[The Jester]

Indeed all maths will tell you 1+1=2, 2+2=4 etc,
And the sampling theory is perfectly, mathematically correct, a signal can be digitised and reconstructed perfectly in a perfectly bandwidth limited signal, but it is yet to be achieved in the physical world, sure there’s reams of research on different filtering methods of achieving a perfectly bandwidth limited signal that has no effect on the signal being converted, and recently with the advances in processing power, it’s closer than its ever been, so then it’s argued that “close enough is good enough” to be “audibly transparent” when the human sciences still don’t know all of how we perceive, process and interpret sounds entering the ears,
Don’t get me wrong, measurements are vitally important but they are not everything, until we “know” everything about everything, and that’s down to science, with most things science asks more questions than it answers.

 

[gregorio]

but couldn't accept that even at inhuman level of measured performance e.g. 2 DACs and 2 amps chain that both have -118 dB THD+N or better (i.e. Holo Audio May KTE paired with AHB2 against a D90 with Topping PA5), those gears will sound subtly different from one another

Unless I’ve misunderstood what you’re saying, there’s two problems with this assertion:

Firstly, THD+N is NOT “measured performance”, it’s simply a measurement of one aspect (metric) of performance. For example, let’s say we measure the horse power of a vehicle at 500hp and another vehicle also produces 500hp. The performance of these two vehicles could still be very significantly different, for instance if one were a 500hp Ferrari and the other a 500hp bus. There are obviously other measurements together with horse power which define performance (power to weight ratio, aerodynamics, etc.). Likewise, 2 DACs/amps with -118 THD+N could sound significantly different, unless the all the other metrics of performance also measure the same.

Secondly, I’m not sure there is anyone who “couldn’t accept that” are there? Maybe there are some extremists who believe that THD+N is the only metric of performance but I don’t recall ever seeing one. For example, power output is also one of the metrics of performance and many/most external consumer DACs have at least some difference in their output power (and sometimes a significant difference). Therefore, many/most DACs will sound at least subtly different or even significantly different! And again, I’ve never heard of anyone within the scientific, pro-audio or “objectivist” audiophile community who doesn’t accept this fact. However, a condition of those who claim the vast majority of DACs do sound the same is that they must be accurately volume matched, if they’re not, it’s likely they will sound different.

G

 

[CiptoH]

If you’re excluding noise-shaped dither then 16bit does not have a resolution equivalent to 20bit, it just has it’s actual 16bit resolution and likewise, DSD does not have a resolution equivalent to 20bit, it also just has it’s actual resolution, 1bit. How many times?

Even with a noise-shaped dither, it will never be possible for a standard 16-bit CD format to provide the original 20-bit resolution on a master recording, the CD resolution still 16-bit !
All the original low-level bits below the 16th bit that was removed during the conversion from 24-bit master recording to 16-bit can never be restored-back when playing back a standard CD !

A noise-shaped dither was added to the 16-bit recording to shift noise out of the audible range, and not to store the low-level bits (from the 17th bit to the 20th bit) onto the CD which meant it couldn't be restored-back again during playback.
The standard 16-bit CD format stores data as is and is not compressed and not then expanded again when playing back, unlike the FLAC file format (or ZIP file on a PCs) which uses lossless compression-expansion techniques that can store audio data whose size is larger than the container-file size.

So, if someone says that using a Noise-shaped dither on a 16-bit CD recording will be able to bring out the original contents of the low-level bits (from the 17th bit to the 20th bit) that are in the master record so that it can be said to have a resolution of 20-bit, that's clearly misleading the public !

 

[gregorio]

Indeed all maths will tell you 1+1=2, 2+2=4 etc,
And the sampling theory is perfectly, mathematically correct, a signal can be digitised and reconstructed perfectly in a perfectly bandwidth limited signal, but it is yet to be achieved in the physical world

True, we do not have perfect amps and we certainly do not have perfect transducers or infinite hearing. For example, let’s take a theoretical amp that’s better than any in the world, with noise/distortion at say -140dB. What would be the point of a DAC with noise/distortion at minus infinity? We could never achieve noise/distortion lower than -140dB in the output of any amp no matter how perfect the DAC. The output would be the same whether the DAC had noise/distortion at -160dB or -5,000dB. But even this is hypothetical because no transducers can respond to such a minuscule signal (-140dB), so there is no difference in the physical sound and therefore any question of hearing ability is irrelevant!

so then it’s argued that “close enough is good enough” to be “audibly transparent” when the human sciences still don’t know all of how we perceive, process and interpret sounds entering the ears,

Firstly, if the DAC artefacts cannot even be converted into sound (by amps/transducers) then obviously it MUST be good enough, without even discussing human hearing!

Secondly, we obviously don’t need to know everything in order to know something. We don’t know everything there is to know about mathematics but we do know that 1+1=2. Likewise, we don’t know everything about how humans perceive/interpret sound but we do know a great deal. In fact there’s an entire branch of science specifically dedicated to this, which goes back over 120 years and which has had countless millions of dollars invested in research because trillions of dollars, national security, world health and life and death has depended on it. Are you really suggesting that after more than a century of research, the science of psychoacoustics knows nothing at all and is completely empty? Simple basics like thresholds of hearing not only are very well known but have been for many decades.

until we “know” everything about everything, and that’s down to science, with most things science asks more questions than it answers.

So by that logic, as we don’t know everything about mathematics, then you’re not sure that 1+1=2? We don’t know everything about the universe, so maybe the Earth is flat? We certainly don’t know everything about gravity, so maybe you can jump off a skyscraper and not slam into the ground?

G

 

[The Jester]

Just the sort of ridiculous, dissected reply I expected,
You go on believing digital audio is “audibly perfect” from simple measurements and never listen to the results of continuing research,
Time will tell, and I’m not about to enter into back and forth rhetoric and add to the off topic nature of the last 6 pages,
I voted Yes,
And I can hear differences, sometimes subtle, sometimes more obvious with several different DAC’s.

J
Out.

 

[gregorio]

All the original low-level bits below the 16th bit that was removed during the conversion from 24-bit master recording to 16-bit can never be restored-back when playing back a standard CD !

As you yourself have quoted, dynamic range is defined by bit depth and as you have also quoted, dynamic range is the ratio between full scale and the noise floor. If you lower the digital noise floor (by moving that noise to inaudible regions of the frequency spectrum), you increase the dynamic range and therefore effectively increase the bit depth.

Noise-shaped dither obviously cannot restore the detail below the 16th bit if those bits have already been removed but that is NOT when noise-shaped dither is applied. It’s applied during the process of converting from 20bit (or higher bit depths) to 16bit, BEFORE those additional bits have been removed!

if someone says that using a Noise-shaped dither on a 16-bit CD recording will be able to bring out the original contents of the low-level bits (from the 17th bit to the 20th bit) that are in the master record so that it can be said to have a resolution of 20-bit, that's clearly misleading the public !

No, it’s not! The content in the 17th-20th bits cannot be revealed in a 16bit recording because the digital noise floor of 16bit obscures it but if we use noise-shaped dither when converting from 20bit, the digital noise floor is lowered and that content is revealed. It’s easy to test this for yourself, with free software! Simply create a signal at say -106dB (10dB below the 96dB limit of 16bit), record it to a 16bit file with noise-shaped dither, then turn the volume up by 50dB or more and see for yourself if that signal exists or not.

G

 

[gregorio]

Just the sort of ridiculous, dissected reply I expected,

Sure, simple logic and basic facts are ridiculous. 1+1=2 but not always because we don’t know everything about maths, got it.

You go on believing digital audio is “audibly perfect” from simple measurements and never listen to the results of continuing research,

I’ll go on believing that digital audio is “audibly perfect” because it’s been proven to be so, I work with it every day and there are no results of research that prove otherwise or demonstrate human hearing is vastly different to what science has already determined. You carry on believing digital audio is not audibly transparent because of audiophile marketing/myths.

Time will tell

How much more time do you want? Digital audio has been around for 70 years, even before stereo vinyl was released to the public!

G

 

[CiptoH]

You go on believing digital audio is “audibly perfect” from simple measurements and never listen to the results of continuing research,

Not all aspects that can be heard by humans have already been measured.
The simplest example is the width, height and depth of the stage from a live music performance (or the result of reproducing music recordings via loudspeakers or headphones) that is felt by the brain, there is no standard measurement yet.

 

[gregorio]

I voted Yes

I voted yes too. I once did a controlled listening test with a filterless NOS DAC. Every one of the other engineers taking the test also got 100%.

G

 

[gregorio]

Not all aspects that can be heard by humans have already been measured.

But all aspects of what can be recorded and reproduced have been measured, by definition!

The simplest example is the width, height and depth of the stage from a live music performance (or the result of reproducing music recordings via loudspeakers or headphones) that is felt by the brain, there is no standard measurement yet.

The simple example of width, height and depth within a stereo recording, is an illusion and we cannot easily measure peoples’ perceptions of this illusion. However, we can of course measure the audio signal which contains the properties which influence that perception (balance, initial reflections, etc.) because of course that’s what a digital audio recording is!

G

 

[CiptoH]

Noise-shaped dither obviously cannot restore the detail below the 16th bit if those bits have already been removed but that is NOT when noise-shaped dither is applied. It’s applied during the process of converting from 20bit (or higher bit depths) to 16bit, BEFORE those additional bits have been removed!

No, it’s not! The content in the 17th-20th bits cannot be revealed in a 16bit recording because the digital noise floor of 16bit obscures it but if we use noise-shaped dither when converting from 20bit, the digital noise floor is lowered and that content is revealed. It’s easy to test this for yourself, with free software! Simply create a signal at say -106dB (10dB below the 96dB limit of 16bit), record it to a 16bit file with noise-shaped dither, then turn the volume up by 50dB or more and see for yourself if that signal exists or not.

It would be incorrect to say that the low bits below the 16th bit that were discarded during the conversion from 24-bit to 16-bit can be recovered exactly 100% of the original bits on the master record when using noise-shaped dither, it never will !

If you say otherwise, I want to ask, where on the CD is the place to store the low bits (17th to 20th bits) on the master record that were discarded during the conversion 24-bit to 16-bit ?

CD format can only store 16-bit data and can't store 20-bit data, noise-shaped dither will not be able to restore the low-level bits that have been discarded during conversion to be 100% the same as the original.

Remember, if the low bits (17th bit to 20th bit) restored during playback are not exactly 100% the same as the original on the master record, this means that a CD resolution with noise-shaped dither cannot be called a true 20-bit-resolution recording, but only fake 20-bit-resolution recordings !

 

[gregorio]

It would be incorrect to say that the low bits below the 16th bit that were discarded during the conversion from 24-bit to 16-bit can be recovered exactly 100% of the original bits on the master record when using noise-shaped dither, it never will !

No, it would be incorrect to say that the low bits below the 16th bit that were discarded during conversion from 24bit to 16bit can be recovered!
See, I can do that too!

this means that a CD resolution with noise-shaped dither cannot be called a true 20-bit-resolution recording

Who are you arguing with? Who has said that a CD with noise-shaped dither is a “true” 20bit recording? I (and Wikipedia) have said that a CD with noise-shaped dither has a resolution equivalent to a true 20bit recording, not that it is a true 20bit recording. Likewise, 1bit DSD also has a resolution equivalent to a 20bit recording but it isn’t a true 20bit recording, it’s actually just a 1bit recording.

It’s clear you do not understand the fundamental basics of digital audio. For example, the relationship between bit depth, resolution and dynamic range (despite having quoted it yourself!), nor the role of noise shaped dither.

G

 

[CiptoH]

No, it would be incorrect to say that the low bits below the 16th bit that were discarded during conversion from 24bit to 16bit can be recovered!
See, I can do that too!

I have said before, original bits information that has been removed during conversion cannot be restored even if you use noise-shaped dither !

Never will CD digital data with 16-bit resolution (65,536 steps of voltage) with noise-shaped dither be able to become digital data with 20-bit resolution (1,048,576 steps of voltage) even if the software is processed with noise-shaped dither !

Don't forget, the original standard for CD player hardware and software only supports digital data with 16-bit resolution (65,536 steps of voltage) and not digital data with 20-bit resolution (1,048,576 steps of voltage), which means CD software recorded with noise-shaped dither when played it will not suddenly change from 16-bit resolution (65,536 steps of voltage) to 20-bit resolution (1,048,576 steps of voltage), because this is not supported by standard CD player hardware !

Who are you arguing with? Who has said that a CD with noise-shaped dither is a “true” 20bit recording? I (and Wikipedia) have said that a CD with noise-shaped dither has a resolution equivalent to a true 20bit recording, not that it is a true 20bit recording. Likewise, 1bit DSD also has a resolution equivalent to a 20bit recording but it isn’t a true 20bit recording, it’s actually just a 1bit recording.

A CD software with a noise-shaped dither can be perceived as having a 20-bit equivalent dynamic-range because the noise is lower (-120dB), but the resolution is still 16-bit (65,536 steps of voltage) and not 20-bit resolution (1,048,576 steps of voltage) because standard CD player hardware and software only supports 16-bit resolution (65,536 steps of voltage) and not 20-bit resolution (1,048,576 steps of voltage) !

It’s clear you do not understand the fundamental basics of digital audio. For example, the relationship between bit depth, resolution and dynamic range (despite having quoted it yourself!), nor the role of noise shaped dither.

Can you please explain how can CD digital data which is 16-bit resolution (65,536 steps of voltage) recorded to a CD disc can suddenly turn into digital data with 20-bit resolution (1,048,576 steps of voltage) when playing CD software processed with noise-shaped dither ?

This is impossible for standard CD player hardware and software.

In other words, if the digital audio data on a CD disc has a 16-bit resolution (65,536 steps of voltage), the CD player transport is 16 bit (65,536 steps of voltage) and the DAC is also 16-bit (65,536 steps of voltage) then the output resolution of the CD player is will remain 16-bit (65,536 steps of voltage) and will not be 20-bit (1,048,576 steps of voltage) even if the CD software is processed with noise-shaped dither.
It is possible to convert 16-bit resolution data (65,536 steps of voltage) to 20-bit resolution data (1,048,576 steps of voltage), for example with the AK4137EQ AKM Sampling Rate Converter (SRC) chip and a 20/24-bit DAC, but this is not pure 20-bit resolution (1,048,576 steps of voltage) anymore.

 

[gregorio]

I have said before, original bits information that has been removed during conversion cannot be restored even if you use noise-shaped dither !

So have I, so shall we both keep repeating it over and over?

G

 

[sup27606]

@CiptoH , @gregorio , you guys should really take this discussion to the audio science forum or PM, otherwise you risk shutting down the entire thread. During heated discussions, we all get carried away sometimes, thats all right, but now it’s time to put a brake.

This is a thread where I personally am more interested in hearing different people’s experiences with diverse DACs, even if its subjective.