", end-to-end, only top and bottom sample coordinates.

Nyquist rate ought to be three times the audio frequency to provide enough detail for wave drawing. 96000/3=48000. 44100/3=14700. CD Audio mangles overtones, which is what violin (and brass/woodwind) players notice.

That a human cannot hear 35 KHz tones doesn't mean the ADC and/or anything in the filter chain can't. Lowpass filters can (depending on construction) cut off the actual peaks of waves, turning sine waves into saw/square-like.

Again, this is something musicians and analogue listeners will notice more readily.

Finally, the 44100 sampling rate is really stressing on the analogue filter stage of a DAC, even more so with the typically companded modern CDs with an average loudness of -20 - -12 dB (Oasis having set a "record for a record" with an average loudness of -12 dB on "What's The Story"). 96000 voltage steps is a lot smoother than 44100, and that's what oversampling helps with when playing CDs, too.

Bigshot is making the classic mistake of mixing up dynamic range and loudness. Dynamic range in digital audio defines the accuracy of reproduction; as a simple example, a Cowon D2 player with 95-dB signal-to-signal-noise ratio will sound crisper and more detailed, with a much better presence, than a portable CD player with an 80-dB SNR. And, it doesn't have to be driven to full 95 dB loudness (actually, it could do in excess of 100 dB with the right headphones/amp - but still with a definition of 95 dB, just like a noise gate at -50 dB actually "thins" out the sound even if it's applied to a recording that'll have the full 96 dB range, silence-to-peak) for that detail to come out - on the contrary, it is that detailedness that makes listening at lower volumes possible. Headphone amplifiers also increase dynamic range, bringing out detail, in the same way.

Digital audio can be a very confusing subject as it requires concentration on the exact procedures used to achieve an end, and noticing (not ignoring, of course, trying to live in an illusion of "bliss") the problems with certain methods. What makes many people give up is the lack of correlation between perception and logic. It is there, it just has to be examined thoroughly by perception of processes as well as noticing of the actual "warmth" of sounding, the transmission of musical energy.

 

[Leporello]

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif Nyquist rate ought to be three times the audio frequency to provide enough detail for wave drawing. 96000/3=48000. 44100/3=14700. CD Audio mangles overtones, which is what violin (and brass/woodwind) players notice.

They do? Evidence, references?

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif Again, this is something musicians and analogue listeners will notice more readily.

They will? Evidence, references?

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif Finally, the 44100 sampling rate is really stressing on the analogue filter stage of a DAC, even more so with the typically companded modern CDs with an average loudness of -20 - -12 dB (Oasis having set a "record for a record" with an average loudness of -12 dB on "What's The Story"). 96000 voltage steps is a lot smoother than 44100, and that's what oversampling helps with when playing CDs, too.

What has sampling rate to do with compressed recordings? What do you mean by "96000 voltage levels"?

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif Digital audio can be a very confusing subject

Quite obviously.

Regards,

L.

 

[hciman77]

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif Some relevant information: Nyquist rate ought to be three times the audio frequency to provide enough detail for wave drawing. 96000/3=48000. 44100/3=14700. CD Audio mangles overtones, which is what violin (and brass/woodwind) players notice.

Can you please provide a serious citation for this assertion, since this is such a big piece of scientific iconoclasm (A scientific revolutuion in Kuhnian sense) I would prefer at least a peer reviewed journal paper or respected conference paper.

A sampling rate of 44,100 will adequately capture everything up to 22050, any thing beyond that is lost, not many people can hear above 22K, for those that can sorry, but there are no musical fundamentals there, the highest note on a Piano is about 4K so we are talking 5th harmonics. These are all at a pretty low level .


Quote:

Finally, the 44100 sampling rate is really stressing on the analogue filter stage of a DAC

I am sorry can you please explain this. How does this stress reveal itself technically.

Quote:

Bigshot is making the classic mistake of mixing up dynamic range and loudness. Dynamic range in digital audio defines the accuracy of reproduction;

Dynamic range is simply the number of different levels that the system can render. If the signal has a dynamic range of 96db then a 16 bit system will be pertfectly adequate for playing it back accurately. It is only when you exceed this limit that you hit problems.

Quote:

as a simple example, a Cowon D2 player with 95-dB signal-to-signal-noise ratio will sound crisper and more detailed, with a much better presence, than a portable CD player with an 80-dB SNR.

The 95db system will be more accurate, but why would you even buy a portable CD player that was that bad ? Even my Creative Zen does much better than that.

Quote:

Headphone amplifiers also increase dynamic range, bringing out detail, in the same way.

This is wrong surely. A normal amplifier applies a level of gain it does not increase the dynamic range as such.

Say an amp applies a 10db gain across the board, an amp with a nominally flat FR will do this. Say a 16 bit device can output signals from 0db to
-96db , a signal that goes in at 0db becomes amplified to
10db a signal that goes in at -96db db becomes -86db The amp will output from 10db to - 86db a dynamic range of 96db. The quieter sounds get louder and the loud sounds get louder but the dynamic range is unchanged.

You could EQ the sound to artificially boost bass frequencies by an extra 10db so in absolute terms you could get from +20db (bass ) to -96(treble) but even here the dynamic range on the bass frequencies would be exactly the same just 20db louder.

And there are dynamic range expanders, but these are a different beast entirely.

 

[hciman77]

Quote:

Originally Posted by Leporello /img/forum/go_quote.gif What has sampling rate to do with compressed recordings? What do you mean by "96000 voltage levels"?

That is a confusion between sampling and bit-depth. A 16 bit system can render discrete voltage levels of 0 - 65535 (65536 levels or 2 ^ 16) a 24 bit system would render 16777216 voltage levels. He is confusing sampling rates and bit-depths. Since we are dealing with powers of 2 you cant get 96,000 with any integer exponent.

 

[Seidhepriest]

Quote:

Originally Posted by hciman77 Can you please provide a serious citation for this assertion, since this is such a big piece of scientific iconoclasm (A scientific revolutuion in Kuhnian sense) I would prefer at least a peer reviewed journal paper or respected conference paper. A sampling rate of 44,100 will adequately capture everything up to 22050, any thing beyond that is lost, not many people can hear above 22K, for those that can sorry, but there are no musical fundamentals there, the highest note on a Piano is about 4K so we are talking 5th harmonics. These are all at a pretty low level .

That humans cannot hear well above 22050 does not mean gear can't.

That humans cannot discern well above 20 KHz also doesn't mean they can't actually "hear" and be affected by such frequencies.

Here's an article on tests of how hypersonic audio affects perception:

Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect -- Oohashi et al. 83 (6): 3548 -- Journal of Neurophysiology

A good sensitive microphone can easily pick up frequencies above the 22050 Hz.

Anything above 22050 gets "sawed off" by the ADC stage filters; more or less tolerable if it's a windowing filter, quite harsh if it isn't. In either case,

In reality, the practical range of CD audio is ~20000 Hz (the remaining 2 KHz are typically used up by noise shaping, and there's not much definition in the 20000-22050 Hz anyway; 22050 Hz already leaves two possible values - "up" and "down", disregarding actual shape of the waveform.

The 44100 Hz sampling rate, by the way, was adopted by the number of scanlines in VHS tape used for early CD master transfer.

Quote:

Originally Posted by hciman77 I am sorry can you please explain this. How does this stress reveal itself technically.

See Digital Sampling and Fidelity Issues - DSPWiki

Quote:

Originally Posted by hciman77 Dynamic range is simply the number of different levels that the system can render. If the signal has a dynamic range of 96db then a 16-bit system will be perfectly adequate for playing it back accurately. It is only when you exceed this limit that you hit problems.

It won't be "perfectly adequate". Real dynamic range of 44.1/16 is ~80 dB. Anything outside of that is already severely limited. Try creating a 16-bit 44100 Hz wave file in any editor (Cooledit/Adobe Audition, Soundforge, etc.) and see how low the editor will even allow a test A440 sine to go. It won't be capable of generating anything meaningful below -80 dB. It will generate at -96 just fine in 32-bit though.

This is why, by the way, the better-sounding CD players have oversampling and 20-bit+ DACs.

Quote:

Originally Posted by hciman77 The 95db system will be more accurate, but why would you even buy a portable CD player that was that bad? Even my Creative Zen does much better than that.

Many CD players have an SNR of 84-88 dB. Apple IPods have an SNR of 90 dB. Cowon players have an SNR of 95 dB, and it shows instantly. The clarity of an expanded dynamic range is quite obvious. This means, by the way, that the internal DA resolution of such a player is better than 16-bit (a 16-bit device cannot produce a listenable 96-dB dynamic range).

Quote:

Originally Posted by hciman77 This is wrong surely. A normal amplifier applies a level of gain it does not increase the dynamic range as such.

The effect an amplifier has for headphones, supplying enough power to properly drive them, bringing out detail and presence, is similar to the effect 32-bit audio has compared against 16-bit.

 

[Roam]

Quote:

Originally Posted by hciman77 /img/forum/go_quote.gif A sampling rate of 44,100 will adequately capture everything up to 22050, any thing beyond that is lost, not many people can hear above 22K, for those that can sorry, but there are no musical fundamentals there, the highest note on a Piano is about 4K so we are talking 5th harmonics. These are all at a pretty low level .

Nyquist's theorem gives the sampling rate required to capture information of a given bandwidth onto the recording media. This is the famous sampling rate equals twice the highest frequency axiom which we're all familiar with.

Unfortunately, the Nyquist theorem which people endlessly quote without understanding tells you nothing about the requirements of converting the data on the media back into an analogue waveform, that part's covered by Claude Shannon's contribution to the Nyquist-Shannon sampling theorem. It states that at 22050Hz in the case of redbook CD, an infinite number of samples are required to accurately reproduce the waveform, which of course is clearly impossible. In other words, as the frequency of the waveform approaches half the sampling frequency, D/A conversion accuracy goes down the crapper.

 

[Publius]

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif That humans cannot hear well above 22050 does not mean gear can't. That humans cannot discern well above 20 KHz also doesn't mean they can't actually "hear" and be affected by such frequencies.

Um. no. If you can't "discern" a sound, I would figure you couldn't "hear" it too? Or do you have any actual evidence that suggests that such sophistry is required?

Quote:

Here's an article on tests of how hypersonic audio affects perception: Inaudible High-Frequency Sounds Affect Brain Activity: Hypersonic Effect -- Oohashi et al. 83 (6): 3548 -- Journal of Neurophysiology

That paper is full of crap and any examination of its mention on audio forums will point out its flaws. It is absolutely meaningless to conclude any audibility of ultrasonics from it.

Quote:

It won't be "perfectly adequate". Real dynamic range of 44.1/16 is ~80 dB. Anything outside of that is already severely limited. Try creating a 16-bit 44100 Hz wave file in any editor (Cooledit/Adobe Audition, Soundforge, etc.) and see how low the editor will even allow a test A440 sine to go. It won't be capable of generating anything meaningful below -80 dB. It will generate at -96 just fine in 32-bit though.

That's evidence more that you can't do the math, and that you can't analyze a signal correctly, more than any actual limitation of the format.

If you create the sine wave in floating point, and do a proper noise shaped dither - hell even a rectangular dither might be enough - you should be able to hit -100db without breaking a sweat.

If you play that back amplified, the sine wave should come through loud and clear. It probably won't show up on the waveform plot, but any decent FFT should pick it out.

Hint: Don't use Audacity. Compose the wav in a math package and use foobar to do the noise shaping and 16-bit conversion.

 

[bigshot]

x

 

[bigshot]

Quote:

Originally Posted by Seidhepriest /img/forum/go_quote.gif That humans cannot hear well above 22050 does not mean gear can't.

Do you mind if I put that quote in my sig file? It makes me chuckle.

See ya
Steve

P.S. I'm going to skip answering all that stuff about how terrible it is that the things we can't hear are all chopped off. You boys seem to have that well in hand.