Surprise findings on Ethernet cables/cleaners/reclockers

[BlackTalon]

Some equipment may not have good galvanic isolation, or may be sensitive to noise, but those are flaws which should be fixed, not mitigated.

You, sir, have inspired me to trademark the name 'Flawed Devices' and start a company producing ethernet products that deteriorates the data stream yet is claimed to enhance the sound a thousandfold. The ironic name alone will allow me to sell the products for deep into five figures per box (US $). I, and my bank account, thank you. I intend to use a small part of the proceeds to purchase BS Audio and Fraudioquest, and close down both companies.

 

[anotherhobby]

Can you name a single audio streaming service that applies to this discussion that uses UDP that's not somebody's open source project on github?

https://fiestajetsam.github.io/draf...rements/draft-gruessing-moq-requirements.html

...points to a draft open source project on GitHub, that also does not directly apply to this discussion.

 

[AudioSceptic]

This was posted on another forum (in German):

Ethernet Infrastructure Measurements - Switches​

https://www.open-end-music.com/foru...von-ethernet-infrastruktur-switches-nur-lesen

Use Google translate if you need help reading it.

Eric, the author of the study, investigates how noise and jitter in an Ethernet transmission can be affected by ethernet cables and various audiophile tweaks, such as switches, EtherRegen and external clocks, isolators, reclockers, and "better" power supplies, such as those from iFi, etc. Unlike most measurements on ASR that are done at the output of a DAC, Eric measures in the digital domain before the DAC.

The last post in that thread contains conclusions so far. Here's a sample (courtesy Google Translate):
View attachment 237865

PS: The title of this post is a bit sarcastic, as I think, most here would not find Eric's findings to be a huge surprise

I hope this is not OT but how about a $3.2 k switch (plus another $3.1k for its PS and cables).
<https://www.stereophile.com/content/nordost-qnet-network-switch-qsource-linear-power-supply>

The first thing I heard after I installed the QNet was that Gens's voice grew in size. Colors were more vivid. As silence filled spaces between notes, the soundstage seemed to expand in all directions. All that from a simple switch?

It seems this insanity shows no sign of stopping any time soon.

 

[voodooless]

I hope this is not OT but how about a $3.2 k switch (plus another $3.1k for its PS and cables).
<https://www.stereophile.com/content/nordost-qnet-network-switch-qsource-linear-power-supply>

It seems this insanity shows no sign of stopping any time soon.

Yeah, we’ve had that one before:

Snake Oil Department, Top This

TAS and SR at the same time.....that's a tough combo to beat. These companies are just profiting from human stupidity!

www.audiosciencereview.com

 

[DonR]

I hope this is not OT but how about a $3.2 k switch (plus another $3.1k for its PS and cables).
<https://www.stereophile.com/content/nordost-qnet-network-switch-qsource-linear-power-supply>

It seems this insanity shows no sign of stopping any time soon.

He neglected to mention the acid he dropped first.

 

[anotherhobby]

He neglected to mention the acid he dropped first.

I was thinking mushrooms, but it could also be acid.

 

[Doodski]

I was thinking mushrooms, but it could also be acid.

At the rate that magic shrOOmz are recommended and becoming used for medical use and the want for them as a recreational drug they will be legal in Canada in several years to 10 years I'm guessing. They are harmless fun lil tasty morsels.

 

[AudioSceptic]

Yeah, we’ve had that one before:

Snake Oil Department, Top This

TAS and SR at the same time.....that's a tough combo to beat. These companies are just profiting from human stupidity!

www.audiosciencereview.com

Sorry, I missed that.

 

[AudioSceptic]

I was thinking mushrooms, but it could also be acid.

There's not much else that could explain their disconnection from reality.

 

[deadwood83]

I hope this is not OT but how about a $3.2 k switch (plus another $3.1k for its PS and cables).
<https://www.stereophile.com/content/nordost-qnet-network-switch-qsource-linear-power-supply>

It seems this insanity shows no sign of stopping any time soon.

Hey man, we can;t just bash switches based on price. Some of us come by our multi-thousand dollar switches honestly. E.g. my home switch has have an active HPE warranty.... It may also have 100gb uplinks....

It handles audio fine if you're not in the same room. But, you know, maybe a tp-link unmanaged 10/100 with a reclocker would be better after all. TP link HAS come a long way in the past 10 years.

 

[kchap]

There is no such thing as a 1000BASE-T loopback. All 4 pairs are bidirectional. Another effect of this is configuration is you cannot send clocking over 1000BASE-T. Clocking, SynchE, is okay on100BASE-T and fibre.

 

[KSTR]

You're forgetting about end station buffering. Every audio streaming app I've tried (several, including Apple, Amazon, Spotify, etc.) buffer at least two seconds of audio. All of the effects you're discussing will be negated by buffering.

You still don't understand and still think in the digital domain only, it seems. Buffering does nothing against common-mode RF noise entering the system, this happens right at the RJ45 input jack and it is an analog phenomenon.

 

[MaxwellsEq]

It handles audio fine if you're not in the same room.

Or if you like listening to recordings of fighter jets at 120dBA

 

[MaxwellsEq]

...points to a draft open source project on GitHub, that also does not directly apply to this discussion.

You didn't read it, then...

 

[mcdn]

It is trivially true that even a buffered digital stream will exhibit timing artifacts on output that depend on the stream contents, RFI, phases of the moon, whatever. There is a whole class of cryptographic attacks based on detailed timing artifacts, and those folks at the NSA and GCHQ have put a lot more work into the field than any audio company.

The web browser you are reading this on contains code to deliberately blur the system clock to any code running inside the browser, precisely because timing attacks are so easy that allowing access to a precise clock is a security risk.

Heck, it’s probably possible to look at the jitter spectrum of a signal going into a DAC and do better than chance at predicting if the music is Kenny G or Metallica.

But it just doesn’t matter. The information content of the timing variations is so far below audibility it‘s like listening for a baseball hit in New York from London. What was the residual jitter on the SMSL500, -120dB?

 

[DonH56]

Seems like a good time to invoke @RayDunzl's famous Shoutometer: https://www.audiosciencereview.com/forum/index.php?threads/the-shoutometer.2555/

 

[anotherhobby]

You didn't read it, then...

I did. Did you? I didn't see anything on a single active audio streaming service that doesn't use TCP. In fact, the article specifically states the following regarding on-demand media (which is what we are talking about here):

Finally, the "On-Demand" use cases described in this section do not have a tight linkage between ingest and streaming, allowing significant transcoding, processing, insertion of video clips in a news article, etc. The latency constraints for the use cases in this section may be dominated by the time required for whatever actions are required before media are available for streaming.

In other words, there is little if any advantage to low latency in the standard on-demand music streaming use case (which was the argument that was made for UDP or other non-TCP based transmission). And every audio streaming service uses TCP, so I honestly have no idea what point you are trying to make here at all. I'm guessing I'm not the only one.

 

[blueone]

You still don't understand and still think in the digital domain only, it seems. Buffering does nothing against common-mode RF noise entering the system, this happens right at the RJ45 input jack and it is an analog phenomenon.

I read the article, as best I could, and it assumes poorly-designed switches, DACs or AVRs, etc. You can get almost any sort of result when good design principles are ignored. If you go looking for rare cases where 1000BaseT implementations have issues you may find them, but asserting that this is the general case would be incorrect.

 

[deadwood83]

Or if you like listening to recordings of fighter jets at 120dBA

I did recently stream Top Gun: Maverick ¯\_(ツ)_/¯. Maybe it felt 'inspired' by the bits passing through. Or maybe I finally need to adequately vent that server closet.

You still don't understand and still think in the digital domain only, it seems. Buffering does nothing against common-mode RF noise entering the system, this happens right at the RJ45 input jack and it is an analog phenomenon.

But PHYs have been magnetically coupled and have had 49.9 ohm load resistors on each pin for over a decade. If you have enough noise coming through to impact the end device ground, you probably don't have a working connection.

I work in IP telephony where mixed TCP and UDP is my day to day. It does not take much to upset TCP. Ethernet will fault (and probably the PHY/controller) before ANYTHING even comes close to impacting the endpoint ground via some sort of potential difference which might cause analog "sneaking in." This is doubly true for PoE devices. They have 0 tolerance for any funny business. Triply true for a 10g link. Want 802.3bt on 10G-baseT? You will know at roughly the 500ms mark if your cable supplier has cut any corners.

But if you are really worried about PHY transmission, there's nothing stopping you from replacing your copper runs with OM1-OM4. It would still be multitudes better value than any audiophile funny business because OM1 is still rated for 10g up to 33m. And before anybody jumps on the latency bandwagon, I get motion sickness very easily, but manage to use fiber-based DP and USB for 8k VR between rooms. You can even run fiber inside HVAC plenums without any special rating. It saves a lot of heartache on drywall and paint. (That's a return plenum, hence why I need to clean beneath the grille and make a custom unit.)

Audio reminds me of a famous Shakespeare piece. I believe the title is Much Ado About Nothing.

 

[blueone]

In other words, there is little if any advantage to low latency in the standard on-demand music streaming use case (which was the argument that was made for UDP or other non-TCP based transmission). And every audio streaming service uses TCP, so I honestly have no idea what point you are trying to make here at all. I'm guessing I'm not the only one.

Isochronous transmission is never used in consumer audio devices. Buffering a few seconds of music is always the cheapest solution, even in the cheapest mobile devices.

IMO, Ethernet really does suck at the link level for high performance data center use, which is why most other interconnects like PCIe, CXL, FC, IB, and HPC interconnects work differently, but digital audio is just not a challenging application for even a 1GbE network. TCP is hardly a state of the art protocol either, but this notion that digital audio streaming is somehow a demanding or challenging application for modern Ethernet/IP networks is just nonsense.