Is Digital Audio Transmission Analog? [video]

[DonR]

Just a silly question: is this thread now about how ultrasonic noise on an Ethernet connection may impact music payload bit integrity and hence sound quality?

I think it is more about RF noise coupling and demodulating to the analog stage but who knows?

 

[antcollinet]

So use a tone.

That was the suggestion.

 

[KSTR]

Just a silly question: is this thread now about how ultrasonic noise on an Ethernet connection may impact music payload bit integrity and hence sound quality?

No, its about the effect of RF noise directly induced into the analog sections of a DAC. We silently always assume bit perfect data and this can be easily monitored if in doubt.

While this kind of disturbance should not happen with competent circuit design and layout, it does happen in practise. Same goes for direct GND voltage drop errors inside a DAC, a very related issue, 8kHz "USB packet noise being" one commonly observed effect. Unless these disturbances create actual malfunction or readily audible noise a device having these problems may still pass EMC tests with regard to robustness.

How relevant could these effects become in actual listening practise? As of yet, I'd say we simply don't know because there has not been a thorough investigation as of yet, to my knowledge.

 

[redstang]

So now Hans is suggesting that network switches can introduce jitter. In the intro he talks about how one audiophile switch sounds better than another. Ugh.

 

[DonDish]

The usual blubbety blubbery. I like Hans, when he stick to the facts! When he wonders about he is not so smart

 

[pablolie]

So now Hans is suggesting that network switches can introduce jitter. In the intro he talks about how one audiophile switch sounds better than another. Ugh.

...

Of course network equipment can introduce jitter and latency, and if poorly planned even packet loss. Completely irrelevant with something as low bitrate as audio, no matter how HD you want to get - worry about that when you support high volume real time trading, where 10ms can mean several million bucks lost, easily. :-D [1]

BTW there's a reason why buffering exists. It corrects jitter if it's relevant to the application. Another thing about jitter: the dumber your network device is -ie. the less queueing/buffering- the less jitter you'll have as a rule. Price to pay: potential packet loss at high utilization.

THERE IS NO SUCH THING AS AN AUDIOPHILE NETWORK SWITCH/ROUTER. Period.

[1] I work for one of the top high-end networking companies in the world. Friendly provider of mega-Terabit routers etc. The deals with financial, government or healthcare institutions call for immense penalties if something goes wrong. You'd think those audiophool network comanies would try to enter such very profitable and large markets... note they don't - and never will. They'd end up bellyup like goldfish in boiling water in a sec... and know it.

 

[Vacceo]

Of course network equipment can introduce jitter and latency, and if poorly planned even packet loss. Completely irrelevant with something as low bitrate as audio, no matter how HD you want to get - worry about that when you support high volume real time trading, where 10ms can mean several million bucks lost, easily. :-D

BTW there's a reason why buffering exists. It corrects jitter if it's relevant to the application. Another thing about jitter: the dumber your network device is -ie. the less queueing/buffering- the less jitter you'll have as a rule. Price to pay: potential packet loss at high utilization.

THERE IS NO SUCH THING AS AN AUDIOPHILE NETWORK SWITCH/ROUTER. Period.

So in essence, they are functionally the same as electricity wires for audio gear: they work or they do not.

 

[pablolie]

So in essence, they are functionally the same as electricity wires for audio gear: they work or they do not.

Exactly. Except with good network gear you get the managment tools to establish traffic trends and act on stuff before it bites you in the derriere. You get all the stats: latencies and jitter and packet loss, warnings with exact pointers to the location etc etc.

A well designed network doesn't suddenly fail if you truly examine what lead to it. If you look at the famous big network outages, it's human error >95% of the time. When hardare fails it's trivial - people have a response plan and spares in place. When software fails, it can go from trivial to being a nightmare that keeps you up at night until it's fixed. If it's user error, it takes very careful case management to not discredit your best customers in front of their bosses....

 

[antcollinet]

Of course network equipment can introduce jitter and latency, and if poorly planned even packet loss. Completely irrelevant with something as low bitrate as audio, no matter how HD you want to get - worry about that when you support high volume real time trading, where 10ms can mean several million bucks lost, easily. :-D [1]

BTW there's a reason why buffering exists. It corrects jitter if it's relevant to the application. Another thing about jitter: the dumber your network device is -ie. the less queueing/buffering- the less jitter you'll have as a rule. Price to pay: potential packet loss at high utilization.

THERE IS NO SUCH THING AS AN AUDIOPHILE NETWORK SWITCH/ROUTER. Period.

[1] I work for one of the top high-end networking companies in the world. Friendly provider of mega-Terabit routers etc. The deals with financial, government or healthcare institutions call for immense penalties if something goes wrong. You'd think those audiophool network comanies would try to enter such very profitable and large markets... note they don't - and never will. They'd end up bellyup like goldfish in boiling water in a sec... and know it.

Jitter (as it applies to audio) only has any effect at the point where digital samples are converted to analogue.

Jitter (such as it exists) in any digitial to digital device has no impact at all, unless it is so severe that it causes data to be corrupted. And then in an ethernet domain, you'd sort of hope error detection and retransmission would eliminate that as a problem.

 

[redstang]

Of course network equipment can introduce jitter and latency

How can it introduce jitter if the clocking is all done by the DAC? The DAC either has the packet to assemble into the bitstream (which I assume is buffered) or it doesn't.

 

[pablolie]

Jitter (as it applies to audio) only has any effect at the point where digital samples are converted to analogue.

Bingo.The only problem with jitter is when the receiving codec is "starved" ie doesn't receive the next packet in time. Buffers eliminate that possibility to zero if well designed for the application.

Jitter (such as it exists) in any digitial to digital device has no impact at all, unless it is so severe that it causes data to be corrupted. And then in an ethernet domain, you'd sort of hope error detection and retransmission would eliminate that as a problem.

Not just Ethernet - *any* network... routing (IP) or switching (Ethernet)... and actually data corruption that isn't corrected is *extremely* seldom, cosmic ray stuff. I mentioned this before, but in my venerable 30+ year networking career I just have *heard* about undetected data corruption through the network (it does happen in static storage though, but they have methods to address that). I have never once experienced it. Packet loss is another matter altogether - that happens, for a variety of reasons. And like you said, retransmission is there.

 

[pablolie]

How can it introduce jitter if the clocking is all done by the DAC? The DAC either has the packet to assemble into the bitstream (which I assume is buffered) or it doesn't.

The DAC clocking is entirely internal to the end device. The end device has a network layer that receives the data without being impacted by *network* jitter (if the designer passed a basic test or two in DAC design). That's why you have a buffer stage *between* the network layer and the application layer (which is what the DAC is).

The OSI model is not entirely accurate in the current world, but it stll represents the design concerns of the different domain responsibilities that need to be in an end-to-end system to perfection. Separation of duties with a layered model is *vital* to any working design.

 

[Ken Tajalli]

Just little common-sense observations:
How many switches, servers, b;ah blah does streaming music go through, before it gets me?
I mean, on something like Qubuz or Amazon HD?
I still manage to get the data in bit-perfect condition, despite all that, don't I?
How much of a difference does one switch, or router at my house, can make?

Edit:
BTW, I am going to blame bad switches and serveres for all my past and future spelling missstakes.

 

[Axo1989]

So now Hans is suggesting that network switches can introduce jitter. In the intro he talks about how one audiophile switch sounds better than another. Ugh.

There must be a corollary to Betteridge's law of headlines. If a headline that starts with "why" the story won't tell you that.

 

[tmtomh]

I like Hans, when he stick to the facts!

When is that?

 

[Laotze]

I think that all analog audio is actually digital!

Only a discrete number of electrons jump across the bandgap of whatever amplification device.

Therefore it's all digital.
/Tongue out of cheek

 

[amirm]

So now Hans is suggesting that network switches can introduce jitter. In the intro he talks about how one audiophile switch sounds better than another. Ugh.

Shame that he spends so much time on stuff that doesn't matter, and so little on what does. It is only towards the end that he claims jitter on Ethernet magically jumps to jitter in a DAC. Where are the pretty graphs to show that let alone measurements?

Fortunately, we have those measurements. I tested EtherRegen which he says is better than any commercial IT switch: https://www.audiosciencereview.com/.../uptone-audio-etherregen-switch-review.10232/

Here is the jitter test using it, or generic switch:

There is absolutely no difference in jitter. Same is true above audio band:

There are more tests there that prove beyond any doubt that the network switch makes no difference whatsoever.

Please link to my EtherRegen review to his youtube and see if he a) leaves it there and b) what he has to say. As far as I know, he deletes such comments so clearly he is not interested in learning as he says in his video.

Bottom line, a bit of knowledge here is dangerous. Yes, jitter is like what he says. But no, it has no merit in this discussion since the two components of the system run asynchronous from each other and are separated by many moats. He would have known this if he had measured. But he didn't despite the equipment sitting behind him which is capable of such.

As to what he has heard, that is obviously done sighted and without controls. So he might as well say the color of the switch changes the sound!

 

[amirm]

As an aside, measuring jitter directly from say a clock requires high-end measurement gear. However, measuring it using FFT spectrum as I show above, is trivial. We are able to use signal processing to significantly reduce measurement noise and then be able to measure jitter down to sub picoseconds (well below resolution of 24 bit audio). So his implication that jitter can't be measured is wrong. It is done in every measurement I make of DACs as you all know.

 

[Laotze]

As an aside, measuring jitter directly from say a clock requires high-end measurement gear. However, measuring it using FFT spectrum as I show above, is trivial. We are able to use signal processing to significantly reduce measurement noise and then be able to measure jitter down to sub picoseconds (well below resolution of 24 bit audio). So his implication that jitter can't be measured is wrong. It is done in every measurement I make of DACs as you all know.

Maybe they are right about you damaging your hearing Amir. I mean, you can't hear femtosecond delays like the self proclaimed golden ears obviously can.

Personally, I switched to the old horned record players. Since the jitter of the (lossy digital) electrons sounds grating to me.

 

[redstang]

Shame that he spends so much time on stuff that doesn't matter, and so little on what does. It is only towards the end that he claims jitter on Ethernet magically jumps to jitter in a DAC. Where are the pretty graphs to show that let alone measurements?

Fortunately, we have those measurements. I tested EtherRegen which he says is better than any commercial IT switch: https://www.audiosciencereview.com/.../uptone-audio-etherregen-switch-review.10232/

Here is the jitter test using it, or generic switch:

There is absolutely no difference in jitter. Same is true above audio band:

There are more tests there that prove beyond any doubt that the network switch makes no difference whatsoever.

Please link to my EtherRegen review to his youtube and see if he a) leaves it there and b) what he has to say. As far as I know, he deletes such comments so clearly he is not interested in learning as he says in his video.

Bottom line, a bit of knowledge here is dangerous. Yes, jitter is like what he says. But no, it has no merit in this discussion since the two components of the system run asynchronous from each other and are separated by many moats. He would have known this if he had measured. But he didn't despite the equipment sitting behind him which is capable of such.

As to what he has heard, that is obviously done sighted and without controls. So he might as well say the color of the switch changes the sound!

I have to think he is deleting any negative comments or even anything that refute his claims because all the comments on there now are 100% positive.