Welcome Tom!
Ethernet automatically accommodates balance error because a legal high or legal low have a range of voltages that are accepted as ārightā so symmetry has to be good enough to meet BER for a worst case Ethernet system. The CUB, capacitance unbalance, limits control the balance error to keep the signal levels legal high or low values at the spec distances.
Yes, shields can add as much of a problem as they solve (worse internal NEXT by about 6dB all things being the same). The advantage shields have to offset the limitations.
There are tests that inject noise into an Ethernet system until the system exceeds BER requirements and show Ethernet, even UTP in industrial settings, is VERY robust moving bits with essentially no errors.
My PS Audio DS DAC and my T+A SD3100HV DAC both are changed when the filters are changed. I have NOT heard changes in my system with USB verses UPnP/DLNA Ethernet using the same files. I do seem to hear better sound when I use the PS Audio DS Memory player through AES/EBU or I2S. My take is that the sound of the player itself is somehow involved and yes, it is technically all digital so I canāt blame the AD and DA as there is none. This is interesting.
Iām pretty much pointing out what we know as fact, and can measure as that, fact. My bank account never seems to fail digitally from anywhere in the world. Ethernet is āaccurateā. AD and DA filters are NOT as accurate as we hear. Slight quantization errors like jitter are time based. We sample the time window amplitude above, below to the left or right or two of them (above to the left or right or below to the left or right). Our brains are time based organs and hard to trick for long.
As far as cable design for Ethernet, the dielectric can alter the ACR, attenuation to crosstalk ratio, and trade-off signal strength for NEXT and cable size. ACR doesnāt care which gets better, attenuation or noise, it is the difference Shannonās law works with. This stuff is well understood looking at BER streams and eye pattern tests.
Ethernet cables are some of the most tested products out there;
Impedance
RL
Attenuation
ACR
PSACR
NEXT
PSNEXT
ELFEXT
PSELFEXT
SKEW
DELAY
DCR
DCR Unbalance
Capacitance
CUB
Best,
Galen
1 Like
I am not talking about BER, but about that the not symmetrical signal can cause irritation in the end device.
Yes you are talking about BER but donāt realize it. The cable symmetry driven by capacitance that sets the impedance and RL properties and even NEXT does not cause BER error changes in a 100 meter channel due to CUB, capacitance Unbalance, unless it is far out of spec. The test data confirms the BER is in spec up to the non symmetry allowed in the two legs to ground by the CUB spec limits. Ethernet is VERY well speced and nothing is left to the imagination. Itās all tested and verified over and over.
CUB is the difference in the capacitance from one lead to ground compared to the opposite lead to the same ground. āSameā is the target. You are correct there is a difference but not that it irritates the end device to fall out of essentially perfect BER in up to 100 meter runs. In God we trust but all else needs to bring data to show the current Ethernet spec causes excessive errors above spec within the variation of ALL spec limits. It wonāt. This has been tested with SPC over 30 years or more.
We donāt have a smoking gun outside of DA and AD to explain audio differences, there are NO tests to even show change in errors that can then be discussed as causing a difference in sound. We keep looking. Thatās the scientific way so Iām not saying analysis should ever stop. All current science is up for changes as we go. Symmetry changes within reason so far havenāt been a smoking gun for bit perfect Ethernet systems.
Best,
Galen
2 Likes
Hello,
please donāt get me wrong
I have the problem that I hear these damn differences - even in the blind test. Iām just looking for answers that havenāt been found yet. Iām about to buy a 1Ghz scope as a hobbyist, just to get an explanation here. I find it absolutely unbearable that all the providers of so-called audiophile network devices and cables do not even announce the effect of their devices on the Ethernet signal - perhaps because they do not measure at all or the result is sobering.
I am absolutely sure that there will be no data change and that sound differences can only arise through the DAC or afterwards. My RME ADI 2 DAC has a bit perfect test and it is always positive, even with the cheapest network cables and network switches.
My theory is not based on a change in the data, but on the fact that asymmetries related to GND are carried on in the end device and interfere with the DAC or the analog signal after the DAC.
but that is just an attempt to explain the inexplicable ā¦
3 Likes
I am very glad you are here and I am really looking forward to the results of your work. I want answers as well.
I look forward to finding out what you learn. You have set up quite a task for yourself.
Good luck!
yes, caused of the inability of the audiophile industry to shed light on the dark.
2 Likes
I have yet to find two cables of two different model or different manufacturers that sound the same. Most ethernet cables sound pretty bad after you hear a better one.
I also found if you change Linear Power Supplies in your streaming chain. The ethernet cable you like best can also change. Mainly because of distortion s is my theory. The more revealing cable then becomes the preferred. The darker sounding cables make the bass to grungy, slow, and thick you end up with a one note bass speaker that doesnāt delineate.
I may slice up a cable just to pull the shielding and insulation to try the trick you uncovered on the German forum.
I do appreciate you stepped to the dark side and can admit science does not explain everything since not everything is known quite yet.
2 Likes
I donāt understand : what is received is 100% what is sent, bit for bit. So how could the sound be different ? This is a very interesting thing to investigate IMHO. The Hi-Fi higgs boson. Very interesting subject. Iāll try to find the post about the german forum right away. Thanx everybody.
Couldnāt find the link to the german forum, or let me know what is the trick from the german forum.
Mother Nature does indeed explain everything but Iāll grant you the idea that only when we understand SOMETHING repeatable we can then start to call it a science verses a theory.
Best,
Galen
4 Likes
Do you mean the German forum i mentioned?
That is aktives-hoeren.de. You have to use the search function to find the right thread and posts.
Still looking for an affordable scope to do measurements - even if there is the risk, that it will show nothing related to the sound differences I think i can hear.
@rower30
What Belden cable with bonded pairs would you suggest, when only using the stripped pairs.
Are the twisted pairs as long as these have the same gauge the same in differenten cable types of Belden?
Regards,
Tom
Hi Tom,
Each category cable will vary the four lays depending on the NEXT parameters. 5e is longer lays than 6 for instance. 6e is shorter yet again than 5e. Generally lays are 0.6" or less. You can visually pick the lays you are interested in.
Not sure how you want to test, but you can use a FREE bonded pair tool to separate the BONDED pairs and test the striped side and / or the solid color stripe. Call and Belden sales can send you one out.
1700A bonded 5e
1872A bonded 6
7851A bonded 6e
1212 unbonded 5e
2412 unbonded 6
3612 unbonded 6
Best,
Galen
1 Like
Hi Galen,
Iāve been following this thread on and off for a while now and it wouldnāt be truthful for me to say that I āgetāā what seems to be some of the more technical points that have been expressed.
I use two Ethernet cables that run from my X1 Gateway ā one goes to an Oppo 105D and the other to an original Bridge in my DS DAC. Iāve used a Cat 7 cable and a Cat 8 from Monoprice and more recently two Cat 7 cables from Pangea. To my ears the three Cat 7 cables sound a little better (more open) than the Cat 8.
Reading through this thread it seemed that Cat 5e and Cat 6 cables were held in some regard so I tried one of each that I had as giveaways from Comcast. The 5e and 6 seemed to sound even more open/detailed but at the same time not as smooth and harmonically correct.
Am I hearing what I want to hear (that is, the 7 and 8 are smoother, richer) while the 5e and 6 are more open and a bit more detailed or am I hearing what are the inherent differences between the categories?
Thanks,
Joe
2 Likes
Hi Joe,
As you will find, I tend to point out how cable is electrically different, and how lumping those parameters into place can help ID cableās āsoundā as it is driven by the electricals.
If interested I have papers that go through all this. It isnāt meant to be a secret and isnāt. It simply how it works. Just ask.
An ISTP cable is VERY different than a UTP cable. When we add a shield we add a lot of TROUBLE if it is not done well, or the cable is damaged during use. The geometry of where the shield is relative to the dielectric alter the capacitance and final electrical of any shielded cable. Itās a part of the equations that determine the capacitance. Nothing magic here.
The higher you go in frequency (CAT7, 8 etc) cables the worse the variation becomes. It a simple matter of WHERE the shield is along the cable. But, if the NOISE is WORSE than the issues derived by using a shield it is a benefit. Go high enough in frequency and long cables the signal is TINY and NOISE is a real issue to extend the BW so the high effort is worth it for what is called ACR, Attenuation to Cross TALK Ratio. Keep the noise OUT and the signal pure helps the cables bandwidth. A bigger copper size to increase the circumference and reduce attenuation is a help, too. The cross section isnāt important at RF, but the surface area is and to do that we make it bigger. Ethernet uses a WIDE frequency range that drops to 1 MHz so we canāt use steel or aluminum inside the wire like GHz RF coaxial cables do.
Shorter cables can be too good! Yes, the RL reflection inside the wire from non ideal reactive impedance values cause some signal to bounce off the pure resistive load. We ideally want to attenuate this out and in short length it isnāt too well attenuated. Usually the signal is still BIG even if the RL is there, so it is swamped by a nice fat signal. Usually. Much of the fear is talk, not fact.
Ethernet patch is a smaller AWG size on purpose to help mitigate near end reflections in 15 mtr worst case channels where the length is shortest and RL reflections will be the worse. Good properly made cable removes a lot of this, though. A failing channel can be changed with a patch swap either way (bad horizontal or bad patch). But the LENGTH is important to what happens and what the failure is; RL, NEXT or other issues.
I can send a UL test set to show all a Ethernet cable has to pass. It is a long list of things. Again, Just ask.
I tell people to get the best patch cable you can as this is where most of the problems are made in Ethernet channels. The closer you are to the end, the more money you should spend.
So whatās that all mean for sound? In my system. I use my PCās USB (JRiver MC28) and UPnP/DLNA Ethernet (MINIMSERVER) both with a T+A SD3100 HV DAC or PS Audio DS DAC. I canāt say sound is different on a given DAC but the DACās are VERY different! So no change in cable but the general sound of the DAC filters are not the same at all and never have been. I hear a difference in the DACās filter compromised and not the USB verses Ethernet delivery. And I use Belden PS6+ bonded pair patch andā¦to test worst case, I use power-line Ethernet modules! Correct, my āpermanent linkā is ROMEX!! Compare this to USB with spec tested USB 2.0 cable. I canāt say I hear the difference but I STILL hear the DACās filter changes. The 3100HV letās you change them in firmware, and the PS DS DAC uses a firmware update flash change. But it changes the filters either way.
My take? Make sure you enjoy the general sound of your DAC first. THAT will be 99% of what you hear with quality USB or Ethernet delivery. Second is your files, same as records. NOTHING trumps the source sad to say. Last is really the Ethernet or USB butā¦last does not mean forget about it. I use AES/EBU with quality XLR for my DS memory player to either DAC and USB from the PC with the 3100HV as this is a more direct connection.
Cable is definitely different and more $$$ to make as we add and optimize SHIELDS and the internal insulation geometry. Can this be heard? It sure canāt be if the cable isnāt better and better as we go! Nice quality Ethernet is made in MILES at a time so the value is exceptional. I havenāt seen cable exceed the quality level of good TIA approved Ethernet cable in real world UL/TIA approved tests. Never. Some as good? Few. Most arenāt meeting the TIA specs and arenāt UL or ETL third party verified. Look for that rating!! If you donāt see it guess what!
I go with the foundation that the basics have to be founded in reality BEFORE and weird ātheoryā is able to exceed standard cable that might, I say might, not have that added āmagicā. Get the basic (UL or ETL approved ratings) before the magic or move on. You deserve to get BOTH for extended performance cable.
Best,
Galen Gareis
2 Likes
I could not agree more.
Excellent advice.
3 Likes
Hi Galen, I am considering a short 0.5 meter Ethernet cable. Is there a minimum length ? If I understand you correctly, the answer is that shorter cables are just as good as long ones. Correct ?
Hi Galen,
Thank you for your response which Iāve read through a couple of times and think I get the general concepts.
The Ethernet cables I use are short in length: either 1M or about 1.5M and run from my router: the Comcast/Xfinity X1 Gateway. The only time USB cables come into the picture are when they connect an external hard drive either to my 105D or to my laptop and in the latter case the files are then mediated by JRMC 28.
Besides shiny discs the other ways the Oppo is fed can be wireless from the router or by Ethernet from the router. I can hear differences between wired and wireless and if the latter I can hear a difference due to what Ethernet cable is in use.
When it comes to the DAC the choices are a little more varied. The signal can go from the Oppo via coax to the DAC with the original signal from the Oppo as I described or via Ethernet from the router to the DAC. All of these avenues sound different to me depending on the structure (wired/wireless) of the āroadā and what material (Ethernet cable category) is used to construct the road.
Considering the shortness of this āroadā I do not think that noise is of concern although I could be naively writing-off this factor. So we come down to shielded/unshielded and other factors such as dielectric and conductor composition (metallurgy and gauge) and their possible effects on sound. Iām assuming this last sentence to be true because they are factors for non-Ethernet cables and maybe that isnāt as true for Ethernet cables.
Iāve been involved with this hobby (in your case profession) of ours for a very long time but it seems there are still things about Ethernet cabling I need to learn. You mention some source material I could read on the subject of Ethernet cables. Is there a specific website youād recommend?
I very much appreciate the time youāve spent helping an old dog maybe learn a new trick or two. 
Kind regards,
Joe
In a shorter permanent linkā¦the stuff that stays in your wall, determines if a shorter or longer patch is āOKā. A real short permanent link likes a longer and 26 AWG patch to attenuate out the end reflections. We have a lot of things that change a total channel (permanent link + Patch);
-
How good is the impedance / RL control in the permanent link cable itself? Connectors canāt improve that just add less damage!
-
The patch total RL value cable is largely impacted by the connectivity installed. Whatās the patch cable test look like? Blue Jeans tests and sends a certificate / test chart with EVERY patch cable.
-
How long is the permanent link? Shorter means there can be stronger RL reflections that donāt attenuate out. Is the Permanent link passing RL requirements?
OK, great. All the above means it is better to error on the side of a slightly LONGER patch. I recommend BONDED pair as it control RL (variation in center to center distance) on a patch cable that is moved a lot. I would never go less than 1 meter.
Best,
Galen Gareis
1 Like
Hi Joe,
Ethernet is DIGITAL so we have the system re-clocking everything to a 0 or a 1. All the Error we hear in analog (wire distortion, dielectric distortion and all that) goes 100% away with digital and why it was even invented. It works, too.
The entire Ethernet system is aimed at reclocking out all analog type errors once the signal pops out of the NIC, Network Interface Card.
Error has to happen in the DA process after that. If the bit stream is inconsistent enough to fall off the clock so to speak. JITTER is the often referenced error in TIME that can force a retransmission in Ethernet butā¦Ethernet is DESIGNED to not pass errors but recognize them and ask for it to be resent if the parity bits donāt line-up (no errors). These MAC address violations (where stuff is going) are called CRC error check violations. Bad cables and stuff can force too many error checks and the link can time-out, if it has already dropped to the lowest but most error free, speed. Ethernet automatically pings the channel to see how good it is and what produces the proper BER and uses that speed. It does not up the speed until it is certain everything allows the higher speed to work to Ethernet BER requirements. And I mean EVERYTHING, cable, connectors, NIC cards.
Most are interested in how NOISE getting past the Ethernetās zero bit error influences the analog sound coming off the DA section. Analog is ALWAYS going to superimpose signals onto the root information as you go getting worse and worse every inch. This is where galvanic isolation comes in. How to get perfect bits but not the RANDOM RF noise the system makes and the room we are using. This includes the PC or streamer. There are isolators you can try to see if this is important in your system. A lot of new equipment is isolating the bit data transfer to prevent noise from grounding any where near the AD and DA sections.
As far as cable goes, it is rated by Shannonās bandwidth laws. The higher the bandwidth the better the ability to move data error free and at higher speeds. The structured cable system Belden uses is pretty simple, it uses the data speed to be the cable part number in gbits/sec. There is a difference between the required standard bandwidth and the sales / marketing BW (usually higher). Thatās the headroom value.
1200 = 1.2Gb/sec or a frequency of 160 MHz
2400 = 2.4 and 200 MHz
3600 = 3.6 and 200 MHz
4800 = 4.8 and 300 MHz
10G = 10 and 500 MHz
All Ethernet cable can be organized into their bandwidth speeds. Nothing new here. Match the cable to the speed of your end hardware. Better 4800 cable wonāt change a 5e NIC cardās effect on the channel speed. Butā¦if the cable is hard to get to, install the best you can so it wonāt limit the future as much.
Donāt forget that the CONNECTORS are the WORST offender in poor tested channel performance and are tricky to install if not impossible without the proper tools. I never make my own above CAT5 and buy TESTED and CERTIFIED after assembly patch cable WITH a true test report. No WalMart of their rack stuff. But the plugs can be good and have poor quality jacks in your equipment that canāt be changed. THAT can be a hidden problem and the manufacturer needs to up their game and get high quality and properly loaded jacks in their equipment based on their BW target. Not all jacks are the same at all as youāll find below. RL and NEXT vary with quality and the cable or plug on your patch canāt fix that jack if it is poor quality.
I would argue with no data, that people are listening to the connectors and not the cable most of the time and donāt realize it. The connector follows the cable and is the WORST link in the chain by far.
To learn more go here Belden Wire and Cable Products
Best,
Galen Gareis
3 Likes