192 khz vs 48 khz poll

[Miska]

Now in regards to music, in hirez widebandwidth material, even those with ultrasonic info, it has always been low in level. I am sure somewhere there is music with it high in level, but it appears rather uncommon. I haven't seen those super steep transients. If there are some they are not an ongoing common signal that will explain general sound quality differences.

 

I just provided example of claves, but pretty much initial impact of any cymbals or similar will generate these transients. Or claves, castanets, maracas, etc...

 

Listen to Spanish music with castanets and you'll have a lot of those transients. There's also plenty of Asian music with strong cymbal content.

 

Spectral output of these as shown in the second link I provided is almost flat as high as the microphone is capable of going.

[One and a half]

Thus many believe that sampling frequency is more important for speed (fast changes) than for the extended frequency range.

 

This is an important point, thank you Teresa. The accuracy of what the OP is after depends on the content the music is recorded to begin with. For much of modern pop music engineering, 16kHz sampling at DR4 is enough, with the correct EQ, it can play through $5 earbuds very well, but not for most that visit this site.

 

Anything else that demands attention of a 'good recording' needs a higher sampling rate to capture transients and fine details. Many a time have I listened to a fading piano tone on CD, and as the note drifts into the black, it rolls down the stairs like a medicine ball, chop chop chop. The same recordings from the masters sampled at higher frequencies (Waltz for Debby) reveals far more detail than from a CD and piano notes fade correctly. This is still PCM, DSD is faster again, so it captures these notes better.

 

To the OP: No I don't have any proof, just a subjective opinion.

[sandyk]

BTW, some of Barry Diament's 24/192 recordings have genuine musical content as high as 57kHz.

See also:

http://www.cco.caltech.edu/~boyk/spectra/spectra.htm

[Jud]

I just provided example of claves, but pretty much initial impact of any cymbals or similar will generate these transients. Or claves, castanets, maracas, etc...

 

Listen to Spanish music with castanets and you'll have a lot of those transients. There's also plenty of Asian music with strong cymbal content.

 

Spectral output of these as shown in the second link I provided is almost flat as high as the microphone is capable of going.

 

We shouldn't think only of percussion. Inharmonic attack transients exist for nearly all instruments. Just think of the difference between playing a trumpet, and gently pressing a key on a synth set to sound like a trumpet tone.

[Mike Mcsweeney]

Mike you are again correct when you say one would not be able to reproduce the composite signal "perfectly" such that nothing is lost, audible or not.

 

However, if we restrict ourselves to discussing a real world signal as it applies to human hearing there are some interesting things we can do. We can now state a limit to the amount of dynamic range we want to consider as well as a limit to the amplitude and frequency content of the signal. Why? Because human beings have a limit to what frequencies they can hear, to how much amplitude they can tolerate or even detect, and to what details they can detect

 

Thanks....

So in your first paragraph above, you are stating the answer to the poll is true, correct?

 

In the second paragraph, you suggest restricting ourselves to discussing a real world signal as it applies to human hearing...

In my example of 1 billion elephants, each with unique music of varying tones, that may not have been real world, but I am not sure that a good pair of ears can't differentiate the characteristics of sound that couldn't be put into ones and zeros?

 

I just listened to a dsd recording where i heard the plucking of a guitar that sounded so precise and detailed, that i have never heard before. It was as though you could see the actual plucking and the "reverberation" (i am not sure that is the correct word) in such detail that i am relatively certain that if the same 5.6mhz signal was down-sampled to 48khz, you would not hear the same fidelity of the plucking and reverberation.

 

My guess is that 20 years ago, engineers would not believe human hearing could detect a difference between a 192K signal compared to a 5.6mhz signal? Is it possible we are underestimating what our ears are capable of, especially when it comes to the infinite number of harmonies that are possible.

 

Do you agree or disagree that we can hear a lot more detail and "space" in a dsd 5.6mhz sampling over what we can hear in a 48k sampling?

[mayhem13]

Anything in the input that exceeds filter's pass-band will result in ringing. Higher the filter order, more it rings. Only first order filter doesn't ring. Music content doesn't abruptly stop at 24 kHz, you can see it in the measurement results I sent, or by analyzing many hi-res recordings.

 

For perfect transient response you should use only first order filter, you can calculate how high sampling rate you would need to avoid aliasing with first order anti-alias filter with 24 kHz f_c.

 

The only problem with this Miska is the speaker system....real world, moving masses in simple motors.....they simply can't express the same impulse response as even some of the worst digital devices. The speaker components will always impart their own flavor of ringing, NOT based on contextual impulse response anomolies but rather due to mechanical and physical properties of the drivers and system. Liken it to feeding a 4k video signal to a 1080p display......there's simply no value in it. We don't have 4k speaker systems........yet.

[mayhem13]

BTW, some of Barry Diament's 24/192 recordings have genuine musical content as high as 57kHz.

See also:

There's life above 20 kilohertz! A survey of musical instrument spectra to 102.4 kHz

 

One day, animal cruelty charges might be leveled against him.lol. If i still lived in my gothic revival home with twin towers loaded with bats, I'd have grabbed a Soundkeeper recording or two. Less cleanup.

[mayhem13]

This is an important point, thank you Teresa. The accuracy of what the OP is after depends on the content the music is recorded to begin with. For much of modern pop music engineering, 16kHz sampling at DR4 is enough, with the correct EQ, it can play through $5 earbuds very well, but not for most that visit this site.

 

Anything else that demands attention of a 'good recording' needs a higher sampling rate to capture transients and fine details. Many a time have I listened to a fading piano tone on CD, and as the note drifts into the black, it rolls down the stairs like a medicine ball, chop chop chop. The same recordings from the masters sampled at higher frequencies (Waltz for Debby) reveals far more detail than from a CD and piano notes fade correctly. This is still PCM, DSD is faster again, so it captures these notes better.

 

To the OP: No I don't have any proof, just a subjective opinion.

 

......and no way of knowing the actual decay of the note being played in that studio environment at the time. You may have a 'preference ' of what you'd like to hear but you can't establish accuracy or a reference to that. Piano is notorious for being recorded in far too many obscure ways compared to the way we actually hear an acoustic piano. Don't know about you, but I don't crawl underneath a grand piano or climb inside either. Those exceptionally long decay notes you speak of.......that's not a naturally occurring content within an acoustic listening session.....maybe a hint of it on the performance floor and directly in front of the player........certainly not fifth row center of even the most intimate chamber or precisely designed acoustic venue.

 

.....but there's no fault in it either. YOU enjoy that type of performance as much as I might like kiwis instead of strawberries. It's a matter of taste. It's when we go looking for reasons why we prefer one taste to another where the speculation begins. Trying to tie tastes to technology only further complicates matters.

[Jud]

We shouldn't think only of percussion. Inharmonic attack transients exist for nearly all instruments. Just think of the difference between playing a trumpet, and gently pressing a key on a synth set to sound like a trumpet tone.

 

Just was thinking about more examples along these lines. Take vocals. The obvious hard consonant sounds, like "t" and "k" and "p," all begin with inharmonic transients. "Tenderly," if the inharmonic transients were badly done, might sound more like "denderly," or if you deliberately minimized or removed them, "nennerly." Even various vowel sounds have subtle inharmonic attacks. Feel the difference in the back of your mouth between the "u" sound in "under" and the "o" sound in "wonderful." The kind of push against the closed back of the throat that you get at the start of "under" is a very subtle inharmonic attack that isn't there with the "o" sound in "wonderful."

 

I wonder what spectral decay plots of various inharmonic transients would look like. Don't know if Teresa (or anyone else who's willing to help out) would have a reference for that two millionths of a second figure, but it would be great to see some information on energy versus elapsed time for various inharmonic attack transients.

 

The other thing about these transients that may make them a bit knottier problem than harmonic tones has to do with what RobbieC mentioned above regarding a workaround for the problem of infinite time - "The DFT requires that the curve be discrete and periodic (meaning repeats to infinity)." Inharmonic sounds aren't periodic. My completely fact free thought of what would be done for inharmonics is that there would be curve fitting, where harmonics would be used to approximate the shape of the inharmonic transient as well as possible. Can someone who knows tell me if this is in fact what happens?

[mayhem13]

Odd order harmonics....we'll refer to them here as distortions for the sake of discussion....propogate in somewhat a linear fashion in regards to frequency. From the fundamental on down f3,f5,f7,f9..........will all 'line up' so to speak in a distortion plot of a driver. So if we consider the driver imparting its own distortion profile ( and not mildly so in comparison) on top of the very small added distortions of complex digital filters, the real world relevance of what we're talking about here should become more clear in perspective.

You can take two midwoofers whose freq response through its passband are near identical (within .5-1db spectrally) who's distortion profiles are completely different( always the case as seen in measurements/much more complex) and the drivers will sound noticeably different from each other. Add that real world variable to digital filter impulse response and then think about what we're actually 'analyzing' in the first place. Again, accuracy has to be relevant to something. If you/we can't establish a reference, the discussions become far less academic in reality and more subjective in context. Some will make the arguement for the 'live' presentation being the reference..........but we could shoot holes in that arguement until next Tuesday. For me, knowing and experiencing what I have with speaker systems is how I've 'developed' my small signal position. And when we examine real world examples of uncertainty in small signal variances, the position of what we're discussing with such high sampling rates becomes even less relevant.

 

I find a peculiarity in the this and other forum bodies where one section of the signal chain is emphasized. The trend is to look so deep into the forest, that the trees get looked past. Your speaker system is the tree in this forest of audio.....it's the foundation of reproduced music,....it's what we hear. What goes in is entirely dependant on the speaker as to what comes out. I often feel that if more people here understood the properties of a speaker system, there'd be far less fretting over digital filters and the like. But it's this forum's area of interest so who am I to suggest otherwise. But at the same time, look at like this. If you print a picture with opacity in its ink over different color paper ( paper being speakers)' your results will ALWAYS vary. And music content is certainly opaque in these discussions as long as there's multiple mediums in the reproductive chain. Ignoring the color of the paper when adjusting the color spectrum of the ink doesn't make much sense, does it?

 

That being said, I certainly appreciate your objective standpoint here on this particular topic.....if I'm understanding your responses correctly.

[Paul R]

I want to try to be precise here. Because Nyquist's idealizing assumptions do include infinite time/samples, it is true even if you have a 2 Hz sample rate applied to a wave of less than 1Hz frequency, that an infinite *number* of samples is necessary to provably specify the waveform completely. But you do not need an infinitely fast sample *rate*. In fact it is provable that given an infinite number of samples in each case, a 1MHz sample rate would not specify a 1 Hz waveform any better than a 3 Hz sample rate.

 

The way this is implemented is a sort of trick- an infinite number of samples is sort of assumed in the math using a Discrete Fourier Transform. Or you could in theory flip that around and say an infinite sample rate, but it is more difficult to implement.

 

And yes, mathematically one can obtain perfect answers. But in the real world implementations, all we get are approximations. Arguably, higher sample rates result in better reproductions. Whether better here has any significance is what is arguable.

 

So to be very precise- does the original question address only theory, only real world implementations, or both? The answer quite possibly (arguably) changes depending upon the exact problem domain addressed. Confusing and it is what leads to some of the more heated arguments.

[Sam Lord]

192 khz sampled digital audio will record and reproduce analog musical signals below 20 khz more accurately than 48 khz sampled digital audio.

 

False. Digital audio doesn't record or reproduce anything, it is data representing audio waveforms. Now that is a harsh treatment of the question, but imprecision was the death of the poll from the start. At this point in history, equipment records and reproduces, and that equipment is burdened with physical constraints. In the case of 48kHz vs. 192kHz, the former requires stronger anti-aliasing filtration than the latter. But the latter operates at higher speeds, increasing decimation error from the reduced settling time of the circuits.

 

At the moment most recording engineers prefer a premium 192 system over an equally-priced 48 system, though that preference changed not terribly long ago. Loop degradation tests have been done using 44.1 vs higher rates and the higher rates have won, so I'll go with 192 or some other flavor.

 

That said, the Nyquist theorem allows for the possibility of mathematical near-perfection of 48kHz circuits for 20kHz reproduction, but physical limitations will remain indefinitely.

[Jud]

False. Digital audio doesn't record or reproduce anything, it is data representing audio waveforms. Now that is a harsh treatment of the question, but imprecision was the death of the poll from the start. At this point in history, equipment records and reproduces, and that equipment is burdened with physical constraints. In the case of 48kHz vs. 192kHz, the former requires stronger anti-aliasing filtration than the latter. But the latter operates at higher speeds, increasing decimation error from the reduced settling time of the circuits.

 

At the moment most recording engineers prefer a premium 192 system over an equally-priced 48 system, though that preference changed not terribly long ago. Loop degradation tests have been done using 44.1 vs higher rates and the higher rates have won, so I'll go with 192 or some other flavor.

 

That said, the Nyquist theorem allows for the possibility of mathematical near-perfection of 48kHz circuits for 20kHz reproduction, but physical limitations will remain indefinitely.

 

Tell me more about loop degradation tests (he asked breathlessly), please?

[Miska]

The only problem with this Miska is the speaker system....real world, moving masses in simple motors.....they simply can't express the same impulse response as even some of the worst digital devices.

 

For speakers (that < 25% of my listening) I am using two-way design with AMT-type folded ribbon tweeters with extremely strong magnets that go flat to around 50 kHz and bass drivers with aluminum coated cones. (to be exact the 0 degree response increases as function of frequency, but that's offset by increasing directivity)

 

My amp can put out 150 amp current and 1.5 kW power to keep speakers under tight grip (10 power transistors per channel).

 

For most of my listening, I am using headphones, (AKG, Beyerdynamic, Sennheiser and Stax) and I'm moving towards using these as my primary headphones:

K812 | AKG

But Stax and Sennheiser HD-800 are also very good.

[PopPop]

Tell me more about loop degradation tests (he asked breathlessly), please?

 

This is one of the best threads I have yet to read in CA. It has cleared up a lot of confusion I have had about perceptions and expectations for high sample rates. That you are breathless with anticipation just cracks me up, Jud.

[Sam Lord]

Tell me more about loop degradation tests (he asked breathlessly), please?

Hi Jud, kinda surprised you are interested...that is, I would expect news to spread in these parts if there were loop tests showing the opposite. Anyway, at Pro Recording Workshop's Mastering forum, they occasionally refer to tests they have done or other audio researchers (non-ME's) have done and comment. I haven't dug up a 48k vs 192k comparison null test of 60 cycles, but here is a trial and discussion of a posted test for degradation involving external word clocks. Chuck Zwicky ran it, and I always read his viewpoints since he's one of the more scientific guys in the bunch (and there are many):

 

ZMIX Clocking Test

 

The discussion is here:

 

Studying the effects of external word clocking after 60 passes through A/D and D/A Converter pairs in The Mastering Room Forum

 

I recall a test done a few years ago, I believe it was also 60 passes, and the observation was the degradation of a 16/44.1 recording IIRC. These folks and AES folks run these tests occasionally. I encourage you to look back on earlier threads, it is a great resource. They aren't hard to do if you automate the passes. I bet Charles Hanson has done some; I know Bruno Putzeys has done quite a few. Most are in-house and unpublished. Dave Collins, the moderator and a good friend, has told me of many tests that he ran without bothering to publish the results. Of course the degradation doesn't correlate directly with perceived sound quality or fidelity in a purely objective, definitive way.

[Miska]

The only problem with this Miska is the speaker system....

 

I checked through bunch of step response measurements and most tweeters settle in way less than 100 µs, probably closer to 10 µs. While typical digital filter takes around 1 ms...

 

Look for example at positive going tweeter step here:

http://www.stereophile.com/content/wilson-audio-specialties-alexia-loudspeaker-measurements

And compare that to for example Filter 1 here:

dCS Vivaldi digital playback system Measurements | Stereophile.com

 

And headphones are much better than loudspeakers...

[Sam Lord]

On the question of high-amplitude transients, it's important to consider air absorption, a ubiquitous fact of airborne sound. It refers to the change in amplitude of the longitudinal waveforms of different frequencies as they pass through air. High frequencies and short transients are attenuated much faster than longer wavelengths. Thus a drum hit with extremely high peak power (almost infinite by some calculations) has its amplitude quickly smashed by the air. The energy and impulse (change in momentum) isn't great to begin with, and it very quickly becomes a normal set of diminishing waves like all sound.

 

Another question is what filters our ears use. We know that most of us can't detect sound frequencies above 20kHz, but we still don't know how we filter sound as it approaches that high pitch. We obviously account for errors of phase either at our ears or brains for evolutionary reasons, but the actual Bode graph of our own hearing acuity is a moving target of investigation. So the quest for accurate playback will require more knowledge of how our ears work.

[mayhem13]

I checked through bunch of step response measurements and most tweeters settle in way less than 100 µs, probably closer to 10 µs. While typical digital filter takes around 1 ms...

 

Look for example at positive going tweeter step here:

Wilson Audio Specialties Alexia loudspeaker Measurements | Stereophile.com

And compare that to for example Filter 1 here:

dCS Vivaldi digital playback system Measurements | Stereophile.com

 

And headphones are much better than loudspeakers...

 

Be careful with looking at just step or impulse.....we're talking a three dimensional function very similiar to how Nordost and their hire lab did (poorly) frequency and amplitude with respect to time. CSD is much more usefull here.......and that's not to say that a speaker can't reproduce the smear of a bad filter.....IMO it certainly can and will be the focus of an upcoming Blog im putting together. What i AM saying is that there's inherent ringing within todays drivers clearly evident within CSD plots based on the drivers design,within a range of response that IS NOT 'derived' from the input signal.

 

I'd use your example earlier about AMT tweeters to 50k and aluminum cones to further my point above. There's absolutely nothing about what you mentioned in regards to drivers thats empirically true.....which isn't to try and insult you in any way about your knowledge of speakers, as i have very little of an understanding of the work you do within digital filters. Which direction we're traveling .....in to out/ out to in......that's a chicken or the egg thing......as long we cross the road at some point. LOL

[mayhem13]

On the question of high-amplitude transients, it's important to consider air absorption, a ubiquitous fact High frequencies and short transients are attenuated much faster than longer wavelengths. Thus a drum hit with extremely high peak power (almost infinite by some calculations) has its amplitude quickly smashed by the air. The energy and impulse (change in momentum) isn't great to begin with, and it very quickly becomes a normal set of diminishing waves like all sound.

 

Another question is what filters our ears use. a moving target of investigation. So the quest for accurate playback will require more knowledge of how our ears work.

 

You skipped over the really important part.....CAN a modern day speaker system reproduce it? As you've described the event, the answer is a universal NO. It's a flawed reproduction at best, certainly not a recreation. If we can get past this, a lot of what we fret abopute in audio becomes far less significant in perspective.

 

Perspective is the key word here.

[Miska]

CSD is much more usefull here...

 

I've made CSD plots of my speakers too, but it's very boring to look at regions outside room modes because the response dies off immediately. Mainly because FFT as algorithm is not suitable for analyzing this kind of behavior. You need to make CSDs using WVD (although Gabor transform is not completely useless either).

 

Step and impulse responses are a good starting point. Here you can see how filter ringing becomes visible through loudspeakers measurements:

DRC: Digital Room Correction

 

I'd use your example earlier about AMT tweeters to 50k and aluminum cones to further my point above.

 

I'm not going to argue too much about speaker design, because it is outside my interest and I anyway use primarily headphones. Based on my experience and measurements I get better transient response with ribbon tweeters (like Elac's JET in this case). While the secondary system's Dynaudio speakers with coated silk domes are quite good too, although don't go as high in terms of frequency.

 

 

P.S. My signal analysis background is from passive sonars (like recognizing engine noises, intercepted active sonar transmissions and human-made transients) and radar systems.

[mayhem13]

Duly noted.......and no arguement intended or desired. It's simply my viewpoint that you and others might be correcting functions that will get undone in the final part of the chain. All though i'm all about speakers, i admit they're by far the weakest link in all of audio as far as technology goes. When we have a single driver that can do 20-20 with low distortion, no breakup modes and constant directivity, i'll rethink my assesment. Sadly we're no where near that lofty goal.

[PopPop]

 

P.S. My signal analysis background is from passive sonars (like recognizing engine noises, intercepted active sonar transmissions and human-made transients) and radar systems.

 

Mine too, to a limited extent. US Naval Facilities on the east coast. Up until this moment, I never connected my listening to music (details) to my "listening" for engine and propeller noises.

[Miska]

All though i'm all about speakers, i admit they're by far the weakest link in all of audio as far as technology goes. When we have a single driver that can do 20-20 with low distortion, no breakup modes and constant directivity, i'll rethink my assesment. Sadly we're no where near that lofty goal.

 

I don't think it's reason not to improve on other parts. And IMO headphones are not bad at all, especially Sennheiser HD800 and AKG K812.

 

For analyzing loudspeaker element behavior, I would probably use LDV. We successfully used it for rotating machinery bearing fault analysis together with acoustic analysis.

Laser Doppler Vibrometer (LDV) for Vibration measurement: Optomet GmbH

[Paul R]

Mine too, to a limited extent. US Naval Facilities on the east coast. Up until this moment, I never connected my listening to music (details) to my "listening" for engine and propeller noises.

 

(grin) So both you guys know what a TACCO is and an Acoustic Sensors Operator huh? That's essentially my background as well and I've used what I learned there is just about every other field I have been involved in. It does tend to bias one as to what can and can not be heard doesn't it? It is why I tend to get a smile around here when someone comes out and absolutely tells me that small periodic noises are not audible...