nelsonvandal said:
This can start a new battle. What sounds closest to a sinewave, a squarewave or a sawtooth.
But at what frequency ? At what frequency are your ears unable to differentiate between sine/square and triangle. Not talking about volume, as you would have to correct for form factor, I mean timbre
The last time I checked my HF hearing range it cut off between 14kHz and 15kHz. I can't repeat that just now because my ears are blocked up with a cold/cough.
I can tell you that I could easily hear the difference between sine/square and sine/sawtooth and square/sawtooth at 10kHz, even though the first harmonic is at 20kHz.
Yes, changes in timbre can be distinguished when the components comprising the difference are well above the hearing range of the listener.
That is what is so wrong with the CD standard. It cannot reproduce the HF components correctly.
I wonder if anyone can post the output of a 16/44.1 DAC that is trying to reproduce HF sines, squares and sawtooths?
I can tell you that I could easily hear the difference between sine/square and sine/sawtooth and square/sawtooth at 10kHz, even though the first harmonic is at 20kHz.
Yes, changes in timbre can be distinguished when the components comprising the difference are well above the hearing range of the listener.
That is what is so wrong with the CD standard. It cannot reproduce the HF components correctly.
I wonder if anyone can post the output of a 16/44.1 DAC that is trying to reproduce HF sines, squares and sawtooths?
Interesting. Using some cheap headphones at really quite a low volume I can hear to around 13+ khz before any really noticeable roll off occurs, beyond that point the roll of is fairly steep and I can just hear 16 khz on a good day, but the level ( voltage out of the generator ) has to be around 25 to 30 db higher than the 13k point. Switching between sine and square, or just listening to square and increasing the frequency, again at low volume all trace of the harmonics seems to go at around 4.7 khz to me.
So where's the proof?
There's not one single manufacturer of hi-speed digital audio that has given any proof for the superiority of high speed PCM. Only words in marketing texts.
That some converters works better at this or that sampling speed is another discussion though. I have for example participated in blind tests where a particular converter sounded cleaner at 192kS/s than at 44.1kS/s but that was with allready bandwith limited material so there was no HF to be captured or reproduced.
Guesswork, wrong conclusions and dogma seems to be common in audio.
/Peter
There's not one single manufacturer of hi-speed digital audio that has given any proof for the superiority of high speed PCM. Only words in marketing texts.
That some converters works better at this or that sampling speed is another discussion though. I have for example participated in blind tests where a particular converter sounded cleaner at 192kS/s than at 44.1kS/s but that was with allready bandwith limited material so there was no HF to be captured or reproduced.
Guesswork, wrong conclusions and dogma seems to be common in audio.
/Peter
you stated "there is no proof". Now prove what you said is true.Pan said:
What's your definition of "perfectly"?
care to divulge the distortion and scope pics of 20kHz & 21kHz sinewaves from a 16/44.1DAC when played at -1dBfs, -20dBfs, -40dBfs and -60dBfs. There's little point in asking to see the -80dBfs and -100dBfs data.
AndrewT said:
Perfectly as regards to perfect waveform to the eye and distortion below audibility. Perfect as you get out what you put in except for frequencies above the nyquist frequency or thereabout.
The noise floor and dynamic range has not much to do with frequency of sampling. The restrictions in S/N is from the bit wordlength of the system. And yes, a -60dBFS sine at ANY frequency in a 16bit PCM system will have clearly visible noise on the scope and feel free to look at analog tape at same level. ;-)
Thing is that at typciall SPL settings the 16bit noise is no problem. Only with dynamic material played back at high SPL the noise become an issue.
I can't proove this to you, all I can do is give you info about listening tests I have done and possibly also point you to published listening tests which have prooven 16/44.1 to be practically transparent in real world listening situations. Ivo Tiefenbrun was very sure about 16/44.1 PCM would destroy the sound of good analog but failed to hear a difference between the direct sound fro analog and the same material that had been fed to a AD-DA chain.
Several tests like this has been published. OTOH there are not one single test that have prooven that we need anything above 20k for faithful reproduction of music. Well, of course there can be such tests out there somewhere but I haven't been able to find one myself.
/Peter
Samplerates
Hello all.
Just to inform I know of one good reason to have high samplerates.
In composing, if you're using MIDI then you usually take a few recordings of an instrument at different notes, and slow or speed up the playback of those notes to make different notes.
This is known as resampling. If you take an instrument with a 22050KHz samplerate and slow it down, it will sound really horrible because the "effective" sample rate has gone way down, whereas if you have a recording with 196KHz samplerate, you can slow it down 5 times and it will still sound fine to the human ear.
- keantoken
Hello all.
Just to inform I know of one good reason to have high samplerates.
In composing, if you're using MIDI then you usually take a few recordings of an instrument at different notes, and slow or speed up the playback of those notes to make different notes.
This is known as resampling. If you take an instrument with a 22050KHz samplerate and slow it down, it will sound really horrible because the "effective" sample rate has gone way down, whereas if you have a recording with 196KHz samplerate, you can slow it down 5 times and it will still sound fine to the human ear.
- keantoken
Guys,
what you bring up above is the most common misunderstanding about digital audio.
When you look at a 20k sine in an audio editor you see a digital representation.. sample points with lines drawn between them. This is NOT how a sine looks at the output of a DAC and it's reconstruction filter. It's true that a 20k sine is represented by a few samples in 44.1kS/s PCM audio but that is also all it takes to convert the data to analog waveform again. The sharp angles you see in some editors is a sign of much higher frequencies than 20k but these are filtered out in the DAC leaving only the 20k info left.
There are however some audio software that show the waveform as it will look on a scope after it has been converted back to analog.
Andrew, your CDP must have been defect or you made some error measuring.
Spend 5-10 minutes reading up on the Shannon-Nyquist theorem and you will get a better picture of digital audio.
/Peter
what you bring up above is the most common misunderstanding about digital audio.
When you look at a 20k sine in an audio editor you see a digital representation.. sample points with lines drawn between them. This is NOT how a sine looks at the output of a DAC and it's reconstruction filter. It's true that a 20k sine is represented by a few samples in 44.1kS/s PCM audio but that is also all it takes to convert the data to analog waveform again. The sharp angles you see in some editors is a sign of much higher frequencies than 20k but these are filtered out in the DAC leaving only the 20k info left.
There are however some audio software that show the waveform as it will look on a scope after it has been converted back to analog.
Andrew, your CDP must have been defect or you made some error measuring.
Spend 5-10 minutes reading up on the Shannon-Nyquist theorem and you will get a better picture of digital audio.
/Peter
Mike,
instead of linearity, I would like to use the term tonal balance here. Our approaches do result in different tonal balance and are therefore perceptually different. One of them is simply inferior.
I don`t know what "perception level" means, but it has probably no relevance whatsoever. Even if nothing of a considerable harmonic distortion is hearable, plays a crucial role in the tonal balance. All harmonics, including those far outside the audible range effect the audible range, being a structural part of (or if missing, by impairing) the tonal balance.
instead of linearity, I would like to use the term tonal balance here. Our approaches do result in different tonal balance and are therefore perceptually different. One of them is simply inferior.
I don`t know what "perception level" means, but it has probably no relevance whatsoever. Even if nothing of a considerable harmonic distortion is hearable, plays a crucial role in the tonal balance. All harmonics, including those far outside the audible range effect the audible range, being a structural part of (or if missing, by impairing) the tonal balance.
Lumba Ogir said:
HiFi gear is not about "tonal balance" it's about not getting in the way of the recorded signal and the instruments true tonal balance. For that you need gear that imposes no signature.. transparent gear. It's not criminal to use or like gear that colors the signal but it has nothing to do with HiFi.
Yes from a HiFi perspective the coloring device is inferior.
No. We have limitations in what we can hear and perceive. Signals of low enough amplitude can not be heard and signals with high enough frequency can not be heard.
There are tons of evidence for this but the things you believe has absulotely nothing in the form of back up.. zero.
/Peter
Andre Visser said:
keantoken said:
As Pan mentions, you need to study sampling theory. I spent 3 days on a Sony course in the early years of CD, and we pulled all this theory apart.
A 20khz sinewave can be perfectly reconstructed from a 44.1k sample, in theory the sample needs only to be 2*F. Hard to believe when you see the "stairstep" presented to the LPF for reconstruction but true.
Andre says the output of his player was visibly and severely distorted at 20khz, but doesn't mention the level involved. Non linearity of D/A converters at lower levels is well known, and if he were testing at -90 db or so I am not surprised at the findings.
This has nothing to do with sampling theory though, it's a different issue. Related to this is "dither" that was introduced to try and minimize this effect.
As to how audible all of this is, well you have only to see a 1 khz square wave off CD to realise we have some serious limitations. On the other hand, given sypathetic equipment, CD can be very, very good indeed, but then we are going in circles again, coloured sound, accurate sound, distortion spectra
I have nothing against 44.1Khz CD sound, certain well recorded CD's prove that it can sound great. I do believe that higher resolution sound better though.
Mooly, I've used a test CD with a sine sweep from 20Hz to 20Khz, measuring on the OP of the CDP. I would say the level was quite high because it compared quite well with normal CD listening levels on the system. The CDP wasn't one of the cheap players but dating from the mid 90's, I guess the newer DAC's would be much better.
I believe it is important to be able to reproduce frequencies higher than 20Khz because there are harmonics on musical tones which will change the shape of the original signal if they are lost.
Mooly, I've used a test CD with a sine sweep from 20Hz to 20Khz, measuring on the OP of the CDP. I would say the level was quite high because it compared quite well with normal CD listening levels on the system. The CDP wasn't one of the cheap players but dating from the mid 90's, I guess the newer DAC's would be much better.
I believe it is important to be able to reproduce frequencies higher than 20Khz because there are harmonics on musical tones which will change the shape of the original signal if they are lost.
Andre Visser said:
Is that beliefe grounded on actual listening tests or from the way you have been thinking on sampling theory? Wouldn't it be cool if you knew that higher resolution was equal to higher fideliy instead of believe? If you find some evidence please tell me about it. I thought the same as you.. until I started looking for proof and I found none.
That said I have a 192kS/s AD-DA converter I use for music recording and it actually has better performance at the highest sample rate, something that couold lead some people to believe that we need +20k info in the reproduced music but as I mentioned earlier this slightly better perofrmance is appearant with bandwith limited material ie a CD as a source which takes ultra high frequencies ouy of the equation.
Exactly. It's easy to make that mistake as well. Investigating one machine which may be a SOTA design or a real piece of garbage tell us nothing about teh format, only about that particular machine.
But we don't hear what a waveform looks like on the scope. A scope can have GHz bandwith but our ears has much more limited bandwith. Also how common is it with wide bandwith microphones? Not very common but sure, they do exist.
Try to put up some tests and see what you learn from it. Use a wide bandwith signal and slap on different LP filters until you hear the difference. Again such tests have been done and there is nothing that indicates that we need much more than 20kS/s for normal music.
Listen to sines, squares and triangular waves at frequencies from a couple of hundred Hz up to 15kHz and see how they sounds more and more similar the higher up you go. Be aware though that nonlinearities in teh playback gear may skew the reults slightly. Same for a el-guitar with overdrive/high gain. As you work yourself up on the neck the buzz dissapear and the notes sound clean. The dist is there on a scope but our ears filter the HF out.
/Peter
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.