Ethernet cable comparisons - a few thoughts

1  Yes

2  Example of 1

3  Timing is determined by the Media Server &  network switch(s) not by the cable

4  see comment about RFI

5  see comment about RFI

6  For cables with asymmetrically connected screens, see comment about RFI

7  For the older Naim streamers the buffer is way too small (the new Unitis may be able to buffer an entire track.

8  Yes I have; I tweaked my network reducing the noise coupling until I could hear no difference.

9  I'm not inside your head, so I can't assess this! 

RFI.

RFI is an electrical noise like signal that's all around us in the environment and also conducted in along the mains.  This electrically couples into anything conductive, and that includes Ethernet cables.  Once inside the cable it gets electrically coupled to the sensitive analogue amplifiers in the streamer.  This degrades sound quality.  Changing the Ethernet cable will change the coupling (degree and frequency selectivity) of this noise, hence subtly changing the sound of the system.

It's interesting all this stuff about ethernet cables. I'm still using wireless from my router to an Atom as it's a bit of a distance away, I did try a bog standard long CAT5e cable as a test but couldent detect any audible difference. I'm going to relocate the router at some point so as its closer to the atom and will experiment with cables again then. I do wonder though about other weak links in the chain and how the final cable can make such a difference -  my broadband arrives in my house via the phone cable, adsl I presume, there's a fibre cabinet on the road about 1.5 miles from my house so it's copper from there up, it then arrives to a connector box on the wall outside which also feeds a few other houses and a feed comes in to my living room via a very standard junction box and then about 3 meters of telephone cable to a connector box which my Huawei router is connected to, surely all these standard krone type connections in the chain are weaker links the final ethernet cable from my router or is it only from the router onwards that's of issue?

Mercky posted:

It's interesting all this stuff about ethernet cables. I'm still using wireless from my router to an Atom as it's a bit of a distance away, I did try a bog standard long CAT5e cable as a test but couldent detect any audible difference. I'm going to relocate the router at some point so as its closer to the atom and will experiment with cables again then. I do wonder though about other weak links in the chain and how the final cable can make such a difference -  my broadband arrives in my house via the phone cable, adsl I presume, there's a fibre cabinet on the road about 1.5 miles from my house so it's copper from there up, it then arrives to a connector box on the wall outside which also feeds a few other houses and a feed comes in to my living room via a very standard junction box and then about 3 meters of telephone cable to a connector box which my Huawei router is connected to, surely all these standard krone type connections in the chain are weaker links the final ethernet cable from my router or is it only from the router onwards that's of issue?

you should use a dedicated switch to your athom: router-lan-switch-lan-athom.  A better quality lan from switch to athom is advised too ( like chord c stream). The router is noisy and it is better to separate tv/phone from streaming audio.

The new Unitis do have buffers big enough for a whole track. I remember reading on the forum that it was a favourite Naim developer party trick to unplug the Ethernet while a track was streaming and the track would keep playing. This buffer is part of their strategy to handle the likes of Tidal over real life broadband. I.e. Get the track over to the Uniti asap and play it from the buffer.

best

David

Huge posted:

1  Yes

2  Example of 1

3  Timing is determined by the Media Server &  network switch(s) not by the cable

4  see comment about RFI

5  see comment about RFI

6  For cables with asymmetrically connected screens, see comment about RFI

7  For the older Naim streamers the buffer is way too small (the new Unitis may be able to buffer an entire track.

8  Yes I have; I tweaked my network reducing the noise coupling until I could hear no difference.

9  I'm not inside your head, so I can't assess this! 

 

RFI.

RFI is an electrical noise like signal that's all around us in the environment and also conducted in along the mains.  This electrically couples into anything conductive, and that includes Ethernet cables.  Once inside the cable it gets electrically coupled to the sensitive analogue amplifiers in the streamer.  This degrades sound quality.  Changing the Ethernet cable will change the coupling (degree and frequency selectivity) of this noise, hence subtly changing the sound of the system.

Huge, the RFI explanation seems very plausible and would also explain why the data transfer itself even if subject to error correction for packet loss etc seems irrelevant, and would also explain the observation that if a file has already fully buffered on a newer system how could the cable affect the sound?

From comment so far and in other threads I've viewed it seems people have felt locally connected media provided better audio quality until various cable or other tweaks for ethernet were introduced playing back identical files from a NAS.

As a test of the noise issue Ill be intrigued to see if a locally attached device playing back a file sounds worse if an ethernet cable is attached, though on the new series unless I can control via wifi this may bugger up actually playing anything! 

In 24 hours or so of testing several cables I am astounded in the differences between them and it really makes me wonder if noise from outwith our sources and pre/power amps is fundamentally an issue with audio quality which even lower end components benefit from.

In my older Naim system I was always astounded at what Hi-Caps brought to the table for CD sources and amplification, and as an aside I can remember a Lingo PSU for my LP12 in the 90's flummoxing my student pals by the astonishing difference over the standard PSU and motor though - even if this never affected my Linn amps at the time, it showed the value of a good power supply.

David Hendon posted:

The new Unitis do have buffers big enough for a whole track. I remember reading on the forum that it was a favourite Naim developer party trick to unplug the Ethernet while a track was streaming and the track would keep playing. This buffer is part of their strategy to handle the likes of Tidal over real life broadband. I.e. Get the track over to the Uniti asap and play it from the buffer.

best

David

Thanks for confirming David - as noted if the whole file buffers relatively quickly ethernet effects cannot be purely due to data transfer integrity as the file will likely buffer in a far shorter time then its playback.

I am not sure naim are in control of how much buffer tidal provides? 

garyi posted:

I am not sure naim are in control of how much buffer tidal provides? 

No of course not, but Tidal will stream the file much faster than it can be played, so if the new Uniti takes the file as fast as it can get it and stores it locally (which is what it does), then this reduces the risk of "buffering" or other stop/starts.

it's an obvious strategy to improve streaming reliability, made possible because memory is relatively cheap now.

best

David

Perhaps David but this is not something tidal we’re able to implement with in their own app when I used it, perhaps they have improved things.

It's not in an app, but in the Uniti hardware. That's why Naim can do something that Tidal can't do in an app. 

I don't have a new Uniti streamer, but anyone else reading this who has a new Uniti and Tidal can give it a try. Get a track started on Tidal, let it get about a quarter way through and then pull the network connection. How long does the track go on playing for?

Best

David

It goes on for 2 minutes and 20 seconds

Last time I tried, it was something like 2 to 3 minutes for a 16/44 stream, and 15 seconds for hi-res - a lot more than the old streamers. 

I personally can't see this is an issue, so long as whatever file format/ sample rate you are wanting to stream the buffer can exceed the running time and your not experiencing drop outs it is of no benefit! 

Regarding the introduction of ground noise from pins 4&5 or shielding on the ethernet cable, getting into the analogue domain of the NAIM equipment, perhaps this can be tested and verified.  

I'm thinking, do not play a track, so there is no bit traffic on the ethernet cable.  Turn off your media server and switch, but leave the cables plugged in.  There should then only be ground noise.  Now, on the receiving end, the D/A conversion is not doing anything, so no jitter or data or anything coming out of the D/A conversion domain.  

That should 'leave' the ground noise only, propagating through analogue domain.  The Golden Ears that are hearing a difference should then be able to hear it.  You can do a cable swap in this mode and do an a-b comparison.

The only caveat here is, does NAIM shut off the analogue when the D/A has no traffic...  I wouldn't think they would bother.

I'm interested in the result, and your thoughts.

The other possibility is that there is noise escaping the ethernet cable, which is inducing noise in a physically adjacent analogue interconnect from pre to amp.  I'm trying here.. I am adamantly skeptical of any data integrity or clocking effects on the digital domain, but I am open to analogue effects for sure..

I started The Great Ethernet showdown thread.  Dealer claimed cable direction was crucial.  Have not personally compared.

SB955i posted:

I am adamantly skeptical of any data integrity or clocking effects on the digital domain, but I am open to analogue effects for sure..

I don’t think anyone has seriously proposed that (lack of) data integrity is an issue in any difference in audio observed between functioning Ethernet cables connected to a (Naim) streamer. I’ve always been puzzled by this insistence, but it’s a perspective thing I guess: data folks think about the digital, and often stop there. The ones-and-zeros are indeed fine, there is no other problem.

Music folks (in this case, but indeed anyone interested in an analog signal taken out of the digital domain) may not always understand the origin of the difference, but can detect it... 

How can this be reconciled? Interference in one area that manifests as noise or an alteration of the signal in another; this is the principle line of thought, and it’s a good one. 

For the computer and data folks, as a thought experiment rather than as an explanation for the interaction that may be at play here, I invite you to consider some of the most recent mechanisms for hacking and breaching security: many involve detecting electromagnetic changes due to different internal states of the system. The hacker extracts the hidden data state of the processor from its external effect on something else; akin to a more sophisticated version of old-fashioned crt freaking. Now ask yourself why it is possible to make very sensitive analogue measurements that expose a hidden signal that depends on the nature of the quickly-evolving data structure encoded in the memory, processor or transmission lines when you go looking for it on (nefarious) purpose, but impossible for this same effect to manifest itself as a (nefarious?) change in the sound of the music, for example,  when you’re not really interested in specifically detecting or decoding it. 

I have no idea how different cables sound, nor how “clear” an audiophile system would need to perform such that these potential but extremely subtle physical interactions could be revealed. I have no faith that the cost of the highest priced cables can be “justified” on any basis other than “what the market can bear”, nor that every product performs as it claims. 

That said, I don’t think it’s fair to use purely digital domain quality arguments (“the file was transferred as a bit-perfect copy”) when discussing consequences in the analogue domain (“it sounds different”). People who make very high precision / resolution measurement run into this all the time: in our lab, we can see a noise component imprinted on an electrical measurement that is related to weather and wave action many thousands of kilometres away...  but since we are not building a detector for this effect, it’s just an unwanted noise component to us. Something similar, at least in principle, is probably going on here. 

Regards alan

Hi Alan33, thank you for shedding light, I now have a few more things to consider.  I wasn't aware of this hacking technique so I have some more learning to do. Very interesting.  

I am however interested in this ground noise experiment to see if that is the noise injection route.  I tried it on my equipment but could not discern a difference, at full volume.  My hearing is not perfect though, as I have a slight tinnitus ( or I'm 'hearing' my own wifi! ).

And would a G figure prominently in your research?

Bryce Curdy posted:

I started The Great Ethernet showdown thread.  Dealer claimed cable direction was crucial.  Have not personally compared.

Hi Bryce

It's you thread that got me interested in this, as I was firmly of the view that surely it's just data and I'd assumed that  there'd be adequate noise filtering on the ethernet inputs of these streamers but it would seem not.

I initially hooked up a Chord C-Stream cable and thought this cable produced a stunning improvement relative to the modest cost even though I found the white/cyan colour a bit garish.  I either forgot or mixed up the directionality and later set it up in the 'correct orientation' but confess I didn't actually listen to the difference straight away - even the wrong direction made a huge difference to my Nova.

For the cost I'd buy one just to have a better built cable to be honest, it's with the more expensive things I suspect it gets interesting.

There must be something to the directionality but I'd not be surprised if it's all dependent on the given environment and devices which will differ.  Another cable I tried was better in some ways but also had some negative effects - I suspect in another system it might have been the other way around and these cables may offer a way of tweaking certain characteristics of the audio that are fixed for a given Naim pre/power and speaker set up.

Apologies f I have detracted from your thread, I just had several things running through my mind that I wanted to specifically query and get opinions on.

With the new Uniti series there is very little you can do to upgrade SQ by design apart from adding in an external amp which I did as I had an old 250 lying around gathering dust.  No option for an external power supply for example (shame as I have several Hi-Caps).

With this in mind, I'm utterly bowled over by the sonic improvement a cheap ethernet cable can bring - the sound can be transformed in my set up for 1% of the cost of the Nova.

Yes in UK we get a Powerline Lite with the new Uniti series - I wonder how much a proper Powerline would bring to the table - I guess I'm saying that while we all might want to do some 'wow' upgrades there may be some that could provide significant improvements to an all-in-one device which could be carried forward to a more sophisticated set up in the future at a relatively cheaper cost than making the jump directly.

SB955i posted:

Hi Alan33, thank you for shedding light, I now have a few more things to consider.  I wasn't aware of this hacking technique so I have some more learning to do. Very interesting.  

I am however interested in this ground noise experiment to see if that is the noise injection route.  I tried it on my equipment but could not discern a difference, at full volume.  My hearing is not perfect though, as I have a slight tinnitus ( or I'm 'hearing' my own wifi! ).

And would a G figure prominently in your research?

 

if you want to isolate your network from noise, i recommend you the network bridge ( 2 little fmc converters with optical fiber).  You have the topic from december on that :  ethernet tweaks.  For around 100GBP.  A wonderful jump in sound quality !

French Rooster posted:

SB955i posted:

Hi Alan33, thank you for shedding light, I now have a few more things to consider.  I wasn't aware of this hacking technique so I have some more learning to do. Very interesting.  

I am however interested in this ground noise experiment to see if that is the noise injection route.  I tried it on my equipment but could not discern a difference, at full volume.  My hearing is not perfect though, as I have a slight tinnitus ( or I'm 'hearing' my own wifi! ).

And would a G figure prominently in your research?

 

if you want to isolate your network from noise, i recommend you the network bridge ( 2 little fmc converters with optical fiber).  You have the topic from december on that :  ethernet tweaks.  For around 100GBP.  A wonderful jump in sound quality !

Must re-visit that thread - does a fancy ethernet cable make any difference after the 2nd fmc converter to the streamer too?

Found this thread amusing in hindsight:

https://www.whathifi.com/forum.../posh-ethernet-cable

So my next queries, as I've not tested yet, are:

1 - If there is so much noise being introduced over ethernet why are we all not just streaming from USB attached devices? * Or do they introduce their own noise, especially mechanical drives vs solid state storage, in which case we might get even better performance over ethernet as we have some control via the ethernet cable over the noise?  I assume there are bespoke USB cables too which purport to offer benefits.

2 - I assume that the source of the ethernet data stream is not important and we may see equally good improvements with a given ethernet cable for Qobuz/Tidal/Spotify WAN streaming as we would for a local NAS stream?

(* Naturally if we have several playback devices on the LAN a NAS makes more sense than multiple USB attached devices, but the latter would tick the backup considerations too to some extent)

SB955i posted:

And would a G figure prominently in your research?

Thanks SB955i -

Not a G, not for us (yet)... the experiment I was referring to is about determining h. 

Glad you’re thinking about things so deeply, it’s cool stuff. I like breaking out of the digital comms thought process to consider everything from its (underlying) analog side; that’s where the action lives here.

Ground plane coupling, or high(er) RF leakage from poorly made twisted pairs, could be implicated, I agree... but there are likely way too many influences to isolate. That’s why (in the other thread started by Bryce) I proposed that someone with an interest in this and who hears differences with different cables might like to take the cable out completely and compare against WiFi , especially in the new Uniti boxes which apparently have separate internal optimization for Ethernet vs WiFi network connections. This would be especially interesting for folks who have found that there are big gains to be had from fibre converters in the “last leg”. 

Regards alan

ps - not sure if it was autocorrect or me, but “phreaking” has a “ph” not an “f”; sorry about that!

French Rooster posted:

SB955i posted:

Hi Alan33, thank you for shedding light, I now have a few more things to consider.  I wasn't aware of this hacking technique so I have some more learning to do. Very interesting.  

I am however interested in this ground noise experiment to see if that is the noise injection route.  I tried it on my equipment but could not discern a difference, at full volume.  My hearing is not perfect though, as I have a slight tinnitus ( or I'm 'hearing' my own wifi! ).

And would a G figure prominently in your research?

 

if you want to isolate your network from noise, i recommend you the network bridge ( 2 little fmc converters with optical fiber).  You have the topic from december on that :  ethernet tweaks.  For around 100GBP.  A wonderful jump in sound quality !

The same can be achieved with a simple unscreened Cat5e coupler (Ethernet is balanced, so that DM noise is normally mostly symmetrical and the CMRR all but eliminate CM noise); when combined with ferrites, in my experience, the noise is reduced to near irrelevant levels.  In fact with the right combination of ferrites, I've been able to remove the unscreened coupler as well, with no degradation of performance.

If you're worried about large differences in the potential of the protective ground connection(s), medical installation grade Ethernet isolators probably have less impact on the signal data timing (both the frame timing and individual switching rise-times) that using fibre converters at each end.