Is Digital Audio Transmission Analog? [video]

[Vacceo]

BTW, if you can stand the acid reflux of browsing Darko´s page (I hate to increase his traffic), there is a letter from Erin next to Amir´s. I will not give Erin crap; he may be more diplomatic, he may have a chrematistic interest and so on, but he does measure and he provides data, thus setting him in a different galaxy than the Darkos of this world.

It´s very interesting to see that, once boiled down, his message tells the same story as Amir´s yet the answer is completely different. To answer what measures can and cannot tell us: they tell us if a product is well designed and executed and what it does; they cannot tell us the amount of bootlicking you want to provide to manufacturers. I understand, we all have to make a living and an apartment in Berlin is not cheap. It´s a city with a long history of looking the other way when con men speak, I guess everything ends up permiating.

 

[JustJones]

If Amirm bans himself from ASR he might get an invite.

As an aside, if you are concerned I’m just an Amir/ASR clone who believes solely in data and nothing else, the answer is “no”. I disagree with Amir on a lot of topics. I’m even banned from his forum.

Me:

Hi Erin

Thanks for your email. I watch your channel and yes, I’d be more than happy to have you on the podcast later in the year. Perhaps we could run with the same theme: what the measurements can and cannot tell us about a loudspeaker’s sound? What do you think?

 

[DonR]

BTW, if you can stand the acid reflux of browsing Darko´s page (I hate to increase his traffic), there is a letter from Erin next to Amir´s. I will not give Erin crap; he may be more diplomatic, he may have a chrematistic interest and so on, but he does measure and he provides data, thus setting him in a different galaxy than the Darkos of this world.

It´s very interesting to see that, once boiled down, his message tells the same story as Amir´s yet the answer is completely different. To answer what measures can and cannot tell us: they tell us if a product is well designed and executed and what it does; they cannot tell us the amount of bootlicking you want to provide to manufacturers. I understand, we all have to make a living and an apartment in Berlin is not cheap. It´s a city with a long history of looking the other way when con men speak, I guess everything ends up permiating.

oh my.... Godwin's Law is never far away.

 

[Vacceo]

oh my.... Godwin's Law is never far away.

Who said I meant certain Austrian corporal? Aren´t you aware of who Comrade Honeker and Comrade Ulbrich were? Any idea about the claims of Wilhelm Reich?

 

[fpitas]

Who said I meant certain Austrian corporal? Aren´t you aware of who Comrade Honeker and Comrade Ulbrich were? Any idea about the claims of Wilhelm Reich?

Sure; but you know who else...oh, never mind

 

[Vacceo]

Sure; but you know who else...oh, never mind

If you mean those guys, Hjalmar Schacht was even more ahead of the game being a con man.

 

[fpitas]

If you mean those guys, Hjalmar Schacht was even more ahead of the game being a con man.

Mine was simply a clumsy reference to Godwin's Law.

 

[simbloke]

That’s it indeed… what an absolutely idiotic thing to do! If anyone still things this dude is genuine and sincere, well… think again

Opened in another browser and it works. How pathetic is he?

 

[audio_tony]

Bloody he'll that response from Noel at hifiworld is just so incredibly incorrect I'm struggling not to believe a junior writes the responses for him.

Surely he doesn't actually believe that high frequency samples are represented by shorter bits of data or transmitted at different speeds. That shows an absolute failure to understand the structure of pcm audio data transfer.

Or it's a deliberate attempt to pander to his advertisers...

 

[PeteL]

Ahem...

I did exactly this around 5 years ago (with the help of @mansr). 3 blind tests (essentially DBT, because Mans was controlling the test using a random generator on his phone, and sitting in a different room to me - I could neither see nor hear him) - the first two non-ABX, and the last ABX. Here are the results:

View attachment 223622

I won't go into more detail because the test has been thrashed to death in other threads, both here on ASR and on AS, with Mans critical of the method, apparatus, and pretty much everything to do with it. He believes the ABX result (3rd test) was a fluke. For my part, I know I heard differences in sound. And what was changed between A and B? Simply buffer settings in the software player. A and B were shown to remain bit-identical throughout the test (we captured the digital input into the DAC).

Bear in mind that:
- I had never been involved in a blind test before, and took some time to get used to it
- we were using spdif into a pretty old DAC, which may well have had an analogue output stage susceptible to noise

I have a few decent modern DACs here now, and might try the test again... one day... if I have the inclination to do so.

Mani.

Sorry it is not clear "A and B are shown to remain bit identical troughout the test". Isn't this exactly the same as we demonstrated that the buffer setting didn't do any difference? Whether it's a fluke or not whether your blind test is valid or not is not what I would argue about, I am ok with trusting your findings but for me, what was the process to "capture the digital input "into" the DAC" what does it mean? How?

 

[earlevel]

[me: Woah, you clearly have not understood what I wrote. Show me where I said it's "the numbers that make it discrete". I went out of my way to separate the terms.]

I was replying to your original statement: "Digital means numbers".

Yes, I said "digital means numbers", so why did you claim I said, "it's "the numbers that make it discrete"? Vastly different claims.

Digital—pertaining to digits, numbers (please, don't tell me it might mean fingers). Hence "digital means numbers"—the numbers are why we call it "digital".

Fundamentally, the difference between digital & analog encodings is that digital uses discrete states and analog does not. The difference is not about numbers, which can be discrete or continuous. In short, we're quibbling about semantics. But I did say that in my first reply...

First I never said the the difference between analog and digital is about numbers. I said there was discrete time, which can be via analog voltage or digital values. I said this first thing. Then I said digital is specifically about numbers. You said it could be anything, not numbers, but I'm still waiting on how you're going to do any signal processing, including a simple gain change.

I'm going to stop, you're just misrepresenting what I say to act like I don't understand the subject or are at least explaining it wrong. Explain it to other people the way you want, I explained it the way I want, and I'm confident I'm not mistaken. I feel like I'm in a discussion on gearspace.

 

[earlevel]

I think you are mixing up the "discrete steps" of a dynamic range limited sampling process, to the "discrete states" which is simply high or low voltage of a digital transmission which is what the OP was referring to. Further more, I am not sure really what you mean with discrete steps being about "resolution", I know we hear it that way sometimes but it's a flawed reasoning. The word length ultimately only define the dynamic range, not the "resolution". 24 bits have a lower noise floor than 16, nothing else, it's not clear from your post if you really understand the sampling theorem and the attributes of a sinc function. We should not link resolution to word lengt, it's a bit more appropriate for and I understand that it's generally accepted to talk about High-Res for high sample rate digital audio but its flawed. The sampling rate limits the bandwith, the word lenght limits the dynamic range, Inside those limit the reproduced waveform have the same resolution at least theoretically there is not one that is more precisely resolved that the other. Or have "higher resolution".

Then you are at odds with the audio industry and the digital signal processing community.

Resolution mean the degree to which something can be resolved. In samples, the degree to which is can be resolved depends on the number of bits. I could bury you with links that reference bit depth as resolution. Here's a codec selector from Mouser Electronics, not that the number of bits is under the heading, "Resolution":

Enter "digital audio resolution" into your favorite search engine if you're still not sure.

And no, I'm not mixing up anything. Maybe you read something incorrectly, feel free to ask me about any specific phrase or paragraph.

 

[manisandher]

Sorry it is not clear "A and B are shown to remain bit identical troughout the test". Isn't this exactly the same as we demonstrated that the buffer setting didn't do any difference? Whether it's a fluke or not whether your blind test is valid or not is not what I would argue about, I am ok with trusting your findings but for me, what was the process to "capture the digital input "into" the DAC" what does it mean? How?

Here's the test setup:



The software player was XXHighEnd, which allows you to vary a whole load of parameters that (apparently) affect the processing load, whilst keeping things bit-identical. We varied a parameter called 'SFS', which appears to be a buffer setting of some sort.

In any event, all 54 digital captures were bit-identical.

 

[TonyJZX]

ironically and quite obviously Johnny Darko is Australian, not Austrian

also I do not know his educational background vis a vis Fine Art.

 

[PeteL]

Then you are at odds with the audio industry and the digital signal processing community.

Resolution mean the degree to which something can be resolved. In samples, the degree to which is can be resolved depends on the number of bits. I could bury you with links that reference bit depth as resolution. Here's a codec selector from Mouser Electronics, not that the number of bits is under the heading, "Resolution":

View attachment 223638

Enter "digital audio resolution" into your favorite search engine if you're still not sure.

And no, I'm not mixing up anything. Maybe you read something incorrectly, feel free to ask me about any specific phrase or paragraph.

I get that it's the sampling resolution, it's is correct in this context. But when talking about a digital signal, or a digital file or a full conversion device. it has a bit depth or word length and it has a sample rate. What I am trying to say is that by using in other context than just saying with how many bits it's sampled, it give the impression that the resulting analog waveform is more precisely resolved, which is not the case, it just have more space between the ODBFS and noise floor, that's all. It allows more amplitude, or more Dynamic range, but not more "resolution" on the end result.

As the other comment, it's self explanatory: the OP said: "Digital is about representing information as discrete states" and you replied: I disagree that digital is about discrete steps. As I explained the are two completely different concept, there are 2 states, there are, 65 thousand steps in 16 bits sampling.

 

[ReinhardS]

Sorry, amirm, but you missed the critical point here: in the SPDIF/AES Standard biphase-mark encoding is used and the clock is derived from the zero crossings. But the zero crossing time depends on how long the high/low periode was before. This means that if the signal bandwith is insufficient (and with higher samplerates it is in most cases) you have data depended clock jitter by the transmission. On the other hand: modern receiver ics should only consider the preamble bits to avoid that problematic. An insufficient bandwith and some noise makes the clock reconstruction more difficult anyway.

It would be interesting to measure the dacs jitter performance with poor signals with limited bandwith.

 

[PeteL]

Sorry, but you missed the critical point here: in the SPDIF/AES Standard biphase-mark encoding is used and the clock is derived from the zero crossings. But the zero crossing time depends on how long the high/low periode was before. This means that if the signal bandwith is insufficient (and with higher samplerates it is in most cases) you have data depended clock jitter by the transmission. On the other hand: modern receiver ics should only consider the preamble bits to avoid that problematic. An insufficient bandwith and some noise makes the clock reconstruction more difficult anyway.

It would be interesting to measure the dacs jitter performance with poor signals with limited bandwith.

May I suggest you use the reply function? There is no way to know which post you are replying to. It is obviously not mine but since I am right above it's confusing. No one knows who you are adressing.

 

[earlevel]

I get that it's the sampling resolution, it's is correct in this context. But when talking about a digital signal, or a digital file or a full conversion device. it has a bit depth or word length and it has a sample rate. What I am trying to say is that by using in other context than just saying with how many bits it's sampled, it give the impression that the resulting analog waveform is more precisely resolved, which is not the case, it just have more space between the ODBFS and noise floor, that's all. It allows more amplitude, or more Dynamic range, but not more "resolution" on the end result.

As the other comment, it's self explanatory: the OP said: "Digital is about representing information as discrete states" and you replied: I disagree that digital is about discrete steps. As I explained the are two completely different concept, there are 2 states, there are, 65 thousand steps in 16 bits sampling.

I'll say this one more time, I don't want to beat it to death. You're saying I'm making a mistake, I'm not. Tell the OP what you want, but please stop saying I'm wrong about things that I'm obvious right about. It's one thing if you presented a valid argument that shows that I said something wrong, but you haven't.

"...how many bits it's sampled...It allows more amplitude, or more Dynamic range, but not more "resolution" on the end result.—Here where you go wrong. Digital audio has a fixed maximum level. That level is the same for 8-, 12-, 16-, 20-, 24-bit—16- and 24-bit obviously the most commonly used, but all these have been used for audio over the years, can name devices at each. Ultimately, they are feeding standard audio equipment, so the maximum output range remains the same. Therefore, the more bits you have, the more resolution. The finer the steps. The greater the dynamic range, and presumably the great s/n.

Referring to it as resolution is so common that if you had googled "digital audio resolution", you would have been buried in links to material discussing the number of bits in digital audio. I have no idea why you think this isn't an actual thing. Have you designed digital audio hardware? Written digital audio signal processing code used in commercial products?

Sorry, not trying to be mean, just absolutely baffled by your assertion that "resolution" isn't a correct term to describe digital samples.

PS—When you stress "end result", I have a feeling that you mean once it has been converted back to analog. But we were comparing digital and analog, and I described resolution as a trait of digital. So your "end result" argument makes no sense—it isn't digital any more if you've converted it to analog. Sure, everyone knows that digital resolution influences dynamic range, both in the digital state and when converted to analog. But to say resolution is the wrong word...not sure where you went to school or what books you read, but you are at odds with the world.

 

[PeteL]

I'll say this one more time, I don't want to beat it to death. You're saying I'm making a mistake, I'm not. Tell the OP what you want, but please stop saying I'm wrong about things that I'm obvious right about. It's one thing if you presented a valid argument that shows that I said something wrong, but you haven't.

"...how many bits it's sampled...It allows more amplitude, or more Dynamic range, but not more "resolution" on the end result.—Here where you go wrong. Digital audio has a fixed maximum level. That level is the same for 8-, 12-, 16-, 20-, 24-bit—16- and 24-bit obviously the most commonly used, but all these have been used for audio over the years, can name devices at each. Ultimately, they are feeding standard audio equipment, so the maximum output range remains the same. Therefore, the more bits you have, the more resolution. The finer the steps. The greater the dynamic range, and presumably the great s/n.

Referring to it as resolution is so common that if you had googled "digital audio resolution", you would have been buried in links to material discussing the number of bits in digital audio. I have no idea why you think this isn't an actual thing. Have you designed digital audio hardware? Written digital audio signal processing code used in commercial products?

Sorry, not trying to be mean, just absolutely baffled by your assertion that "resolution" isn't a correct term to describe digital samples.

PS—When you stress "end result", I have a feeling that you mean once it has been converted back to analog. But we were comparing digital and analog, and I described resolution as a trait of digital. So your "end result" argument makes no sense—it isn't digital any more if you've converted it to analog. Sure, everyone knows that digital resolution influences dynamic range, both in the digital state and when converted to analog. But to say resolution is the wrong word...not sure where you went to school or what books you read, but you are at odds with the world.

OK, Indeed I should let go of this semantic debate, it's not fully wrong to talk of resolution, all I wanted to clarify is that a waveform will be reproduced exactly the same way in 16 bits or in 24 bits. This waveform will only be closer to the noise floor in 16 bits. That's what I was not sure you where fully getting.

 

[DonH56]

Increasing the number of bits not only reduces the quantization noise floor but also provides better time resolution (more bits can resolve smaller time differences).