ok..i guess that ends it.

Quote:



Not quite, the story goes both ways.

The Japanese researchers published 2 papers (in 2000 and 2006) than another paper was published by another group later (2007) to show that the result in previous two research can not be reproduced - However, all three research employed speakers (air conduction) as their primary testing equipment to see if ultrasonic will affect music perception / enjoyment, the so called 'hypersonic effect'. However, none of the three researches disputes that human can hear ultrasonic under certain conditions, such as when the ultrasound is fed directly to part of the skull (bone conduction) to bypass middle ear (which believed to serve as a low pass filter in hearing). In fact, there are papers dated back more than 50 yrs or so on human ultrasound hearing. More recently, there are also research and proposals published on using bone-conducting ultrasound as treatment for tinnitus and speech perception (for the hearing impaired / deaf). Given that bone conduction is far more likely to happen with headphone usage (as sound transmits way more efficient in denser media (like plastic, metal, water, and bone) than in lighter media (like air)), there is too early to tell whether ultrasound doesn't play a role in headphone music reproduction as well.

Quote:



Yes, of course - that is true for commercial CD. However, my 'almost' covers 'all CD, lossy and lossless music', and there are lossless music produced these day that has 48kHz or even 96kHz sampling rate. Depends on how you rip your music, those lossless can be transcoded to lossy with higher than 44.1kHz sampling rate. This is especially true to music produced independently and published solely on the internet. Beside all these, there is also the CD compatible but least-noticing HDCD format owned by Microsoft, which can encode 24bits worth of data (96kHz) into 16bits CD track (44.1khz) while still playable in normal CD player.

Quote:



Yes, again you are right. However, as I have pointed out, this guide is only meant for 'IEM user in portable setting' - and I mean IEM user with a portable DAP. Most DAP on the market only have 16bits DAC chip that are very inefficient in oversampling, and many only do well up to 48kHz (some lower models only do well up to 44.1kHz). I am not talking about whether 16bits is enough for encoding, but whether the hardware in your DAP is efficient on handling music that is over 16bits / 44.1kHz, like 24bits / 96kHz or 24bits / 192kHz for instant. I hope that clear things up a bit.

Quote:



Then there is no point in upsampling and higher bit rate either then because the reason for higher sampling rates is to reproduce frequencies that are also beyond what we can hear. 44,1khz 16bit already contians audible frequency range so what is the point in using 96khz 24bit?

Quote:



If you refer to Kotelnikov theorem (for some reasons also called Nyquist-Shannon theorem ) you'll see that 44.1KHz would be enough if the signal didn't contain any frequency component higher or equal to 22.05 KHz. The problem is that signal does contain such frequency components, so they should be filtered out. If you're limited to 44.1KHz and want to reproduce frequency components of up to 20 KHz you'll need an analog filter with very steep slope that's very hard to achieve. The worst thing about it is that such filter would add lots of phase distortions. The only solution is to use upsampling, downsampling and digital filters.

Quote:



All true, but even 20K is not really often necessary for music. If you apply low pass filters to music files you find that you can be really quite draconian without making an audible difference, few MP3 files have much above 16K yet can be near impossible to tell from uncompressed music. I did some experments myself and with some music samples a 13K lowpass applied to wav files was undetectable (in blind tests) even allowing for my aged ears ( nothing above 16K these days) this does sugest that perhaps we get hung up on high frequencies. Analog FM stereo for instance has nothing above 15K.

As for phase distortion I have not found any papers that suggest that the levels of phase distortion found in commercial kit are audible and in Meyer and Moran's paper they used a 16/44.1 AD/DA loop with no detectable difference.

The bigger margin you have the easier the filter would be. 16K vs. 22K is less than octave. With 8x oversampling (the minimum we have in contemporary DACs) there are 3 octaves.

Why do some people say no upsampling sounds better and set their media player to 44.1 16bit if upsampling is supposed to sound better?

Quote:



Because we don't have the same gear and set of ears I presume.
Hence why audio quality is subjective.

ClieOS - You are working under a number of misconseptions. First of all the question is not about what frequencies are percievable. If I play you a 50kHz signal @ 200dB you won't be able to hear it, you will be able to percieve it though, just before you drop dead with a fried brain!!

The vast majority of recording equipment cannot reliably record anything much over 22kHz and even if it could, 99% of what you would record above 22kHz would be noise. So perceiving it isn't the problem, it's decerning the 1% of usable sound which may exist behind the noise. Is this starting to sound silly to anyone yet? Remember also, that as no one can hear above 22k the recording engineers and producers can't mix what they can't hear!!

"there are lossless music produced these day that has 48kHz or even 96kHz sampling rate. Depends on how you rip your music, those lossless can be transcoded to lossy with higher than 44.1kHz sampling rate. This is especially true to music produced independently and published solely on the internet."

There are some potential benefits of recording at 96kF/s sample rate, namely making it cheaper to implement a decent sounding digital anti-alias filter, not because there are any usable frequencies above 22kHz. It is possible therefore under certain situations that there may be a slight perceivable improvement if the material was recorded at distributed at 96k but none if any part of the chain has a 44.1k conversion. 44.1kF/s already has all the freqs above 22kHz removed, so upsampling is a waste of time. Generally, internet only releases are made by project studios which cannot afford really high quality 16bit/44.1k converters, so a 96k converter may provided a very slight improvement on original 96k material. However, a 16bit/44.1k recording from a high quality studio is still going to sound better quality. Bare in mind I'm only talking about sample rates here, bit depth makes no difference whatsoever. In a 16bit recording you are hearing an absolute maximim of about 10bits of music, the other 6 bits (at least) are just noise, with 24bit therefore you still only get the 10bits max of music but you now have 14bits of noise. In answer to what is the difference between 16bit and 24bit the answer is absolutely nothing except an extra 8bits of noise!

"Then there is no point in upsampling and higher bit rate either then because the reason for higher sampling rates is to reproduce frequencies that are also beyond what we can hear. 44,1khz 16bit already contians audible frequency range so what is the point in using 96khz 24bit?"

Absolutely, you've got it! There is absolutely no reason or benefit whatsoever, except to those who sell and market 24bit/96k consumer equipment!!

Shoereck "The only solution is to use upsampling, downsampling and digital filters."

Which is exactly what has been done in all professional ADCs for at least a decade and many ADCs have been doing this for more than 15 years!

G

Quote:



Cough Balinese Gamelan cough .....

But not just that cymbals and trumpets go way above 22khz and still above noise floors, you cannot hear it though but it is there.

Strangely enough, I know Balinese Gamelan pretty well. My ex-wife owned a gamelan orchestra and I mixed a live broadcast from Jakarta of one of the world's top gamelan orchestras several years ago, even included it in the odd film score.

I'm not saying there is nothing there above 22kHz, just that mic's and speakers aren't very good at capturing/playing it back. Some mics claim 40kHz as their limit but most are much lower and look at the roll off on their freq response charts. Beyond 22kHz the vast majority of what the mics do pick up is blown away by HF noise from the circuitry. Also none of what it picks up can be cleaned or mixed because the recording engineers and producers can't hear it. Lastly, having higher sample rates does not greatly improve the frequency response. Higher sample rates largely just give more space for smoother anti-alias filters so whatever is up there, is:

A. Likely to be almost entirely noise.
B. Largely filtered out by anti-alias filters and
C. Is completely inaudible anyway!!

G

Quote:



As has already been said, the more extended (...into infra- and ultrasound) the frequency response, the more even the frequency response in the audible range. High- and low-pass filters such as the natural drop-offs shown by sound transducers don't have infinite steepness and ultra-sharp slopes, so every bandwidth restriction comes with a quite large adjacent area suffering from nonlinearity.

And let's not forget that every filter introduces transient distortion -- the steeper and the sharper the slope, the more so; and the closer to the usable audio band, the stronger the impact.

As to the CD cutting off ultra-high frequencies: That's exactly a case of a steep and sharp-edged low-pass filter as described above.



Quote:



That's not exactly an adequate test: The sample used has already passed an extremely sharp low-pass filter. What you do is switching another (most likely smoother) low-pass filter in series. Now take into account that there are some CD Players with switchable filter characteristics, some of them largely corresponding to your filter -- with the goal of reducing the HF ringing (Gibbs phenomenon) introduced by the anti-aliasing filter -- and apparently some success in this regard.

I have tried a similar filter, the Meier «Analoguer» circuit, in my headphone system. The sound change it produced with CDs wasn't glaring, but an absolutely enjoyable smoother sonic variant with a tad less sparkle and harshness -- all in all neither better nor worse. With SACDs, OTOH, it blatantly annihilated their sonic advantage over redbook CD in the form of higher high-frequency resolution, detail and airiness. For me this test result is a clear hint that a sharp high-pass filter at ~21 kHz is bad for preserving detail and transient accuracy within the audio band.


Quote:



Not true! I have always wondered why FM stereo didn't exhibit the often cited (and to my ears existing) «digititis» of the CD format, particularly the first CD player generations. And I realized what the fundamental difference was: Apparently the radio sound is much less clean, it has lots of distortion, even at the best reception conditions. And distortion is exactly what prevents the FM radio's bandwidth limitation from sounding sterile -- in contrast to the CD format. In the latter case all overtones are highly affected by the filter ringing -- they lack accuracy and sharpness, are heavily smeared around the time axis --, whereas with FM radio «time smearing» is less of an issue, firstly because the filter is less sharp, secondly because beside smeared overtones there's still a considerable amount of unsmeared overtones in the form of harmonic-distortion products (up to 20 kHz and beyond!) -- making for a more organic sound. The same applies to some tube output stages with some CD players and DACs.
.

Jazz "As has already been said, the more extended (...into infra- and ultrasound) the frequency response, the more even the frequency response in the audible range."

What you are asserting here may or may not be true. Just because cans have an extended range does not necessarily mean they are going to have an even frequency response in the audible range.

"In the latter case all overtones are highly affected by the filter ringing".

This is obviously incorrect. Lets say we have a note with a 200Hz fundamental, the first overtone is at 400Hz, so you are saying that this 400Hz overtone is going to be highly affected by filter ringing from a filter set at 22kHz?! Let's not forget that the vast majority of what is recordable above 20kHz is just noise anyway and distorted noise is .... um .... noise!

Whereas your assertion regarding steep filters is true as far as smearing and transient distortion in the analogue domain, it's much less true with modern filtering algorithms in the digital domain. According to your post, the solution for transient distortion is to add tube distortion and that all this distortion makes the sound in some way "organic"?!

How old are you Jazz? If you are older than your teens, it's unlikely you can hear much above 18kHz and certainly any experienced producer isn't likely to hear much above 16kHz, so the frequencies affected are beyond any mixing capabilities. Lets not forget that even with excellent hearing the human ear can't even tell the difference between a 12kHz square wave and a 12kHz sine wave. So, I re-assert what I said before.

BTW, why do you think SACD is better than red book CD? I'm not saying it's not better, I'm just interested in why you think it sounds better.

G

Quote:



Of course not. But without other variables and in the sense of the thread topic my above statement holds true.


Quote:



You are correct. But don't take «all overtones» literally; I meant all high frequencies above ~8 kHz (without exception) -- which merely consist of overtones.


Quote:



Obviously you don't get my point. I wasn't contemplating ultrasonics at all, just the effect of filters on the audible range.


Quote:



That's not correct. Digital filters cause the same ringing -- they just favor symmetrical ringing (including pre-ringing) instead of pure post-ringing. The only way to minimize the ringing to almost zero is a Spline filter à la Wadia -- at the expense of a -3.5 dB drop-off at 20 kHz (doesn't matter if purely analogue or digital + analogue filtering). But then again, there's still the anti-aliasing filtering from the ADC in the recording studio in the signal path...


Quote:



It's not a solution, just a cure helping to hide the negative effects -- with certain not so welcome side-effects. Please read my statement again; I'm not sure if you've really understood it. In my understanding the filter ringing is mainly responsible for the «digital» coloration attributed to redbook CD in that it smears all overtones (!), and there's no distortion which would help to hide the smearing. FM radio on the other hand doesn't show a similar coloration, despite the steep low-pass filter it also incorporates. The reason is explained in my previous post.


Quote:



I won't say it's better in every way, but it's certainly better with high-frequency transients, making for an airier sound and much better definition. Overtones sound more real, less smeared. In comparison CD overtones sound glassy and like wrapped in a hazy halo. -- This advantage only holds true with SACDs produced from a high-rez master; otherwise the DSD layer often sounds worse than the redbook layer. Which shows that DSD isn't perfect either.
.

Jazz - I've never seen or heard any evidence that freqs of 8kHz are negatively affected by the anti-alias filter.

Generally I've found SACD to sound better than CD. However, I don't necessarily believe it's anything to do with the format. DSD recordings are generally made in higher quality studios (due to cost) and the products made in DSD format specifically target a more audiophile demographic. This is why they sound better in general. However, I believe it is possible to create just as high a quality using 16bit/44.1k.

"the often cited (and to my ears existing) «digititis» of the CD format, particularly the first CD player generations."

Agreed. But ADCs have come a long way in nearly 30 years. At this point I've come full circle. The audio quality available from 16bit/44.1k can be as high as any other format on the market. It requires top class ADCs (and a top class signal chain in general) and engineers and producers who really know what they are doing.

G