Is Digital Audio Transmission Analog? [video]

[amirm]

A new argument has developed that transmission of digital audio is really analog. And for this reason, everything digital can be subject to audible difference from digital audio cables to digital output of streamers. This was emphasizes in a video by Darko Audio saying this explains his subjective opinion of streamers sounding different. I address this in this video and how there is some validity in what he says but his end conclusions are incorrect:

 

[Deleted member 2702]

Amir is a man not accepting BS (which is a rarity among hi-fi reviewers). Gotta love it. Science is a winner.

 

[Fidji]

Never bites the hand, that feeds him.

 

[Ken Tajalli]

I just read the title and laughed!
Whoever claims it, has no conception of digital data and the transfer.
I strongly recommend looking up the meaning of the phrase Analogue .

 

[Vacceo]

Amir, if you ever travel around Old Iberia or I happen to be around the West Coast, let me invite you to diner and a crate of you spirit of choice!!!

You sir just did the video both my brain and heart needed! Thank you, so, so much!

 

[Madjalapeno]

But did you include any IKEA door stops in your analysis?

 

[testp]

I just read the title and laughed!
Whoever claims it, has no conception of digital data and the transfer.

actually digital audio transmission (in-between devices) could be considered analogish square forms,
but end device translates it from low/high potentsial or rising/falling edge to 1s and 0s.. at least that's what i gathered from computer university course

 

[Koeitje]

actually digital audio transmission (in-between devices) could be considered analogish square forms,
but end device translates it from low/high potentsial or rising/falling edge to 1s and 0s.. at least that's what i gathered from computer university course

Yeah, digital audio is transferred by voltage swings if you use a coaxial connection. USB probably works differently (but in the end its still based voltage changes). Optical has no voltage swing of course.

I'm pretty sure Darko understands this, but he is just peddling nonsense because that can make him more money.

 

[Palladium]

So we can already read 7GB/s of any data flawlessly from an very error prone storage medium (NAND flash), but clearly our technology can't do it for digital audio working at 8 magnitudes slower.

Yeah sure thing, bro.

This is the same idea as spending tons of $$$ on external audio interconnects, while ignoring the equipment manufacturers themselves are most likely gonna to use the cheapest cables they can get away with in their own boxes regardless of sticker price.

 

[garbz]

USB probably works differently (but in the end its still based voltage changes).

USB works more like balanced audio. At those speeds you need differential signalling to make any kind of sense of the signal arriving at the other end or it looks like noisy garbage.

Optical has no voltage swing of course.

But it does. Voltage swing that drives a little LED which swings in brightness read by a light dependent resistor on the other end which converts it back to a voltage swing. The result of this conversion also looks pretty nasty when viewed on an analogue scope.

 

[Koeitje]

So we can already read 7GB/s of any data flawlessly from an very error prone storage medium (NAND flash), but clearly our technology can't do it for digital audio working at 8 magnitudes slower.

Yeah sure thing, bro.

This is the kind of speed you can get in consumer hardware. We can read 4 drives at the same time and combine the data without any faults. This is equal to streaming 136243(!) redbook cd's at the same time. Ok technically you wont be able to do that, because its not a sequential task, but you get the point.

USB works more like balanced audio. At those speeds you need differential signalling to make any kind of sense of the signal arriving at the other end or it looks like noisy garbage.


But it does. Voltage swing that drives a little LED which swings in brightness read by a light dependent resistor on the other end which converts it back to a voltage swing. The result of this conversion also looks pretty nasty when viewed on an analogue scope.

Yeah, of course the LED is doing things and that needs to be interpreted at the other end but I was talking about the cable.

 

[Vacceo]

I just saw Amir´s comment about Darko declining to let him talk on his channel. I don´t know about the veils, but the smell of piss on the legs reaches the Mediterranean.

 

[dartinbout]

!!@!!!!!!!! "Worthless as a fortune cookie"

 

[Ralph_Cramden]

Ah, Dorko. Slick production, shallow content.

 

[earlevel]

At first I felt <shrug> so the he wants to refer to it as analog. But OK, words have meanings, let's think about why this is wrong.

Digital means numbers. It doesn't mean sampling or discrete time, although digital lends itself to discrete time and not to continuous time, so we can assume digital implies discrete time.

Analog implies something that is analogous. The rise and fall of the voltage level is analogous to the compression and rarefaction pattern of sound in the air. Clearly, it pushes a speaker to create those exact compression and rarefaction patterns.

The problem: The guy is equating continuous time with analog. It's not. Note that we can't really store numbers, they're abstract. We can write down symbols on paper to represent them (the world settled on a system developed by Indians, spread by Arabs; Mayans, Romans, and others did it differently), we can store "analogs" of the numbers (more to follow). If we want to transmit them over a wire, we encoding them in a continuous time signal. We design the encoding so that it's robust and will survive some degradation.

Encoding notes: Computer memory stores levels (per bit), which will vary and can produce errors (rarely), they have built-in error correcting. SSDs might encode more than one bit per cell, using more than two levels. CDs players have to measure reflectivity in pits, which will not reproduce identically for every "bit", and it may be read at different rates. In fact, there is so much chance for errors (less rarely than memory) that there is robust error encoding built in. There are many ways to do it over wire (SP/DIF, USB, etc.), but the bottom line is usually that it's usually either with an inherent clock, or one that's recoverable from the signal, and it's discrete levels that gives noise immunity.

Such a signal over wire is not an analog of the sound. It does contain analogs of the numbers, though. While I'm not mad at anyone who wants to call a continuous signal "analog", it's pointless and causes confusion. Since digital is always encoded in some way, I don't think it's helpful to call it analog when it's a voltage over a wire and digital when it's not. It's still encoded digital. And since digital is always encoded, drop it and call it a digital signal.

 

[beagleman]

What is the Title of his Actual video called??

 

[lc6]

Excellent video, @amirm . Darko perfectly illustrates the definition of ultracrepidarian.
A self-admitted "non-engineering type," he supports his claims by blindly quoting engineers who either peddle expensive and unnecessary pieces equipment or spit out obvious technical problems that have long been solved:
"Vitorino asserts that one of the biggest challenges in designing good sounding USB outputs is minimising their electrical noise leakage — noise that can disturb the downstream DAC’s clock accuracy [get a better quartz oscillator, then]; that galvanic isolation might prevent noise from seeping along a USB connection’s power lines but not its data lines [the whole point is to optically isolate the data lines and not use USB +5V and GND unless separately converted by a SMPS]; that TOSLINK, which offers a 100% electrically isolated connection, is unfortunately many times higher in jitter than a well-designed USB (or coaxial) output…and will often sound worse [that is where the PLL comes in; and if you have bit losses despite having one, then the output doesn't just "sound worse," it is outright unacceptable, as @amirm said]; that transceivers converting a TOSLINK connection’s light back to an electrical signal can introduce their own noise [yeah, that is why, again, PLLs are used]; a finding that, in turn, has implications for an Ethernet connection’s optical isolation [which is his useful segue into promoting this "dual-clock repeater" device]."

 

[amirm]

Here is Darko's video:

 

[beagleman]

At first I felt <shrug> so the he wants to refer to it as analog. But OK, words have meanings, let's think about why this is wrong.

Digital means numbers. It doesn't mean sampling or discrete time, although digital lends itself to discrete time and not to continuous time, so we can assume digital implies discrete time.

Analog implies something that is analogous. The rise and fall of the voltage level is analogous to the compression and rarefaction pattern of sound in the air. Clearly, it pushes a speaker to create those exact compression and rarefaction patterns.

The problem: The guy is equating continuous time with analog. It's not. Note that we can't really store numbers, they're abstract. We can write down symbols on paper to represent them (the world settled on a system developed by Indians, spread by Arabs; Mayans, Romans, and others did it differently), we can store "analogs" of the numbers (more to follow). If we want to transmit them over a wire, we encoding them in a continuous time signal. We design the encoding so that it's robust and will survive some degradation.

Encoding notes: Computer memory stores levels (per bit), which will vary and can produce errors (rarely), they have built-in error correcting. SSDs might encode more than one bit per cell, using more than two levels. CDs players have to measure reflectivity in pits, which will not reproduce identically for every "bit", and it may be read at different rates. In fact, there is so much chance for errors (less rarely than memory) that there is robust error encoding built in. There are many ways to do it over wire (SP/DIF, USB, etc.), but the bottom line is usually that it's usually either with an inherent clock, or one that's recoverable from the signal, and it's discrete levels that gives noise immunity.

Such a signal over wire is not an analog of the sound. It does contain analogs of the numbers, though. While I'm not mad at anyone who wants to call a continuous signal "analog", it's pointless and causes confusion. Since digital is always encoded in some way, I don't think it's helpful to call it analog when it's a voltage over a wire and digital when it's not. It's still encoded digital. And since digital is always encoded, drop it and call it a digital signal.

Funny enough. I learned all this while in college, in a course called "Digital signal theory"

This is WELL known stuff, even in the mid 80s!

 

[diddley]

Darko keeps that machinery going, that makes this a very good video to rebuttal him with solid and (re)measurable measurements.
Therefor thank you Amir for another intresting video and a go at Darko and his b.s.