Exactly why do Ethernet cables sound different?

My sympathies with typos!

I once typed "Century" with the first vowel replaced with a "u" ... oh dear ... and that got me moderated!

I still could not see the error as I tend to read what I intend rather than the actual! When Adam coloured the offending u in red I could see exactly what I had done in error!

ATB from George

It just shows you're human.

To err is human.  To mess up the same way ten thousand times in one second ... takes a computer!

Could you make a test cable from any individual cable without fear of damage to the equipment. Just to be extreme, if you could use 8 individual cores of twin and earth would this work in the sense that it wouldn't damage the NIC cards.

It would be interesting to make a few up but would not want to damage anything in the process. Yes i'm aware a would not get twin and earth to fit.

The NIC cards will operate as intended on send and receive as long as the 4 paIrs (or 2 pairs for 100mbps) are balanced. So as long as your cable uses balanced pairs and is of approx the correct impedance the NICs shouldn't care. Feel free to experiment.  BTW I think it is unlikely a faulty or incorrectly designed cable will actually physically damage a NIC, but you might want to get a throughput data analyzer for your tests or some other means to see what is actually happening.

Simon

Thanks Simon

If nothing else it maybe interesting to try a really badly designed/made cable to see how it sounds. Then again maybe ill hit on something, but would think the former rather than later

I'm pretty sure the 'sound' of the cable has nothing to do with data transmission quality, unless there are network errors (which would degrade the sound), the reason is this:

The ND5 XS I have always uses the 100BaseT communication protocols.  When they establish a connection, the other NICs on the network auto-negotiate their communication protocols*.  They ask the ND5 what protocols it supports and find out that it will only talk 100BaseT, so they all agree to talk to it using this protocol.

100BaseT only requires Cat5, all the higher Cat cables do is to allow faster communication protocols to be used if the network devices support it.  Even if I use a Cat7 cable (which I do!) the NICs will still only talk 100BaseT to talk to the streamer, because that's all it uses, the cable is then irrelevant to the data transmission.

If you were to get significant numbers of network errors then the buffer would start to empty while multiple attempts are made at re-sending the packet.  If the network was so ropey, it may not catch up in time, causing drop-outs.  The network stats of the Router/Switch would record this, usually as errors and packet loss.

I can hear differences when I swap network cables, even though the router logs show zero packet errors and zero packet discards.

* On a managed network the auto-negotiate can be turned off and the protocol to be used can be specified by the administrator.  For a home network auto-negotiate is almost always enabled.

I know, this is what has me perplexed, too.

It can only be the RFI/EMF from the cable affecting areas of the streamer other than the ethernet data communications card/board.

I have Ferrite beads applied, as per the Texas Instruments Paper, Simon_in_Suffolk provided the link to, but would love to give a set of the Chords or Audioquest cables a try out. Do the manufactures provide a set of demonstration cables through the dealers for home trial?

Simon

I too use ferrites*, but I've still heard differences between cables (not just between 'audiophile' cables, but also different constructions of 'data quality' cables).

Ferrites will just increase the impedance seen by high frequency common mode signals travelling down the cable, they attenuate the RFI but don't block it completely.  They are also limited in their frequency range.

I also think there may be effects other than common mode RFI at work here and I'm currently investigating one possibility (going to talk to some other people about this tomorrow).

Trialling cables (at least here in the UK) is purely a matter of finding a friendly dealer.  I was luck in Bath (I've been offered a chance to try some Audioquest cables, but not done so yet).  I've bought a Chord C-Stream as it's relatively inexpensive (and unlike the Chord TA cables it has the normal appropriate construction) and I wanted it to test one of my hypotheses.  In my system and environment, it works well and does benefit from the presence of ferrites.  Indeed the sound from the NAS drive via the modified Chord C-Stream cable is, to my ears virtually indistinguishable to that when reading files directly from a USB memory stick.  I wouldn't be certain that there actually is any difference (without the ferrites there's definitely less resolution).

*   Originally I has the word ferrets here, but even after extensive training they still weren't delivering the data packets fast enough!

Huge - have you tried one of the Ethernet isolators. I've had one since my Linn days that sits at the end of a 30m run of CAT5 before my amp. It's just an industrial medical grade isolator, 

James

I very much agree - and the occasional dropped frame will not/ should/not be detectable if using TCP

James,

Yes I do see the potential for this to alter (and quite possibly improve) the sound of a system.

However, in theory, for my system, I shouldn't need an isolator:

I don't have significant potential differences between connected equipment.  My streamer, amp, router and NAS drive all have to be on the same ring main (not ideal for mains borne differential interference one to another, but, in practice, I can't do it any other way).

Similarly, the configuration I use doesn't have an earth loop.  The NAS drive has a shielded enclosure that connects to the screen of the SFTP cable between itself and the router.  The ND5 also has a shielded enclosure that connects to the screen of the SFTP cable between itself and the router.  The router however doesn't connect to the shields of the cables, breaking the ground loop.  Each Ethernet cable is in balanced configuration with the screen connected single-ended*.  Any low voltage common mode interference will be attenuated by the ferrites and also subject to the NICs' CMRR.

I'm not rejecting the idea out-of-hand, but I'll try some cheaper options first (or any potential solutions that I can borrow!).

*  Single-ended screening works OK in Balanced configuration, but usually doesn't work properly if used for unbalanced connections.

Same here, no errors or discards. Tried so far 3 standard cat5e/6 cables, better cat6, AC Vodka & Chord C-stream... testing continues.

HH,

Have we given a satisfactory answer to Ethan's original question?

I'm aware this has strayed a little, but there is a good amount of information in the thread as well.

I'd hate to see Halibut on the menu tonight!

Hi Huge

Good description of your network, you are taking some care. I'm convinced various noise sources inducing into streaming systems is a major detractor from performance levels. 

One suggestion - to check if your router case is grounded, and try this if not (assuming its metal).  I have my Netgear switch on a TP psu (which should be low-noise) but floats the DC rails.  A grounding wire brought a significant improvement, and I found some sensitivity as to how this was connected to mains earth. Similarly for my Unitiserve.

I haven't tackled my NAS yet but suspect its a major remaining noise source.  I note Chris Bell has both his NAS and Unitiserve on linear supplies, and Murray Harden has just written how he streams from a USB drive (w. linear supply) on his Unitiserve and disconnects the NAS (which he only uses as a master/backup store).

Totally agree

Not applicable, it's plastic.  I don't think this is a problem, see below

Likewise, however I have disconnected the NAS and the router and compared the sound when the ND5 is just reading files from directly a USB stick.  There was very little discernible difference in the sound - so little I couldn't be certain there was any difference.  To me that suggest that I'm not going to be able to get much more improvement by adjusting the network hardware.

This is the first logical explanation I have seen as to why there may be differences in sound quality between ethernet cables.

It is possible that I completely misunderstand how streaming with a Naim streamer works. This is how I think it works:

The ethernet cable is the means of transmission for the *file* which contains the music data. This is quite different from a coaxial or optical connection, which would transmit the actual PCM audio bitstream. The file (be it .wav or .flac or whatever) is sent from the storage device in a packeted form. These packets may not all arrive at the streamer in the correct order, but this is of no issue. The streamer will collect all the packets of the file and then once it has enough information to fill its buffer, it "opens" the file and from there begins to reconstruct the PCM audio bitstream.

If the above is a correct representation of the fact, then I have to admit my mind refuses to comprehend how an ethernet cable can affect the sound quality. Huge's comments above provide me with the first logical reasoning. That said...

You would think that wireless transmission would solve this and yet I have seen many comments about inferior sound when streaming by WiFi.

For now, I am satisfied with the sound of my system with basic grade ethernet cables. I find that greater improvements can be made by spending my money in other areas of the system.

Best.

EastEnder,

Your understanding of the sequence of operation is essentially correct.

WiFi introduces another problem though - you're opening up a radio receiver inside the streamer case, and hence opening a path for the entry of RFI.  I've found using wired connection (with ferrites on all potential points of RFI ingress) and removing the WiFi antenna to be beneficial.

I've also found Chord C-Stream (with ferrites) to be particularly effective, and not too expensive.  Failing that a good SFTP or SSTP Cat6a (also with ferrites) can work almost as well.

As I have said several times over recent years on this forum, Ethernet cables will create imbalances and therefore produce a resultant potential difference in the sender or receivers as well as reflected energy. This can couple and modulate the ground or radiate. The only debatable bit is to what extent this makes a sonic change in connected audio equipment - and the evidence here appears  that it does appear to be audible in many people's systems.

I have also shown a link of TI's design guidelines / whitepaper to mitigate EMI from Ethernet patch leads and their NICs - I'll show it again

http://www.ti.com/lit/an/snla107a/snla107a.pdf

I really see no magic about this - its regular RFI/EMI design mitigation in highspeed electronics and balanced transmission lines.

But mitigation is not the same as removal and so there will be varying influences of EMI from different connected cables.

Simon

Simon,

Compare that white-paper to the Chord TA cables...

... hmm!

And before anyone else gets a different impression, I believe TI know what they're talking about.

If anyone wants to properly understand how Ethernet cables can have an influence on the rest of the electronics, just read and make sure you understand the paper to which Simon referred.

Somewhat off topic, but still relevant in the wider context...

Here's another paper for those who want to gain some knowledge of how the physical layer of Ethernet works.

users.ecs.soton.ac.uk/sqc/EL336/10GBASE-T.pdf

It starts by giving a simple easily understandable overview illustrating how the 10Base-T, 100Base-TX and 1000Base-T protocols transmit the data on the wire.  If all you want is an overview then this is all you need to read.

However if you have the interest, it then goes on to more complex subjects such as the mathematics behind the data encoding for 10GBase-T and information on the implementation of the NIC's transceivers.

Huge - the link doesn't work?

This one does: 

http://users.ecs.soton.ac.uk/sqc/EL336/10GBASE-T.pdf

Of course, that makes sense and would explain why so many have been disappointed with the performance via WiFi.

Huge, can you tell me at what points you use ferrites? Do you use a specific type/brand? Is there a specific way to use them?

Best.

D'oh...   Thanks Aleg

I really should test links after posting!

Thanks for the link Huge

The guys name says it all really "Gottfried Ungerboeck". Read this and your brain certainly will have Gott Fried, in fact so much you can't pronounce his surname

Graeme