Audiophile Ethernet Switch

Status
Not open for further replies.
It may appear to fly in the face of logic, but different ethernet cables do make audible changes in SQ.

The reason is the same reason that causes different power supplies or audio distributions to change the sound of network streamers: Noise signature.

Ethernet cables do transmit data but in order to do that they also transmit EMI/RFI noise. This noise eventually finds its way into our dacs.

Different cables make for different filtering of this noise. It is that simple.

The same applies for active networking devices (switches, routers, etc). The more quiet their power supplies and circuits, the less noise they emit.
Thank you.
Now, coming back to the original question: did someone ever tweak his switch?
 
Now, coming back to the original question: did someone ever tweak his switch?

Predictably, the usual suspects have talked the usual nonsense in the usual haughty manner but seem unaware of the fairly obvious engineering reasons why you're not talking nonsense. (If some of your naysayers had bothered to visit the AQVOX web site, they might have made fewer damn-fool points. I stress the "might".)

However, I've spent a deal of time on the topic over the last year or so and can confirm that "tweaking" a switch can make for significant differences in SQ if done right. (For the record, I'm linking a music "server" to an endpoint with no Internet connection.)

Quick and easy tweaks include:

1. Your AQVOX switch is 10/100/1000 MB/sec; try running it at 100 or, even better, 10 MB/sec. (After all, it's not expensive to try . . .)

2. Running at 10/100MB/sec means that:

* You need only two pairs, not four, in your LAN cable. (Gigabit needs all four.) Removing the redundant pairs (4&5, 7&8 - keep 1&2, 3&6) makes for a surprising SQ difference. First, try a length of industrial-grade LAN cable (I use Excel CAT6 UTP as it's easy to get locally) and snip out the redundant pairs. If that works, try replacing the stock connectors with the likes of Telegartners. What matters is less whether the cable is CAT 6, CAT 8 or even CAT1024 and more how well the pairs are twisted and thus how well the all-important impedance spec is met.

Compare a length cut from an Excel cable with one from a typical piece of Amazonic crap to see what I mean - I'd wager that the Excel will typically outperform any cheapo CAT-whatever. There's a good explanation of this on Meicord's web site. See also "Is Your Cat 6 Cable a Dog?" on the Blue Jeans site. Measurements, even.

* You can get surprisingly good results from cheapo 10/100MB/sec switches. I use Zyxels but upgrade the on-board Vregs to an LT3045 device and drive them with Ciunas Audio LiFePO4 battery PSUs.

I've tried mid-priced LAN-audio-exotic cables (Meicord, Viablue); though both are better than stock, neither matches a DIY'd Excel/Telegartner at a fraction of the cost.

Is that all? No, but it's a start.

Dave
 
@Dave: Why did you not try the obvious, which is not to use a switch at all?

I'd been using the obvious for about ten years before I started experimenting.

Recently removing many of the upgrades to help track down an intermittently faulty PSU reminded me just how much difference they made.

Besides, my comments on LAN speed and cable configuration apply equally to a direct (no switch) connection. Obviously . . .

HTH

D
 
Last edited:
Status
Not open for further replies.