Upsampling 16/44.1 collection a good idea?

[terryforsythe]

This does not alter the quantisation noise already embedded in the input signal.

Dithering is used to address that. (Dither is an intentionally applied from of noise used to randomize quantization error.)

 

[CapMan]

I think that's exactly why ASR exisits.

A lifetime in business suggest the accountants have more say than the engineers!

 

[antcollinet]

Dithering is used to address that.

That dithering can only be applied at the time of quantisation - IE during the ADC For the original digital file. Any shaping done by a DAC is only shaping the quantisation noise IT is creating as part of its DSP. Once noise is added within the bandwidth of a signal it cannot be subsequently removed.

Again - I’m done here.

 

[CapMan]

I hope as customers, we can also have some say (with help from ASR)

Me too

 

[pkane]

Sorry for my ignorance. Does it mean that we can only discuss an actual problem or issue but not how to improve the current solution for an existing problem?

Looking for a solution to an existing problem implies a problem exists and isn't solved satisfactorily. Nearly everything discussed in this thread is a solution looking for a problem, not the other way around.

 

[terryforsythe]

That dithering can only be applied at the time of quantisation - IE during the ADC For the original digital file.

I believe that is an incorrect assumption. If you have a reference that supports that contention, I'll look at it. But, I found numerous references that state the opposite.

Again - I’m done here.

That's fine.

 

[CapMan]

Looking for a solution to an existing problem implies a problem exists and isn't solved satisfactorily. Nearly everything discussed in this thread is a solution looking for a problem, not the other way around.

Yep - intense navel gazing !

 

[pkane]

I am not satisfied with the DAC's internal oversampling and trying to look for a different solution.

And I'm suggesting that the problem is most likely in how you evaluate equipment, not in the equipment itself. Until you learn how to properly compare two devices for real audio differences, you'll be stuck in a loop, looking for solutions to imaginary problems.

 

[pkane]

I agreed the highlighted part. To me (again my personal experience), I am not satisfied with the DAC's internal oversampling and trying to look for a different solution.

As a prerequisite, I believe that I need to find out more how the existing DAC works in order to find a different approach to the solution.

And I'm suggesting that the problem is most likely in how you evaluate equipment, not in the equipment itself. Until you learn how to properly compare two devices for real audio differences, you'll be stuck in a loop, looking for solutions to imaginary problems.

Something to get you started:

 

[MaxwellsEq]

I guess you never used a particularly good lens then..
But yeah we're not there technologywise that we can capture and display images in all dimensions that exceeds our eyes capabilities, so I am talking about theory mostly, and also me working in 3D I see it from that "perfect" perspective as well. But the digital sampling theorem still very much applies just the same and it's all constrained by our human senses, which is what I have been trying to explain and compare.
Anyways, I don't really have neither the time or every to keep on with this conversation, I have explained my point of view that I still fully stand by and the person I was mostly talking to seem to have understood it so

27.6 Limits of Resolution: The Rayleigh Criterion – College Physics

pressbooks.online.ucf.edu

 

[tmtomh]

Sorry for my ignorance. Does it mean that we can only discuss an actual problem or issue but not how to improve the current solution for an existing problem?

No need to say sorry, and I apologize if my original comment came off as a bit cranky. Of course you, @terryforsythe , and anyone else who's interested should feel free to discuss whatever audio-related issues you like, to enjoy, and to learn. I'm just a rank amateur myself, and a good deal of what I know - and more specifically a good deal of what I now know with a bit more clarity and precision than I used to - has come from reading here, learning from others here, and asking questions here.

One thing I have learned here is that while the line is not always entirely clear and cut-and-dried, there are quite a lot of areas of hi-fi reproduction, especially in the digital realm, that are solved problems at this point: the level of performance achievable with competently designed, decently built gear, including digital sources and DACs, is sufficient to exceed the limits of human hearing.

One other thing I have learned - and that I often observe - is that "a little knowledge is dangerous," as the saying goes. Of course not literally dangerous in this case, since this is just a fun hobby. But just because we know that high-res files have smaller quantization errors aka lower noise floor than CD-quality files does not tell us anything about whether software resampling of digital sources before they enter the DAC is an actual issue for the hi-fi reproduction of music.

So putting these two points together, there's nothing at all wrong about asking and learning about resampling - that was the OP's question of course. But I and others like @pkane are going to point out if and when folks assume that it's a problem, and then all the learning and question-asking is about how to "solve" that problem. I'm not saying this is necessarily what you are doing here; just trying to explain where my comments have been coming from.

 

[Tell]

27.6 Limits of Resolution: The Rayleigh Criterion – College Physics

pressbooks.online.ucf.edu

Yes diffraction is a thing, but you still don't seem to understand the point of my post. Never mind, have nice day etc.

 

[danadam]

View attachment 364897
Source: Oversampling - Wikipedia

The majority of the article describes reasons for oversampling in ADC. Then, in this last paragraph, it says that the term "oversampling" is also used to describe a process in DA conversion. I don't see any basis to conclude that the reasons described earlier apply here too? It's rather clear that the main reason for oversampling in the context of DA conversion is to allow for less complex analog reconstruction filter.

I can also agree that it improves SNR but only to the extent that @antcollinet described it, i.e. it will improve results of DSP calculations but it won't affect the noise baked in the input signal.

Delta-sigma (ΔΣ; or sigma-delta, ΣΔ) modulation is an oversampling method for encoding signals into low bit depth digital signals at a very high sample-frequency as part of the process of delta-sigma analog-to-digital converters (ADCs) and digital-to-analog converters (DACs).

In the context of DACs this only means that you can take 16/44k input, convert it to "low bit-depth/high sampling-frequency" format and still retain the 96 dB SNR (in 0-22k band) of the original 16-bit input. You won't get more than 96 dB SNR (in 0-22k band) even if you convert to "high bit-depth/high sampling-frequency". You can't magically recover information that was lost when 16/44k file was created.

I believe that is an incorrect assumption. If you have a reference that supports that contention, I'll look at it. But, I found numerous references that state the opposite.

Rather you interpret those references as stating the opposite. How about an example instead of reference?

• File 01 is -20 dBFS 1 kHz tone generated in 8/44k.
• File 02 is the file 01 upsampled to 8/352k.
• File 03 is -20 dBFS 1 kHz tone generated in 8/352k.

Is the noise reduced through oversampling? No, because that's physically impossible. Neither file 01 nor 02 has the same audible noise floor as 03:

 

[terryforsythe]

Rather you interpret those references as stating the opposite.

Please provide your interpretation of the references or, better yet, a scientific paper that supports antcollinet's contention "That dithering can only be applied at the time of quantisation - IE during the ADC For the original digital file."

____________________________________________________________

https://www.renesas.com/us/en/document/apn/rl78-dac-dithering-application-note-rev100

"This app note describes a method to increase the 8-bit DAC resolution to 12-bit effective resolution using a DAC amplitude dithering method..."

"When implementing the DAC dithering method shown in this app note..."

"By modulating or dithering the DAC output values over a longer repeat period, with different dither patterns, more equivalent resolution bits can be obtained."

____________________________________________________________

https://ispg.ucsd.edu/wordpress/wp-content/uploads/2017/05/1993-ISCAS-I.-Galton-One-Bit-Dithering-in-Delta-Sigma-Modulator-Based-D-A-Conversion.pdf

"One-Bit Dithering in Delta-Sigma Modulator-Based D/A Conversion"

____________________________________________________________

https://www.analog.com/media/en/training-seminars/tutorials/mt-017.pdf :

"OVERSAMPLING INTERPOLATING DACS"
...
"Also, since the quantization noise is spread over a wider region with respect to the original signal bandwidth, an improvement in the signal-to-noise ratio is also achieved. By doubling the original sampling rate (K = 2), an improvement of 3 dB is obtained, and by making K = 4, an improvement of 6 dB is obtained..."

_____________________________________________________________

https://en.wikipedia.org/wiki/Delta-sigma_modulation :

"Noise shaping: moving noise to higher frequencies above the signal of interest, so they can be easily removed with low-pass filtering."

 

[pkane]

Please provide your interpretation of the references or, better yet, a scientific paper that supports antcollinet's contention "That dithering can only be applied at the time of quantisation - IE during the ADC For the original digital file."

Dither can (and should) be applied any time quantization error is introduced, usually when fitting a result with more bits of precision into a smaller sample size.

 

[danadam]

antcollinet's contention "That dithering can only be applied at the time of quantisation - IE during the ADC For the original digital file."

I believe he said more than that, both in the message you quoted and in earlier messages.

https://www.analog.com/media/en/training-seminars/tutorials/mt-017.pdf

"Also, since the quantization noise is spread over a wider region with respect to the original signal bandwidth, an improvement in the signal-to-noise ratio is also achieved. By doubling the original sampling rate (K = 2), an improvement of 3 dB is obtained, and by making K = 4, an improvement of 6 dB is obtained..."

The next sentence after above quote says:

"Early CD players took advantage of this, and generally carried the arithmetic in the digital filter to more than N-bits".

So the increased SNR was used for more accuracy in DSP calculactions. As discussed earlier.

And the sentence after that: "Today, most DACs in CD players are sigma-delta types.", it relates to sigma-delta, so see below.

https://ispg.ucsd.edu/wordpress/wp-content/uploads/2017/05/1993-ISCAS-I.-Galton-One-Bit-Dithering-in-Delta-Sigma-Modulator-Based-D-A-Conversion.pdf

"One-Bit Dithering in Delta-Sigma Modulator-Based D/A Conversion"

https://en.wikipedia.org/wiki/Delta-sigma_modulation :
"Noise shaping: moving noise to higher frequencies above the signal of interest, so they can be easily removed with low-pass filtering."

These describe inner-workings of Delta-Sigma converters. Those converters are 1-bit or "a few"-bits converters. This means that the bitdepth of the incoming signal, which is usually >= 16-bit, has to be reduced. To preserve the SNR of the input signal (preserve, not improve), dither and noise-shaping has to be used. This is the second quantization or requantization, which I think also was mentioned before.

https://www.renesas.com/us/en/document/apn/rl78-dac-dithering-application-note-rev100

"This app note describes a method to increase the 8-bit DAC resolution to 12-bit effective resolution using a DAC amplitude dithering method..."

This one is strange. It doesn't seem to be related to audio. In fact the DAC is part of a microcontroller for "Motor Control, Industrial and Metering Applications". They describe how to obtain more than 256 levels from 8-bit DAC and they call it increasing resolution to 12-bit. In the audio-world we know that an 8-bit file can have any number of levels and we still call it an 8-bit file. So I'm rather confused and I'm not sure how it relates to the current discussion.


So my interpretation is that yes, the dither and the quantisation applied in ADC determine the noise level in the input signal. Any further operations/requantizations of the signal can only increase that level, they cannot reduce it.

 

[tmtomh]

Do we know how many bits are employed by the delta-sigma modulator in modern DAC chips?

It varies. Why the eye roll?

 

[terryforsythe]

I believe he said more than that, both in the message you quoted and in earlier messages.

He did, but I believe the confusion was based on the portion I quoted. Specifically, I interpreted the statement "That dithering can only be applied at the time of quantisation - IE during the ADC For the original digital file" (emphasis added) to mean he asserts that dithering only can be applied during ADC, and cannot be applied during DAC. That is incorrect. The references I cited were to support that position.

These describe inner-workings of Delta-Sigma converters.

Delta-Sigma modulators are used in audio DACs. E.g.,

https://liu.diva-portal.org/smash/get/diva2:1745070/FULLTEXT01.pdf

Discrete R2R dac vs. sigma delta dac • Magna Hifi

magnahifi.com

This one is strange. It doesn't seem to be related to audio.

Again, that reference was cited to support the position that dithering can be applied in a DAC.

This one is audio specific: https://liu.diva-portal.org/smash/get/diva2:1745070/FULLTEXT01.pdf
Also, note that this paper also discusses using a digital delta-sigma modulator in a "DAC for Hi-Fi Audio", and in the delta-sigma modulator implementing a noise shaping loop using dithering. (See section 4).

 

[tmtomh]

The eye roll because I am thinking (sorry if you have a different interpretation of if). By the way, what do you think of the meaning of "the eye roll'? I just want to be in sync with you guys if possible.

Thanks for your reply. My understanding of the eye roll is to communicate skepticism, not thinking. But now that you've clarified what you mean by it, of course feel free to keep using it that way if you like!

 

[terryforsythe]

So my interpretation is that yes, the dither and the quantisation applied in ADC determine the noise level in the input signal. Any further operations/requantizations of the signal can only increase that level, they cannot reduce it.

I overlooked responding to this last comment. Your interpretation is inconsistent with the evidence. See section 4.1 of "Designing and Evaluating a Delta-Sigma DAC for Hi-Fi Audio" (https://liu.diva-portal.org/smash/get/diva2:1745070/FULLTEXT01.pdf). Specifically, that section describes noise shaping in the DAC, and Figure 4.1 shows the resulting reduction in quantization noise in the audio spectrum (fs/2). Those results correlate with the various other references I cited explaining how noise shaping is used to shift quantization noise to a higher frequency so that it then can be filtered out. Here are the relevant passages: