Do you use a high quality (expensive) Ethernet cable from your modem to your router AND from your router to your streaming device, or do you just use one high quality Ethernet cable from the router to your streamer? Any comments/advice appreciated, TIA.
From the external of the house to my router I have fiber.
From the router on, use the best cable/filter/switch you can afford to address noise as much as possible.
My guess is that like power cords are critical from the wall to a conditioner/component, the last meter can be the most important part to focus on.
1 Like
Excellent, thanks. This is what I was afraid of … ($$$$$$$$$$$$$$$)
1 Like
What brand of cable/switcher do you like ?
Ethernet cables: STEALTH and Pink Faun.
Filter: Muon Pro from Network Acoustics.
Switches: Network Acoustics Tempus and Innuos PhoenixNET.
2 Likes
I have an English 8Switch, it is connected to the router supplied by my ISP. The router gets its data from the coax cable modem
If the switch is powered by an external power supply, it makes a difference which power supply, fuse and cable.
A Furutech ncf Lan offers amazing performance at an even more reasonable price. A very affordable option at an acceptable price, especially for longer lengths, is the Linkup cat 8 cable.
Greetings Andrew
2 Likes
I use an Audioquest Cinnamon ethernet cable from a NAS to a modem/router and the Audioquest Diamond ethernet cable from the modem/router to a PS Audio Airlens. Using quality ethernet cables provides a huge benefit once they’re broken in. I started with cheapo versions from Monoprice, then tried Pangea, then Audioquest. Audioquest provided the best performance and I recommend them.
2 Likes
The Audioquest Diamond is a great cable. However, I would rate the Furutech cat 8 ncf above the other Audioquest Ethernet cable in terms of sound. The vodka may be on a similar level but has a higher price than the Furutech. For the price, the Furutech is excellent. It offers peace and a clean sound that is well above its asking price. The plugs are simple but should contain ncf. The conductors are OCC silver plated. It’s not very thick. Its purple color is striking.
The choice also depends on preference and the entire system. The Furutech doesn’t have the richest bass range. It suits me well. It offers a silence and certain Fineness known from ncf Furutech. My J-cat lan signature gold plays very musically with a larger and organic sound quality. The Furutech surprises with a certain precision with quiet, fine tones that can be lost with many cables.
Greetings Andrew
1 Like
I have fiber from the street to a little converter box. From there, about 2 ft of Belkin CAT6 to my router. From there, it’s about 8ft of straight-shot clean wifi to my Auralic streamer.
Zero issues, zero noise, and sounds better than CAT6 to the streamer.
3 Likes
None, if you put in an optical fiber ethernet “air gap” before your streamer. Then don’t worry about the upstream copper. I use Small Green Computer but PS Audio Airlens has a similar approach.
People are confusing USB ground plane noise from a PC (can be pretty bad) to Ethernet, that could care less about that “noise”. You can use a galvanic isolator in-line with your USB to block this RF PC noise path. Audioquest sells a series of dragonfly ™ isolators that does this. The USB noise current “moves” because the ground plane differential is “broken” and not at the same DCR at each end. The FMV creates an isolation path that let’s the signal, but not the RF, through. It breaks the noise signal path. I use one, the FMV version, as PC’s are terrible at ground plane variation. Stereo quality server PC’s fix this issue as rule number one.
Ethernet has auto error correction and a BER, bit error rate, that is superb and the shorter the cable, the easier it is as the signal is HUGE over the noise. In a near 100 meter length, yes, the most signal RL reflections (not ambient external noise) are from the patch cable interface (last meter). The ambient noise is easily managed with the balanced cable configuration common mode rejection. Unless you are running like 300 feet the ACR, Attenuation to Cross-talk Ratio margin is plenty for error free Ethernet.
The idea to use better Ethernet cables applies to super long lengths where the improved ACR is a benefit. When the ACR is above zero it is a cinch for Ethernet’s error correction. Ethernet can even work to spec with the ACR below zero (noise is larger than the actual signal) as far as -4 dB. Amazing yes.
What we hear for digital, are the ringing errors in the AD and then DA circuits as they all make decisions as to where the ringing happens and the benefits to those concatenated decision at each end. My DAC has five different DA “filters” and all with different non linear error (ringing) properties and no Ethernet cable will fix that. Your choice of the DAC’s or tranceiver’s FILTER are the signal’s source errors of what we hear. Jitter and all that stuff impact the AD and DA process.
We simply do not hear the bits between the cable ends using Ethernet, we hear the DA and AD error that define the bits we send down the cable. They arrive error free after the AD makes the estimated digital singal and all the errors it has. The errors are included in the stream but the AD made them, not the Ethernet cable. At the opposite end the DA again makes estimations as to that signal streams conversion to analog. No two AD’s or DA’s will be the same but the bits off the Ethernet cable are the same. Ethernet, once fed a stream of bits, moves it error free really, really well. Once we have those bits defined, we can reconstruct and resend them error free forever. Just make sure what you send is correct at the AD stage! Once that circuit is “done” encoding the analog, it is permanently the data. There is no going back. Same at the DA, that filter makes “errors” converting to analog, the DA filter is not linear. Look at your filter responses. We have errors being made on both ends, no matter how minute but this is where they are made, not along the Ethernet cable itself.
There is absolutely zero difference in the “sound” of a 5, 5e, 6, or 6e cable until the ACR is so bad that the error correction faulters and we hear a “blank” POP sound as the circuit gives up on retransmission and we have a true CRC error.
“An Ethernet CRC error, or Cyclic Redundancy Check error, occurs when a device receives an Ethernet frame with a CRC value that doesn’t match the value the device calculated. This indicates that the data may have been corrupted during transmission.”
I’ve made Ethernet cable for twenty years and many patent to go with that work and trust me, the errors are ZERO using equipment to measure those errors that cost more than your entire stereo. It doesn’t miss a thing. Not one bit.
Ethernet cable is tested to the Nth level of the law. It is really good stuff. I could go over all the test but there are, if I remember thirteen items that every cable lot has to pass (6a 10G has even more);
DCR
DCR unbalance
RL (into a fixed load and how the cable will be terminated)
SRL (into a matched to the cable load)
Impedance (variation tolerance)
Fitted impedance (too high or too low)
ACR
PSACR
NEXT
PSNEXT
ATTN
ELFEXT
PSELFEXT
I hope this helps you decide where to make viable improvements. The DAC DA filters and the DA transceiver filters. For USB it can’t hurt to pop in a galvanic isolator between the PC and your DAC like the Audioquest FMV. That device really works in testing. Best, it is the cheapest thing you’ll buy in awhile!
Best,
Galen
4 Likes
That took some time to write, thank you.
If you ask him what time is it he writes a similar amount.
1 Like
If I understand some of what he’s written, then properly designed Ethernet cables should sound the same regardless of Category. I own BJC Ethernet Cat 5e and Cat 6a cables and, in fact, I can’t tell the difference by sound between the two types. Perhaps if I were a gamer the higher transmission speed (if that’s the way to phrase it) of the 6e might make a difference in game performance.
That having been said, I do hear differences between brands of Ethernet cables and have a preference for some of them. This could be all in my head (listener expectation bias) or it could be some of the eight or so I’ve tried are in- or out-of-specification.
If you change the DA or AD that will alter what you hear as the filters that do the conversions aren’t the same. We are making estimates as to how to chop-up the analog into pieces at one end and how to put it all together again at the opposite end.
Here is a digital DA filter that rings after the electrical pole. This produces an analog slope that we see below. This is what we hear. If the FILTER is changed, the analog slope will change. The digital signal fed to the filter isn’t changing at all. What we digitize at the front end is the digital stream. So we hear those changes, too. But the signal, once digitized and stuck on Ethernet, never changes, ever. Until we exceed the correction circuit and get CRC errors we have an essentially perfect digital train.
Bandwidth is different than Ethernet errors. Higher bandwidth needs to change the PAM, Pulse Amplitude Modulation voltage steps used to achieve that higher bandwidth. This requires a more sophisticated AD and DA circuits to properly read the added voltage steps. Distortion to the signal on the Ethernet cable will limit the distance we can go and get the achieved and rated BW. The spec is 90 meters plus the pach on each end for 100M total. But higher BW’s can limit the distance shorter. Specialized circuits can make limits much shorter than 100 meters for audio digital circuits.
To get high definition signals from analog we need to chop-up the signal more and more and need higher and higher BW to send it on it’s way. DAC’s accomodate this with different methods. The T+A SD3100 HV uses a dedicated NAA, Network Audio Adapter, interface to get the high resolution digital but limits the cable length and sharing. For streaming it needs a DEDICATED 1000baseT Ethernet line. No shared BW.
If you can’t get the BW, the system down converts it until you can send it on it’s way. Can’t do 512, do 256 as an example. And yes, less of the true analog to digital conversion is available at lower resolution. We can hear that.
Best,
Galen
1 Like
I would be less than truthful if I said I understand everything you explain all the time. It isn’t your explanation just ny ignorance when it comes to a lot of stuff that you discuss but some of it does seep in and for that I’m grateful. Thank you for taking the time and effort in your responses.
…and to think Al hears no difference in ethernet cables. Ha!
Galen thanks for taking the time to share with us.
Makes one wonder, as insertion of a quality opto isolator in the ethernet path is reported to clean things up, resulting in slight sonic improvement.
1 Like
We can improve things and it makes no difference with digital. As an example, we can bombard a 5e link with external RFI and the cable’s balanced common mode rejection maintains the BER of 10^10 bit without error. his is tested this over and over to evaluate the headroom as Ethernet has been advanced in IEEE studies. None of this is an accident. The channel length and errors are monitored to establish the acceptable channel lengths at the BW specified be it 5, 5e, 6 or 10G.
Can removing the RFI be seen, and sold, as an “improvement”? Yes. Did it do anything to the error of the digital stream? No. This requires the end user to not understand how all this works.
This is the beauty of Ethernet, it reduces the complexity of getting ideal performance. That’s not a problem to be “fixed” but applauded. We achieve 10^10 BER as specified, not fixes above that need apply.
To those complaining about the longer tech posts, it takes twenty times or more this amount of study to encapsulate it for you. So be fortunate that the assignment isn’t yours! I’m just trying to help you spend the bucks where it helps you enjoy the music better. Better DA and AD points? Yes. Better Ethernet? Only if the BW requirement isn’t high enough to stream, once there you’re done. 10/100baseT won’t work if you need 1000BaseT. Ethernet cable is comparatively cheap, so it won’t financially hurt to satisfy your doubts, even if it does nothing.
One more hint on cable. 10G is WORSE than high-end CAT6e designs for measured ACR. 10G is what is called mitigated CAT6 to 500 MHz from 250 MHz and switches the PAM to a higher number of voltage steps;
PAM-3: Used in 100BASE-T4 and BroadR-Reach Ethernet
PAM-5: Used in 1000BASE-T Gigabit Ethernet
PAM-16: Used in 10GBASE-T 10 Gigabit Ethernet
10G ONLY will be 10G if the NIC’s (network interface Cards) are all 10G and utilize PAM16. Also, for 1000BaseT CAT6e the ACR is superior to CAT6 design cable. But since 6e does not have ALIEN cable-to-cable cross talk mitigation, it is worse only if you use 10G. CAT 6e does have improved ACR out to 625 MHz. You can improve ACR by improving signal level, NEXT, or both.
Summary - if you use 1000baseT Ethernet use 6 or 6e. If you use 10G use 10G design cable for absolute length reliability.
What’s with 10G? It needs added Alien NEXT or cable-to cable noise supression. Now the added rub to this, if you do not have bundles of cable for ALIEN cable-to-cable NEXT a 6e may do 10G for 100 meters.
There is even a IEEE rcommendation on how FAR standard CAT6 might run 10G before it craps out in a cable tray with lots of cable. - The IEEE P802.3an Task Force established a minimum operating range objective to ensure 10GBASE-T operation over a reasonable percentage of the Class
E UTP installed cabling. The working group converged on a minimum operating
range of 55 meters based on contributions from multiple sources. Figure 10
shows the percentages of installed cabling channels versus channel length con-
tributed by several cabling manufacturers illustrating that for the lengths in-
vestigated 70 percent were less than 55 meters.
Alien NEXT can’t be cancelled like NEXT as it is unpredictable. It has to be rejected by other cable design parameters. 10G includes alien NEXT enhancement. Standard Ethernet, enhanced or not, doesn’t so it is unreliable in 90 meter permanent link+patch distance for 10G.
Best,
Galen
1 Like
2 Likes