Ethernet Cables and Sound

Galen, where can we find these 3600 and 4800 series cables?

I know you asked Galen, but if you check locally for low voltage cabling outfits, I’m sure one of them will sell or likely give you some for the asking. It’s sold in 1000’ boxes and cablers can’t afford to waste labor on a box that has too little left to guarantee it will finish the pull. You’ll have to terminate them to length.

I like the bonded stuff for my large data installations but I know the guys terminating really don’t, adds a bunch of work but completely ensures the pairs will always remain in the same physical relationship.

BJC can probably make up patch cables using 4800

Well, this (reply and subject of discussion) is interesting. I am going to have to revisit some assumptions I have made about optimizing the signal chain. Your point, if I understand correctly, dictates that so-called “galvanic isolation” of the signal (e.g., converting ethernet to optical and back again) prior to feeding the signal to the PSA DSD Sr. DAC via the Bridge Ethernet cards would have no benefit. Interesting…


[Clarification: Noise-reduction benefit, that is…]

Go to the site and then…

Products, Resources, Technical Resources, Product Brochures

Then in filters;

Product brochures

Enterprise Copper

You will find;

Belden® DataTwist ®
1200 Cables - PB294

Belden® DataTwist ®
2400 Cables - PB289

Belden® DataTwist ®
3600 Cables - PB 295

Belden® DataTwist ®
4800 Cables - PB 296

1200, 2400, 3600 and 4800 all exceeded the TIA568C.2 cable standards for 1GBaseT Ethernet. The series part number end in a 12 or a 13; 2412 or 2413 meaning riser CMR NEC flame rating or Plenum CMP NEC safety rating. The electrical performance is the same, just where you can safely install, the cable.

1200= 1.2 GB/s

2400= 2.4 GB/s

3600= 3.6 GB/s

4800= 4.8 GB/s

All, are tested to TIA certified methods for Ethernet digital data transmission, and that’s what our beloved audio Ethernet is, same as your bank statement. Each level has a higher Shannon law bandwidth for theoretically fewer and fewer errors.

Galen Gareis

Hi Scott,

The NIC bridge II card Re-clocks the data to remove jitter, or inconsistent PWM pulse widths from the incoming data stream. This isolated the data into the NIC from the data going into your DAC.

The DAC never ever sees the Ethernet cable directly once the data is re-clocked It does not matter if it is copper or fiber, the signal is re-processed and clocked by most audio DAC electronics.

Ethernet repeaters rebuild an Ethernet signal, and that’s it. It just makes it a new mirror image of what left the transmitter way back there. We can do this thousands of times with zero errors. And, test have done just that. This is why digital is so nice.

Our PSAudio DAC takes the incoming Ethernet onto the Bridge II and re assembles the PWM data all over again, adjusting the pulses that divide the sine wave up such that they are exactly the same, or theoretically far more so. That is what jitter is, the uneven width of the pulses that define a section of a sine wave when it is chopped up as a PWM signal in our red book audio case

This digital “number” that defines each pulse, and there are a bunch of them, get formatted to Ethernet protocol to becsent along the way by a NIC card circuit. At the far end, after the data is processes through the Ethernet pkrtion of the NIC card, we take the raw digital data and rebuild the PWM stream by re-clocked pulse periods to get rid of jitter.

Noise is NOT part of the digital ones and zeros, so it is left behind. The circuit looks only for a one or a zero within the noise. As long as it can see the right data, that’s all that moves forward. In this way, we ISOLATE the signal from the noise with digital.

Analog COMBINES noise into the signal via superposition, and thus we just get more and more noise every time the signal move anywhere. Digital tosses the noise, until it can’t see a one or zero, then it simply stops working, so it is more an all of nothing type system.

Digital TV is a perfect picture or nothing. You get blank blocks all over the screen if that portion of the data stream gets too noisy or corrupted.
Digital can have errors as small as a single pixel on a TV. For streaming 4K, you need about 10,000 of them to see anything. For cables like 2183P and 2183R we limit errors to like 10 on average!

Analog is a great picture and then increasingly more full of artifacts the farther it travels and picks up noise. The picture never goes BLANK anywhere, but gets full of SNOW artifacts and multipath distortion which is uneven latency signals arriving after they should, Making a shifters ghost image.

To prove how good the Bridge II is at isolating my Ethernet network, I built the WORST one I could! Ethernet overpower lines with a standard W10 PC streaming my music from upstairs to downstairs, it sound glorious. The “voice” of my network is not in the picture…pun with the TV anology done on purpose!

I should add that digital has perfect PIXELS or nothing…not the whole picture. Depening on the encoding and compression, BLOCKS can come and go.

Galen Gareis

PS Ted Smith types can further describe how we go from encoding a sine wave to formatting to Ethernet and back to PCM or DSD. This is pretty complex, actually. I just described the BIG BLOCKS in the diagram of the circuit. My brain is just noise after that…junk to be ignored.

1 Like

Yes, to truly get above 5e, proper termination is CRITICAL. It may be working, but at 5e Bandwidth. The right connectors and termination procedures are a must. Ethernet is so good at auto-negotiation of the link that it will work, but at reduced speeds based on where the bottleneck is.


Thank you for the detailed explanation.

For an encore, should you so choose, why is USB notoriously “noisy” then as a digital transmission line? Is it because it’s not Ethernet or not fully digital some how?

I have probably conflated something here somewhere.

Thanks again,


This seems to be a rather sweeping statement, so let me be sure I am understanding the full implications of what you’re saying.

We agree that, with the DSDac, jitter is reduced to that inherent in the DAC since the DSDac re-clocks the input with no reference to external clocks.

You are also saying:

That is, the ethernet connection to the Bridge II is perfectly isolated from noise.

If that is true, and if we also confirm continuous bit perfect transmission of data (which we can on the DSDac with PS Audio supplied files), then the following must also be true:

  1. The ethernet cabling should not affect SQ.
  2. Galvanic isolation is unnecessary.
  3. The software used to create the bitstream should not have an affect on SQ (Roon vs. mConnect vs. J-River, etc.).
  4. The method of creating and transmitting the ethernet stream should not have an affect on SQ (computer vs. Nucleus+, routers, switches, and any of the power supplies along the way, etc.).
  5. Any other DAC input could possibly equal the ethernet input, but it could not surpass it.

Am I understanding you correctly?


Some links that don’t use error detection and retransmission or error correction can have issues. There are links that combine power and data, both. This can cause jitter issues and other signal distortions.

The Ted Smith’s of the world can explain how these pwered links are

Problematic. PSAudio uses a special HDMI to address the noise by taking the power out if the cable.

Remember, digital is many block diagrams end to end before we see analog, and amazingly, digital doesn’t lose bits even as they are repackaged to move about by all the protocols.


1 Like

No sweeping is being done. Ethernet digital ignores the noise constructing ones and zeros. It is what it does withroper ACR values. No matter the noise level, unless the data is buried in noise, a bit is a bit.

A PCM bit stream is a little different in that jitter CHANGES the actual PCM time period from what it should be. Ethernet doesn’t work like a serial bit stream. In PCM if the width is changed by jitter, the amplitude possibly will change, too, depending on the data is quantized, and that will change the analog out of the D to A block.

Ted Smith can bury you in how high speed serial and Ethernet differ. Ethernet is designed to move packets and frames of data from A to B and ignore noise. Not so serial bit streams, which are raw digital pulses. Sample rate windows, and bit depth assignments can get all messed up in serial streams.

Going from Ethernet go PCM changes things. We were talking about NIC card to NIC card Ethernet. PCM is another topic.

We are apparently talking past one another. I apologize for not asking questions more clearly.

Yes, PCM is another topic. I was only talking about NIC card to NIC card ethernet. Nothing I wrote directly referenced PCM. One small point (#5), was that logically, if ethernet is bit perfect, has no jitter, and has no noise (as you are claiming), it can’t be bettered on the same data by any other digital medium.

Great in theory. Not true in practice. A combination of active and blank blocks all over the screen isn’t a perfect picture, and it isn’t nothing. And not just in digital TV reception. The quality of HDMI cables has an obvious affect on picture quality, from perfect picture, to no picture at all, and everything in between.

Read what I said, the data is THERE or it isn’t with digital. And yes, this creates blanking you reference. There is no “partial” blanking on a screen. A pixel is on or off. But this is high speed serial, not Ethernet. Not the same thing at all.

Analog does not read an absolute, it reads a VARYING superimposed electromagnetic signal across the pass band. Analog will always give you “something” for every pixel, and all the time.

You’re fine. PCM is a good subject for FUTURE issues! I’m not anywhere near an expert at serial…but it does share the fact that digital needs a defined value, or you get zip. In analog you see the noise when the signal goes away, digital gives you a BLANKING type default. It never considers the noise a signal.

Bumping this thread in the hopes of getting more comments/participation…

I expected Galen’s posts to to garner more interest and discussion.


This subject is of interest to me too but I have nothing to contribute and just sit in the sidelines as I’m learning here. My NAS with all my music resides in a different room with about 100 feet of Cat5e cable in the walls with routers and switches and other patch cables in between to get to my DSD. I’ve read about galvanic isolation and noise produced by the switch mode power supplies running the switches, routers and modem. I wonder about the relative improvement vs cost in addressing these items vs a plugin music streamer with hard drive direct into the DAC.

1 Like

I’m building a new conference center for my company. The general contractor hired a data center / cabling designer with RCCD certification. He recommended 6A for our wireless access points that only support multi-gig, which tops out at 5Gb/s and we will likely only give them dual 1Gb/s connections.

Based on Galen’s notes about 6 vs. 6A I likely won’t use 6A for this use.

I use optical fiber where ever possible for networks and at home for audio

I have been trying a slew of ethernet cables. So far best is the BJC CAT6A from NAS to router and 30AWG/4P UTP CAT.5 Enhanced from Router to Ultrarendu. Getting great bass, deep soundfield, best yet imaginh with the combination. I have sonmeCAT7A coming Tuesday with 23AWG along with foil and then a mesh shield. Trying cheaper off the shelf alternatives first

May yet try the SOtm black iSo CAt y and dcbl- CAT 7 that Jay in his audio bacon articles liked. If I can find a 30 day trial.

But just playing with combinations sure can improve or degrade more so than I ever imagined. It is like throwing another pair of Iconoclast wires or IC in the mix when combinations hit stride . Just striking improvements are really possible not helping my sleep with the increased listening that beeter combinations elicit. It is like crack.

Read this performance comparison article. I may just order another CAT6A from them. It came with spec comparison. Nearly every other off the shelf cord has had sound limitations…The BJCs perform. I can attest. As far as 6A not for audio. Not my experience yet.

1 Like

Here is comparison. Shocking what folks sell.