An interesting "Linus" video on the "Audio Quality Ethernet"

Solid work, Graeme, thank you.

It is a rare thing for such a well thought out test to be conducted and for findings to be shared.

The results are quiet enough to drown out the loudest contrarian claims. :blush:

An analogy I proferred elsewhere when discussing Ethernet and audio and trying to show what really happens:

** fireworks ignite and burn in the night sky **
Crowd: “ooooh! MAGIC!”
Bystander: “Certainly looks that way, but it boils down to chemistry. Copper-based compounds give off the blue light, brightness and whites are enhanced by Magnesium compounds, Strontium and Barium make it red. Various oxidisers are used to enable the reactions to occur!”
** blank stare **
Crowd: “MAAAAGIC!”

People would be less dogmatic and more open if the networking product claims in much of the marketing bumpf of kit claiming audiophile performance had a shred of engineering plausibility.

1 Like

Thanks Tel for your kind comments.

I love your analogy.

I have to confess, the “silent” test track wasn’t my original idea - I’m sure I read a thread here recently about using one to make sure a DAC wouldn’t mute it’s outputs with no input. If I find it again, I’ll credit the original poster.

It was a fun and relatively easy experiment to conduct, made all the easier by the capability of Benchmark Media’s SoTA noise performance.

Until today, I would never have believed you could push a -120dBFS track into audibility. It turns out you can, and not only that, it’s even possible to pick out a 1kHz test tone.

I’ll have to repeat it with my digital 'scope across the amp’s outputs. I’m curious to see what the 1kHz square wave actually looks like coming out of the amp. It won’t be pretty though, I’m sure…

There’s so much hot air around hi-fi and audiophilia that has been debunked by science, yet despite all of the evidence, there are still those who believe in MAGIC!

1 Like

It’s plausible (well, you proved it). With enough windings and enough juice, you could amplify amoeba flatulence into audibility.

This is an experiment that I’d be really interested in, the results would quantify what any extraneous noise looked like in relation to the signal.

2 Likes

Nicely worded. Also nice cable managment. :slightly_smiling_face:

1 Like

Ah. The “barber theory” springs to mind here:

In a town with two barbers, you have to decide which one you will go to in order to have your mop trimmed. One is immaculately groomed with perfect hair, the other looks like he had an altercation with a cactus in a mattress-stuffing factory. Although it is counter-intuitive at first glance, you eschew the smart and well turned-out barber and go to get your hair cut by the guy who looks better suited to professional dumpster-diving than cutting hair.

This is the reason I am not showing my network. Not the stateful firewall, the routers, any switches and not the cabling. Nope. :rofl:

In all seriousness, Graeme, that looks pretty damn good. There are some parts of my OCD I can ignore (cable management in my own home is one of them - barring prudent separation measures for function rather than form); it seems you are a completist :slightly_smiling_face:

2 Likes

That has probably been me here, though not some special achievement to be credited, but you took it to a whole new level…
:vulcan_salute:

… you bet!

3 Likes

@Graeme_Finlayson - Thanks for your test. No sure I understand everything - but I am commercial guy, not a “Mad Scientist” :slight_smile:

If there is no noise from the ethernet why do Network Isolator exist?

and

PSaudion says “Improving Digital Streaming with a Gigabit Ethernet Media Converter System”:

“I firmly believe that inserting an optical connection into a copper-wired network chain will make a significant improvement in any digital streaming system’s musicality. Period. Best $100 I ever spent!” (Link)

Than you have Lumin X1 that uses “Optical Network provides complete isolation from network digital noise” (Link)

And Sonore opticalModule Deluxe v2? (Link)

Please, this not meant to provoke you - It is just me trying to understand things.

Torben

1 Like

@ffk @ElTel Thanks for the kind words. The network tidiness/cable management is purely out of necessity. The switches ports are 12 x PoE+/12 x PoE/24 non-PoE, so if I disconnect anything for testing or fault finding, I need to know where to plug it back in! I had the whole lot out on Saturday to prep for the new FTTH going in today (and to finally terminate the 4 backup runs of Cat6 I installed out to the man cave) when we built the new patio and I had to duct all of the cables at SWMBO’s behest

I even have a list of device IP and MAC addresses by network port and patch panel ID, though I confess that’s probably a little OCD though :grin:

@Marin_Weigel Thank you for planting the seed of the idea. It took a little head-scratching to work out how to make a silent track, then I remembered that you can save test tracks in REW. Duration is limited to 100s, though.

1 Like

Hi Torben,

No provocation perceived, but very valid questions.

Apologies in advance, this will be a little “wordy”

A quick explanation of the test method might help you to fully understand. In assessing the “noise” coming via the streamer/Ethernet connection, simply playing nothing to the DAC can make some DACs mute their outputs.

In order to prevent this, you need to keep the DAC “awake” by sending it a signal. We don’t want to hear the signal, so we make it “silent”. dBFS is a way of quantifying a signal’s amplitude relative to full scale output. So the maximum signal level that can come from your DAC is 0dBFS, for a balanced connection, this is typically 4 Vrms. It is generally agreed that -120dB is below the threshold of human hearing (-120dB is 1 millionth as loud as 0dB). In fact, in reality it’s way below. It would be like listening out for the sound of a pin dropping to the floor whilst standing trackside at a Formula 1 race!

-120dBFS relative to 4 Vrms is only 4 microvolts, which is a very low signal level indeed. In fact it’s about as low as the inherent noise level from even the best performing hi-fi components.

Benchmark Media’s kit is a little out of the ordinary. It’s origins are in Pro Audio, where much higher signal levels are common. The DAC3 HGC has a calibrated 0dBFS gain (it’s how mine is set to feed the HPA4) of +24dBu which is 12.3 Vrms output, over 3 times a “normal” DAC.

The HPA4 maximum output level is +28dBu which is 20 Vrms. This means that at maximum volume, it can amplify a -120dBFS track to 20 microvolts signal level. An extra ~14dB of gain. But it also does so at vanishingly low noise levels.

When I looked at the specs again, it’s pretty obvious why I could (potentially) hear the 1 kHz tone. Between the DAC3 HGC and the HPA4, I have 135dB of gain at my disposal.

However, if the claims around ethernet noise were to be true, I wouldn’t be able to hear the test tone because it would be lost in the “noise” from my Ethernet connection and the “noise” from my RPi’s PoE SMPS.

The reality is that I could hear the 1 kHz tone, which means that claims of noise from ethernet polluting the audio are unfounded. It also means that the RPi’s SMPS has no influence on audible noise either.

I hope you’ve been able to follow so far.

As to the PSAudio link,

Really? I’ve just proved that it can’t (provided it’s properly implemented, of course)

Which brings me on to the next point. “Audiophile” ethernet cables are almost invariably shielded. They claim Cat7 and Cat8. Few would test out to the rigorous specification targets needed to achieve 10GbE (Cat7) or 25GbE/40GbE (Cat8) transmission rates. Cat7 was never a proper EIA/TIA/IEEE standard and has been superceded by Cat8. Cat8 has no place other than in data centres (this is one of @ElTel 's ares of expertise)

Shielding provides a ground link between components which if you used Cat6 (as you should), wouldn’t exist.

So your fancy snake-oil “Cat7” or “Cat8” has introduced a ground path which shouldn’t be there. One way to break that ground path is via optical isolation “optical fibre”.

Your fancy ethernet cable has introduced a problem you now need to solve… An in the true audiophile fashion, you have to spend even more money.

Optical fibre solutions are solutions to problems which shouldn’t exist and that prey on FUD.

One of the best ways to make money in audio is to create a problem, then sell a solution. It doesn’t even have to be a real problem. Just a suggestion of an issue affecting sound quality that can be “solved” by buying your product.

The reason high-end equipment manufacturers are now including optical capability is, I believe, purely a marketing strategy. It gives them something to set them apart from the competition.

The Sonore optical module is nothing more than a fibre media converter. However it’s an “audiophile” FMC, so they can charge you more for it…

9 Likes

Trust the ones that agree with you :smiling_imp:

@Graeme_Finlayson ,

Your scientific test is nonsensical, and attests a misunderstanding of how digital audio is converted to acoustical sound.

You streamed -120 dB of silence to the DAC over a noisy LAN.
That means that the DATA of the -120 dB sound entered inside the DAC together with the electrical noise from your LAN.

The DAC processes only the DATA into acoustical sound, it won’t process the electrical noise, because the electrical noise does not contain any DATA.

The sound of the DATA was silence, that’s why you heard nothing when you listened to your speakers.
There was no reason at all that you’ll hear any acoustical sound coming out of them.

There are few sites I trust - I can only list 3:

Audioholics is oened and operated run by Gene DellaSala. His resumé speaks for itself.

1 Like

@Dandou There’s no need for the tone.

The last thread you were involved in with me received multiple flags and was ultimately closed. I have no issue with you disagreeing with me, but please keep it respectful.

No, it isn’t and it doesn’t. I understand how digital is converted to analogue.

The audiophile networking aficionados claim that noise from the ethernet can pollute the output of the DAC. Remember this post?

and this one?

These statements would appear to be contradictory. Are you suggesting that the noise interferes with the audio DATA in some other manner?

I didn’t stream silence to the DAC and my LAN isn’t noisy. I streamed -120dBFS pink noise and a -120dBFS 1 kHz square wave to the DAC. Both were clearly audible at the speakers when amplified enough, the 1 kHz tone was only to confirm my suspicions that I could actully hear a -120dBFS signal.

4 Likes

Sorry if I offended you.
I was not flagged in the previous thread.

There are no contradictions at all with my past posts.
Electrical noise does not contain any DATA, so it is not processed nor converted by the DAC into an acoustical sound.

The DATA alone is converted into an acoustical sound. The presence of the noise does not change the DATA. But the presence of the electrical noise inside the DAC degrades the quality of the analog conversion of the DATA. The sound is less resolving, has a lesser clarity, a narrower sound stage, and so on…

If you’ll stream music, you’ll hear that the sound quality of the conversion with the same gear is better when you stream in a silent LAN compared to a noisy LAN.

You streamed silence. You heard silence. That’s all.
If you’ll stream a violent sonata, you’ll hear a drop in sound quality.

@Dandou - Buit how do you secure that you have a silent LAN?

You posted in another thread that you set an optical isolation in your setup, and that you heard a sound improvement. The reason for that is that the optical conversion filtered all the noise of the LAN.

That’s a first step. The second step is to secure silence for the devices of your setup that come after the optical isolation. You do it with shielded RJ45 cables, and with good gear that is powered by linear or regulated power supplies.

2 Likes

The noise can’t “degrade” the analogue conversion of the data - it can only add itself to the analogue signal coming out of the DAC. In this case, had it been there, the noise would be audible.

In my experiment it wasn’t.

I didn’t stream silence as I already explained. -120dBFS is not silent - it’s below the realms of audibility at sensible volume levels and below the limits of resolvability of a lot of audio components. But it’s still a signal and the 1 kHz test tone was clearly audible as such.

Th fact that it was is testament to the exemplary engineering capabilities of Benchmark Media.

1 Like

The DAC is a very sensitive device.
The electrical noise provokes interferences that degrade its performance.

-120db is not silence, indeed, but it’s not a sound at a level that you can hear.
So, you can not say if the presence of the electrical noise rendered its conversion less resolving, reduced the clarity of the sound or its depth, and so on…

You shouldn’t use shielded cables for reasons @ElTel and I have clearly stated previously. For clarity RJ45 (Registered Jack 45) is not a cable, it is a connector:

I disagree.
I base my conclusions on practical tests. You base yours on theoretical considerations.
I conclude that you reached your conclusion without taking into account all the necessary factors.