Jump to content

Ethernet switches for audio - Part B: why a regular switch will suffice.


Recommended Posts



5 hours ago, recur said:

whoah Tex. That's a lot of questions. Take a breath, and relax. 

 

A lot of the best practice guidelines for network segmentation for latency sensitive traffic is well documented and understood. We've been building networks to deliver high frequency trading and voice and video for years and they are much more problematic (and $ intensive in terms of application and output) than audio to transmit. 

 

A quick rule of thumb on things to look at:

- Consider traffic prioritisation if you have a network boundary to cross (eg routing between subnets) to make audio go first (refer QoS with low latency queuing for a starting point for your research)

- Segment the network into different VLANs / subnets if you are worried about collisions on your network. Probably unnecessary on a home network, but you might go there. I use a separate VLAN for all IoT devices so I can firewall them from talking to other devices or to the internet. I've got no audio issues on a home network gigabit backbone.

- Don't route traffic in a CPU bound device (like a router) unless absolutely necessary. Routers are software devices and can't forward packets over subnets at line rate. Buy a good layer 3 switch with at least a static routing license for this.

- Consider enabling jumbo frames if all of your audio devices in the chain support it (unlikely).

 

 

RE: above, a unified gateway is a router/firewall. Use it to replace your home firewall / router. I'd only look at the 1RU unit for performance reasons.

 

Less ports doesn't equal less "noise", but "noise" isn't an issue regardless. I can't be bothered getting into an OSI model discussion and why the protocols that carry the data ensure this is not an issue as it will likely devolve into some audiophile related discussion related to people having golden ears that is of no particular merit.

 

PoE is only enabled if you configure it or a device requests it to power the device. 

Worth a sticky.

Link to comment
Share on other sites

As those Ubiquiti switches need 24v DC input, here's an option for LPS ...

https://www.sbooster.com/botw-pp-eco-24v/

 

Any suggestions how to ground them? There's the Synergistic Research grounding cables

 

My modem benefited quite a bit from LPS and Akiko tuning stick (plugged into a usb port).

 

As for vacant ethernet ports, read this  ... http://www.acoustic-revive.com/english/pcaudio/lan_terminator.html

Link to comment
Share on other sites



10 hours ago, dbastin said:

As those Ubiquiti switches need 24v DC input, here's an option for LPS ...

https://www.sbooster.com/botw-pp-eco-24v/

The US‑8‑150W model doesn't, it's IEC C-13 so built in power supply. I haven't looked at any of the other models with more ports than this so not sure if they do too.

 

10 hours ago, dbastin said:

My modem benefited quite a bit from LPS and Akiko tuning stick (plugged into a usb port).

I don't see how it would, but okay.

Link to comment
Share on other sites

10 hours ago, dbastin said:

As those Ubiquiti switches need 24v DC input, here's an option for LPS ...

https://www.sbooster.com/botw-pp-eco-24v/

 

Any suggestions how to ground them? There's the Synergistic Research grounding cables

 

My modem benefited quite a bit from LPS and Akiko tuning stick (plugged into a usb port).

 

As for vacant ethernet ports, read this  ... http://www.acoustic-revive.com/english/pcaudio/lan_terminator.html

Your first comment is wrong. It’s a 240V AC switch with an IEC plug in the back. They can serve 24V DC out a port for PoE purposes.

 

your product link reads as gibberish. Vibrations from networking equipment? That product is nothing but what we in the industry call a “solution in search of a problem”. 
 

this thread is about will a regular switch will suffice and why. If you want to start playing with nonsense products like this, post it in one of the other Ethernet related threads you’ve made.

Link to comment
Share on other sites

5 hours ago, recur said:

Your first comment is wrong.

Umm, correct, sorry, I was partly wrong.  Most Ubiquiti switches with 8 or less ports need 24vDC input, the rest have inbuilt power supplies.  I was looking at switches that provide the option to use a different/linear psu.

Are the switches with inbuilt psu's better in any way?

 

I think its relevant to cover enhancements to regular switches in this thread.  My openning post did state ...

 

" ... discuss pro/cons of upgraded power supplies, power cables, grounding, etc".

5 hours ago, deafenears said:

I don't see how it would, but okay.

As I"ve said in Part A ... This stuff sometimes makes no sense but does improve sound quality ... and that's all I need to understand.

So, approach with an open mind and spirit of experiment and adventure to hear what there is to be discovered.

Link to comment
Share on other sites

On 06/01/2020 at 2:05 PM, rmpfyf said:

An 'audiophile' switch is at best a band aid.

Firstly, a preamble.  My system is definitely serious.  It is now significantly reliant on Ethernet for connectivity for the 2 X NASs, Roon Core, DSP Box and a network player before anything reaches the DAC.  All are connected with a mix of Audioquest Diamond Cat 7 and or some CAT 8 cables.  As well there are two optical converters in the mix plus 3 Acoustic Revive filters.  As a next step I have become interested in whether a network switch would be a benefit.  There are many things that I do not understand about this topic.

 

 

Today I had the opportunity to try both the two levels of Paul Pang switches.  First the less expensive 8 port D link based switch with TCXO and a serious additional power supply.  A very very impressive benefit.  Far exceeded my expectations.  Then it was the turn of the 16 port OCXO dual switch.  No power supply need as it was IEC in.  It was just gobbing smacking absolutely bloody amazing.  I was stunned.  How come?  The owner of the switches was even surprised.  His take on the switches was that with Tidal there may be benefit but not what we experienced this afternoon.  I do not use Tidal as everything for me comes off the NASs.

 

 

Up until this stage the Optical converters were not being used.  We put them in.  Another step up.  Noticeably more so than my system with out either of the two switches.  I do not know the whys or wherefores but for me these two audio switches were a total surprise.  The one thing we did not try was using the two switches in tandem which is supposed to be even better than just one on its own.

 

 

I now know three things.

Audio quality Ethernet switches do provide a benefit and it was significantly serious for my system.

I don’t just want one I need one!

There are a number of different makes and models out there and I have to make choice.  It will be interesting

 

 

 

John

 

 

 

Edited by Assisi
word
  • Like 6
Link to comment
Share on other sites

I should have said in the post above that I am now listening to my system with just my low level Netgear Switch.  I am sure that most people would say you do not need anything more.  as the owner of the two switches said this is a great system.   I know what I heard this afternoon and I will not stop until I have it again.  It will be like a big itch that needs to be scratched.

 

John

  • Like 2
Link to comment
Share on other sites



Great news, thanks for sharing. Sorry to hear you now have a serious itch for a serious switch.

 

I think the 2nd switch you tried is in fact the dual switch approach, in one box.

56 minutes ago, Assisi said:

There are a number of different makes and models out there and I have to make choice.  It will be interesting

In case you havent seen this, go here.  If I missed a switch in the list, let me know.

From what I've read, the upgraded SOtM is hard to beat, except if you connect 2 with optical link. But of  course M12s are better for higher price.

Edited by dbastin
Link to comment
Share on other sites

@Assisi

 

I have no doubt what you're hearing is legitimate. None whatsoever. I run some Ethernet conditioning myself, and for the tweakers out there I have some nice good Rakon 25MHz OXCOs that could work a treat for many switches.

 

All I'm suggesting is to consider what they're actually doing and whether there are more direct means.

 

Lower noise - conducted from source and EMI - and better packet timing. That's it. If I've missed any other mechanisms of operation it'd be a great discussion point.

 

The noise bit - nothing is going to beat optical. Ever. Physically not possible. There may be bits to optimise at the device ends (what it plugs into) and some people just can't do it as they've devices that simply won't accommodate it. But I can't in good conscience spend megadollars on Ethernet cabling when a solution exists for this. It's a super bandaid. Most of us use CAT8 and other STP forms poorly for this.

 

The timing part sure, I get it. It's the same fundamental mechanism that drives audiophile software builds - jitter. Others eg @recur suggest industrial grade switches with separate PHY per port and dedicated ASICs are the way. They are not wrong - similar principles are used on audiophile USB transports, and on the etherREGEN. You could crack one open and go for better clocks or linear supplies I guess, but the Paul Pang method - putting the same on consumer grade hardware - whilst directionally correct is simply akin to putting lipstick on a pig. Not bad but there's better.

 

The etherREGEN is designed so as not to be perfect but to reflect the usual compromises associated with a home system. I'd go optical into it simply as to resist any temptation to spend silly money in that cable, and then as short a run as possible into whatever streams. A quarter metre, literally.

 

If you're going into a customisable PC it's optical all the way! But most of us... Aren't.

 

I take it you dropped speed to 100Mbps for your tests?

Link to comment
Share on other sites

I’m just baffled at how many people really don’t get that a switch is typically sending audio to a device with a significant buffer for playback. The traffic traversing the switch is not “real time” and as a result, anything these audiophile switches are doing, it’s not on the data stream.
 

There are error correction capabilities in everything from the Ethernet frame up through the transmission layer (tcp/IP) and often at the application layer. If there was anything amiss in any of these communications, you’d be able to see it in a simple promiscuous port capture using open source software like wireshark.

 

if audio was being streamed and real time performance was needed (in the rare and extreme case that there were no buffers on the streaming devices), you wouldn’t use tcp to transmit it, you’d be using udp (much like we do when transmitting telephone calls over IP).
 

Networks were built to reliably send data from one point to another. I’ve yet to see anything in the audio world that hasn’t already been solved to send real time video, telephony or super low latency transactions (go read Flash Boys by Michael Lewis to get an appreciation of the speed and $ involved) without impact.
 

Worrying about what’s happening on a hardware switch within 2 feet of your streamer, with a very low amount of local network traffic is definitely NOT where I’d start or even end my audio optimisation journey. 
 

I’ve worked in networking and security for well over 20 years and the common refrain for poor application performance is “it’s the network”.
 

The most complicated part of any application is the actual application. Networks are often set and forget and run for years without tinkering. 

 

These manufacturers are making Ethernet the boogey man as so few lay people have a good, solid understanding of networking. Reading the data sheets of these new audio optimised switching devices is hysterical. 
 

I’m semi tempted to get some white box Broadcom switches and pump them out to this audience. At least my switches will be fast, ultra low latency, highly configurable and the data sheet would reflect some semblance of reality.
 

 

  • Like 5
Link to comment
Share on other sites

40 minutes ago, recur said:

I’m just baffled at how many people really don’t get that a switch is typically sending audio to a device with a significant buffer for playback. The traffic traversing the switch is not “real time” and as a result, anything these audiophile switches are doing, it’s not on the data stream.

 

I'll reiterate - it's not the quality of data. Data gets there just fine. It's jitter. Of all measurements possible, OS jitter is prolly the best (albeit indirect) measure of any possible differences. The buffer really doesn't matter to these ends. When I reassemble my rig it'd be interesting to compile a kernel with appropriate tracers and run a few experiments to these ends. 

 

The simple measure of proof is to pull the Ethernet cable and listen for a difference. If there's going to be any effect whatsoever it should be audible under this condition. Now... there's a range of possibilities and configurations therein. A lot of people won't be able to pull the cable because their streamers will stop streaming... etc etc. 

 

Unfortunately there's a ton of snake oil out there from various audiophile vendors to the wrong ends. Wireworld makes a nice cable but (a) their suggestion that data integrity is somehow at stake is BS and (b) it shouldn't take a special cable. Just... no. Ethernet over optical has existed for eons for long distance work and even short runs (in very high EMI industrial environments) just for this purpose. It's engineered, it works. Again, we get into the possibilities and configurations - people with $10k servers running copper cabling only are going to have issues. 

 

43 minutes ago, recur said:

I’m semi tempted to get some white box Broadcom switches and pump them out to this audience. At least my switches will be fast, ultra low latency, highly configurable and the data sheet would reflect some semblance of reality.

 

It'd be a welcome addition to any experiment to these ends - I'd run it. 

Link to comment
Share on other sites



54 minutes ago, recur said:

I’m just baffled at how many people really don’t get that a switch is typically sending audio to a device with a significant buffer for playback. The traffic traversing the switch is not “real time” and as a result, anything these audiophile switches are doing, it’s not on the data stream.
 

There are error correction capabilities in everything from the Ethernet frame up through the transmission layer (tcp/IP) and often at the application layer. If there was anything amiss in any of these communications, you’d be able to see it in a simple promiscuous port capture using open source software like wireshark.

 

if audio was being streamed and real time performance was needed (in the rare and extreme case that there were no buffers on the streaming devices), you wouldn’t use tcp to transmit it, you’d be using udp (much like we do when transmitting telephone calls over IP).
 

Networks were built to reliably send data from one point to another. I’ve yet to see anything in the audio world that hasn’t already been solved to send real time video, telephony or super low latency transactions (go read Flash Boys by Michael Lewis to get an appreciation of the speed and $ involved) without impact.
 

Worrying about what’s happening on a hardware switch within 2 feet of your streamer, with a very low amount of local network traffic is definitely NOT where I’d start or even end my audio optimisation journey. 
 

I’ve worked in networking and security for well over 20 years and the common refrain for poor application performance is “it’s the network”.
 

The most complicated part of any application is the actual application. Networks are often set and forget and run for years without tinkering. 

 

These manufacturers are making Ethernet the boogey man as so few lay people have a good, solid understanding of networking. Reading the data sheets of these new audio optimised switching devices is hysterical. 
 

I’m semi tempted to get some white box Broadcom switches and pump them out to this audience. At least my switches will be fast, ultra low latency, highly configurable and the data sheet would reflect some semblance of reality.
 

 

Whilst I agree with you 100%, you're wasting your breath here.

  • Like 1
  • Haha 1
Link to comment
Share on other sites

1 hour ago, recur said:

These manufacturers are making Ethernet the boogey man as so few lay people have a good, solid understanding of networking. Reading the data sheets of these new audio optimised switching devices is hysterical. 

So few even within the IT industry, to be honest.

Link to comment
Share on other sites

46 minutes ago, PositivelyMusicallyGeared said:

So few even within the IT industry, to be honest.

im only quoting the last comment here but im am referring all previous. disclosure: i sell these switches and in particular Paul Pang manufactured in Taiwan. And of course there are many other brands.  i stand by my products. i do not sell anything that i dont believe is a good product. i do not sell anything that i dont believe is good value for money. these are modified for audio switches not IT INDUSTRY switches. i am not qualified in any IT or engineering area. i am qualified in audio. that is my business, without giving a technical explanation i can categorically state that these switches do improve audio. you can quote all sorts of reasons why they would or would not work but at the end of the day its what comes out of your audio system that is the proof in the pudding. the 2 switches i put into @assi 's system are: 1/ Paul Pang D-link TCXO based switch and 2/ the Paul Pang OCXO based dual switch. The technical specs along with photos are available on Paul's website. Yes optical is the way to go and this will take over from the current ethernet network set up eventually however we are not there yet in the audio world unless you are prepared to spend big. i am willing to demo these switches anywhere in melbourne and you are welcome to come to me for a demo as well. @assi is a serious audio lover and the quality of his system shows that. the comments he makes in reference to his experience with yesterdays demo are exactly as he described. my philosophy is the basis for a good digital rig is good power and good ethernet. both of which i can easily demonstrate.

  • Like 1
Link to comment
Share on other sites

It's crazy that we have intelligent people on this thread hellbent on making it a 100% one-way-or-the-other argument.

 

There are so many things that change system-to-system. 

Link to comment
Share on other sites

10 minutes ago, rmpfyf said:

It's crazy that we have intelligent people on this thread hellbent on making it a 100% one-way-or-the-other argument.

 

There are so many things that change system-to-system. 

correct impact varies hugely on system to system ...... what i say is that audio grade switches should have more of an impact on an internet network than just a local one however i was quite surprised at how well the switches worked on @Assisi's local network.....as was he. and again i am willing to demonstrate these switches anywhere in melbourne and i will leave the decision to your ears ......should it be no impact positive impact or negative impact. its just my opinion that they absolutely make a difference.....

Link to comment
Share on other sites



20 minutes ago, rmpfyf said:

It's crazy that we have intelligent people on this thread hellbent on making it a 100% one-way-or-the-other argument.

Conversely it's crazy that we have intelligent people on any thread here hellbent on believing that everything makes a difference.

  • Like 3
Link to comment
Share on other sites

Just now, Ittaku said:

Conversely it's crazy that we have intelligent people on any thread here hellbent on believing that everything makes a difference.

I admit, I was getting fed up with it, and started reading a certain other, more measurement oriented forum, but, not being content, some of the above mentioned have migrated there as well.  So annoying :) 

  • Like 1
Link to comment
Share on other sites

Just now, Dacman said:

correct impact varies hugely on system to system ...... what i say is that audio grade switches should have more of an impact on an internet network than just a local one however i was quite surprised at how well the switches worked on @Assisi's local network.....as was he. and again i am willing to demonstrate these switches anywhere in melbourne and i will leave the decision to your ears ......should it be no impact positive impact or negative impact. its just my opinion that they absolutely make a difference.....

 

I can understand how some people could feel aggrieved that people selling these items have no idea how they actually work and where they might be applicable therein. I'm not suggesting this is you, though there are some disingenuous types pushing all manner of snakeoil as though simply more=better. That's a BS argument to part people from their hard earned. 

 

I'm not comfortable with the notion that they 'absolutely' make a difference. In a suitably designed system they should make zero difference. It's possible to engineer accordingly. 

 

The test is simple - if you can pull the cable and hear a difference, dig deeper. If you can't, whatever Ethernet could do is being dealt with - spend no more.

 

Meanwhile at the other end of the argument...

 

1 minute ago, Ittaku said:

Conversely it's crazy that we have intelligent people on any thread here hellbent on believing that everything makes a difference.

 

Con, you have an extremely expensive DAC very specifically engineered to deal with these sorts of issues. If I'd paid what you paid for yours and a $4 Ethernet cable way upstream made a snot of difference, I'd be an extremely aggrieved customer. This said the market for people with your DAC is small. 

 

A bit like the market for people with a relevant understanding how buffers, interrupts and timing circuits actually work and can interact pending any number of configurations being small - the market for people being able to point out these components and say 'a ha! A buffer! Everything else is bullshit and none of you know what you're talking about, you're all stupid!' is considerably larger. 

 

But if you want to believe that your experiences are absolute, or that an Ethernet NIC having a buffer (or a USB receiver or whatever) kills any possibility of anything upstream doing anything downstream re timing, go for it. 

 

I'm just not sure (aside from the usuals hi-fiving each other on likes) that what everyone loves on a discussion forum is having discussion replaced with their being told they're idiots. 

Link to comment
Share on other sites

6 minutes ago, rmpfyf said:

I'm just not sure (aside from the usuals hi-fiving each other on likes) that what everyone loves on a discussion forum is having discussion replaced with their being told they're idiots. 

 

No-one likes that.   I have discovered that someone thinking a particular product works, and spending their hard-earned on it, is NOT what actually is annoying to me.   What IS irksome is some of those people posing as experts, and arguing and evangelising on forums.  It also makes me very wary, when I am seeking advice in areas where I have no expertise of my own.  

  • Like 2
Link to comment
Share on other sites

4 minutes ago, rmpfyf said:

 

I can understand how some people could feel aggrieved that people selling these items have no idea how they actually work and where they might be applicable therein. I'm not suggesting this is you, though there are some disingenuous types pushing all manner of snakeoil as though simply more=better. That's a BS argument to part people from their hard earned. 

 

I'm not comfortable with the notion that they 'absolutely' make a difference. In a suitably designed system they should make zero difference. It's possible to engineer accordingly. 

 

The test is simple - if you can pull the cable and hear a difference, dig deeper. If you can't, whatever Ethernet could do is being dealt with - spend no more.

 

Meanwhile at the other end of the argument...

 

 

Con, you have an extremely expensive DAC very specifically engineered to deal with these sorts of issues. If I'd paid what you paid for yours and a $4 Ethernet cable way upstream made a snot of difference, I'd be an extremely aggrieved customer. This said the market for people with your DAC is small. 

 

A bit like the market for people with a relevant understanding how buffers, interrupts and timing circuits actually work and can interact pending any number of configurations being small - the market for people being able to point out these components and say 'a ha! A buffer! Everything else is bullshit and none of you know what you're talking about, you're all stupid!' is considerably larger. 

 

But if you want to believe that your experiences are absolute, or that an Ethernet NIC having a buffer (or a USB receiver or whatever) kills any possibility of anything upstream doing anything downstream re timing, go for it. 

 

I'm just not sure (aside from the usuals hi-fiving each other on likes) that what everyone loves on a discussion forum is having discussion replaced with their being told they're idiots. 

i didnt say i have no idea how they work....i do....however what i did say is listen before you make a call... like it or not leave it to your ears. again i stand by my products and if you listen to them and dont like them then so be it....

Link to comment
Share on other sites



  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...
To Top