High resolution audio. The science, or lack of...?

[manicm]

I have read this thread with interest. As a Sound Engineer you have all missed the most important link in the chain.... The microphone. As far as I am aware, no matter how much you spend, there are no microphones that you can put in front of a source, ie piano, violin, voice, guitar, that will record beyond the range of human hearing. Even if they can, they are usually rolled off steeply after the 20khz point. Microphones have different voicings, I agree, which gives a different tone, even to the same source.

However, if it is not recorded above 20khz with the very first link in the chain, then how on earth can you hear it later on. No matter how much how much one spends on kit.

I recently worked in a studio where they were mixing down on a pair of PMC BB5 XBD actives. I think that the installation cost in the region of £45k for the speaker set up alone. Accurate no doubt but these monsters would not look good or sound good in your average house. Whilst there I went through their microphone selection, I found one of my favourites, the Neumann M150 Tube Omni, this costs £4k retail, and is superb for recording most sources. It has a dynamic range of 119dB and frequency response of 20hz to 20 KHz. It doesn't record beyond this, thus it cannot transfer information to the recording medium in excess of this. Consequently, you will not have anything beyond that to listen to. Add this to the aforementioned proven Nyquist theory, then in my honest opinion there is no point to HD beyond the maximum 48 kHz CD rate. I can see the reasoning behind 24 bit against 16 bit, however, unless you have the hearing of a child, or dog, then again there is no point.

I simply prefer to buy kit that excites me and makes me smile, because it makes me happy.

I really like this post. You can't add what isn't there in the first place. As for your last line I think it defines what a system should be about IMO - getting you excited about the music, allowing you to engage with the music.

:cheers:

True, but PEAYEL unwittingly or not failed to mention there are microphones able to capture frequencies above 20khz - note the link I provided in my reply to him. There was one such microphone in 1998.

Also, the bit depth is not just about dynamic headroom, but more importantly how much information can be stored in any one slice of time, which doesn't require massively powerful amplification to take advantage of.

 

[steve_1979]

Just as a point of interest.......

Would anyone like to guess at the number of bits that a decent digital playback system (transport/computer/streamer and Dac) might be capable of resolving in a home environment......... :?

Come on don't keep us hanging. What's the answer?

 

[manicm]

Some of Naim's latest digital devices, and probably other makers too, have so-called 32-bit DACs.

 

[Andrew17321]

I have no intention in reading through all this topic, but I feel that I must correct a misunderstanding of the maths mentioned. Nyquist's Theorem is correct. However, like all mathematical theorems there are conditions which apply, and unfortunately, the conditions for applying Nyquist's Theorem are not satisfied in audio systems.

If the signal was an unchanging note; if the analogue signal was measured perfectly; if the electronics could reconstruct these measurements perfectly, then one could claim that Nyquist's theorem applies, and that the signal from the DAC was a perfect copy of the original analogue signal. Unfortunately, these are not the case.

In fact, the higher the sampling frequency the nearer the 'unchanging signal' aspect is to being satisfied; the more bits in the digital measurement, the more accurate the measurement. So, theoretically, Nyquist's theorem is a better approximation for higher sampling frequencies and higher numbers of bits.

Also the higher the sampling frequency, the better the filtering can be done. Filters at 22 KHz are necessarily a compromise which distorts the analogue output.

Sorry if this has already been mentioned.

Andrew

 

[PEAYEL]

Yes, but it still rolls off beyond human audio spectrum. Interestingly I recorded elephants' sub sonic conversations using mics that go down to 4hz and then some. I could see the sound on a screen, but I or the others that were with me couldn't when they were actually communicating. Strangley we could feel the communication when we amplified the screen signal with an array of 10_kW amps and huge speakers, these were big enough to produce the sub sonic waves. Similarly, in my earlier days, I worked on the problem of rice burn, caused by insects that communicate through the substrate in rice fields. They use frequencies beyond our hearing range to do this and also vibration through the plants. I couldn't hear them when my ears were right next to them. However, I was able to record them using 'Bat' microphones and by taking a CD laser mechanism out of a player and using this to record the vibrations/movement on the plant that they were resting on.

So I will stick my ideas originally, that anything outside of CD 48khz 16 bit is not worth the expense or effort.

 

[PEAYEL]

Just a note, and it is not a domestic system but in my house, but I have a 3kw per side PA system that I still do gigs with. This will bang it out at that level all night long. Unfortunately, I'm not allowed to as it causes damage to your hearing and the bass frequencies tend to make you wet yourself if you stand next to the bass bins. I seem to remember that a jet passenger aeroplane has a dB level of around 120 dB when taking off. If you like and can stand your music that loud then good luck to you and may I suggest that you start learning British Sign Language.

 

[oldric_naubhoff]

So, theoretically, Nyquist's theorem is a better approximation for higher sampling frequencies and higher numbers of bits.

tell that to those that claim they red and understood the N-S theorem. I wonder if they would listen :twisted:

Also the higher the sampling frequency, the better the filtering can be done. Filters at 22 KHz are necessarily a compromise which distorts the analogue output.

Sorry if this has already been mentioned.

Andrew

it has been mentioned but it's a good thing you chip in. maybe some people will finally understand that digital audio is not only about N-S theorem. :cheers:

 

[pyrrhon]

First if ou want to be scientific you got to admit that there is no sound, no smell, no colors. The sound you ear is something of the consciousness, outside its airwaves and nothing more. When you ask if someone ear the sound you ask if the waves got put at the counsciousness level. But you need to be aware that the data in the ear is recomposed in sound AND emotions. I would just point that we are affected by smells without ever being councious of the smell. Your mood also plays a big deal and things are far more complexe then we can understand with our measures. My belief is that we can not use counsciousness to describe the whole experience of music. The ear might be like the nose, getting tons of detail but not channel all to counsciousness but to some other spectrum like emotions. Keep an open mind, science is still very young !

 

[pyrrhon]

First if ou want to be scientific you got to admit that there is no sound, no smell, no colors. The sound you ear is something of the consciousness, outside its airwaves and nothing more. When you ask if someone ear the sound you ask if the waves got put at the counsciousness level. But you need to be aware that the data in the ear is recomposed in sound AND emotions. I would just point that we are affected by smells without ever being councious of the smell. Your mood also plays a big deal and things are far more complexe then we can understand with our measures. My belief is that we can not use counsciousness to describe the whole experience of music. The ear might be like the nose, getting tons of detail but not channel all to counsciousness but to some other spectrum like emotions. Keep an open mind, science is still very young !

 

[jcbrum]

there are microphones able to capture frequencies above 20khz

The microphone in my bat detector can hear squeaks well above 20kHz.

Some bats can squeak very loudly, - over 130dB.

I haven't tried putting recordings on CD. I use a flash card.

JC

 

[jcbrum]

Over on the AVI forum, discussing the same topic, one contributor suggests study of this (click on it) document as being helpful.

JC

 

[lindsayt]

All this talk about theoretical dynamic range is all very well, but what about some real life measurements? Well here's someone that's measured the noise floor of the same albums on CD and vinyl: http://www.audioholics.com/audio-technologies/dynamic-comparison-of-lps-vs-cds-part-4/dynamic-comparison-of-lps-vs-cds-part-4-page-2

Chariots of Fire CD

Chariots of Fire LP

For most of the frequency range, the noise floor on LP is comparable to or lower than CD. It's only below 500 hz that noisefloor increases to about -50dbs.

The noisefloor of vinyl does depend upon the engineering quality and condition. The LP measurements were made from a $1 thrift store 2nd hand LP.

Here's his noisefloor measurements of a well engineered LP, Three Works For Jazz Soloists & Symphony Orchestra

I'd expect 12" singles to be able to measure better than LP's.

Also these measurements were made with a Rega P3. A lot of the noisefloor hump below 500 hz will be down to motor and bearing noise produced by the turntable. Whilst the P3 is fine for what it is, it is just a budget belt drive turntable. I would expect a world class turntable to have significantly better noisefloor measurements below 500 hz than the Rega P3, reflected by better overall signal to noise ratio measurements and better subjective sound quality.

These measurements indicate that at the the frequencies where the ear is most sensitive, that vinyl has a comparable and possibly better noisefloor than CD. They also indicate that engineering quality of the vinyl pressing and the turntable are important for getting the best noisefloor results from this medium. It's a gross over-simplification to say that CD has a greater dynamic range than vinyl, because it depends upon what frequency we're talking about. If, for example we're talking about 2.5 khz then vinyl is measureably better than CD.

 

[altruistic.lemon]

Over on the AVI forum, discussing the same topic, one contributor suggests study of this (click on it) document as being helpful.

JC

Ah, so this thread is another attempt to educate the clueless, is it? Explains the refusal to check various websites, plus the use of ashleymatics to calculate amp wattage

 

[andyjm]

All this talk about theoretical dynamic range is all very well, but what about some real life measurements? Well here's someone that's measured the noise floor of the same albums on CD and vinyl: http://www.audioholics.com/audio-technologies/dynamic-comparison-of-lps-vs-cds-part-4/dynamic-comparison-of-lps-vs-cds-part-4-page-2

Chariots of Fire CD

Chariots of Fire LP

For most of the frequency range, the noise floor on LP is comparable to or lower than CD. It's only below 500 hz that noisefloor increases to about -50dbs.

The noisefloor of vinyl does depend upon the engineering quality and condition. The LP measurements were made from a $1 thrift store 2nd hand LP.

Here's his noisefloor measurements of a well engineered LP, Three Works For Jazz Soloists & Symphony Orchestra

I'd expect 12" singles to be able to measure better than LP's.

Also these measurements were made with a Rega P3. A lot of the noisefloor hump below 500 hz will be down to motor and bearing noise produced by the turntable. Whilst the P3 is fine for what it is, it is just a budget belt drive turntable. I would expect a world class turntable to have significantly better noisefloor measurements below 500 hz than the Rega P3, reflected by better overall signal to noise ratio measurements and better subjective sound quality.

These measurements indicate that at the the frequencies where the ear is most sensitive, that vinyl has a comparable and possibly better noisefloor than CD. They also indicate that engineering quality of the vinyl pressing and the turntable are important for getting the best noisefloor results from this medium. It's a gross over-simplification to say that CD has a greater dynamic range than vinyl, because it depends upon what frequency we're talking about. If, for example we're talking about 2.5 khz then vinyl is measureably better than CD.

Absolute nonsense. The report compares chalk, cheese, apples and pears. Just take a step back, think about the technology, do you really think that the noise floor of a mechanical system, with a motor, bearings and a stylus scraping along a piece of plastic is going to have a noise floor lower than a digital CD system?

 

[steve_1979]

All this talk about theoretical dynamic range is all very well, but what about some real life measurements? Well here's someone that's measured the noise floor of the same albums on CD and vinyl: http://www.audioholics.com/audio-technologies/dynamic-comparison-of-lps-vs-cds-part-4/dynamic-comparison-of-lps-vs-cds-part-4-page-2

Chariots of Fire CD

Chariots of Fire LP

For most of the frequency range, the noise floor on LP is comparable to or lower than CD. It's only below 500 hz that noisefloor increases to about -50dbs.

The noisefloor of vinyl does depend upon the engineering quality and condition. The LP measurements were made from a $1 thrift store 2nd hand LP.

Here's his noisefloor measurements of a well engineered LP, Three Works For Jazz Soloists & Symphony Orchestra

I'd expect 12" singles to be able to measure better than LP's.

Also these measurements were made with a Rega P3. A lot of the noisefloor hump below 500 hz will be down to motor and bearing noise produced by the turntable. Whilst the P3 is fine for what it is, it is just a budget belt drive turntable. I would expect a world class turntable to have significantly better noisefloor measurements below 500 hz than the Rega P3, reflected by better overall signal to noise ratio measurements and better subjective sound quality.

These measurements indicate that at the the frequencies where the ear is most sensitive, that vinyl has a comparable and possibly better noisefloor than CD. They also indicate that engineering quality of the vinyl pressing and the turntable are important for getting the best noisefloor results from this medium. It's a gross over-simplification to say that CD has a greater dynamic range than vinyl, because it depends upon what frequency we're talking about. If, for example we're talking about 2.5 khz then vinyl is measureably better than CD.

Are you really trying to argue that as a format LP's have more dynamic range than CD's? :roll:

There are two words that explains those differences:

Different Mastering.

 

[davedotco]

I feel that a lot of this discussion is unhelpful due to a lack of true understanding on the part of those involved. I know a little about digital recording and playback but I am no expert. What I do have is a scientific education and a reasonable grasp of how 'science' works.

This leads me to the following conclusions, in scientific terms digital recording and playback to Red Book Standard works. Nyquist-Shannon and other scientific work that takes in such factors as anti-aliasing filters, oversampling and noise shaping make it clear that this process is essentially transparent.

You do not have to take my word for it, digital theory has been tested by peer review time and time again, by people with a far greater understanding of these things than (I suspect) anyone on here. The core theory is so sound that it should be accepted as fact by anyone, you do not need to understand the science, just the basic idea of scientific method. It would take a conspiracy of gigantic proportions for this not to be true.

In addition we have empirical data that is pretty conclusive, pretty much everyone has heard at least one CD that sounds great, showing that the system does work when the record industry can be bothered and there are plenty of documented tests that shows the introduction of digital processing into an otherwise analogue playback chain to be undetectable. Some years ago I heard a demonstration using a CDR to record vinyl playback, it was quite impossible to hear the difference, despite a very fine playback system.

In reality the debate should be moved on to the more important topic of why the record industry produces such poor recordings and how they look to exploit the situation by charging more for quality that could and should be the norm.

Regarding an earlier post,

Measurement of budget (sub £1k) cd players supplied with regular mains power, in a normal shop or home situation, typically produce a s/n ratio of 100 - 110dB below full output, this equates to a bit depth of 17 bits. High end processors, without the noise of a transport, usually do a little better, some resolving 18bits, computers and many streamers fair slightly worse due to increased digital noise.

In Lab conditions with better mains supply and low ambient electrical (RF) noise you might be able to squeeze an extra bit from the best processors.

 

[oldric_naubhoff]

I seem to remember that a jet passenger aeroplane has a dB level of around 120 dB when taking off. If you like and can stand your music that loud then good luck to you and may I suggest that you start learning British Sign Language.

while this is true that hi res audio offers higer DR this is not the main advantage of it. the biggest advantage is you greatly lower quantization error going from 16bit to 24 bit format. here's some numbers to back it up. say you're recording a musical event of dynamic range of 18dB (for ease of calculations). you digitize the input signal into 16bits and 24bits using full scale. in case of 16bit audio that means you use 14,15,16 bit of resolution and in case of 24bit 22,23,24 bit. in case of 16bit format you get 2^16-2^13= 57 344 of discrete volume levels to digitaly capture infinitly varying analog waveform. in case of 24bit format you get 2^24-2^21= 14 680 064 of discrete volume levels, meaning that the quantization error will be much lower.

the fact that higher bit depths offer more discrete amplitude "steps" to represent the analog waveform manifests in its higer DR over Red Book standard. but predominantely this is not the aim of hi res audio to be able to record events of 120dB DR! this is just a less significant by-product of the format. I hope you grasp the idea.

whether this better representation of analog signal as offered by hi res audio format is audible is a completely different matter. various tests show it isn't. but the truth is that 24bit audio is superior to 16bit audio in that it capturs analog waveform more faithfully.

there's one more thing about claimed DR of Red Book. I brought this issue some time before but seems few people grasped it. 16bits equal to roughly 96dB of DR. true. the DR limit lies where quantizing error artifacts completely overshadow input signal at which stage it won't be distinguishable from white noise (this is how quantizing error manifests itself in analog domain). the problem with quantization error is that it's in a way similar to distortion, that you'll be able to hear it along with the reconstructed signal if the bit depth gets shallower. you'll be able to hear quantization artifacts as soon as the signal drops to 6bits of resolution (no dithering applied). that means Red Book standard has only about 10bits of usable DR. applying of dither will grant you a few more bits of DR. but this will not get you more resolution. the purpose of dithering is to mask quantization error within lower noise floor.

by the very same token 24bit audio has easily 18bits of usable DR, no dither applied.

 

[oldric_naubhoff]

Regarding an earlier post,

Measurement of budget (sub £1k) cd players supplied with regular mains power, in a normal shop or home situation, typically produce a s/n ratio of 100 - 110dB below full output, this equates to a bit depth of 17 bits. High end processors, without the noise of a transport, usually do a little better, some resolving 18bits, computers and many streamers fair slightly worse due to increased digital noise.

In Lab conditions with better mains supply and low ambient electrical (RF) noise you might be able to squeeze an extra bit from the best processors.

this one easily offers 21 bits of resolution.

http://www.stereophile.com/content/auralic-vega-da-processor-measurements

and there's another one:

http://www.stereophile.com/content/benchmark-dac2-hgc-da-processorheadphone-amplifier-measurements

and they are not the only examples of so well engineered domestic DACs out there (running on domestic mains power supply ).

 

[NHL]

Regarding an earlier post,

Measurement of budget (sub £1k) cd players supplied with regular mains power, in a normal shop or home situation, typically produce a s/n ratio of 100 - 110dB below full output, this equates to a bit depth of 17 bits. High end processors, without the noise of a transport, usually do a little better, some resolving 18bits, computers and many streamers fair slightly worse due to increased digital noise.

In Lab conditions with better mains supply and low ambient electrical (RF) noise you might be able to squeeze an extra bit from the best processors.

this one easily offers 21 bits of resolution.

http://www.stereophile.com/content/auralic-vega-da-processor-measurements

and there's another one:

http://www.stereophile.com/content/benchmark-dac2-hgc-da-processorheadphone-amplifier-measurements

and they are not the only examples of so well engineered domestic DACs out there (running on domestic mains power supply ).

but the errors are random and not systematic. How would you test the influence?

 

[davedotco]

Regarding an earlier post,

Measurement of budget (sub £1k) cd players supplied with regular mains power, in a normal shop or home situation, typically produce a s/n ratio of 100 - 110dB below full output, this equates to a bit depth of 17 bits. High end processors, without the noise of a transport, usually do a little better, some resolving 18bits, computers and many streamers fair slightly worse due to increased digital noise.

In Lab conditions with better mains supply and low ambient electrical (RF) noise you might be able to squeeze an extra bit from the best processors.

this one easily offers 21 bits of resolution.

http://www.stereophile.com/content/auralic-vega-da-processor-measurements

and there's another one:

http://www.stereophile.com/content/benchmark-dac2-hgc-da-processorheadphone-amplifier-measurements

and they are not the only examples of so well engineered domestic DACs out there (running on domestic mains power supply ).

I am pleased to see that the state of the art has improved since I observed the measurements I detailed above, about 10 years ago.

However the Auralic manages just 20 bits (almost 21 is still 20) and the Benchmark 21 bits, so we are looking at an improvement of one or two bits in processors that are acknowledged as close to the state of the art. My data is still correct for more affordable processors and other digital devices.

From the same source, a Naim Uniti, a real world product, is limited by noise to 17 bits resolution. This is not a dig at Naim or anything, just an example of what we can reasonably expect from more affordable products.

I no longer read Stereophile on a regular basis, but when I did they made quite a big thing of maintaining a good quality, transformer isolated power supply to ensure 'a level playing field' for the product under test. I would be surprised if that was not still the case, for the sake of consistency if nothing else.

 

[lindsayt]

Are you really trying to argue that as a format LP's have more dynamic range than CD's? :roll:

There are two words that explains those differences:

Different Mastering.

Sorry, this is one time that you can't blame different mastering for the areas where vinyl is better.

Read the whole link again.

The measurements were made during the quiet bit just before the start of the title track of that album. The mastering has almost nothing to do with how quiet this quiet bit is. Different mastering kicks in once the music starts with the dynamic contrast between the quieter bits of music on each track compared to the transient peaks. He did different measurements for that and includes his results in the link.

Look at his measurements. Above 2khz his measurements indicate that vinyl has more dynamic range than CD due to the lower noise floor. Below 500 hz CD has more dynamic range than a Rega P3. How much better it would be below 500 hz with a Pioneer Exclusive P3a remains to be measured, although I would still expect CD to measure better at those frequencies.

Are you trying to deny the evidence of your owns eyes? Evidence you can get by looking at the graphs? And, more importantly, are you trying to deny the evidence that you could get with your own ears by comparing CD's and vinyl on a properly engineered record player? Have you ever actually compared CD to vinyl, using a world class record player?

 

[Clare Newsome]

In reality the debate should be moved on to the more important topic of why the record industry produces such poor recordings and how they look to exploit the situation by charging more for quality that could and should be the norm.

Not so much a debate as basic business economics. As I noted earlier in this thread, the record industry is in massive decline.

Why invest time/money in high-quality recordings when most 'product' is destined to earn pennies from music streaming services. If you're lucky you'll comparitvely strike gold with some downloads, radio play fees or even some disc sales. To achieve such revenues you'll probably have to have invested a sizeable sum in marketing, too: that leaves precious litte (if any) profit.

This 'value range' approach to mainstream music is no different to other industries like food or clothing: it's being produced to hit a price-point and be consumed casually.

However, there are consumers that want more - more affluent audiences with particular tastes. That could be classical music fans chasing the ultimate recording of their favourite work; it could be a hipster wanting vinyl. And just like the clothing companies and supermarkets offer premium ranges for their customers, the record industry serves up premium product to those willing to pay.

It's not rocket science - or even mathematic theory - but it is business reality (for good or bad).

 

[jcbrum]

but the truth is that 24bit audio is superior to 16bit audio in that it capturs analog waveform more faithfully.

This statement is false. The analogue waveform reproduced by the converter stage in the DAC is equally accurate in both cases. It is simply that noise is higher (but still inaudibly low) in the 16bit sample, due to greater quantisation error.

Bit depth affects the noise floor, and hence dynamic range, not resolution, in terms of the analogue output of a DAC. 16bits is adequate for human hearing requirements.

the purpose of dithering is to mask quantization error within lower noise floor.

This statement is false. Dithering increases the noise floor, but it is still at a level which is inaudible to humans.

by the very same token 24bit audio has easily 18bits of usable DR, no dither applied.

This comparison is false. All of the bits are used, and indeed are necessary, in the converter stage of a DAC. There are no 'redundant' bits in either case.

The NS Theorem makes this plain and obvious.

JC

 

[lindsayt]

Absolute nonsense. The report compares chalk, cheese, apples and pears. Just take a step back, think about the technology, do you really think that the noise floor of a mechanical system, with a motor, bearings and a stylus scraping along a piece of plastic is going to have a noise floor lower than a digital CD system?

Yes. Above 2khz. No below 500hz with a Rega P3. Based on these measurements and my own listening tests.

The report measured noise floor. Have you done your own noise floor measurements? Do you know of any other reputable source that has done this kind of measurement where the noise and frequency are plotted?

Also think about what generates the signal in a stereo cartridge. It's movement perpindicular to each V shaped groove. If there's no movement there's no signal. If the vinyl is smooth, which it will be during the silent bits there will be no movement caused my the vinyl. If it's clean and unscratched where the tip is tracing the vinyl there will be no movement from dirt or scratches. It's then down to the quality and execution of the engineering as well as isolation from external vibration that will determine the noise floor. A Rega P3 will be OK, but not state of the art in these respects. A well isolated Pioneer Exclusive P3a will be.

What's with all this closed minded dogmatism? Keep an open mind, and if you can, go out and compare CD to vinyl yourself with your own ears, preferably using a world class record player.

 

[jcbrum]

Clare's point has some relevancy.

Thirty or more years ago the likely recording method was DAT 16/48, and distribution was by Compact Disc 16/44.

These days recording is usually 24/96 or greater, and distribution is by download, more often than not.

So if you have made a recording at 24/96, and can distribute it at 24/96, then why not do so ?

The question is a commercial one, whether you can charge more for it.

If you can convince people to pay more, that is perhaps to be seen as a desirable commercial objective.

JC