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

[Guest rmpfyf]

11 minutes ago, TDK said:

What ethernet does is not down to opinion I'm afraid!

Not arguing with you on that one. Ethernet is robust, data certainly gets to the NIC and audio isn't a bandwidth-critical application generally. 

 

That doesn't define the ability for Ethernet to affect SQ. Just means that any differences therein aren't a function of data getting there or not.

[Guest rmpfyf]

6 hours ago, Ittaku said:

You almost seem smug in your attempt to make me spell out what I believe, when what I said was just a string of facts - ethernet clocks are unrelated to DAC clocks and one never drives the other. You know how I feel about ethernet and hifi and are intent on trying to extract it from me here based on that comment, but I have no interest in being dragged down into that debate. 

 

Somewhat misplaced in that no one here actually suggested that whatever clocks an Ethernet network drives a DAC IC... and you came down like a ton of bricks when an easier interpretation might have just been 'yes, plenty of DACs take an external clock signal'. 

 

If we know how you feel about Ethernet and hifi then... why are you here?

 

On 22/12/2019 at 2:29 AM, dbastin said:

This is where people can share their views about why ethernet switches specfically made for audio should not make any difference to sound quality, and also the other side, why they should and do.

 

This thread is a place for these debates to occur and therefore minimse these debates cluttering up other threads.

 

Posts in this discussion may suggest or recommend particular switches, and discuss pro/cons of upgraded power supplies, power cables, grounding, etc

 

I'm kinda hoping we get something out of this thread, and that your experience there's room to disagree and allow for possibility too. 

 

2 hours ago, Ittaku said:

Stick to what I said, and try to find any science or evidence that ethernet clocks have anything to do with DAC clocks instead.

Interesting use of pronouns both stated and implied. 

 

Have plenty of science and evidence to these ends though wouldn't think this qualifies me in any way. How does the above comment make this a place to share?

 

Anyone wanting to talk down to the great unwashed that hear differences in cables/switches/etc / theorise how that could be so / etc can head over to ASR where lines of thought such as 'I don't get how that could work' can morph into 'anyone that disagrees is an idiot' and like minds can spend endless pages of rant high-fiving each other ad nauseam. I'm glad we're not that. 

 

That Ethernet anything does anything to my SQ frankly drives me up wall. Can't deny it though. Have a spare router here with SFP and some spare OXCOs of the right frequency. When I've some spare time, who knows. I'm here to learn and share. If useless keen to avoid the time spend. Would think I'm not the only tightwad here interested in a ghetto way to make better. 

 

At any rate would like to think all turned up with better intentions than telling others they're not hearing things. 

[gonefishing999]

Maybe putting these under a switch will help.

Cardas Golden Cuboids Myrtle Wood Blocks

[andyr]

3 hours ago, gonefishing999 said:

Maybe putting these under a switch will help.

Cardas Golden Cuboids Myrtle Wood Blocks

 

In a listening session with my 'golden-eared team', a few years ago, I put 3 of the Cardas Myrtle blocks under my miniDSP unit (instead of relying on the small rubber feet).

 

We heard an improvement - so they are still there. 

 

Andy

 

[dbastin]

My current switch sits on a couple of damping disks from Walker Audio Valid Points, which made a small but noticable improvement.

[gonefishing999]

I use Tasmanian Myrtle it works a treat.

I have been thinking that maybe the Cardas Logo and name would make it any better.

Edited December 25, 2019 by gonefishing999

[dbastin]

Just discovered this ... Comparison of 7 switches, in Dutch, but use translator and see diagrams.

 

Tested were:

 

standard Netgear (baseline)

Fidelizer EtherStream (€399)

Ansus switch (€1999)

Fiber (€250)

Netgear standard (€149)

Bonn N8 (€ 395)

SoTM - basic model (€999)

Cisco Meraki enterprise 8 port with fiber and PoE (€799)

 

Concludes top 3 being:

- Cisco Meraki

- SoTM

- Silent Angel Bonn 8

https://www.alpha-audio.nl/review/zeven-switches-voor-streaming-audio-getest-blind/

 

Take a look at the methodology, I note:

- all switches used with their supplied power supplies.

- Short term listening. 

- Changing switches may not have enabled proper warm up.

 

Interesting they didnt include the Aqvox from their region.

[deafenears]

On 22/12/2019 at 5:48 AM, LogicprObe said:

Well, you really want as little delay as possible for audio.

Want as little delay as possible?

 

- move to a place that has NBN FTTP rather than ADSL or FTTN/VDSL. That's ~30ms down to ~2ms.

 

- find a switch that does cut-through forwarding (vs. the usual store-and-forward which buffers). Likely you want a device that does both switching and L3 routing so your NBN connection plugs/terminates directly into it.

 

- set up QoS / DSCP / TOS to prioritise packets to and from your streamer.

 

- if your streamer is a PC, find a NIC from the Mellanox ConnectX adapter family. Let's you tune IRQ Affinity and pin to specific CPU cores.

 

- if your streamer is a PC, also tune the TCP stack for low latency.

[deafenears]

Oh also, switch to a business grade Internet connection and with a reputable ISP that won't route your traffic all over the country before making it's way out via International transit links (in order to reduce cost).

[F18]

On 23/12/2019 at 5:47 PM, dbastin said:

I'd speculate that some companies making switches for optimal audio have or obtain the expertise to achieve that objective.

 

Take Melco/Buffalo for instance ... check out what they're in to ... https://en.m.wikipedia.org/wiki/Melco

 

I would think, aside from tweaking existing switches, development of a switch optimised for audio would involve experts its all aspects.

 

I don't think they have the right knowledge in what they are trying to apply to because anybody who knows how network and switching works will know that no matter how fancy the ethernet cable is, it will not make a lick of difference as long as its built right to the standards set and used for the past 30 years. Digital data transmission be it at a layer 2 or layer 3 switching/routing level doesn't care about the metallurgy or laser medium it is travelling on as long as its there. There are protocols implemented and programmed into the device to re-transmit lost packets, fragments of packets which carry fixed sized blocks of data. Analogue cables operate more differently and can be affected by length, capacitance, resistance, inductance, RFI/EMI design etc. 3 of the mentioned does not affect digital transmission.

 

If these so called companies had experts with knowledge that stood out of the crowd they would be putting in a white paper design for approval by IEEE/ICNP because everybody would be benefiting from it.

 

As a qualified EE and Network Engineer, it is an insult to me to see people over pay for crappy products that don't make a difference or work. Why weren't these so called audio switches released 10 years ago? Switches have been around for decades, 10 years ago we had network streamers and network DAP's. The answer to this is because it is a money grabbing opportunity because it is more popular nowadays. 

 

Also I'd avoid buying any of the Pang jitter modified switch. Overpriced, poorly built and does nothing, if you must need a low jitter clock attachment to a generic $50 switch for whatever reason one can be made with $20 worth of parts on a protoboard.

Edited January 5, 2020 by F18

[BugPowderDust]

I'm always amazed that people think audio is a difficult thing to send on a data network.  The issues with getting traffic reliably and without error between two endpoints were solved many moons ago. The data centres that feed your gaming, social media, streaming and the like shunts more traffic than you could imagine, at extremely low loss rates and have done for decades.

 

With gigabit interfaces and appropriate network segmentation (this is even questionable in a home environment), LAN transmission is a doddle for low bandwidth (see Audio) applications. Moving to a WAN brings it's own challenges, but that's not the problem these audiophile switches are solving. 

 

If we can solve the issues of high frequency trading, delivering audio isn't an issue. Extreme latency, jitter and packet loss performance metrics and the like can be delivered reliably using TCP/IP over quality ethernet switches by someone with a cursory knowledge of networking. 

 

If you are really hell bent on spending your $ on a top shelf switch, the brands to look at are Cisco, Arista or any of the Broadcomm based chip set switches that run the Cumulus OS (or similar). 

 

If you're routing traffic on a software based, CPU bound switch you're potentially doing more damage than good.  

[Guest rmpfyf]

1 hour ago, recur said:

I'm always amazed that people think audio is a difficult thing to send on a data network.  The issues with getting traffic reliably and without error between two endpoints were solved many moons ago. The data centres that feed your gaming, social media, streaming and the like shunts more traffic than you could imagine, at extremely low loss rates and have done for decades.

 

With gigabit interfaces and appropriate network segmentation (this is even questionable in a home environment), LAN transmission is a doddle for low bandwidth (see Audio) applications. Moving to a WAN brings it's own challenges, but that's not the problem these audiophile switches are solving. 

 

If we can solve the issues of high frequency trading, delivering audio isn't an issue. Extreme latency, jitter and packet loss performance metrics and the like can be delivered reliably using TCP/IP over quality ethernet switches by someone with a cursory knowledge of networking. 

 

If you are really hell bent on spending your $ on a top shelf switch, the brands to look at are Cisco, Arista or any of the Broadcomm based chip set switches that run the Cumulus OS (or similar). 

 

If you're routing traffic on a software based, CPU bound switch you're potentially doing more damage than good.  

 

And again, that's not the point. An 'audiophile' switch is at best a band aid. The problems solved are not in TCP/IP. No one here running half speed logic would assume theres any issue in getting data to an endpoint where audio is concerned. 

 

Whilst appreciating your dig at software-based switches, most audio rigs owned by most here cost less than a Cisco running Cumulus. 

[Stereophilus]

Would the people here advising against audiophile switches also conclude that Ethernet-optical converters placed upstream of the Hifi server/streamer make no difference to the sound quality?

[BugPowderDust]

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

Whilst appreciating your dig at software-based switches, most audio rigs owned by most here cost less than a Cisco running Cumulus. 

um, no Cisco switches run Cumulus.

You could easily get a 24 or 48 port gigabit Cisco switch (refurbed) for circa $500 though. That's less than most people here spend on a DAC or cables and it's utilitarian, being able to be used to speedily connect a house, as well as provide wire speed inter vlan routing to isolate clients.

[BugPowderDust]

ha, I just looked on gumtree: https://www.gumtree.com.au/s-ad/st-james/other-electronics-computers/cisco-catalyst-3750g-switch-24-port/1236308272 

 

So for $100, you get a switch that new would have been circa $15k and would have demonstrably better stats on every front than any of these audiophile switches could hope for.  

[Ittaku]

On 06/01/2020 at 8:02 PM, Stereophilus said:

Would the people here advising against audiophile switches also conclude that Ethernet-optical converters placed upstream of the Hifi server/streamer make no difference to the sound quality?

Speaking purely theoretically, it's entirely possible the optical receiving component at the far end generates more electrical noise and earthing issues than the wired ethernet would have.

[Guest rmpfyf]

45 minutes ago, recur said:

ha, I just looked on gumtree: https://www.gumtree.com.au/s-ad/st-james/other-electronics-computers/cisco-catalyst-3750g-switch-24-port/1236308272 

 

So for $100, you get a switch that new would have been circa $15k and would have demonstrably better stats on every front than any of these audiophile switches could hope for.  

 

That's awesome. Always keen to try something new - I'm down. Can I ask a dumb question then - is there anything similar that'll fit a 300mm deep rack?

[BugPowderDust]

Yeah, Cisco make a bunch of smaller form factor switches with 12 ports that would fit a 300mm deep rack. 

See: https://www.cisco.com/c/en/us/support/switches/catalyst-3560cx-12pc-s-switch/model.html 

 

In terms of their older stuff, a 3750G-12S could do it: https://www.cisco.com/c/en/us/support/switches/catalyst-3750g-12s-switch/model.html though it's just a tad deep (326 mm) so might sit a little proud depending on your rack.

 

The amount of gigabit ethernet L3 switches from Cisco in the used market is crazy. Every medium enterprise up used all of these switches, so plenty to pick and choose from on the used market.

Edited January 8, 2020 by recur

[BugPowderDust]

Could also look at a Juniper EX2300: https://www.juniper.net/assets/fr/fr/local/pdf/datasheets/1000579-en.pdf 

 

[Guest rmpfyf]

Thanks much @recur - i'm going to take a serious look. I have two switches here and could probably stand to replace one with something better, bonus if it supports >9 ports of PoE. 

[Guest rmpfyf]

On 31/12/2019 at 11:32 PM, deafenears said:

- if your streamer is a PC, find a NIC from the Mellanox ConnectX adapter family. Let's you tune IRQ Affinity and pin to specific CPU cores.

 

- if your streamer is a PC, also tune the TCP stack for low latency.

 

This is sage advice - affinity and latency tuning works, particularly if whatever does playback is elsewhere. 

 

A ConnectX adapter is the correct way to do fibre.

[dbastin]

On ‎6‎/‎01‎/‎2020 at 5:02 PM, Stereophilus said:

Would the people here advising against audiophile switches also conclude that Ethernet-optical converters placed upstream of the Hifi server/streamer make no difference to the sound quality?

Try reading up about Gigafoil v4 ethernet filter.  It was reviewed by Audiostream and it has quite a following ... but certainly not your basic FMC.

 

Also, someone I trust, a Waversa Dealer, said that Waversa tested FMCs and concluded that conversion both to and from optical creates noise, errors and jitter.  Gigafoil may be an exception.  And perhaps Sonore Optical Module is not as bad as generic FMCs.  The dealer concluded to avoid FMCs unless absolutely necessary.  However he did note that SOtM switches optical connections seems to have solved problems usually created by conversion to/from optical.  Hence, hi end networks feature 2 x SOtM switches connected and isolated from one another by fibre ... big bucks tho.  So some switches intended for audio isolate 2 sides of their connections (eg. EtherRegen and I think Melco, Waversa Systems WRouter).

Edited January 8, 2020 by dbastin

[Stereophilus]

4 hours ago, Ittaku said:

Speaking purely theoretically, it's entirely possible the optical receiving component at the far end generates more electrical noise and earthing issues than the wired ethernet would have.

Yes, it is possible.  No dispute there.
 

The point of my question was that if we can agree that electrical noise and earthing issues in Ethernet transmission to a hifi streamer/server can affect sound quality, then we should dispense with the whole “perfect data integrity = no affect on SQ” argument.  Data integrity, in my view, has never been in question.  The core of this thread is about what else in Ethernet streaming of music could affect SQ?

[Stereophilus]

1 hour ago, dbastin said:

Try reading up about Gigafoil v4 ethernet filter.  It was reviewed by Audiostream and

it has quite a following ... but certainly not your basic FMC.

 

Also, someone I trust, a Waversa Dealer, said that Waversa tested FMCs and concluded that conversion both to and from optical creates noise, errors and jitter.  Gigafoil may be an exception.  And perhaps Sonore Optical Module is not as bad as generic FMCs.  The dealer concluded to avoid FMCs unless absolutely necessary.  However he did note that SOtM switches optical connections seems to have solved problems usually created by conversion to/from optical.  Hence, hi end networks feature 2 x SOtM switches connected and isolated from one another by fibre ... big bucks tho.  So some switches intended for audio isolate 2 sides of their connections (eg. EtherRegen and I think Melco, Waversa Systems WRouter).

In my own somewhat basic trial with FMCs, I found adding iFi iPower power supplies to power the FMCs made a noticeable improvement in SQ.  As is the norm these days, the thread I started providing this information was shut down.

 

I think the EtherRegen device provides a much more elegant method for potentially isolating noise from inbound Ethernet.

[Assisi]

2 hours ago, Stereophilus said:

The point of my question was that if we can agree that electrical noise and earthing issues in Ethernet transmission to a hifi streamer/server can affect sound quality, then we should dispense with the whole “perfect data integrity = no affect on SQ” argument.  Data integrity, in my view, has never been in question.  The core of this thread is about what else in Ethernet streaming of music could affect SQ?

@Stereophilus

 

I have a reasonable quality pair of Ethernet to optical converters, an optical cable plus Cat 8 cables.  My perspective is that there is a definite difference with the above configuration compared to without.  What I am not sure about is whether the difference is a benefit or not.  One other person has heard my system and his comment is that the difference is not beneficial as there is a to him a very small but discernible loss of dynamics. 

 

His comment is that whenever there is a conversion happening as there is in this chain, there can be a detrimental impact.  The conversions that happen may impact the SQ.  Minimal is better than more. 

 

I thought that last week when we did the comparison that at the end of the comparison that the Ethernet/optical was removed.  I thought that I would try them again today.  Actually, they were not removed and I have been listening to them for the past 7days.  They are now removed and I now think that the SQ is probably slightly better without them.

 

It is a bit confusing.  It may be worthwhile having a small GTG somewhere to try to work out.

 

John

Edited January 8, 2020 by Assisi