Playing with ethernet switches and cables

[brucew]

Dave Ryelands recommended to me the Zyxel ES-105A switch over the Netgear GS105E. Further, he recommended three of them in series for the benefits of each reclocking the signal. Further, he recommended upgrading the internal 3.3v voltage regulator with an LT3045-based regulator. Why not? I can get the switches used for £10 and see.

 

For the Meanwell switching PSU (12v) on the Netgear switch, I had already connected the DC neg output to the Earth pin per Swenson's recommendation, and found that to make a noticeable improvement in sound, in the order of detail and texture for both HF and LF. So I was surprised to get a little improvement over that using the Zyxel switch with their stock wallwart, in the order of better depth and dimensionality of instruments and space. I then put a Meanwell 5V with the same DC neg-Earth wire on the Zyxel and found further improvement (detail & texture). I added a second and third Zyxel switch and found improvement with each (dimensionality). Subtle, but noticeable, especially when you remove one.

 

I ordered some LT3045 3.3v voltage regulators from www.LDOvr.com. I got the 7805 config (TO-220) and just turned the Vreg backwards and crossed the Grd and Vout to match the SOT-223 of the LD1117 vreg I was removing.

Again further improvement in sound. I wondered if I could go back to just one switch with the cleaner Vreg, but the dimensionality was deeper with multiple switches in series.

 

One surprise

With three switches stacked vertically, it was a little unstable. I added a 1 KG doorstop on top which helped, but I was looking for something that would hold things together better. I added 2 x 2.5" strips of velcro between each switch, which held things together nicely, but within a couple hours, I suspected I'd lost some HF detail, space, and harmonics. So I ripped off the velcro, and voila, it came back.  It seems that the adhesive strips were damping the metal enclosure of the switches and thus damping the sound. Conventional wisdom says this may happen in an analog environment but should not happen in an all-digital environment.

 

I added back small tabs of velcro just at the corners and got a result that still dampened the HF a bit too much. The feet on each switch are dots of foam rubber and the velcro was thick enough to render them irrelevant as the velcro had become the sole basis of connection and isolation. So, I pulled them back off. I also noticed that the 1KG doorstop weight had a dampening effect due to its rubber base. But things were a bit too lively with just the foam rubber feet and no doorstop, not to mention that the stack was physically unstable. I pulled off the rubber base and put the doorstop back on the stack. This seemed about the right balance sonically.

 

Again, I've experienced the sonic affects of dampening an analog component, but did not expect to experience it with digital switches. But it was easily repeatable, and takes a bit of experimentation to find the right balance sonically. Then again, I also got hold of some Belden 1303E CAT6A which is double jacketed in rubber (2 layers). In between the switch and DAC that also had the effect of sounding too damp and lacking energy.

 

Ethernet Cables

I also tried several non-boutique ethernet cables wondering if I'd hear any difference. I heard a difference between almost every cable, more easily discerned than between different USB-DAC cables. I don't know why, they just did. For the main ethernet cables, the best sounding were Excel Cat6A 23 AWG solid core with foil shields (UFTP, PIMF). Best HF and LF detail, musicality, and dimensional depth and width.

 

But when I put those between the daisy chain of switches, the effect was a bit heavy handed. I tried deconstructing them and just using one pair, with heat shrink tubing over them. That sounded lively and detailed but lost nuance and subtlety. Between the switches I found that the best sound was with 0.5m CAT.7 cables (round, not flat) by 1attack.de, an amazon/ebay seller. Most basic cables sounded different to each other, different brands, different CAT types. Generally, unshielded or lower CAT types sounded worse, especially if any WiFi, TV or radio devices were in the room. Each brand sounded different, and the two I thought more reliable, Lindy and Excel factory terminated, did not sound best. Between switches the 1attack.de CAT6a or CAT.7 sounded best with dual screen and foil shielding.

 

For the main ethernet runs, the 23AWG that I terminated sounded better. Excel CAT6A foil shielding only (UFTP) sounded better than other Excel cables,  better than the Belden CAT6A FFTP 23AWG, and defintely better than Belden 1303E CAT6A SFTP with double rubber jacket. That wasn't expected because FFTP should be better shielding than UFTP, and SFTP should be better as well. But I was letting my ear be the judge and it was easily demonstrable and repeatable.

 

BTW, I'd wondered about the serious looking Telegartner plugs. Research showed that they were "field assembly" plugs needing no crimping tool and were metal for durability. I wondered if they'd sound better. I also used VCE CAT6A shielded crimping plugs SJ650, and both they and the Telegartners are the few plugs that fit the insulation of 23AWG solid core. The VCE actually sounded a little better, a little more natural. I wonder if this is because of Telegarter's use of a PCB and multiple connections between wire and plug pins. The Telegartner is certainly more user friendly and fits 23AWG, but looking at the Metz field connector, I wonder if it would sound better due to few connections. But by that time, I'd spent enough money and time to not be motivated to buy a couple Metz. So I called it a day and stuck with the VCE crimp plugs. 23AWG solid is certainly easier to get aligned properly than the 26AWG stranded of most patch cables.

 

In all cases, I found the ethernet cables to sound better and more detailed with the screen wire attached to the plug shielding at both ends. I know some here prefer unshielded or floating shields, but in my testing I consistently preferred the sound of shielded at both ends.

 

YMMV.

 

Thank you to Dave Ryelands for his recommends.

 

My config is:

 

NUC server feeding switch and bridged wireless router on 1st floor. (brick interior walls so no cabling between floors)

Bridged wireless router on Ground floor 12 feet away from HiFi (sounded better than next to HiFi).

6m of Excel CAT6A 23AWG from router to Zyxel switch.

Three Zyxel ES-105A switches in series. With LT3045 3.3v Vregs. Fed by one Meanwell 5V with DCneg-Earth wire. Switches connected with 1attack.de CAT.7 0.5m cables.

1m Excel CAT6A 23AWG from last switch to microRendu.

microRendu fed by Uptone Audio Ultracap LPS-1.2 and Meanwell with DCneg-Earth wire.

Calyx 24/96 DAC with PS Cap upgrade, fed by iFi iPower.

[elan120]

Very good looking mod.  I did similar mods short while ago with ZyXEL GS-108B V3 with good result also, and here are the links to two of the posts.  I have not try using multiples in series, but certainly be interesting to find out how that works in my system.

 

 

[brucew]

Nice work with your own board and SMD resistors. That much micro work with SMD is too fiddly for my eyes and hands! But the idea of adding the clock is attractive. I assumed the switches in series was meant to be another way to approach better clocking.

 

14 hours ago, elan120 said:

Very good looking mod.  I did similar mods short while ago with ZyXEL GS-108B V3 with good result also, and here are the links to two of the posts.  I have not try using multiples in series, but certainly be interesting to find out how that works in my system.

 

[Puma Cat]

I find it interesting just how noisy all these Ethernet switches are. Anyone know of a good "consumer"-grade one that is reasonably quiet? My TP-link pretty much drives me out of the room; its pretty ghastly. I can always add a Jameco Reliapro LPS to improve the el cheap wall adapter that comes with them.