DIY Ethernet Cable

[TomJ]

Hi Dan,

 

this is really interesting.

Are there any other information available (material, gauge, price etc?).

 

I have tried also an approach with single twisted pairs and separated shielding, which was not successful regarding sound quality compared with the unshielded version. 

 

 

P.S.: Just found this info: https://artsexcellence.com/crystalconnect-monet-network.html 1.899$ for 1m!

 

[TomJ]

Hi Don,

 

the Switch is an IP67 industrial switch which ist completely encapsulated with resin and only on 100baseT level.

 

I have also tried the JSSG shielding, but also here, the unshielded version has more details.

 

 

Regarding your cable: Do you use both wire (red + black) for JSSG?

I also read about cables that use 2 GND systems for shielding to improve EMI resistance:

The outer shield is only connected to earth (PE) on the switch side and the inner shield of the twisted pairs is connected to the GND system of the end device.

Ever heard of this?

[TomJ]

Hi Dan,

 

Have you ever tried to remove the shielding from the viablue cat cable?

 

[TomJ]

On 6/4/2021 at 3:52 PM, LondonDan said:

No I haven't tried this and I don't have any cable left or would be trying it now for sure, the cable is quite nice quality, it slightly better than SUPRA in that it brings a bit more detail and a bit more air to everything.

 

I have in the "to do" list a DIY test on ethernet cables which is why this topic you started interested me and I would like to join the conversation. I would like to try making up a Neotech silver and gold 24awg terminated into Telegartner for the last section of the network from switch to Dac / Amp. I had put it off due to the expense of 8M of cable but if its only 4 cables for 2 twisted pairs its much more palatable a cost (still a lot ££ but I already have the Telegartners already).

https://www.hificollective.co.uk/wires/neotech-AG-GD-silver-gold-wire-teflon.html

 

I am settling / burning in a whole network change for the time being so it will need to wait as I want to explore that big change with everything I have at the moment. But when I do get round to it in a month or so I will definitely try without shielding first, then with 3M copper tape covered by Nylon (the tape holds the twists in place so they don't move (giving a little effect of the Beldon bonded I guess but not as secure as the beldon), then with copper braid over the top and a nylon outer to finish, something like the Crystal Monet.

 

I see sense in the first 2 out of 3 network cables in my chain from ISP modem-Router-Switch-Dac could be shielded because of how they pass power cabling, PSU's and other electrical equipment but the 3rd and final ethernet to the Dac has a bit more space around it and here maybe its worth experimenting to see if unshielded or low shielded works or doesnt.

 

One of my purchased ethernets (not DIY by me) is an odd one - its silver 95% copper 5% (basically a sterling silver which is not regarded as an audiophile material but I was open to experiment with it).  The twisted pairs are wrapped in copper tape and then a carbon "infused" braid over the top with cat8 Telegartner Rj45s. The constructions seems to be lighter on shielding (the carbon is probably only an 80% coverage) for the very reason you discuss in that it may be beneficial not to over shield an ethernet as long as the cable is not passing through a "noisy" area. I don't have a pure silver to compare it too though. As "the last cable" I like it, do I love it? I don't know yet, but if you daisy chain 2 of these sterling cables they become too bright which I expect is down to the sterling silver so:

 

Router (silver ethernet) Switch (silver ethernet) Dac / Amp = too bright but lots of detail

Router (Viablue ethernet) Switch (silver ethernet) Dac / Amp = good tonal balance and good detail... well for the time being as I would like to experiment more here once the network is broken in, I expect I have some gains to be made in getting the cables right. I would like to match all cables by manufacturing as much as possible in the network. I am thinking Neotech OOC copper for the first 2 and the final Neotech ethernet a DIY OOC silver/Gold and experiment with shielding here on the DIY one.

 

http://wp.neotechcable.com/ethernet-cable_up-occ-copper/

 

I realise some of this is a little off topic with ethernet cable matching rather than shielding but it all comes together. Its unlikely we just have one single ethernet cable and so with small, medium or large networks this type of 2 twisted pair unshielded or shielded cable might only work in one area of the network or it may not be good at all for some.

 

Hi Dan,

 

to try the stripped cable approach you can just make a test with a 5 € CAT5 cable by stripping the outer isolation and cut the unused twisted pairs.

I practice this only in my music room, as I have horizontal CAT6a in the wall for structured cabling in the house.

 

Have you ever used an ethernet isolator?

[TomJ]

3 hours ago, LondonDan said:

No I’ve not used one (yet). for isolation I turned my ISP hub into modem mode going to edge router where WiFi access point also connected then from there into Ediscreation Extreme network switch. This is the link I use the Viablue JSSG Ethernet. Then from there into dac /amp. The ediscreation has a grounding on/off switch per socket on the switch but even if I un ground the viablue cable I don’t notice any difference. It seems the already ungrounded diy Ethernet is already doing a good job ( or the grounding on / off is marginal impact for me and I can’t notice ). Pic below - I can’t say for sure how the on off grounding works and it’s still new so haven’t experimented much yet.

 

Tom I was interested to try a short fiber section as an isolation too. Your inline Ethernet isolator  looks like a cost effective trial though
 

As for the stripped back Ethernet I will try it, I am for away couple weeks now though. 
 

did you order your 100ohm cables yet or try another diy?

Interesting Switch - never heard of this.

 

The Gore 100 Ohm cable should arrive this weak - will report.

Regarding the isolator, you should, if you use this, try this little mod: https://www.aktives-hoeren.de/viewtopic.php?f=6&t=12225&p=205858&hilit=delock#p205858

Sorry, its German - please use google translate ;-)

You should ground the isolator with the grounding of the switch.

 

As more I read and experiment about the topic of improvement of ethernet streaming, i come more to the conclusion that it is not the more precision time frames of the clocks, that improves the sound, but more the accuracy of the signal regarding voltage stability. 

If this is the case, filters can do a lot here. 

 

Here is a test of an filter / cable combination, which seems to outperform the EtherRegen:

 

 

 

 

[TomJ]

Hi Dan,

 

the JCat is (in my opinion) a relabeled EMO EN-1005+: https://emosystems.de/en/shop/emosafe-network-isolators/self-enclosed/en-1005/

This is a medical grade isolator.

 

Yes, the shorter the cable is, the stronger the improvement with the Delock. You can also place several Delocks in your chain: one Delock in front of each node. I use a short CAT cable stub - you can see in my mod.

The problem with fiber is, that you need mostly a converter back to ethernet and here we are again. It can help to cancel noise of your home network, but the converter will once again produce noise. So an isolator makes also sense in a fiber setup.

I was in contact with a DAC developer and he mentioned, that fiber converters produce much noise on the ethernet - it was only a statement without details.

 

Regarding filtering: The transformer chips used by the Delock has per line a common mode chokes (in the medical grade isolators there are no common mode chokes) and a transformer. The common mode choke is a filter regarding time symmetry of TX/RX+ and TX/RX-. With the earthing of my mod, differences in the signal (Ethernet 100BaseT is differential) will be lowered and the voltage symmetry will be improved. I think it will not be eliminated due to the tolerances of the transformers.

I think that these differences are the main reason regarding sound ,because these will travel inside the endpoint and influence the following components.

 

The Delock is by fare not perfect - there and in most other isolators are resistors for termination missing, which could negativ influence the signal. There is much potential for improvements. But regarding the step you can do with this device regarding the effort, it seems to me most interesting thing at the moment.

 

By the way - I am developing currently a filter which is a few steps above the Delock.

 

Kind regards,

 

Eric

 

 

 

 

 

 

 

[TomJ]

Hi Dan,

 

really nice description of your journey!

At the moment I am a bit busy - therefor I will keep it short - but more to come!

 

Regarding the single 100 Ohm Gore cables I have to say, that these play pretty well. 

 

Regarding impedance, I think, that there is no shift because of the shielding and they are not close to each other.

 

But regarding resolution and details the stripped 2x2 23AWG cable is still better.

Here a picture (but with 2x RJ45 instead of M12 connector). The twisted pairs are from a CAT6 cable.

 

But why?

• Is it the missing outer isolation with its dielectric?
• Less crosstalk? 
• Reduction of capacity of the removed pairs?

As I learned in the PS Audio thread from grandmaster of the Belden CAT cable Galen Gareis, the structure of the cables also has an effect on the impedance. As Dan has already mentioned, a twisted pair that is set up as 4 pairs still has 100 ohms, when it is free in the air it has almost 125 ohms.
So by leaving out the other wires, I increased the impedance, which of course affects the signal quality. Is the impedance a main reason for the differences in sound between the cables?

I also learned in the PS Audio thread that building a CAT cable is not that easy in terms of impedance. The individual twisted pairs influence each other, but the shielding also has a considerable share.

Many of the audiophile cables on the market are not sold with test protocols. What if they are all imperfect in terms of impedance and the differences in sound result from this?
The structure of the shielding is also not very easy with regard to internal reflections and can have an influence if used naively.

 

When thinking about how the sound changes come about, I first thought of asymmetries in the voltage on the data line. Ethernet is a differential signal, so the difference between the two data lines of a twisted pair should always be zero. However, due to EMI effects and inaccuracies, this is usually not the case. It was suspected that these differences would become "noise" in the end device.
I built several filters out of Ethernet transformers, studied their circuit and built mean derivatives of the voltage differences to GND with the success that the sound became more and more spatial.

What I then learned, however, is that the differential Ethernet signal in the end device is processed into a signal by a differential amplifier, so that the theory of voltage differences had to be discarded.

 

The sound difference may be due to the fact that the high-frequency Ethernet signal, which has approx.31.25 MHz at 100baseT, is influencing the audio signal under certain conditions and then, although the frequency is itself inaudible to the ear, at a overlay has an effect. I had an experience with a power bank that had an effect similar to that of my 2x2 cables. More space and more depth. In terms of current noise, the power bank is pretty much the worst thing, because it uses a very cheap switching regulator that also generates high-frequency spikes. I havent seen any audio signal measurement with resolution up to 30 Mhz yet. 

 

What if the actually perceived improvements are actually caused by a deterioration and peaks in the signal?

 

Incorrect impedance with Ethernet cables results in a bad signal. Perhaps the wrong impedance of cables, as well as reflections caused by the insertion of an isolator / filter or wrong connectors, distort the signal in such a way that more spikes/peaks are generated and a sound improvement is perceived.

 

Since I want to get to the bottom of the matter and find it very difficult with snake oil theories, I have currently bought a high-resolution oscilloscope with 1Ghz resolution and will now start to make measurements - results will follow.

 

But as I said, I'm a bit busy right now, so this won't be today or tomorrow.

 

Regards,

 

Tom Eric