Digital audio explained (even I grasped some of it!)

[E-Stat]

The sampling rate is chosen to avoid aliasing, it's not chosen to "bring out the highs", etc. The audio spectrum of the digital recording is pretty much determined by what kind of low pass filtering is applied to the original analog prior to sampling. Thoughts?

The primary reason for a higher sample rate is to avoid the ringing issues caused by necessarily steep slope brickwall filtering. The thirty year old Redbook standard (forever in computer storage terms) must provide 100% signal attenuation in the space of a mere 2 kHz to achieve flat response to 20 kHz. There have been AES papers (Craven 2004, etc.) on ways to reduce this effect but still offer compromises.

Using a higher sample rate allows the use of simpler non-ringing filters.

 

[sprins]

...If you can hear at 25 kHz...

Yeah, well... My ears go upto 15kHz max. (my stereo goes upto 11 that's why).

 

[bshorey]

An organ can produce fundamentals into around 7khz. Depends on what you're defining as a musical instrument. Why is this getting restricted to music? I'm just talking about recording and playback period.

I get what you're saying but even then we're getting a little past the point of what my argument even was. But there is this:
http://www.overgrownpath.com/2013/08/how-classical-music-was-covertly-dumbed.html

Interesting discussion. Ok, here's my next question. If the speaker can reproduce the fundamentals of any given instrument, then by definition wouldn't it also result in similar harmonics as those produced by the original instrument?

For the most part, most recording mediums fall off below 20 kHz, most electronics produce up to 20 kHz, and most speakers produce up to 20 kHz. I understand that there are exceptions. So if you're producing content for that 99.99% of the population, it doesn't really make sense to chew up 5-10 times the storage space to extend from 20 kHz to 28 kHz, does it?

If you've got the entire equipment chain to support production to 30 kHz, then you do have the option of capturing digital at higher rates. I'd guess that the people who are producing high end vinyl stampings could probably be convinced to produce high end digital files too.

bs

 

[E-Stat]

Is the higher sampling rate intended to preserve these frequencies, i.e., prevent them from being aliased?

That is simply a bonus of handling a broader range of instruments with response above 20 kHz. The audible effects of the ringing, however, are heard well below that range.

Measured instrument frequency range

 

[rifftrax]

Interesting discussion. Ok, here's my next question. If the speaker can reproduce the fundamentals of any given instrument, then by definition wouldn't it also result in similar harmonics as those produced by the original instrument?

Oh no, absolutely in no way. Instrumentation or object resonances are a property of the actual materials and construction of whatever is creating the sound. A speaker transducer's breakup modes or distortion properties are completely unrelated to the recording/source material it's producing precisely because there are no instruments made out paper cones suspended between a set of excursion limits besides... well, other speakers. You have to consider the mechanical properties of why a thing actually creates a sound to begin with (whether bowed, blown, struck or actuated in some other manner).

For the most part, most recording mediums fall off below 20 kHz, most electronics produce up to 20 kHz, and most speakers produce up to 20 kHz. I understand that there are exceptions. So if you're producing content for that 99.99% of the population, it doesn't really make sense to chew up 5-10 times the storage space to extend from 20 kHz to 28 kHz, does it?

Recording mediums? What? Why would a recording medium simply fall off at some arbitrarily defined point? Maybe you've selected an ADC that's limited but, well that's just due to convention and the type of interface you decided to pick.

"Most electronics". This is really funny to me. I think the problem here is you're not really defining things from an engineering standpoint, you're defining them as marketing has taught you to define them as (which is the most damaging part of the pseudo-sciencey bullshit of people just repeating "the world ends at 20khz").

Bob Stuart of Meridian is generally acknowledged as one of the experts on digital audio and audio perception.

And yet, arguments by authority may be the precise reason why I am one of the few people (potentially the only one I'm aware of) that's suggested a readily verifiable test for destroying the commonly accepted "wisdom" of human hearing limitations.

 

[tppc]

In the video he talks about a bandwidth limited signal. For regular consumer type music CD's what is the actual cutoff frequency? Isn't this the real issue regarding high frequencies. If you can hear at 25 kHz but the analog is cut off at 20 kHz no sampling rate no matter how fast is going to recover the lost information.

The sampling rate is chosen to avoid aliasing, it's not chosen to "bring out the highs", etc. The audio spectrum of the digital recording is pretty much determined by what kind of low pass filtering is applied to the original analog prior to sampling. Thoughts?

Correct. There won't be any high frequency content beyond ~22.5 kHz in 44.1 sample rate material due to band-limiting (anti-aliasing) filter prior to ADC input. The filter cutoff is determined by the output sample rate of the ADC.

For example if output rate is 44.1 then filter cutoff is ~22.5. Similarly if output rate is 192 then filter cutoff is ~96 which allows for much more bat/dolphin HFC.

Note that some of the newer ADC chip designs have intrinsic anti-aliasing and thus there is no need for an external analog filter. Exciting stuff!

 

[E-Stat]

And yet, arguments by authority...

It would make an interesting conversation between Stuart and another well respected engineer like Ed Meitner of EMM Labs or Charles Hansen of Ayre Acoustics who have a different take on the issue.

 

[bshorey]

Recording mediums? What? Why would a recording medium simply fall off at some arbitrarily defined point? Maybe you've selected an ADC that's limited but, well that's just due to convention and the type of interface you decided to pick.

"Most electronics". This is really funny to me. I think the problem here is you're not really defining things from an engineering standpoint, you're defining them as marketing has taught you to define them as (which is the most damaging part of the pseudo-sciencey bullshit of people just repeating "the world ends at 20khz").

Hmmm, if you turned down the level of condescension, it might make for a better discussion. I was trying to engage.

Wrt some recording medium falling off at some arbitrary point, on the analog side of things, I didn't think it worked that way. My understanding is that the engineers worked within budgetary and physical/mechanical constraints. So on a tape deck, for example, the upper frequency limits would be determined by such things as tape speed, tape width, head quality, electronics, etc. That's not arbitrary. I don't think anybody set out to produce products that topped out at, say 14 kHz, but if that's where they ended up for a mid fi deck that had to be produced and sold for a profit at some price point, then that's where they ended up. In a parallel to this digital discussion, I know that there were cassette and reel to reel decks where you could double the tape speed and raise the upper end of the frequency response, but not a lot of people found the need to do that, or felt it was worth the doubling of the storage capacity usage. Much like the trade off between digital sampling rates and file size..

As for my 'falling for the marketing', you're wrong. Manufacturers don't build equipment with no upper limits of frequency response and then spec it to 20 kHz just for chuckles. It takes engineering effort and $$$ to design and produce equipment that can reliably produce those frequencies at the end of a production line. When companies invest that effort and command those prices, they are happy to advertise it, I know of more than one audio product that is spec'd to beyond 30 hkz or even 40 kHz. But I think that decades of experience has shown that the frequency of diminishing returns begins right around 20 kHz.

bs

 

[rbbert]

It would make an interesting conversation between Stuart and another well respected engineer like Ed Meitner of EMM Labs or Charles Hansen of Ayre Acoustics who have a different take on the issue.

Actually, if you bothered to read what they write and/or say, you would know that they don't disagree with Stuart at all. None of them think recording should be limited to 20 bits/60 kHz sampling, only that that is all that is necessary for accurate playback. Notice that even though Charles has made very clear that he prefers 24/192 PCM to DSD, he has still added DSD to his new DAC, just as Bob Stuart's products all do 24/192 PCM.

BTW, to demonstarate that your "dog dazer" is being heard, you would need some brainstem audio evoked potentials in response to being exposed to it. I'd be willing to bet that "sensing" a 115 dB 25 kHz sound has nothing to do with hearing it...

 

[E-Stat]

My understanding is that the engineers worked within budgetary and physical/mechanical constraints.

I agree entirely. The 700 MB Compact Disc most certainly rendered constraints when the Redbook standard was written in 1982. In order to maintain LP playback times, engineers had to sacrifice something. If you recall, the very first digital recorder from the late 70s, Dr. Thomas Stockham's Soundstream unit, sampled at 50 khz and was commercially used by Telarc and other labels.

Much like the trade off between digital sampling rates and file size..

Fortunately, that situation changed in 1995 when an inexpensive 4.7 GB storage medium became available. Today, you can buy a spindle of a hundred for around twenty bucks.

You may be aware that some music is available using 50 GB Blu Ray discs.

 

[E-Stat]

None of them think recording should be limited to 20 bits/60 kHz sampling, only that that is all that is necessary for accurate playback. Notice that even though Charles has made very clear that he prefers 24/192 PCM to DSD, he has still added DSD to his new DAC, just as Bob Stuart's products all do 24/192 PCM.

My mistake. I looked up this Stereophile interview and found the following quote from him:

"The CD specification was laid down by Philips and Sony, and like I said, it wasn't quite good enough. If we'd had the chance, it would have been a lot better at 48kHz than 44.1kHz. If it could have been 20 bits at a 55kHz sampling rate, it would actually have been good enough for ever, in my opinion. "

Some may argue where that magic sample rate should be (especially given editing losses) but 44.1 k is clearly insufficient.

So, I should have substituted "Xiph Monty" on the left with these three guys on the right. Apparently he doesn't grasp the concept as these gentlemen do after years of engineering actual products for the market.

BTW, to demonstarate that your "dog dazer" is being heard, you would need some brainstem audio evoked potentials in response to being exposed to it. I'd be willing to bet that "sensing" a 115 dB 25 kHz sound has nothing to do with hearing it...

I think you'll find that commentary was with a different poster.

 

[PAGS]

True story...in the early 90's when I lived in Boise, I noticed that when spending any extended period of time at the back of my house that I'd get these terrible headaches. I also noticed that, when orienting myself 'just right', I was sort-of 'hearing' something...a weird pressure on the eardrums that was difficult to describe. I thought I was going nuts.

Weeks later, in the backyard clearing brush and pulling weeds, all the while 'hearing' this noise, I noticed some metallic box over the fence and in the garden of the elderly lady who lived next door. I immediately realized that this device was the source of the noise and headaches...some sort of animal deterrent thingamajig. I thought it very odd that I was hearing it, but relieved that I had found the source of my misery. I went to my neighbor and told her about my experience, and after looking at her documentation on the device and seeing that it was very directional, we simply rotated the thing 90° so it wasn't pointing my direction. Problem solved.

I do not know what frequency this thing was emitting...I had 'tested' my own ears some years earlier and determined that my own hearing dropped rapidly after about 16KHz. If we assume that the device was emitting in the band somewhere above 20KHz, I have no explanation as to why it caused me such misery, except to assume that the SPL level emitted must have been huge.

Anyway, this whole 'dog whistle' thing is detracting from what is otherwise a good OP and discussion. I suggest that if we'd like to discuss such tangent subjects, that a separate thread be created. My own thoughts are that ultrasonic energy is a waste of sampling bits and amplifier power.

Umm, if I were you, I'd avoid full moons.

 

[fred soop]

Sometimes, during a discussion such as this, it is worthwhile to start over and look at the basics.

First, high fidelity audio was developed to reproduce music in an environment different from the original performance in such a manner as to attempt to recreate the illusion of that live performance. The key word here is "music". Dog whistle is outside that definition, except maybe to the dog.

What do we have in music? Both the piano and violin go to C8 which is 4186 Hz fundamental. My ancient information, which may not be accurate, is that the 5th harmonic of a violin has nearly the same amplitude as the fundamental and the 7th harmonic still has significant energy. That would put us at 29,302 Hz. The piano has only slightly less energy in similar harmonics.

Here is a link to a paper that I am assuming was written by students for a thesis. They measured (maybe accurately, maybe crudely) harmonics from various instruments.

http://courses.physics.illinois.edu...hurman_Ivy_Thomas_Phys199pom_Final_Report.pdf

Graphs are down in the document a few pages. They even determined that a trombone, of all instruments, had measurable harmonic content up to 20,000 Hz. It appears that they did not go beyond 20,000 Hz.

So, maybe the frequency response of a "perfect" system should be extended to 30 kHz, but the next problem would be, "Who are you going to get to perform your music?" Arturo Toscanini (not my favorite by any means) was acclaimed by some to be the greatest conductor of the 20th century. Sorry, but you won't find any of his work even in stereo unless it's been artificially reprocessed. Many of the other great conductors from the era when they were still "performing" rather than "putting notes on a recording" (to be further manipulated in the back room afterwards) are simply dead! So, until something new is invented, the good performances are on analog tape masters complete with limited frequency response and tape hiss.

As for the dog whistle, that is a stand alone pure tone, and is not in the context of music. Frequencies above 20 kHz that are part of the music will usually (not in our violin playing C8) will be masked by lower frequency content to the extent that their existence or absence will not be (readily) noticed.

Even for those that can't actually hear the dog whistle but are affected by it, out of range frequencies can affect in range frequencies. Let's look at the other end of the spectrum: We can't hear DC, but if we put enough DC into a woofer to drive it to the limits of its suspension, anything else sent to it will be very badly distorted.

 

[chicks]

Another great, facts-based article on digital audio that references Monty's:

http://www.sonicscoop.com/2013/08/2...t-you-knew-about-bit-depth-is-probably-wrong/

 

[E-Stat]

Sharp fellow that made these vids. He also wrote an article on the futility of 24/192 downloads here.

Can't really debate that one as it's over my head but theres a hundred page thread on headfi he contributes to

It is over your head.

Ask numerous engineers who actually produce recording and playback equipment liike Stewart of Meridian, Meitner of EMM Labs, Hansen of Ayre Acoustics, etc. who don't fall for his simplistic approach that is outside of the real world.

They just smile and continue to produce superior equipment.

 

[rifftrax]

BTW, to demonstarate that your "dog dazer" is being heard, you would need some brainstem audio evoked potentials in response to being exposed to it. I'd be willing to bet that "sensing" a 115 dB 25 kHz sound has nothing to do with hearing it...

Then how would you explain the problem of it only being sensed when it directly on-axis with the ear-drum?

I think what you're proposing is adding a layer of test complexity that is simply needless. Occam's razor has very pertinent application here.

 

[Lavane]

The video did a nice job of clearing up a few questions I had about sample rates. Personally I can't really hear a difference between 16 bit and 24 bit recordings. One of my CD players has an 8, 16 and 24 bit selector switch. The 16 bit recordings sound just as good as the 24 bit recordings. I honestly can't hear a difference. Our own brains sample the sound waves passing through our ears. I wonder what the sample rate is for the human brain? :scratch2:

 

[rbbert]

Then how would you explain the problem of it only being sensed when it directly on-axis with the ear-drum?

I think what you're proposing is adding a layer of test complexity that is simply needless. Occam's razor has very pertinent application here.

Not when carefully done tests for hearing show that it is unusual to hear above 20 kHz and almost never above 24 kHz. OTOH, it is well documented that ultrasonics are "sensed" (not heard) at high enough intensities.

 

[RobInONT]

It is over your head.

Ask numerous engineers who actually produce recording and playback equipment liike Stewart of Meridian, Meitner of EMM Labs, Hansen of Ayre Acoustics, etc. who don't fall for his simplistic approach that is outside of the real world.

They just smile and continue to produce superior equipment.

Fair enough, I read some writings from these fellows.

JR Stuart seems the most knowledgable and accepted of the above. He doesn't advocate for 24/196 downloads though. He feels that 20/48 would optimal. 20 bit word length for a little more dynamic range and 48 khz sampling rate for frequency reponse up to 25 kHz for the tiny percenatge of the population that can hear above 20 kHz. (My take on this article: https://www.meridian-audio.com/ara/coding2.pdf)

Not sure what to make of Charles Hansen. He talks about the "stair-step output waveform of a DAC" in this interview: http://www.audiostream.com/content/qa-charles-hansen-ayre-acoustics
That makes my ears perk up after watching Motgomery's video, as it is often a misunderstanding. Not sure of his context in this interview though.

Can't find any writings from Meitner, though I didn't spend a lot of time searching.

As far as recording in longer word length and sample rates, all agree, Montgomery included, that it is desirable.

It's unfortunate that audio has suffered from so much snake-oil and misinformation over the years. It makes people look at manufacturuers claims with a jaundiced eye. Ever heard the saying "Lotteries are a tax on people who are bad at math!" ? Our desire to pursue the ultimate sound accompanied by our not-so-great technical understanding can often be (and regularly is) exploited for personal gain. I'm not accusing any of the above of this but videos like Montgomery makes help us less-technical types make informed decisions and for that I appreciate his efforts.

 

[rifftrax]

Not when carefully done tests for hearing show that it is unusual to hear above 20 kHz and almost never above 24 kHz. OTOH, it is well documented that ultrasonics are "sensed" (not heard) at high enough intensities.

And I'm telling you otherwise. And so would everyone in my office.

I don't have any reason to grind this axe other than "this is my legitimate experience". I mean, my whole job is testing these things.

Seriously. What's the difference between "sensing" something that only works when it's aimed at your ear-drum and uh... hearing? Sounds like we're talking about the same thing here.