Ethernet connection for best sound quality

Any CAT-5 is fine. All this "audio grade" is pure snake oil. I say that because being a computer systems architect and system admin, ( retired) I know how Ethernet works.  You see, there are these things called "packets" that are buffered and assembled.  Now it is possible to get noise and ground loops over Ethernet, but that is solved by proper practice, not magic. Use shielded if for some reason you have very long runs in horrible environments. Doubtful.  In the data center world, we use fiber to totally eliminate noise and ground loop issues.   Now, as far as errors, TCP/IP will correct any error. Absolutely 100%. If running UDP, then you could get a bit error in the transport layer, but an application layer if well designed will detect that.  You would be surprised how much traffic goes around the world, sat hops and who knows how many switches in UDP with no errors. 

I disagree with "streamer vs PC" as a blanket statement.  Sure going through the Windows audio stack stinks, but ASIO or WASAPI-exclusive, asynchronous with the server configured at - 3dB so the oversampling and filters do not cause digital clipping and a sufficient buffer,  then bits are bits until they reach the DAC.  One can even set the process priorities so when Microsoft decided upgrades are the most important thing in the world, or McAfee takes over the Kernel, then you can get prevent a skip.   To say all streamers are better than all PCs is nonsense. There are some crap streamers out there.  There are some very very good ones if you can stand the tiny UI. 

Yes, USB was designed for keyboards and mice, but they have beat it to death over 20 years and it is now very good.  Don't hold biases that are obsolete.  Computers putting out PCM put a lot more of the issue to manage jitter on the DAC and there is no buffer. You are basically real-time from a source that is not a real-time operating system or IO.  USB and IP solve that. 

Actually, "digital" is a mathematical concept. There is no such thing as digital in the physical world. 

But in the  transport area, bits are bits.  What you do with those bits is where the differences in sound come from. I might suggest a review of the IP stack and some basic systems understanding would enlighten some readers here. I only have over 40 years CS experience so what do I know?  :)

Above CAT-5 is ridiculous for audio.  You are not building a 10G system. CAT-3 is fine but having the pairs with consistent twists and bonded as in CAT-5 is good practice. Most consumer grade Ethernet hardware does not even have provision for the shield ground.  Look at the socket. If it is not plated, then shielding is useless. 

To be clear,  I am a hard core "listening is what counts" audiophile and have been for longer than most here have been alive. But I am also an engineer and understand how some of this technology works.  Physics are the same for everyone regardless of belief.  

What this combination tells me is that as good as our classical measurements are, they are either incomplete or we are not interpreting them in a way to enlighten us. There is no measurement of "musicality".  A scientist looks for the differences and tests to identify them. A believer just puts their head in the sand and ignores any facts or viewpoints that do not align with theirs. 

A curious feature about humans: If we "believe" our brain may well skew what we hear to align with our belief.  If that makes your music sound better, then great as it is the music that matters.   

Tony,  Believe in your CD eraser. No argument other than there is no law of physics that supports it as last I checked, aluminum and styrene were non-ferris.  Is it possible that the extra handling is draining a static charge that could bias the photodiode?  Doubtful, but within the laws of physics in this universe. An old "Zero-Stat" may do better. They were critical back in the low humidity of Colorado for LPs.  I will keep using my old RS bulk eraser left over from my tape days to de-gauss chisels and screwdrivers. 

Yea, can't go wrong with a Cisco.  On the cheap end Netgear is usually reliable. For home use, a dumb hub should suffice. Maybe if you have a couple of gamers in  the house you may want a managed switch. 

OK, streamers.   Not everyone wants a big PC in their living room so a "stereo" looking box and combining with preamp functions is useful for the use case. Most can be used with an external DAC. 

Just remember, while in the digital domain, the bits you start with are what was on the source file.  There is no way to add any information, only mess it up by poor timing, bad DSP algorithms, etc.  A good example is understanding digital filtering overshoot and how that can cause "digital clipping" with all these loudness wars CDs and how the oversampling interprets peaks.  Digital filters overshoot just like analog! So a 0 dB peak run through the filter could cause clipping and the harsh distortion that causes.   So, now to the argument for and against NOS and analog filters or adding a tube buffer stage to add various combinations of masking distortion.  :)

PS: Under the hood, A Ferrari ( I have friends that had them) is nothing but an unreliable Fiat with an attitude. Terrible reliability, not much fun to drive around town, and a Tesla will out-run it. Inexcusable that for the price, they are not more reliable than a Corolla. I'll stick with my Stag.  I can weld tubes better than a 308!

I DID NOT CLAIM SUCH!

Yea, static on a CD is grasping at straws, but it does at lease exist in the realm of possible in this universe. I did not say it was the answer, I only said some possibility may exist. I don't personally believe any effect, but have not tested it. All my CD's are in boxes in the closet. I RIP with bit-to-bit verification so it is a non-issue. Old CD players worked in real time so a lot of little issues could pop up. Not today. 

If you "believe" than as I said, be happy.  There is a big difference between "unknown" and "impossible"  I do happen to know how IP works. I made a pretty good living at it. 

I suggest you examine the IP stack from layer 1 to when you get the application layer providing a steady clocked PCM to the DAC. Maybe some crappy streamer did not do a decent job on their NIC and some noise is getting to the system. Just like early USB that were not addressing the reality of the input.  Band aids can mask bad design. I prefer to fix the problem than cover it up. 

Yes, I do expect almost all parameters of audio, analog and digital, to improve. I expect measurements will improve.   

Correct pretty much: 

Incoming, I assume fiber or coax to your service provider MODEM. Please use the provided RG-6 COAX as it will be double shielded and meet FCC emissions requirements. At this level, bits are bits. 

Modem to a router with Ethernet cable:

Most use the provider router. Provider Routers are fine for wired, but WIFI performance is often very poor as far as bandwidth and range.  Not a problem for most uses but if you have range issues or hard core gamers, you may want to look into higher performance routers.  Irrelevant for audio quality as it is still a packet based data delivery service and bits are still just bits at this level. The data rate is trivial.    You can cable Ethernet or consider WIFI to the computer or streamer. I prefer to cable anything that does not move, but WIFI  can carry HD stereo bandwidth just fine. Still packet based delivery. Bits is bits at this point. TCP will correct any errors.  If running UDP, shame on you but the higher levels of the stack may still correct any error. 

Again, a shielded Ethernet cable does NOTHING unless the ports on both ends are plated to provide the shield ground.  An Ethernet cable  does not have a pin for shield.  Why? Because it is not necessary.  There are some applications in the Medical field where emissions could cause errors in other equipment so shielded cables are required. These locations usually specify isolated power grounds. It is not a good thing for your IV pump to glitch.  You can buy consumer level routers, switched, and hubs with plated connectors if you believe you need it. 

CAT5+ had the sole advantage of bonding the pairs for more consistent impedance when going to Gig-E.   CAT-3 is all you need for audio bandwidth. 

If anyone would like a better factual understanding, might I suggest, rather than believing some You-Tuber or subjective reviewer sponsored by snake oil manufactures, ( or me) you review a little more reliable information.  Here is a good place to start on WIKI.

https://en.wikipedia.org/wiki/Ethernet

If you want to dig deeper, MIT has their entire engineering course selection on-line for free. 

A streamer is nothing but a stripped down computer loosing all the services not needed. They attempt to make a computer an appliance as they should be. I use a PC as a music server and for the limited streaming I do as I like the bigger screen for my UI and don't own a tablet.  It has about half as many services running as an out-of-the -box PC and many configuration changes.  Audio quality was in fact improved by PC and Music server application configuration for how the processing of those bits was done to assemble the USB output. Stream or disk file.    Bits are still bits, but different bits.  Subjective reviewers suggest a Mac  is a better platform but I don't own one so I do not know or if using ASIO or WASAPI has the same advantages as the Apple music stack.  A streamer appliance should have these taken care of these questions so you don't need to understand them. 

Eventually, those bits get delivered to a DAC. Internal or external.  How the DAC input DSP processor handles them and how they are clocked does seem to make quite a bit difference in sound.  Some well understood and measurable, some maybe not fully understood.  Next D 2 A conversion. That is the easy part.  This is followed by the integrator and analog stages. There is gets harder again and can influence the quality. Much understood, some not, and tons more opportunity for snake oil salesmen. ( Did you know you can bias an op-amp into class A?)

Now, to confuse things further, Ethernet CABLE can be used for other transports. I used a CAT-5 run between baluns to send low quality background music from my cable provided audio ( before I canceled cable).  I did not have any noise issues, but as it was audio over the cable, it was subjected to degradation though the baluns probably did more harm.  Ethernet cable is used for proprietary links in automation and I believe in professional audio production.  I do not know anything about that other than the cable is cheap and the connectors are excellent. 

Ignoring the sound better magic or not, I was checking up on the specs that have emerged since I retired ( we used fiber almost exclusively anyway)

7,  7a and 8 standards are proposed but will use proprietary connectors. Not RJ45s

Prices on cables have come way down. ( beware of source) 

Plain old CAT-5, UTP is of course sufficient for anything in audio unless you need magic .

5e is a little better with bonded pairs.  6 and 6a have a center core for more consistent impedance. 6a is also a little smaller.  CAT 5 to 6a may or may not have an overall shield. 6 and 6a may have individual pairs shielded.  CAT 6 connectors fit RJ-45 but are staggered pins in the crimp.  You are still looking at 10 ee-13 error rates or less with layer 2 fixing any that may happen.  TCP/IP is absolutely reliable bit for bit.  As with anything, I would not trust Amazon/E-Bay etc that the cables actually meet spec. Unfortunately BJC cables are not shielded but they are tested to spec.  So I have no guaranteed source for shielded cable. 

So for the paranoid, or want it because it is available, or bragging rights, whatever, 6a shielded is reasonably priced now.  I noticed most of the Netgear switches all have shielded ports now.  If you believe you need fantastic speeds for your streamer, you can get managed switches so the Internet to your streamer has priority.    You can even get a D-link or Netgear switch with a sticker on it for 20 times the price that says "audio quality"  It may have a better power supply. If you buy a data-center class switch from Cisco, it may be cleaner. 

I noticed the older unmanaged 5 port usually used 5V, but most 8 port and manages switches use 12V.  A couple turns around a big fat ferrite will knock any RF off those cables feeding the switches.  

I may upgrade as I have a WI-FI repeater and going PoE reduces the wiring mess.

We can measure anything in a data cable. We can't tell you which measurements may cause an audible difference. We can't tell you how well your device ports were executed. We can't tell you how good your Internet MODEM and router are but they can be measured.   This is the same in analog cables. We can measure, but that does not tell us what we need to target. Too many unknowns on both ends.   Without measurement I can absolutely tell you, no cable can add any detail that was not in the original source. That is magic.