• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

High Resolution Audio?

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
Resolution and Accuracy are terms that are often interchanged when the performance of
an ADC is discussed. It is important to note that Resolution does not imply Accuracy nor
does Accuracy imply Resolution.
The resolution of ADC is determined by the number of bits it uses to digitize an input
signal. For a 16-bit device the total voltage range is represented by 216 (65536) discrete
digital values or output codes. Therefore the absolute minimum level that a system can
measure is represented by 1 bit or 1/65536th of the ADC voltage range.
The accuracy of the A/D converter determines how close the actual digital output is to the
theoretically expected digital output for a given analog input. In other words, the
accuracy of the converter determines how many bits in the digital output code represent
useful information about the input signal.
As explained earlier, for a 16-bit ADC resolution the actual accuracy may be much less
than the resolution because of internal or external error sources. So for example a given
16-bit ADC may only provide 12 bits of accuracy. In this case, the 4 LSb’s (Least
Significant Bit) represent random noise produced in the ADC.
There are DACs with 24 bit precision. As in can a signal be adjusted by the least significant digital accurately. Can I raise and lower the volume at the output by 1/16,777,215 th. of full scale. Yes, that can be done. Can they output the 4 LSB's, yes they can, I've already shown you that, and the other link on the RME shows you that. Will that output be swamped in noise? Yes, but that is due to inherent noise in the analog side of things which cannot be avoided. The DAC itself is doing the job it should. As for 16 bits, if a DAC can't do better than 12 bits accuracy it is a lousy DAC these days. All those you linked over at SBAF are ladder DACs. Lousy DACs. Guess why delta sigma replaced ladder DACs? Because of those problems with LSBs in ladder DACs.

I'm beginning to agree with SIY here. You obviously don't understand this, but also refuse to listen. You've put all kinds of objections that DACs can't do this or can't do that. When they can, you just weren't aware of it. What knowledge or experience has you so dead certain that most DACs are inaccurate and can't do what it is they are supposed to do?

And what is the thing with wanting them to do it at 40 khz? You just moved the target because you think it is more difficult to accomplish. It isn't despite what you think.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
@solderdude
I don't say you need that.
I say that high res provide this precision but if no dacs can use it there is no need to ask if the format is good or not as no dacs can render the extra precision provided by the high res format. (At least not fully)

First we should find the limit of the dac.
If the format give more than the limits there is no need of the format.

It's known that delta sigma looses precision as frequency is higher.
Modern one does good up to 20khz it seems but how at 40khz?
How does it do at 40 khz? Just fine.

Here is what is more worth thinking about. Limits of what is recorded. And limits of what humans hear. If the format exceeds both of those then why a higher res format.
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
There are DACs with 24 bit precision. As in can a signal be adjusted by the least significant digital accurately. Can I raise and lower the volume at the output by 1/16,777,215 th. of full scale. Yes, that can be done. Can they output the 4 LSB's, yes they can, I've already shown you that, and the other link on the RME shows you that. Will that output be swamped in noise? Yes, but that is due to inherent noise in the analog side of things which cannot be avoided. The DAC itself is doing the job it should. As for 16 bits, if a DAC can't do better than 12 bits accuracy it is a lousy DAC these days. All those you linked over at SBAF are ladder DACs. Lousy DACs. Guess why delta sigma replaced ladder DACs? Because of those problems with LSBs in ladder DACs.

I'm beginning to agree with SIY here. You obviously don't understand this, but also refuse to listen. You've put all kinds of objections that DACs can't do this or can't do that. When they can, you just weren't aware of it. What knowledge or experience has you so dead certain that most DACs are inaccurate and can't do what it is they are supposed to do?

And what is the thing with wanting them to do it at 40 khz? You just moved the target because you think it is more difficult to accomplish. It isn't despite what you think.
I just say that high res audio is meaningless if the dacs cannot exploit the added resolution in frequency and in level compared to a cd format. What's wrong? Obviously you don't want to say that 32 bit 764khz is pure marketing. Maybe 20bit until 20khz is more close to the truth. Then no need for high res format if the dac is limited to hd cd quality.
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
How does it do at 40 khz? Just fine.

Here is what is more worth thinking about. Limits of what is recorded. And limits of what humans hear. If the format exceeds both of those then why a higher res format.
Well no test shows that any dac has such precision at those frequencies.
If format cannot be completely decoded by the dac there is no need to use it or to wonder if we can hear it.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
Well no test shows that any dac has such precision at those frequencies.
No you mean you aren't aware of any that do. So if we show you such results is everything okay then? What is your next complaint, that we can't undo the laws of physics to get to 24 bit levels of low noise?
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
Two reasons to avoid high res.
- dacs can read it but cannot exploit much more than equivalent of hd cd quality
- if dacs could exploit all the data most human hearing won't notice the difference

Imo second point maybe some sensitive human could detect the difference
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
No you mean you aren't aware of any that do. So if we show you such results is everything okay then? What is your next complaint, that we can't undo the laws of physics to get to 24 bit levels of low noise?
I read some paper on ess and how they could have good resolution until 20khz and how it was difficult.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
I just say that high res audio is meaningless if the dacs cannot exploit the added resolution in frequency and in level compared to a cd format. What's wrong? Obviously you don't want to say that 32 bit 764khz is pure marketing. Maybe 20bit until 20khz is more close to the truth. Then no need for high res format if the dac is limited to hd cd quality.
I'm not pushing hires. However the reason it isn't needed is not because there are no DACs that can perform with those formats.

And yes, I most definitely would say 32 bit 784 khz is pure marketing. There is an edge case to be made for 96 khz. I don't see one at all for more than that for music use. 96/16 with noise shaped dither would give you all you can make use of down to about -120 db.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
I read some paper on ess and how they could have good resolution until 20khz and how it was difficult.
Don't know what you read, but ESS must have figured it out. You seem confused on resolution in bits and frequency range.
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
@Blumlein 88
If we step further into the ES9018, we see that each of the 6-bit converters are actually an array of 64 equally-weighted 1 bit converters. Using an array of 1-bit converters eliminates the linearity errors that always occur when using binary weighted elements.

The maximum 6-bit code is represented by turning all 64 of the 1-bit elements on. The minimum 6-bit code is represented by turning all 64 elements off. The half-scale (code 32) is represented by turning 32 elements on.

While it is easier to match the sizes of 1-bit elements than it is to match a more -conventional binary-weighted ladder array, it is still impossible to match 64 individual elements with enough precision to deliver 20 to 24-bit accuracy. 21-bit accuracy would require a matching of 1 part in 2,097,152 (0.00005%). The ES9018 solves this problem byrandomly selecting which of the 64 1-bit converters will get turned on. This random mapping changes on every clock cycle and this eliminates the linearity errors that would have been caused by slight mismatches in the size of the 1-bit elements. The matching errors create a small amount of noise above 1.7 MHz instead of producing in-band distortion products. The 1.7 MHz out-of-band noise is easily removed by the analog low-pass filters in the DAC2output stage. The result is an output signal with extremely low THD+N.

The net result is that the 6-bit converters have the near-perfect linearity of a 1-bit converter while achieving an 18 dB reduction in noise (due to the 64:1 parallel structure). This improvement delivers a 6-bit sigma-delta modulator that has an 18 dB noise advantage over a classic 1-bit sigma-delta converter (such as that used in DSD).

The array of 1-bit converters also allows native DSD conversion with digital volume control. This combination of features is very unusual, but the ES9018 provides a unique solution to the DSD volume control problem. Normally, it is very difficult to implement a digital volume control (or any other form of digital processing) in a 1-bit DSD system, but with an array of 1-bit converters, we can set the volume by controlling how many DSD converters are turned on.

Time for a little math: Each channel of the DAC2 uses 4 fully-differential converters. Each half of one differential converter has 64 1-bit converters. This means that each output on the DAC2 is derived from 64*2*4=512 1-bit converters operating at megahertz sampling rates. These 512 1-bit converters are the equivalent of a 9-bit converter with perfect linearity. The DAC2 essentially has a 9-bit sigma-delta converter for each channel. The sigma-delta modulation is running at a very high oversampling ratio and is driven by 32-bit data. In the DAC2 this brute-force redundancy is followed by precision analog processing. Together these digital and analog elements give the DAC2 its industry-leading THD and noise performance.

https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing
 

edechamps

Active Member
Joined
Nov 21, 2018
Messages
167
Likes
380
Location
London, United Kingdom

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
@Blumlein 88
If we step further into the ES9018, we see that each of the 6-bit converters are actually an array of 64 equally-weighted 1 bit converters. Using an array of 1-bit converters eliminates the linearity errors that always occur when using binary weighted elements.

The maximum 6-bit code is represented by turning all 64 of the 1-bit elements on. The minimum 6-bit code is represented by turning all 64 elements off. The half-scale (code 32) is represented by turning 32 elements on.

While it is easier to match the sizes of 1-bit elements than it is to match a more -conventional binary-weighted ladder array, it is still impossible to match 64 individual elements with enough precision to deliver 20 to 24-bit accuracy. 21-bit accuracy would require a matching of 1 part in 2,097,152 (0.00005%). The ES9018 solves this problem byrandomly selecting which of the 64 1-bit converters will get turned on. This random mapping changes on every clock cycle and this eliminates the linearity errors that would have been caused by slight mismatches in the size of the 1-bit elements. The matching errors create a small amount of noise above 1.7 MHz instead of producing in-band distortion products. The 1.7 MHz out-of-band noise is easily removed by the analog low-pass filters in the DAC2output stage. The result is an output signal with extremely low THD+N.

The net result is that the 6-bit converters have the near-perfect linearity of a 1-bit converter while achieving an 18 dB reduction in noise (due to the 64:1 parallel structure). This improvement delivers a 6-bit sigma-delta modulator that has an 18 dB noise advantage over a classic 1-bit sigma-delta converter (such as that used in DSD).

The array of 1-bit converters also allows native DSD conversion with digital volume control. This combination of features is very unusual, but the ES9018 provides a unique solution to the DSD volume control problem. Normally, it is very difficult to implement a digital volume control (or any other form of digital processing) in a 1-bit DSD system, but with an array of 1-bit converters, we can set the volume by controlling how many DSD converters are turned on.

https://benchmarkmedia.com/blogs/application_notes/inside-the-dac2-part-2-digital-processing
1560117553629.png


Do you realize that very nearly all delta sigma DACs are a made with a few bits. One bit types haven't been used for quite some years alone. Most DAC chips are 4-7 bits with various architectures to handle the issues with how they work.

Do you realize that DSD as in DSD64 is not the same as Delta Sigma converters?
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
You see there are lot of tricks. So we don't know if tests show all and if those tricks have drawbacks or how they perform with high res.
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
View attachment 27444

Do you realize that very nearly all delta sigma DACs are a made with a few bits. One bit types haven't been used for quite some years alone. Most DAC chips are 4-7 bits with various architectures to handle the issues with how they work.

Do you realize that DSD as in DSD64 is not the same as Delta Sigma converters?
I know dacs are multibit delta sigma and that dsd is like 1bit ds.
You say that modern dacs can exploit all the data of 24bit/192khz file??
I just say it's more like 20bit up to 20khz. Upper i got no idea how they do the job.
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
I know dacs are multibit delta sigma and that dsd is like 1bit ds.
You say that modern dacs can exploit all the data of 24bit/192khz file??
I just say it's more like 20bit up to 20khz. Upper i got no idea how they do the job.
So you say 20 bit up to 20 khz and then say you have no idea how they do the job. That is okay, nobody knows everything. But why are you so set on the idea these DACs can't do above 20 khz correctly. They can, I could do the tests for you, but it is a bother and you still won't accept it from what you've demonstrated here so far.
 

Calexico

Active Member
Joined
May 21, 2019
Messages
269
Likes
17
It can be (and has been) readily measured. And posted here repeatedly. Not that this will have much meaning for you.
No dac can do more than 21 bit so do not pretend that the 24 bits of high res are exploited. I guess for 40khz there won't be 21bit. If i'm wrong i would be very surprised
 

Blumlein 88

Major Contributor
Joined
Feb 23, 2016
Messages
5,698
Likes
4,709
No dac can do more than 21 bit so do not pretend that the 24 bits of high res are exploited. I guess for 40khz there won't be 21bit. If i'm wrong i would be very surprised
Do you know about Johnson_nyquist noise or thermal noise in resistances? And do you see why it limits the level of noise possible? Do you see that a DAC can output perfectly all the way down to the 24th bit and you can do nothing about the fact that output is in much higher levels of noise? And you do realize 21 bits is 126 db of dynamic range so exactly how is this limiting what a human can hear? And if you understood the measurements posted show when the noise is filtered at least some DACs are putting out the correct signal down to 23 or 24 bits.

https://en.wikipedia.org/wiki/Johnson–Nyquist_noise
 
Top Bottom