Ethernet Cables and Sound

There appears to be little on here regarding Ethernet cables in awhile. I recently acquired a Small Green Computer sonic Transporter i5 and a Sonore ultraRendu. The sonic Transporter came with a ribbon type Ethernet cable. I ordered a cat6e from blue jeans cables. I then spent time swapping cables. The SGC cable was first between the sonicTransport and routet. The BJC between router and ultraRendu buried vocals. A cheapo cat5 cable in same location damped the outputs of all.

The BJC between the sonicTransort and Router and the SGC ribbon cable router to ultraRendu so far best sound balanced vocals and instruments and widest soundtage The cheapo Ethernet cable gave a good impression until same song comparison.

I am considering SoTM cables that audio bacon reviewed. Really might try another BJC 6e cable since they are only manufacture that ensures cable is meeting spec and less likely to drop . Anyone compare the BJC to SoTM? The BJC Ethernet cables are almost as stiff as Iconoclast XLR ICs.

Also interested in @rower30 opinions for Ethernet. Have you did streaming cable measuring and tuned your iconoclast with Ethernet?

I would have never thought data cables were as sensitive as analog but spent time with swaps and both wife and I heard differences that are significant depending what cable is where in the chain. Differences are no way subtle.

https://audiobacon.net/2017/07/09/sotm-iso-cat6-special-edition-the-flavors-of-audiophile-ethernet/5/?amp=1

It appears that is one of the few brave enough to compare ethernet cables

…which has resulted in Shunyata building their own Network cables.

I use Shunyata power cables, but ethernet cables? My audio system starts with a 30m long CAT5 cable running from my modem, over the roof and through a wall into a router next to the audio. I’ve temporarily run a CAT7 internally down the stairs and makes no difference.

I’m not advocating one way or the other for after-market ethernet wire. Shunyata have been working on it and they’ve come to the conclusion they can build something better than $2/ft Cat X/Y/Z cable. Knowing how pricey their wire can be, a 30m cable is gonna be a big number.

1 Like

Ethernet systems are different. They collect bits that tell the receiver all the data has arrived. Devices, like the Bridge II, cache the data and reclock it for identical pulse widths and amplitudes, reducing the JITTER, or the random pulse width genetated using the incoming Ethernet steam.

So a recaching device listens to the awkward story told by one voice, but the electronics relistens to the story by the second voice, who in my example would be a professional story teller! It never hear the first voice. ALL the first voice needs go do is get the whole story to the second voice who is really the voice in charge. Ethernet resends data until it is accepted at the destination, so even poor cables work.True, the network through put is slower as data can be re-sent, wasting bandwidth.

For the best performance, you want a cable with the best Shannon law bandwidth. This means either type system, reclocked or not, will get data more consistently. Re clocked systems won’t really see a benefit unless the original cable was pretty bad (loss of data) this is rare as most links are SHORTER, and thus much improves the signal over the noise. This improves the BW significantly.

You can go whole hog, and use CAT7 ISTP. The 22 AWG wire, and shields that pretty much remove NEXT to -100 dB can’t be beat. The signal has no real noise to compete with. But, connectors are expensive, as is the cable, and it is larger on size and more fragile as it is a FOAM dielectric.

For UTP, look at cables like DT600E or 4800. These are the highest shannons law bandwith UTP ethernet cables Belden makes. Both are over twice the spec requirements for Ethernet.

The SOUND of Ethernet depends on the JITTER figures reconstructed in the D to A process. In my case, I use ethernet over a wall wart to the Bridge II. It reconstructs then re clocks all the data so the external network isn’t “heard”. I detect zero skips or issues at all using cat6 to and from the wall warts.

Oh, don’t use 6A UNLESS you really have 10G. 6A is designed as cat6 internal, not as good as 3600, 4800 or DT600E. The reason is the special and unique designs to reduce EXTERANAL ALIEN NEXT impact the ability to reach optimum internal NEXT figures. Only 6A needs a balance of internal to external crosstalk properties.

6A will NOT be as economical or even as good on 10/100/1000 Ethernet. It isn’t made for that. CAT7, by using ISTP design can manage both tecnologies as external and internal are reduced with the shield over each pair. Overall shields called FTP around the group of 4 pairs are for external noise, but not internal noise shielding. And, outer FTP shield reduce internal NEXT performance as the EM wave are coupled into the pairs more aggressively. But, the external FTP shield reduces EXTERNAL or alien NEXT, called ANEXT. There is a balance of internal to external noise mitigation.

Shannon law depends on, as we read, noise reduction and signal level. Using shields ALSO INCREASES attenuation. This mitigates the shannons theoretical limits since the signal is weaker due to using the shield. However, if you lose 1 dB of signal, but gain 3 dB better noise levels, it is still a 2dB improvement in signal to noise. CAT7 uses 22 AWG with shields to offset the loss where cat 5, 5e, 6 and 6A use 24 or 23 AWG in UTP or FTP. CAT 7 is a brute force design on attenuation and shielding, both. The closer a shield is to the pair, the higher the signal loss. Overall shields aren’t as bad as individual shields. Use bigger wire to compensate and less lossy foam dielectrics in CAT7.

Most so called audio grade data cables won’t, again, give you any data to determine their capability. A value called ACR, or attenuation to cross talk ratio, should go as high in frequency as possible and stay above 0 dB, where the noise and signal are the same, if a better Shannon law bandwidth is to be achieved.

This is all real data on how it is tested. I have no magic theories with no supporting data to determine if it is in any way an advantage. The signal needs to be larger than the noise. Anything that loses signal impacts the BW, such as poor impedance match to the load called return loss. And, anything that allows internal cross talk noise. The ACR value combines all this into one trace, nice!

So that’s how this is really working. You can certainly go with unproven theory that magic materials and designs are supposed to improve, but with no adherence to maxing out the fundamentals, how can the cable be better? If a theory is valid, the design should also adhere to good practice to improve ACR, does it?

Best,
Galen Gareis

2 Likes

I just switched to Cat 8. Thoughts? I thought the sound became slightly more open. But it could be my imagination?

As it embarasses me a little, I felt a pretty big difference when using a Kubala ethernet cable.

I was on the 30-day trial, and I was so certain that it would not matter that I did not bother to install it up until the 20th day. I was preparing myself to send it back without even hearing it. Then the thought: “what the heck? just plug it in”.

And voilà. More detail, more open, more alive. Even felt louder. Pretty big differences. My brother-in-law, my father and all the women in the house. A/B with them blindly. 20 times. They got it right all of them.

My setup comprises fiber up until the last switch, and then the ethernet to UltraRendu.

Hope it helps. It sure did not help my bank account.

Cat7 is 600 MHz
Cat7a is 1.2 GHz
Cat8 is 2.0 GHz.

It is highly doubtful that the BER will move at all with shorter Ethernet links. using only the extreme end of 100 meter channel link could any Ethernet cables show a true advantage. This is distance and x-mission speed both.

Heavy industrial uses mostly UTP ethernet as the noise rejection properties are so good, and they aren’t dumbies knowing the link length is so short ACR is excellent. They even use just 2PR UTP for 10 base T as that is all you need for that.

But, the cost to use a three foot patch cord with cat8 won’t keep clothes off your children. Some links can be too short, and if there is a severe RL reflection, it won’t be attenuated out. Short channels and permanent links can be tougher than longer ones for this reason. Most permanent links want to be 15 meters long to avoid this RL issue.

The RJ connector is a really hyper stresses part and goes WAY past the original spec. Better cable and RJ jacks with compensation parts get the job done to 5e 6 and 6A. Most hi-fi equipment DO NOT use the best JACK to approach 6 and 6A speeds. Special connectors need go be used. Or, you get essentially 5e as the jack is a bottleneck. Most jacks are 5e, 5 is pretty old school. Ask the equipment many what jack they use.

The WHOLE channel, permanent link, connectors, and patch cables need to be optimized if top of the world test values are to be reached.

Galen

4 Likes

Know that the test agencies UL, ETL, ETC ONLY test to the STANDARDS in place, not above it. The additional headroom has to be self certified by the manufacturer.

When a cat6000, made that up, cable is UL verified, it is to the the 7A standard, for instance, the highest current standard.

And what comes to sound i have listened tens of different ethernet cables from cat5 to cat8 and from many manufacturers. They all sound different and not related to cat standard. Cat6 can sound very good and cat8 can sound not so good. Don´t look at that too much :slightly_smiling_face:
It is the same as with any other cable,power or ic or digital…you need to find what suits your system and taste.

1 Like

With the DS senior and the Bridge II network cable influences the SQ like any other system cable (IC, power, etc.), I currently use Sotm CAT6 cable, I preferred it to Wireworld Starlight CAT8, both very good with sounds but the Sotm is still superior.

Reinforcement, be sure to pay attention to the network cable with the DS + Bridge, this step can for everything (or almost) to lose.

1 Like

BER is 10E-10 for ethernet. But a very rare error won’t change the continuous transfer of data and thus the sound. Errors aren’t the issue, if there even is one. Until a bit sees a DA, it has no sound.

To change the sound, the PWM definition would have to change. But the pulse width is fixed, so I don’t see where the external network that should be reclocked removing jitter, by most devices, can change the sound.

Certainly the AD encoding sets the initial level of perfection and thus the final DA sound can’t be better than that, but the initial quality stays intact until the DA firmware filters can make audible changes.

3 Likes

As an “RF Guy” working in IT because it pays the bills. I’ve seen some of the most piss poor cabling in main distribution frames (MDF) & intermediate distribution frames (IDF). How do they successfully transport all that data and yet Audiophiles insist on megabuck Ethernet cables because it makes a difference. I just use good 'ole CAT 6e shielded. CAT 7e is the best but only if you really need that much shielding.

Rant over.

3 Likes

Back when I was a real RF Guy, I used to test 16 bit telecom MUX/DEMUX chips @ 10E-6 short term BER and 10E-10 @ 9.95328GHz, 10.664GHz/10.709GHz (WRAP), &12.249GHz/12.5GHz long term BER. I can’t remember what I worked on yesterday but I still remember those data rates from 20 years ago !

“Thanks for the memories” ! :grin:

2 Likes

Analog cable has several ares where changes in the cable can alter measurable attributes that superimpose go distort the signal in various ways. This is the nature of analog,

Digital, too, has measurements that define how much the “analog” signal in the guise of a square wave is distorted, I agree. But, digital. Isolates the signal to a one, or a zero. We only need the fine, answer to be the correct one or zero that it was when it left the far end of the wire.

This is what is so nice about digital. We break up the signal into easily identified pieces. Sure, the smaller the piece the better it is built back. Those fundamental pieces identified right REMOVES the cable from the “sound”.

Sure, if we mis identity the bits, we hear clicks and pops, or error circuits fills the gaps with what was being played immediately before. That does not change the tonality and timber of the music though. A persistent and continuous analog “distortion” does that.

Moving bits from A to B is pretty transparent to the external cable. A change in sound means SIGNIFICANTLY different PWM bits are comprising the finalsignal. To digital, this is a BIG change, and would trigger all sorts of detection and correction errors. The DA filters can, and do, change the musical timber and tone. But this is the analog domain now.

How are we changing the data between point A and B such that a CONTINOUS change in tone and timber are now encoded, and go undetected as errors in the Ethernet link to be corrected?

I’ve designed digital Ethernet cables for 25 years, and kind of know how to make sure the bits are the same at each end. Same, same is really good, and why digital can be so good, too.

Differences in theoretical errors on cable, yes, but differences in the sound? What are the underlying reasons we think we KNOW digital is as fragile as analog for signal integrity?

Don’t say, because (same as no answer), or jitter (We reclock the data to re-set the pulse width).

Trying to define real reasons for the signal at A turning into, digitally, an entirely different signal at B. Noise is not it. Noise won’t, and can’t, change the PWM modulated initial signal encoding into a different animal. Errors, yes, but those aren’t a continuous change in the signal.

It is BER. Same data means the same sound UNTIL the DA filters. There we can have some discussions.

5 Likes

Ah, but what if the “noise” gets transferred through the digital cable to the analog circuitry of the DAC?

I’ve noticed an SQ improvement with adding a fiberoptic segment to my Ethernet connection just before the Bridge.

In addition, the differences in software running the stream (say mConnect vs Roon), are minimized in this set up.

1 Like

Well, the “noise” doesn’t get re-clocked and is isolated at that point. The circuit that re-builds the bit stream ignores “noise” with digital where analog can’t. This is how repeaters work, we take the signal just before we can’t tell a one from a zero, and rebuild it just like brand new!

The major issue is ACR, the signal simply falling too far BELOW the noise to be rebuilt, and sending it on it’s way again, either into your NIC in your DAC card or down another length of cable. The ambient noise stays the same, but the signal ACR is restored.

A different sounding PAM encoded signal is TREMENDOUSLY different digitally and would trigger error detection for sure and it should, Ethernet is designed to see ANY differences.

Better cable do indeed improve ACR so there is slim chance of an error. But until an error happens, the bits are bit perfect.

1 Like

The same company that I tested those MUX and DEMUX chips also produced CDR (clock and data recovery) modules used in the transatlantic fibre optic systems (underwater). They also incorporated parts of the CDR circuitry in a later version of the DEMUX chip.

Galen,
So far a BJC CAT6A from streamer to Router and a CAT7 thin ribbon from Router to Ultrarendu and Inakustik USB to DirectStream is really lively and widest and tallest soundstage. Best percussion and vocal to instrument balance tone and timbre. I need to spend time and figure out howdeepsoundstage goes.

I would like to try the CAT7 with the foam I saw some raw cable but nobody doing patch cords yet. Any ideas where to source those. Also going to try another type CAT7 with round wires.

I have another half dozen plain vanillas to try. I am going to try to find something great that sounds audiophile without the price tag. But every cord sound different as does the position. A cheap one can gives lots of depth but output is not goosebump territory. It reminds me of usiing really long XLR ICs on the DirecrStream.

You will find 3600 and 4800 series cable is superior to 6A for ACR with 1000 Base T Ethernet. That’s the facts. 3600 and 4800 is less likely to see a Bit Error than using 6A for 1000 BaseT. 6A has special external noise verses internal alien noise cancellation that is important to 10G. The alien noise technology REDUCES the ACR for 1000 BaseT but allows 10GBaseT by extended BW and noise filtering.