Basic switch....should I upgrade?


I'm using a D-Link switch into a EtherRegen (with an AfterDark PSU & Clock) via a Signature Ethernet Cable. Coming out of the EtherRegen I use a basic fiber optic cable into my streamer. (All connects/cables not mentioned are upgraded.)

Simply put, should I upgrade the switch or does the fantastic EtherRegen clean up the switch's audio shortcomings?

Please don't tell me to get a TT!  

128x128wsrrsw

Showing 21 responses by fredrik222

@jerrybj not my opinion, facts. You don’t have to take that into consideration, for sure, but do not call my knowledge opinions. That is offensive when you make up stuff to do that is in the manual for every Ethernet based device NOT to do. 

@fastfreight as I am sure you are aware, Cat8 has better shielding than Cat6/a, so if you are worried about interference, good choice. just don’t attach batteries to them to try and introduce interference! :)

 

 

@sns you are applying ANALOG terms to a digital connection. Completely irrelevant. Noise floor in Ethernet only means did it transmit a clear enough signal to discern a 1 or a 0, and did the frame add up in the end to match the CRC check.

so, for someone who knows what is actually going on, it is very simple. All the stuff you are talking about is just things that simply doesn’t have an impact at all, never ever. 

You can’t “clean up” a bad network switch. Bad network equipment causes packet loss, jitter, and other bad things. 
 

think of it as have a super clean and transparent window, and shine light light through it, the light looks exactly like intended. But if you but a dirty old  window between the source and the clean window, the clean window can’t do anything at all to “clean up” the source. 
 

that said, the transfer protocols at the layers above take care of quality, with a crappy network switch like d link, you may experience enough issues that it could potentially impact sound quality, and there is nothing you can add to fix this. Now, “audiophile” switches are typically not better from my research, you really need an enterprise switch, from juniper, extreme networks, or arista for example. 

@wsrrsw  I highly doubt you will find a 5 port enterprise switch, but you can probably find 8 and definitely 12. 
 

In theory, cleaner power improves poor circuitry more than good circuitry, as good circuitry is made to operate with a much more narrow acceptable performance range and higher performance threshold, regardless of the input. Is audible however, maybe? 

If you want really clean, most enterprise network equipment can be made to order with DC input. 

@wsrrsw the angel n8 doesn’t have any actual switch performance metrics, so pretty hard to tell. Switch performance metrics include things like forwarding plane through put, switching method, over subscription rate, buffer size, etc. 

@ghdprentice depends on how bad the switch is. D Link are among the worst consumer equipment out there, so there is a real possibility that the streamer chache and tcp/ip can’t cover for it.

What is clear is that you can’t plug use a bad switch, and then add another device to clean up the bad switch output. That’s not how things work in networking. 
 

Also, Audio Science Review have some thoughts on this topic too.

 

 

@duckworp The Cisco 2960 is a low end enterprise grade switch. Very very few home users will ever be able to push it to where it will be remotely a concern for streaming. 
 

 

@antigrunge2 he is 100% correct in his review of these products, because you can measure them, and actual network switch performance metrics are not listed on these websites. They are, 100% snake oil. 
 

@wsrrsw i’m in it security, but I have 20+ years of network experience. 
 

I use juniper ex2200-c, with check point 1550 firewall, and extreme networks 801.11ax wifi. Cat6a cabling, newly redone. 
my streamer / server is NAD M50.2, AES/EBU to Classe Delta Pre, and Classe Delta Stereo power amp driving Audiovector R3 Arrete. 

@steakster if you are calling my 20+ years of enterprise networking experience is lack of experience, what experience do you have?

@wsrrsw going to and from optical involves signal conversion. While I subscribe to that you should not be able to hear any difference from an audiophile switch compared to an enterprise switch, similarly if everything works ok, you shouldn’t notice any difference from optical transmission​​​​​​, I am sure someone is going to argue that moving from/to optical degrades the signal. 
 

personally, I have to fibers run to my detached garage due to the length of the run from my switches, but no audiophile grade stuff in the garage, just a SVS prime and some B&W speakers for workouts. Fiber is expensive however, at least 10 times more.

@b no, adding to fix a bad link in the chain never works. Remove the crap out of the chain instead. 
 

basically you are saying that using your iPhone DAC and source into another significantly better DAC will fix the issues with the phone DAC. Not possible. 
 

a switch operates on layer 2 in the tcp/ip stack, and will only check for crc frame errors. A switch doesn’t do anything to correct timing (jitter) issues. That is just not how a switch works, period. 

The amount of things you guys are trying is astonishing. But none of it works, and a lot of it is counter to how Ethernet works. For instance, for audio listening (not two way like voip), latency is irrelevant. The change in latency matters however, and if packets arrive out of order. Another things, the only possible outcome of taping batteries to an Ethernet cable is that you introduce CRC Frame errors, which the will force the switch to discard the frame, and higher level protocols to trigger a retransmission. 
 

pro audio applications use enterprise equipment, not made up cures but introducing voodoo into the chain, and adding things. 
 

maybe it is just me, but in general for my audio setup, I like to Keep It Simple, the fewer components, the better, since adding will only add to the signal, which means you have modified the recording in transit from source to speaker, and that is not audiophile grade. 

@thyname at a high level, yes, there is nothing you can do. If you have a low quality switch, you will lose packets if you load it, that will impact sound quality. If you tape anything generating magnetic fields to the Ethernet cable, you will likely lose packets, also improving the sound quality.

 

as for your email to China. While a completely different protocol at a much higher level, similar principles, if you have punctuation issues, that is all about the what’s in front of the keyboard. 100%. I have unfortunately had to prove that one too many times to think it is funny. 

Meant to say impacting not improving SQ. 
 

here is a very high level overview of how networking works: 

 

 

@dbastin of course the digital signal is transmitted using analog electrical signal over copper, however, it doesn’t change that it is digital, and noise floor only is relevant to make sure that the frame made it across intact. It is pretty simple, each Ethernet frame has a CRC check, if it passes the check, the frame is accepted, if it doesn’t, the frame is discarded. So, if you have enough interference, for instance, by taping batteries to your cables, you will have a higher rate of discarded frames, that the higher level protocols need to correct for, typically by asking for a retransmit.

https://en.m.wikipedia.org/wiki/Ethernet_frame

@kennyc while I personally don’t like Cisco, you can typically find cheap Ciscos on eBay, like the 2960G mentioned above.

the current version of my switches are Juniper EX2300-C

https://www.juniper.net/us/en/products/switches/ex-series/ex2300-c-fanless-switch.html

@thyname well, the article is way off the mark. Especially the first paragraph where it says it applies to all digital connections. What they are talking about can apply under some circumstances to asynchronous digital connections without error detection and correction. 
 

Ethernet is synchronous, and has error detections, but that is just one layer, the next layer also have error detection, call IP, then you have TCP, also have error detection and correction, and lastly, you have the application layers, also have error detection and correction. So it doesn’t apply.

Second, no human card here a different of 1 ns. You need to get into 300 to 400 ms, or 300,000 or 400,000 times larger difference before human hearing can register an issue. 
 

and, clock rates don’t work like that in switches. If you want to learn about that very low level detail, have fun, here is a place to start: 

 

@wsrrsw you should always have security. Too much shit going on to trust your network provider to catch all.