Changing Switch-Router from Cat5 to Cat7

Thanks Simon. As I see, the WS-C2960-8TC-Lswitch is 8 x 10/100 ports and one uplink? 

As well as Cisco's PSU engineering, LAN speed - 100 or 1000baseT - represents a possible dimension of variabiliy.  I'd previously proceeded on advice that 'GiGE is good', and anticipation that the cable length sensing/power level adjustment might help.    I think the streamers are limited to 100baseT?, whereas the Unitiserve supports 1000baseT.

I will order one for comparison, the SFP uplink also presents a further option......

As I've been changing several factors, I went back to quickly check NAS vs USB replay with network on, using 192/24 WAV conversion of "Where do the children play?"  I found both similar in frame, but with the local USB displaying more delicacy in the guitar and clear definition between the strongly-struck guitar notes and the bass tracking each other.  The variation was clearly arising from loss of low-level information on network replay, rather than any other obvious distortive factor.  However the NAS was quite enjoyable in its own right, a big improvement. 

I'll do another run later, comparing network disconnected - Unitiserve library rescan time was a bit of a deterrent.

I will try chokes too, but have so far been focusing so far on noise elimination rather than suppression as my primary approach....

Yes, the uplink connector module  is absent on my switch... So I use the regular 10/100 ports. The streamers are currently limited to 10/100. 

it could also be that 100baseT is less troublesome with RFI than 1000BaseT as only two pairs of wires are used for duplex communication.

I use about 8 metres of cable between switch and NDX, with the closest end of the cable to the NDX being shielded Cat5e, and the rest of the cable to the switch is flat Cat5e so I can route under carpet, door threshold etc.

Simon

Interesting observation - I knew that but completely failed to spot the potential significance.

looking quickly at http://www.10gea.org/whitepapers/gigabit-ethernet/ it states 1000basetT employs pulse shaping to occupy the same spectral density as 100baseT, and scrambling to avoid spectral spikes from repetitive data.  This would lead one to think they might be equivalent.  Recent switches can sense cable length and reduce power under 10m too, which should be advantageous.  Hopefully higher-speed devices might pay more attention to noise immunity?

I'll order the 1000baseT version and see how it turns out.  I'll also see if I can demo a good glass optical cable between Serve and DAC, I think Jan has good experience with Wireworld and feels its equivalent to DC1. My experience with Optichord is "OK for the TV..."

1000BaseT uses four twisted pairs for communication, 100/10BaseT full duplex uses two twisted pairs for communication.

A GigE switch can only talk 100 BaseT to a Naim network player currently, so only two twisted pairs will be used. The encoding format and speed is established as soon as the link comes up. 

Simon

If I may add my own anecdotal "evidence", changing unshielded cables (flat Cat. 6 UTP) for shielded ones (flat Cat. 6 STP) brought an improvement in clarity, details, slam and microdynamics. And it was not at all subtle. I'd go as far as saying that this was, to this day, in my system, the absolutely essential step towards revealing what it was capable of (as far as network tuning is concerned, that is).

My very limited experience with ferrites has been mainly negative so far, no idea why.

I may have the same kind of electrical arrangement Aleg described, and there's nothing else except the stereo in my listening room (if I turn the lights off). Maybe the current is clean enough, maybe RFI is limited, maybe a switch with a good power supply (Cisco 3560 Series) helps, I don't know. So many parameters. I admire you guys for trying to sort all that out, while I lazily awaits your conclusions !