Thread: Do I really need an " Audio Grade Network Switch "?

I think it's been a quite controversial topic for years, like what's the difference between normal network switch and an audio grade network switch, the price difference is certainly obvious though...
Anyway, I've done some researches, most audio reviewers say that under this " new digital streaming era " that an network switch is a must for an audio system, which is understandable for me, I mean because if I wanna play TIDAL or Qobuz or Spotify, I gotta use network so I can stream these online services, so yeah I get that if the network quality is good enough, it can possibly level up the music performance.

But anyhow, I'm new to this area, so I don't like to spend big bucks on my first purchase hahaha... there's a very wide range of the prices though, the top one is Ansuz Power Switch I think, the inner circuit and design look pretty sharp, and surely over my budget lol

So I'm choosing between Bonn N8 and SW-8, these two both got good reviews, and the prices seem so darn much friendly to me as I'm looking for an entry level switch now, do any of you have any insights to share?
or should I just go for the higher level ones?

Best,

You could go the fiber route if you're concerned with interference or other noise on a copper cable. Either a fiber SFP port on the switch and a media converter on the device side, or two media converters. If you're not familiar, essentially this converts the ethernet data transmission to light, then back again closer to the audio device, where you can use a short high end "audio-grade" ethernet cable that in theory protects the signal from badness like rmi/efi or near-end cross talk etc.

No matter what you do the last yard of signal for any streamer that only accepts copper ethernet (Cat 5/6 etc) is going to be a copper ethernet cable. Some devices are starting to include SFP ports to go all-fiber, like Lumin X1 which to me seems the best design.

If you're converting the signal to fiber I can't really see a need for anything other than a normal ethernet switch from a reputable manufacturer to pump out the 0s and 1s.

(Thought this was interesting, although I can't vouch for it's scientific accuracy: Top 2 Things to Consider When Running Ethernet and Power Cable)

Originally Posted by jmusica

...If you're converting the signal to fiber I can't really see a need for anything other than a normal ethernet switch from a reputable manufacturer to pump out the 0s and 1s.

What about between the incoming Ethernet signal (typically coax from a cable provider) and the conversion to fiber?

If you are streaming then I would say probably worth it. If you are not streaming then there is no purpose what so ever for it.

Originally Posted by rbbert

What about between the incoming Ethernet signal (typically coax from a cable provider) and the conversion to fiber?

Converting to fiber would happen after your internal/home ethernet switch and in theory takes care of any mess brought in before it, whether the cable modem or your switch. The design key with the fiber conversion theory is that you leave only a very short ethernet run into your streamer and you control for any ethernet mess as close to your streamer as possible. In a related scenario, same logic would apply for a direct connection from a NAS to the ethernet port on a streamer, in case you're not streaming from the internet or your home ethernet switch won't be in the mix.

I love the "in theory" . I do have fiber conversion, short lengths of various Ethernet cables, an NAS and an EtherRegen with LPS, and have done a little experimentation with putting each of those at different parts of the chain, with no clear conclusions apparent to me (yet). I don't do any critical listening via Internet stream, only from local files on the NAS (or the servers internal SSD)

I have never done comparisons between switches. I did try an Ether Regen, they offer a return, I kept mine, not sure what's in the box but the improvement in sound quality was worth keeping it. This would only impact streaming music from the internet.

Originally Posted by rbbert

What about between the incoming Ethernet signal (typically coax from a cable provider) and the conversion to fiber?

In my system replacing a long run of Ethernet with fiber optic "may" not have improved the SQ but it definitely contributed to what has been a rock solid Roon connection for several years now. I could never get a GigaFoil and Keces power supply to work in my set-up. I inserted a switch with a fiber optic output that sits in a upstairs closet where my routers and SGC i5 are located. A log run of fiber optic then goes down to my 2 channel room and terminates in a optical module just in front of my Upsampler. There is a short run of Shunyata Sigma ethernet that connects between the optical module and the Upsampler. Great sound and a rock solid Roon implementation ever since.

Good point about the physical advantages of a fiber run, particularly over a longer distance or near anything environmentally challenging. Theoretically fiber (in appropriate conduit, run by professionals) could be used between between floors, or even between buildings, eg a mother-in-law suite, etc. Copper ethernet has distance limitations (100m) that could affect a larger, multi-floor house.

Question: my switch is used to connect a streamer (Metrum ambre) and a SGC sonictransporter (Roon core) to my router - so it comes into play for both “internet” streaming (e.g tidal through Roon) and for local USB drive files through the SGC. I would think the switch equally affects both audio feeds is that correct?

Originally Posted by roncagg

Question: my switch is used to connect a streamer (Metrum ambre) and a SGC sonictransporter (Roon core) to my router - so it comes into play for both “internet” streaming (e.g tidal through Roon) and for local USB drive files through the SGC. I would think the switch equally affects both audio feeds is that correct?

i am not going to get into the debate whether switch's can impact the SQ of your audio. As noted in my post above I added a switch that allows me to convert to optical and remove the potential for noise via Ethernet. It sounds great and I am not going to A/B it to a higher end switch. If it impacts the SQ I would think it would do so on both.

I appreciate the sonic advantages of switches and fiber-optic (any copper Ethernet cable I might use is less than 60 ft). Both Internet streaming and local streaming follow the same path to my server, which also has a (generic) SSD. I can batch load local files (one or several albums) into buffered RAM prior to play (or play from the SSD directly), which is not possible when internet streaming (although the next file can go into buffered RAM while one is playing), and the sound quality difference is noticeable. It might be that the extra CPU work loading a file into RAM buffer while one is playing could affect the sound quality but that hasn't been my experience (no difference if I buffer just the next file vs. no buffering). Have those of you using Qobuz or Tidal as your primary digital source not noticed this?

My internet connection is pretty solid 600 Mbps (70+ MBps) down.

I find SQ of local digital streamed from NAS (or Lumin L1, which is a proprietary NAS for Lumin streamers) generally better than internet-streamed, even from hi-res sources like Qobuz. I don't really have enough info or expertise to venture an (educated) guess as to why, but I tend to doubt it has to do with the switch- both sources go that route.

I recently read an article on What's Best Forum where a very detailed reviewer insists he heard a difference when replacing his switch with the Netgear S8000, which is a gamer-oriented switch, and coincidentally the switch I've used for years and just replaced with a Cisco Meraki, because the Meraki has an in-built SFP port, whereas I had to use 2 converters with the Netgear. Can't say I've detected a difference.

I guess at the end of the day, if pressed my input would be spend your $ (and time) elsewhere in the audio chain; I'd think any difference would be more detectable elsewhere than a network switch.

Any audible improvement by using a switch, optical break, etc. will not be due to the primary purpose of the device. That’s well understood.

It will be a secondary effect, most likely the reduction/elimination of electrical noise that ends up inducing jitter in the upsampler/DAC.

Jitter measurement is a well known and solved problem, yet manufacturers of these switches cables and isolation devices have yet to show that their devices quantitatively reduce jitter. It’s crickets out there, apart from a hand-waving white paper from one popular device manufacturer.

It’s too bad, because enough people whose opinions I value are hearing differences. Are these differences in these individual’s cerebral cortexes?

Originally Posted by W9TR

It will be a secondary effect, most likely the reduction/elimination of electrical noise that ends up inducing jitter in the upsampler/DAC.

This is the bit I have trouble connecting the dots - or in this case, bits - on. How does a series of 0s and 1s induce - or conversely, solve - jitter? May well be the case, but I sure don’t understand it. And as you’ve said, there seems little-to-no scientific/empirical proof of said.

Originally Posted by jmusica

I find SQ of local digital streamed from NAS (or Lumin L1, which is a proprietary NAS for Lumin streamers) generally better than internet-streamed, even from hi-res sources like Qobuz. I don't really have enough info or expertise to venture an (educated) guess as to why, but I tend to doubt it has to do with the switch- both sources go that route.

I also doubt it is due to a switch, or to any particular switch; I was more curious as to how those who apparently use streaming as their primary digital source (as opposed to a secondary digital source that is a good way to find new music or for less critical listening) reconcile that with its lesser sound quality, or do they find the sound quality equivalent, or just don't care. And if the sound quality is equivalent, how do they make that happen?

Originally Posted by jmusica

This is the bit I have trouble connecting the dots - or in this case, bits - on. How does a series of 0s and 1s induce - or conversely, solve - jitter? May well be the case, but I sure don’t understand it. And as you’ve said, there seems little-to-no scientific/empirical proof of said.

The theory of the case goes something like this. All digital signals are really analog signals that have finite rise times and fall times. At some point in the chain, a device, usually the DAC, has to interpret what is a 0 and what is a 1. This is done with a device called a voltage comparator, which is set at a fixed voltage somewhere between digital 0 and digital 1. If this comparator is perfect, it will correctly determine the TIMING of the digital 0 to 1 transition. However, if there is noise on the ground or power supply of this comparator, it will change the TIMING of the 0 to 1 transition, inducing what we now call ‘jitter’. Switches, cables, isolators and the like strive to reduce or eliminate this noise to create more perfect transitions and more precise TIMING, thus reducing or eliminating the resulting uncertainty, or jitter.

Now all of this is measurable to sub-picosecond accuracy with modern test equipment. Yet no device manufacturer has chosen to do so. I wonder why?

Originally Posted by Puma Cat

Read this paper by John Swenson carefully (it's pretty deep): URL Shortener. John Swenson, who is the designer and lead engineer for the UpTone Audio EtherREGEN, IsoREGEN, Sonore OpticalModule, UltraRendu and OpticalRendu, was a professional Ethernet electrical engineer that spent his career at Broadcom and Cisco developing Ethernet-based networking technologies, so you're getting this info from the best source in the audio industry.

As a design engineer, I would have to understand the mechanisms causing the issue. I would experiment with different components and topologies to determine what worked best. As an engineer/scientist, I would rely on measurements in addition to my ears to optimize the design. Yet no device or cable manufacturer has chosen to share any objective evidence that their products actually reduce the impact of the distortion mechanisms they have so eloquently described.

Originally Posted by Puma Cat

The actual signal that is sent from a server to a switch to via then to a network bridge/streamer, etc. is an analog voltage in the form of a square wave

Thanks for all the additional info, it's helpful. Now that you've said it, I vaguely remember this from some network engineering course, and it starts to make more sense. (and helps explain why streaming from a local network device sounds better than an internet stream, and why Lumin allows you to connect their NAS directly via ethernet to the streamer)

Stephen,
I’ve read it and understand it. John has done an excellent job describing the distortion mechanisms involved, and how they may impact a digital playback chain. It’s a fine piece of work. I have made changes to my system to reduce the kind of errors he discusses in the white paper. I’ve noticed sonic improvements that are substantive.

What I see here is a missed opportunity. I’m a member of the Audio Engineering Society, and there have been dramatic improvements in the capability of test equipment used to measure the performance of digital systems. AES17-2020 describes test methodologies that, with the right gear, will allow measurement of jitter spectra orders of magnitude below the noise and distortion floor of the equipment being measured. This is not a case of ‘you can hear what you can’t measure’.

So these devices continue to be considered audiophile nonsense to some, even though they may have merit.

That is the missed opportunity.

Ok, still processing (no pun intended). Help me with this part:

The 0s and 1s are sent as voltages, in the form of an ethernet packet (frame). The packet is expected to take a very specific form - header, source, destination, data, frame check etc. So are you saying that in the data payload of the ethernet packet, the 0s and 1s can basically be "corrupted" by noise etc? Wouldn't that lead to the packet being malformed and rejected and show up as errors and retries in the switch? Or is it more subtle than that and the data payload is basically just not properly representing the original 0s and 1s contained in the music file? (in which case it seems like a miracle that internet-streamed music gets to the endpoint sounding like music at all, with all the network hops/equipment and cabling and datacenter environments etc it has to traverse before it even enters your home)

Originally Posted by Puma Cat

Got it. Thanks, Tom!

Ok ha guess I have to go read this. Feels like work.

Originally Posted by jmusica

Ok, still processing (no pun intended). Help me with this part:

The 0s and 1s are sent as voltages, in the form of an ethernet packet (frame). The packet is expected to take a very specific form - header, source, destination, data, frame check etc. So are you saying that in the data payload of the ethernet packet, the 0s and 1s can basically be "corrupted" by noise etc? Wouldn't that lead to the packet being malformed and rejected and show up as errors and retries in the switch? Or is it more subtle than that and the data payload is basically just not properly representing the original 0s and 1s contained in the music file? (in which case it seems like a miracle that internet-streamed music gets to the endpoint sounding like music at all, with all the network hops/equipment and cabling and datacenter environments etc it has to traverse before it even enters your home)

I’m saying the opposite. The data will get there uncorrupted. But along with that data comes a bunch of noise - from power supplies, from the data itself, from switches, cables, and other network gear. This noise appears on the grounds of audio gear connected to your network, and this noise then corrupt how your DAC interprets the timing of the data - not the data itself. That creates jitter, which we know to be audible. That’s the mechanism.

It seems that one issue is people do not understand how digital data is transmitted. It is transmitted as an analog signal that is then turned into digital in the receiving device. Because the analog signal can be changed, the point where a 0 or 1 is determined can be slightly different from what was originally transmitted, which leads to jitter.

Originally Posted by BlueFox

It seems that one issue is people do not understand how digital data is transmitted. It is transmitted as an analog signal that is then turned into digital in the receiving device. Because the analog signal can be changed, the point where a 0 or 1 is determined can be slightly different from what was originally transmitted

Definitely the part that I'd forgotten. Even when I learned it, was sort of like grammar. "Meh, rules, why would I ever need to know that, I just need to know how to check for errors on a switch port and have some idea what things could cause it ". Live and learn, as they say.

This has all been an interesting and useful (re)education. For me at least.

But now, got me looking at audiophile ethernet switches dammit.

Originally Posted by Puma Cat

Just get an EtherREGEN from UpTone Audio. Easy. Only $640. Has UpTone's proprietary ADIM and a really good clock, the Crystek CCHD-575 oscillator.

EtherREGEN – UpTone Audio

Mehbe. Would you say this, or the SoTM sNH-10G and just replace my current switch?

With the REGEN I could still go fiber out on the in-built SFP port, to the X1, correct?

Update: I see, I would do it this way: "For the few people who have an endpoint with optical input, one can “turn around” the EtherREGEN and feed that DAC-connected endpoint from the optical cage, while connecting the lone ‘B’-side port to the network. Thus ‘B’ >’A’. "

Pretty cool. I used to do this for USB, and the results were clear to me, so in concept (and reputation), this makes sense to me now also. Guess I just had to stitch the pieces together.

Originally Posted by jmusica

With the REGEN I could still go fiber out on the in-built SFP port, to the X1, correct?

Yes.

Regardless of fiber or copper Ethernet, the recommended connection is this:

WiFi router (or switch) -> side B - EtherREGEN - side A -> Lumin T2 / X1 / P1 / U1 (MINI) / M1

WiFi router (or switch) -> side A - EtherREGEN - side B -> Lumin S1 / A1 / T1 / D1

Make sure you use unshielded CAT-6 network cable with EtherREGEN.

Originally Posted by Puma Cat

That's not accurate, either. The Ethernet switch, it's cheap-*ss clocks, Ethernet receivers, power supply, and transformer cores in the receivers can all have an impact on the audio quality. You can also get high-source leakage impedance current between the RJ45 Ethernet receivers that can have an audible impact, as well.

And then, there's common-mode noise, which can...also be in the Ethernet cables. Doh! 😜

Both jmusica and I noted that both local and internet-streamed files pass through the same switch but still sound different (local are better). Since the switch is present and the same in both instances, why would the switch account for the difference in sound quality.

Originally Posted by Puma Cat

The SOtM doesn't have an SFP cage...

The one I see on their website does. It's more expensive than the Uptone, and both get great reviews. I wonder if it's possible to hear a difference (at least for me, my ears, with my system - opposed to someone with a 100k setup and better ears than me).

Originally Posted by rbbert

Both jmusica and I noted that both local and internet-streamed files pass through the same switch but still sound different (local are better). Since the switch is present and the same in both instances, why would the switch account for the difference in sound quality.

One thing that comes to mind for me is what is being referred to as local. If you are talking local files being on a NAS and sent through a local network then they would indeed go through the switch. However, if local is files that are stored internally to the server, which is what I refer to as local, then the signal would not go through the switch at all.

Originally Posted by Puma Cat

Here's my recommendation, J.

Buy the one that is less expensive.

Done. Won't have it til February tho. Worth buying a master clock?

I like this one. Brand new. Started another thread with details.


Sent from my iPhone using Tapatalk Pro

Originally Posted by Puma Cat

Sure. But, master clocks can be really pricey. If you find an affordable one, please let me know.

Also, I don't what you're using for power distribution, but something that made a notable improvment for me was connecting a Shunyata Venom CGC cable from the ground post of my EtherREGEN to one of the four GP-NR ground posts of my Shunyata Everest power distributor. I don't know why folks don't use that GP-NR system on the Everest, Denali, Alpha, Sigma or Venom power distributors more, because it can provide a really notable improvement in audio quality.

Found this one, no idea if it's any good as I'm not familiar with the master clock market, never had one. https://ctronicmusic.com/home
Seems they're tying themselves to EtherREGEN marketing-wise. Pretty inexpensive.

I'm not sure if my power distributer (Ayre) has grounding posts... I'll have to check.

Originally Posted by Randy Myers

One thing that comes to mind for me is what is being referred to as local. If you are talking local files being on a NAS and sent through a local network then they would indeed go through the switch. However, if local is files that are stored internally to the server, which is what I refer to as local, then the signal would not go through the switch at all.

Previous posts indicated most "local" files are from NAS, and to my ears the EtherRegen allows the sound to match that on the server/player's internal SSD (if the NAS tracks are buffered into RAM before play). As noted, Internet streamed tracks can't really be buffered into RAM before play, which I suspect is part (but probably not all) of the reason they don't sound as good.

And though it is not obvious from the website, Totaldac now has entered the Ethernet switch marketplace