The inaudibility of Ethernet

[Kal Rubinson]

First, thank you for taking the time and effort to describe your experience in some detail.

The first thing I noticed was the noise floor, I use a Naim Nac52 preamp and barely needed to get to 8 o'clock for loud.

It seems that the core of your story is the lowered noise level which you say is audible in some way. I have no doubt that lowering the audible noise floor can have a generally positive effect on the sound in many ways. Can you hear the difference in noise level from the speakers when the Ethernet is connected/disconnected from your amp and preamp? Can you hear a difference in noise level from the speakers when you change that critical cable?

I have swapped in the old cables, it is not subtle and highly repeatable. Measureable, not sure, though someone on the PSaudio forum has some interesting scope pics. Happy listening.

OK. If you can hear it, surely it can be recorded with a microphone and measured without any fancy equipment.

 

[Mikem53]

Listening is always the best way to discover these things. It’s always about our own contexts.

I agree, Listening is the final arbiter in the debate.. but it’s more difficult to explain and verify on paper. I spent most of my life as a techie in IT. I have many paradigms to overcome still.. Yet when it comes to audio, I have learned to accept some artistry as science. I know that there are physical properties and such that occur that are undetected, unknown as well as other Factors which affect the sound, not to mention our interpretation of what we as individuals hear.. So subjective.. But I’m going to try it and see, once I overcome some of my prejudices..

 

[microstrip]

The problem with Ethernet is not packet loss. It’s jitter. The jitter changes as the packet goes through each device. The end result for text data is none, but for music the jitter is audible.

Unfortunately we can't completely separate jitter from noise. Noise can cause jitter. Ethernet is asynchronous - if it is not for noise it would be extremely simple to re-clock it to minimal jitter.
We had a very interesting exposition from Emile (Taiko Audio) explaining how noise can be carried in jitter-less bits.

 

[microstrip]

(...) OK. If you can hear it, surely it can be recorded with a microphone and measured without any fancy equipment.

Yes, the differences should show on a recording. But we would need proper tools to analyze the files. Just listening to them is not enough - two channel recording is very limited.

Unfortunately our main subjects of discussion are not interesting to audio researchers - we have now access to excellent tools that could explain a lot of what we say we hear. But high quality stereo is so dependent on listener that it can become impredictable - and no one wants to spend time and resources researching such aspects.

 

[Kal Rubinson]

Yes, the differences should show on a recording. But we would need proper tools to analyze the files. Just listening to them is not enough - two channel recording is very limited.

Why is two channel recording necessary? I was asking for a recording of the noise only, without a music signal. This can be done with a single mic close to any one of the drivers. It is either there or it is not.

But high quality stereo is so dependent on listener that it can become impredictable

That's a tough one for me.

 

[rDin]

The problem with ethernet audio seems to be related to conducted electrical noise into the downstream DAC. People, myself included, are making huge gains with better designed switches, better power supplies and, crucially, fibre ethernet connections which completely isolate the network electrical noise from the downstream audio equipment. The differences in sound quality, once you address the issues, are not at all subtle.

 

[kleinbje]

The problem with ethernet audio seems to be related to conducted electrical noise into the downstream DAC. People, myself included, are making huge gains with better designed switches, better power supplies and, crucially, fibre ethernet connections which completely isolate the network electrical noise from the downstream audio equipment. The differences in sound quality, once you address the issues, are not at all subtle.

This! Try it then talk about it. It's a huge improvement.

 

[kleinbje]

First, thank you for taking the time and effort to describe your experience in some detail.
It seems that the core of your story is the lowered noise level which you say is audible in some way. I have no doubt that lowering the audible noise floor can have a generally positive effect on the sound in many ways. Can you hear the difference in noise level from the speakers when the Ethernet is connected/disconnected from your amp and preamp? Can you hear a difference in noise level from the speakers when you change that critical cable?
OK. If you can hear it, surely it can be recorded with a microphone and measured without any fancy equipment.

Appreciate your reply. I may have mislead with the term noise floor as I don't notice more noise when I put the old cables in, I just need to turn it up louder to achieve same perceived and enjoyable loudness. What is most apparent when switching cables to the Vodka is the time I can enjoy preferred volume without constant modulation. I am 1 month in, and each night is the same, I haven't listened to digital this much in many years. I saw this on SHF tonight and it seemed to describe best what I gained with new ethernet cables. As for your measurements I fear the ad/da conversions in such recordings could induce the same errors the AQ cables are preventing.

*being a speaker builder/ hobbyist, I have acquired great appreciation and awareness on transient response, that all so critical attack time. (phase/ time coherent in the treble) It is the difference between turning my head to where an unexpected sound comes from (somewhere outside the speaker) or not being fooled by it.*

Jeff

 

[Kal Rubinson]

Appreciate your reply. I may have mislead with the term noise floor as I don't notice more noise when I put the old cables in, I just need to turn it up louder to achieve same perceived and enjoyable loudness. What is most apparent when switching cables to the Vodka is the time I can enjoy preferred volume without constant modulation. I am 1 month in, and each night is the same, I haven't listened to digital this much in many years. I saw this on SHF tonight and it seemed to describe best what I gained with new ethernet cables. As for your measurements I fear the ad/da conversions in such recordings could induce the same errors the AQ cables are preventing.

OK. Thanks, anyway.

 

[Blackmorec]

Why is two channel recording necessary? I was asking for a recording of the noise only, without a music signal. This can be done with a single mic close to any one of the drivers. It is either there or it is not.

That's a tough one for me.

I’m guessing there are 2 types of noise in play here. The noise that disturbs the streaming and D to A processes and the noise that results from those disturbances, so at least in theory, logically there’d be no noise present when there’s no music playing as the processes that generate the noise aren’t running. What would be present is the noise that causes the disturbance but that’s HF, way above audibility

 

[Kal Rubinson]

I’m guessing there are 2 types of noise in play here. The noise that disturbs the streaming and D to A processes and the noise that results from those disturbances, so at least in theory, logically there’d be no noise present when there’s no music playing as the processes that generate the noise aren’t running. What would be present is the noise that causes the disturbance but that’s HF, way above audibility

I understand what you are saying. But since I have not heard it nor its absence myself, that's still a tough one for me.

 

[Robh3606]

I’m guessing there are 2 types of noise in play here. The noise that disturbs the streaming and D to A processes and the noise that results from those disturbances, so at least in theory, logically there’d be no noise present when there’s no music playing as the processes that generate the noise aren’t running. What would be present is the noise that causes the disturbance but that’s HF, way above audibility

If it's way above audibility why would it be audible?? As far as measurements all you would need to do is take at an FFT measurement right at the amp and compare them. If there is audible noise you can most certainly see it in a measurement. Just look at a 1k sine wave spectrum and look for changes as you switch cables.

Rob

 

[Blackmorec]

If it's way above audibility why would it be audible?? As far as measurements all you would need to do is take at an FFT measurement right at the amp and compare them. If there is audible noise you can most certainly see it in a measurement. Just look at a 1k sine wave spectrum and look for changes as you switch cables.

Rob

Read again what I said. The inaudible noise impacts and modulates the processes, for example D to A conversion, that produce music...so to hear the effect the inaudible HF noise is having, you need to have music playing. You are not hearing the HF noise, you are hearing the effects the HF noise has on processes that produce the music.
For example, lets say that a major dose of RFI reduces the overall resolving ability of your D to A process. You won’t hear the RFI but you will hear the loss of resolution.

 

[tony22]

The problem with ethernet audio seems to be related to conducted electrical noise into the downstream DAC. People, myself included, are making huge gains with better designed switches, better power supplies and, crucially, fibre ethernet connections which completely isolate the network electrical noise from the downstream audio equipment. The differences in sound quality, once you address the issues, are not at all subtle.

@rDin, I agree with what you’re saying here, but it always raises a question in my mind. Wouldn’t there be no difference in what kind of network infrastructure is being used, then, if a single Ethernet / Fiber / Ethernet conversion is done directly at the DAC input? Would using an audiophile switch etc really matter if this fiber conversion was done right at the DAC?

Hmm, I guess jitter might still be an issue.

 

[Blackmorec]

@rDin, I agree with what you’re saying here, but it always raises a question in my mind. Wouldn’t there be no difference in what kind of network infrastructure is being used, then, if a single Ethernet / Fiber / Ethernet conversion is done directly at the DAC input? Would using an audiophile switch etc really matter if this fiber conversion was done right at the DAC?

Hmm, I guess jitter might still be an issue.

What you’re talking about presumably is taking a fibre feed directly from the modem to the DAC with no wire or switches to pick up RFI\EMI? And if the only problem was RFI\EMI from the modem onwards you’d certainly avoid any antennae-like functions. But you’d also be potentially listening to a $20 source driven by a $l0 SMPS (or better named, a switched mode noise supply), supplied by whatever is coming in from the copper or cable of your ISP. In my experience, raw internet signals and consumer electronics do not good audio make.
But unlike analog audio, the digital signal can be improved. The digital signal has 2 components....a digital structure, which defines the music and an analog section of voltage pulses, which transports the structure. Due to their very different physical structures these 2 components can be separated and the analog section can be rebuilt from the digital section. One is the voltage representation of the other. So at each reclocking you can essentially rebuilt the analog section, which of course means that you can apply all the typical ways of cleaning and improving analog signals....Of course the best way to have something clean is to not get it dirty in the first place.....building a new copy based from a dirty signal is not going to give you the perfect signal, but it Is going to be an improvement. And a series of improvements is going to move the needle massively.

 

[tony22]

What you’re talking about presumably is taking a fibre feed directly from the modem to the DAC with no wire or switches to pick up RFI\EMI? And if the only problem was RFI\EMI from the modem onwards you’d certainly avoid any antennae-like functions. But you’d also be potentially listening to a $20 source driven by a $l0 SMPS (or better named, a switched mode noise supply), supplied by whatever is coming in from the copper or cable of your ISP. In my experience, raw internet signals and consumer electronics do not good audio make.

No, that’s not what I’m talking about. A good cable modem like an SB8200 fed by its own LPS, to a good but not especially audiophile router like a Ubiquiti fed by its own LPS, to decent but not especially audiophile Ethernet to Fiber converters fed by LPSes. One of these converters landed right near the DAC so a very short run of Ethernet for that connection to the DAC. What can audiophile level components bring to the table that a configuration like this cannot?

 

[Blackmorec]

No, that’s not what I’m talking about. A good cable modem like an SB8200 fed by its own LPS, to a good but not especially audiophile router like a Ubiquiti fed by its own LPS, to decent but not especially audiophile Ethernet to Fiber converters fed by LPSes. One of these converters landed right near the DAC so a very short run of Ethernet for that connection to the DAC. What can audiophile level components bring to the table that a configuration like this cannot?

@rDin, I agree with what you’re saying here, but it always raises a question in my mind. Wouldn’t there be no difference in what kind of network infrastructure is being used, then, if a single Ethernet / Fiber / Ethernet conversion is done directly at the DAC input? Would using an audiophile switch etc really matter if this fiber conversion was done right at the DAC?

Hmm, I guess jitter might still be an issue.

OK, fair enough, but that’s not ”No difference in what kind of network infrastructure is being used“, rather its a simplified audiophile optimized network.....so lets follow this train of thought. Obviously adding the good cable modem, a good router and FO has already given you significant improvement over run-of-the-mill consumer components running on cheap as chips SMPSs. That much we recognise. So your questions is then, why would FURTHER improvements help. And the answer is at the same time both straightforward and complex. All further improvements are doing is carrying on in the same direction as your initial improvements i.e refining the data stream. Add for example a Buffalo switch with the Pink Faun clock mod and a really good LPS and you’ll hear a marked uplift in SQ. There are plenty of threads on the subject talking about just that. Why? Well here’s where we get to the complex bit. All I can say is that each time we resynthesise the analog part of the digital signal using quieter power supplies and improved clocks, we hear an improvement in SQ.....so the only conclusion is that resythesis of the stream works better the more perfect (less noise, less jitter) the incoming stream, which of course provides the template for the resynthesis.
Further illustration? Let’s say that you use a Sean Jacob’s dual rail DC3 LPS to power the Buffalo Switch and PF Clock. Upgrade that LPS to a DC4 and you’ll hear a major jump in SQ. Then swap out the regular DC cables for Mundorf Silver/Gold and again you’ll be rewarded with a large improvement in SQ. How swapping a DC cable generates such an improvement is at least for me still a mystery...however I’ve no doubt whatsoever that the improvement is real and not some figment of the imagination....again there are lots of testaments to the efficacy of this upgrade.
I wouldn’t disagree with anyone who feels the above is counter-intuitive. I only disagree with those in denial that the improvements are real.

 

[tony22]

Thanks @Blackmorec. As a long time audiophile I know there should be further improvements with purpose built audio networking, but I’ll admit it still makes me rub my forehead. It’s not clear to me what makes it through the Ethernet-fiber-Ethernet “barrier”, if you will, that causes the difference. I can only think of jitter.

 

[Blackmorec]

There’s no doubt that the more technically accurate and refined the components you use to produce, transport and reproduce the data stream, the better the end result, basically music that is more accurate, pure and refined and therefore more natural and musical...the relationship is quite clear. When you upgrade components you mainly get a more natural presentation, greater rhythmic interplay and drive and better time and space resolution. The music is the digital part, the realisation of the music is still analog.....initially hf waveforms, where timing at a specific frequency is king and anything that disturbs that timing the enemy, so system power supplies and clocks along with excellent vibration control are key areas for extracting the most from your network stream.
The other bugaboo of digital audio is RFI and EMI.....essentially streams of high frequency energy that can impinge oin your system and cause some fairly major disturbances, so it must be stopped to as great a degree as reasonably possible. This also means protecting every part of your system. Almost certainly devices intended for audio will offer protection but the how and what are what sets them apart.

The best way to hear the impacts of upgrades to your network is via low resolution transmissions. One of the benefits of high resolution is that it sounds really good and does compensate in some way for some of the damage done with transmission.
But low res has nothing to fall back on. It’s already missing a bunch of data so it’s doesn’t sound so great when it looses more in transmission. But get your network to the point that it ‘massages’ and ‘cleans up’ the incoming data stream, with less noise, greater stability, better, more accurate, lower error clocks and less ‘disturbance’ from RFI , suddenly low res starts to sound pretty good. Not stellar like high speed uncompressed data streams but nevertheless extremely engaging, musical and beautiful with most of the hi-fi attributes.
In any serious listening session I may listen to an hour’s worth of beautiful, new (to me) classical music with one or two favorites thrown in all courtesy of the Swiss (Swiss Radio Classic)

Their radio is a lot like their public transport and health system. Runs like clockwork, to a very high standard, with music as beautiful as their country. (Swiss)

 

[Chrk]

the first page of this thread was really refreshing. I think the first thing any audiophile should measure is its own ear response.