Is there an audible difference or are we looking at more marketing specs?

Any blind tests?

Solid engineering answer: "It depends"
You have not given any application details. Modern DSP algorithms do just fine with Nyquest rates, but older analog filtered converters did better with higher rates.
It does no good to have resolution greater than the source. High rates are handy for understanding the ultrasonic breakup and distortion modes of tweeters. Usless when looking at mids. So on and so on.

OK. Let's be more specific. I believe we do have some sound processors in the market that claim to process 196KHz. Recently we are also seeing some blu-ray sources encoded at 196.

Will the sound necessarily be better or will the difference be only a number game since we are limited by our "mortal" ears. I read 92KHz is already in the border of our hearing, but I am not sure either way. So just wanted to see what others felt.

I read 92KHz is already in the border of our hearing, but I am not sure either way.

The absolute best a human can hear is 20KHz.
These sampling rates are multiples of 48KHz....they double to 96KHz and 192KHz.

Don't confuse the processing clock rates with the resulting output coding. Higher processing rates mean they can do more "stuff" while keeping up in real time. The output is likely to remain 48K (pro) or 44 K (consumer). Some of the "super" formats are higher, but one would have to get me to believe the source was higher too.


Only some imaginary female in her 20's can hear to 20K. Most men, before going to Who concerts stoned, never heard over 18K. If you are an iPod/ car stereo genX something, you probably can't hear over 12K.

Don't confuse the processing clock rates with the resulting output coding.

I'm not, the OP is.

It's very unlikely that 192khz will sound better than 96khz with all other things being equal.

Some of the "super" formats are higher, but one would have to get me to believe the source was higher too.

Then what is this?

http://www.electronichouse.com/article/chesky_releases_192khz_24bit_audio_titles/

Then what is this?

http://www.electronichouse.com/article/chesky_releases_192khz_24bit_audio_titles/

Someone trying to sell recordings by means of advertising large numbers? ;-)

Is there an audible difference or are we looking at more marketing specs?


I think you meant 96 KHz versus 192 KHz.

Those correspond to highest possible recorded frequencies of 48 and 96 Khz.


Any blind tests?

In a way. The current battle is over 44.1 KHz versus anything higher. The jury may be still out about that, but they've already come back with "no difference" any number of times.

Someone trying to sell recordings by means of advertising large numbers? ;-)

So you think the record company is outright lying?

I think you meant 96 KHz versus 192 KHz.

Yes sorry for the confusion. I was trying to type fast and they all sound similar :)

http://www.2l.no/epost/news2008may.html

What are you asking?

What are you asking?

My actual question was if there was any audible difference between 192KhZ and 96KhZ. I wasn't sure if anything beyond 96KhZ was beyond human perception or not.

Then both arnyk and tvrgeek said that there were no 196KhZ sources, so I just gave 2 examples.

Any blind tests?
This (http://www.aes.org/e-lib/browse.cfm?elib=14195) might be of interest to you.

So you think the record company is outright lying?
No, it's engaging on marketing hype. There's a lot of that going around.

Then both arnyk and tvrgeek said that there were no 196KhZ sources, so I just gave 2 examples.

You gave two examples of recordings....there's no mention of the source resolution.

By sources, I meant processed at that resolution from end to end in the consumer audio field. What I do see is a few efforts at 24 bits, 48 or 96K. Not sure that is much of an advantage either as 16 bits gives over 100dB of dynamic range. Can we hear more? maybe through really good headphones. Do we need it? I don't. If I were running a studio, I might consider recording at 24/96 just to be future proof as storage is pretty cheap. Jack Web made a lot of money by filming DRAGNET in color long before anyone had color TV's. He understood reruns.
My Sheffield, direct to disk, split to tape, 2 channel 16/44.8 CD's sound pretty darn good compared to most "digital studio" recordings, so it leaves me to believe the performance and the care matter more than bits.

[QUOTE=mcnarus;16446244]This (http://www.aes.org/e-lib/browse.cfm?elib=14195) might be of interest to you.[QUOTE]

Yes, this is exactly what I was asking. I wonder what they mean with "very high volume"

Women and children may be able to hear a frequency of 20kHz. Most men have a limit of 16kHz to 18kHz.

Technically, high sampling rates (the 96kHz and 192kHz numbers) will make a 20kHz (or 16kHz or 18hHz) sine wave smoother. However, since most pro audio equipment is recording the original at a lower level (24/96 is max for most studios), there is no advantage to increasing it. You'll find very few "sources" that were recorded in digital at 192kHz.

Another factor is bit depth (the 24 number)...

http://www.tweakheadz.com/16_vs_24_bit_audio.htm

Think of it this way:

176.4khz/20bit (88.2khz frequency response, 120 dB of dynamic range) goes beyond what even the best equipment can actually do (speakers, amplifiers and microphones). Anything beyond either of these two thresholds is almost certainly just marketing hype.

Anything below...well that's up for debate I suppose, assuming one's equipment can actually reproduce an amount of it. The blind tests that have been done strongly suggest that if there is any audible improvement, it's likely to be very subtle only. Claims of big or obvious differences all the time can be pretty safely be dismissed, IMO.

One more thing to take into account. While 192kHz is/maybe overkill and mostly marketing hype it costs nothing extra to use over 96kHz. If the source is a 196kHz master then you skip the down sampling step that may/or may not degrade the SQ. It's a fact that 192kHz will not sound worse that 96kHz (or make your popcorn taste stale) so what is the problem with it being used?:confused:

I wonder what they mean with "very high volume"

at very high volume when there is no (or very little) signal you might be able to hear the difference in the background noise if your preamp and amplifier were quiet enough. If your speakers make audible hiss (caused by amp or preamp noise) then probably not. As soon as the music starts you quickly run for the volume knob to preserve your hearing.

Recording at the highest resolution and bitrate possible makes sense since it gives some room for manipulation when mixing.

Addressing the very high levels, ( ones that would damage your hearing anyway) BITS would be more important than sample rate. I see a better excuse for 20 or 24 bits than higher sample rates. This is not a new argument . HP and TEKTRONIX have been arguing about this for 30 years. I forget which was which, but one said ore bits, the other more samples.
So,Nyquest tells us what sample rate we need, but dynamic range tells us how many bits we need.

And marketing says 'I hope they'll buy this new line of shite so's we can jack the prices up once again and sell them more of the stuff they already own that's crappily mastered anyways with more bits and higher sample rates.'

So you think the record company is outright lying?


I'd call it being poorly informed.

posts removed: there is no need to insult other members

As far as I know, the sampling rate (48kHz or 96kHz) has nothing do do with our hearing.
The sampling rate is basically how many samples of the audio has been taken per second, so, the more samples per second, the better quality of the audio.
While the difference is not that noticeable, it does make a difference if you are dealing with professional audio.
Also the file sizes are large for highly sampled audio, so people prefer to use smaller file sizes with highly compressed audio such as mp3.

I think that the DACs in most AVRs nowadays are spec'd to do 192khz if they are fed that material. And, yes, there is 192khz source material. Stereo DVD-A tracks are often 192khz, for example. I do not know what BR disc tracks, stereo or multichannel, might be.

Isn't the purpose of all this to get the best recordings and then transfer them to the best media that are systems can reproduce? But, on top of all that, don't we want the best performance, more so than the best presentation, making the whole process somewhat moot as to what sampling rates we are listening to?

Sampling has (simplistically) two parts: speed, and resolution.

The speed, i.e. number of samples per second, relates to the maximum information content that can be extracted. Higher rate, more bandwidth. Some extra bandwidth is good because real filters have finite roll-off (though high-order digital filters get pretty close to a brick wall response). Beyond a certain point it should not matter, though lower-order filters can be used which are easier to make (and process faster).

Resolution is set by the number of bits (and linearity, and noise floor, and jitter, and...) It relates to dynamic range and so generally the more the better. However, 16 bits gives ideally about 98 dB of dynamic range, far more than is useful for normal listening. More bits allows a lower noise floor, and in the studio provides headroom for mixing multiple tracks.

Both higher sampling rates and higher resolution increase the size of data files.

My hearing testing good to about 21 - 22 kHz in my youth (exceptional) but is around 10 to 12 kHz now.

Some of the biggest issues with early CD systems were the low (44.1 kHz) sampling rate and type of data converters at the time. Without oversampling, and without all the power of modern digital signal processors (DSPs), high-order analog filters were required that caused phase shift and roll-off well down into the upper midrange. Oversampling devices solved that problem for both recording (ADCs) and listeners (DACs) by allowing low-order analog filters followed by high-order (precision, lossless) digital filters. Better processing, techniques, and architectures also did wonders for resolution, not just in the number of bits, but mainly in reducing their nonlinearities (distortion) and reducing the noise floor.

HTH - Don

Just posting my 2 cents here. Seems everyone thinks sampling the frequency at just above hearing threshold has no effect on the frequencies below that. Please check out this link http://en.wikipedia.org/wiki/Spatial_anti-aliasing and pay particular attention to the section "Signal processing approach to anti-aliasing". In order to sample a signal at 44.1KHz, an anti aliasing filter needs to be used (see here http://en.wikipedia.org/wiki/Anti-aliasing_filter). Any audio filter causes phase shift, the greater the cutoff (db per octive), the stronger the phase shift. This is the reason to use the highest sampling frequency you can, to reduce the effects of the anti aliasing filter in the audible range.

Is there an audible difference or are we looking at more marketing specs?


96 KHz audio is all about numbers for the sake of numbers. In your words: its more like marketing specs.

Both DVD-A and SACD failed in the mainstream marketplace. Serious attempts were made to sell them, but they only briefly got off the ground and almost immediately crashed. They may live on in tiny niches, but that is about it. The reason is that there never was a sonic advantage to the new technology that they were based on. The only actual audible good that they may be responsible for is the remastering of some old recordings that had never been "trated right". No technical changes in media format were required for those remastering projects to be effective.


Any blind tests?

Yes, and with no exceptions all reasonable blind tests show that the CD format makes no audible channges on music recorded and played back via it.

When doing listening tests like this, there are subtle problems such as audio gear with excess distortion at ultrasonic frequencies that can lead to false positive results. Also, some of people do a little marketing of their own and try to pass off evaluations with inadequate controls as "blind tests".

JAny audio filter causes phase shift, the greater the cutoff (db per octive), the stronger the phase shift. This is the reason to use the highest sampling frequency you can, to reduce the effects of the anti aliasing filter in the audible range.

This is simply not true in any practical sense. Audio filters can produce a wide range of effects ranging involving frequency response and phase response, but not necessarily both or either. Now that much audio filtering is done in the digital domain, the range of practical options has increased tremendously.

These days virtually every quality anti-aliasing (ADC) and reconstruction(DAC) low pass filter in digital converters is designed to be "Linear Phase". Only very low quality digital audio gear (such as the audio CD playback feature on a CD-ROM drive) is built any other way.

That means that a good modern converter's phase shift is identical to the phase shift that would be produced by a time delay. Since pre-recorded music is by definition already time delayed, the additional time delay they cause is practially moot.

Therefore there is no necessity that a sharp cut-off filter in a music player have any sonically meaningful phase shift at all. This denies the underlying logic behind many claims that sample rates appreciably higher than 44 KHz (CD format) have any sonic benefits at all.

Some of the biggest issues with early CD systems were the low (44.1 kHz) sampling rate and type of data converters at the time.


The effective sample rate of audio CD players has never changed from 44.1 KHz, and that sample rate was never an audible problem.

Oversampling does not change the effective sample rate of a converter. Oversampling needs to be understood as a means to move more of the function of converters into the digital domain, and not any kind of fundamental change in overall performance.



Without oversampling, and without all the power of modern digital signal processors (DSPs), high-order analog filters were required that caused phase shift and roll-off well down into the upper midrange.


The actual problem with analog filters is cost and difficulty in producing them consistently. If you're willing to pay the stiff price very nice analog filters can and were made. The attraction of digital filtering is the ability to do the same thing or better for factions of a penny per dollar spent on analog filters.

Furthermore, phase shift is not the boogey man that some people make it out to be. Great amounts of phase shift can be applied to both channels equally with zero audible effect on sound quality. Of course this is now moot because of the price/erpfrmance of digital filters.

Furthermore, DSPs are not required to build the finest digital filters. Digital filters are basically quite simple and require only the simplest of calculations be used. They are ideal for implementation via inexpensive and simple digital circuitry and have long been effectively implemented that way.



Oversampling devices solved that problem for both recording (ADCs) and listeners (DACs) by allowing low-order analog filters followed by high-order (precision, lossless) digital filters.


Actually, the above is only true for ADCs. In an oversampling DAC, the analog filter *follows* the digital filter.


Better processing, techniques, and architectures also did wonders for resolution, not just in the number of bits, but mainly in reducing their nonlinearities (distortion) and reducing the noise floor.


Only price/performance improvements in ADCs and DACs has benefitted audio. ADCs and DAC implemented via ca. late 1960s technology were sonically transparent. They just cost six figures!

The technology that did the most for ADC and DAC price performance in recent times was Sigma-Delta technology which facilitated moving almost the entire converter into the digital domain. This is BTW completely independent of DSP technology.

It is questionable whether or not the best modern converters which overkill the actual requirements for transparent audio by orders of magnitude could be implemented with ca. 1960s technology, but be sure that nobody is likely to try! ;-)

1. The problem wasn't the CD standard sampling rate itself, but the fact that oversampling converters were impractical then, thus "data converters at the time". I understand the sampling process and sometimes over-simplify since I assume this audience is largely non-technical. An assumption often proven wrong...

2. A good analog filter is neither cheap nor easy; we agree. We are talking consumer gear here, yes? Non-linear phase shift, and group delay problems, can cause audible issues. That was shown back in the late 70's on the early CD systems (as well as a host of other problems unrelated to the standard -- I remember hearing and seeing some ugly test results back then). However, that was long ago, and I do not know what types of filters were used (but probably not nice linear-phase ones). As for the DSP, yes, the filters are simple to implement, but the consumer DSP chip wasn't around and it was hard to find a computer to do real-time processing even at audio rates that wouldn't 10x the price of a CD player. My point wasn't that you need a DSP, but that you need some sort of cheap, fast digital processor, and (audience assumption again) "DSP" is what most people think when they hear "digital filter". I tend to use it as "digital signal processing/processor" meaning a collection of circuitry to do it. That said, I design data converters but am primarily an analog guy -- long time since my grad DSP (digital filter, whatever) courses.

3. Yes, of course, good catch. Anti-alias filter -> PCM sampler -> digital filter for an ADC; other way around for a DAC. I knew that, duh!

4. As I have said, I was taking consumer audio into account, not the ones I and others helped build for the military and test labs, and of course those few high-end audio folk... I would argue that delta-sigma* and digital filter theory (not necessarily stand-alone DSPs, but the concept of digital signal processing) had to grow together. All delta-sigma data converters include a modulator and a digital noise filter, so far as I know, and yah not always in that order.

With your last statement I am in complete and total agreement!

Thanks for keeping me honest - Don

* Gabor Temes put it that way, saying an early technical paper transposed the words for the difference and integrator -- summer -- blocks; he told me that in class long ago, and it's in one of his books. On the block diagram of a DS modulator, the difference (delta) circuit comes before the integrator (sigma).