Installed a new switch today

Today I have finished installing a new switch in my system. A brand new, factory sealed, Cisco Catalyst 2960-L gigabit 24 ports + 4 SFPs.

It is working flawlessly from the start. So I was very curious how it would sound. The same of course as my previous one, that retired after 14 years 24/7 in labour.


It sounds better: darker background, better stage dimensions, more sparkling…

How is that possible? It cannot be the data. So what can it be?

And then I came across this:

How a Network Switch Affects Audio Playback – An Extreme Deep Dive!

It is NOT the data. In our measurements we see a direct relationship between low-frequency noise from the network port of a switch and phase noise on the clock in the streamer/ DAC. This is very well audible. As far as we are concerned, the listening tests show that switches with a lot of noise from the port sound less good.

Translated from:


…for such statements one is usually burned in this forum! :rofl:


I know :wink: and that is why I hesitated to post. But I was really surprised by the different sound coming from my Rossini Apex DAC. And then I started Googling, and found the above linked article.

I don’t want to open a can of worms. But maybe someone @Anupc ? can comment.

1 Like

Usually switches of that kind have a fan. Don’t you think than when the fan is running there is noise ?

Did you consider “audiophile” switch ? Their selling point is often the low noise capability.

My new switch is fanless. It is dead silent. No need for an audiophool switch :stuck_out_tongue:

Erno, remind me please, what switch did the new one replace? I’m a big fan of the 2960/3560 Catalyst switches.

1 Like

Hi Greg,

My old switch is a HP Procurve 1810G-24. It is also a 24 ports gigabit switch, with 2 SFP cages.

It was working fine, and I was not even on the lookout for a new one, until I came across this offer of the 2960-L. Brand new, and under € 500.

1 Like

That’s a good deal. For the switches in my home network, I use just one brand, and for the switches into the audio and video segments, it’s a 3560. I prefer a Level 3 switch for administrative consistency. I don’t get the big ones, just 8 or 12 port.

1 Like

Curious if anyone from the dCS Tech team would care to comment on this :slight_smile:

Seems to me that’s like asking if the university president wants to attend frat-house parties. :wink:


Not sure, but dCS team posted a comment on that kind of discussion in another thread. If I remember well they said that you can use whatever switch you want to, but you have to pay attention with professional switch that are “managed”…With sophisticated firmware which perform some operations that can impact either the sound or the operation of Mosaic.

Sorry for being so vague, but may be someone can find that thread and explain better than me…

1 Like

Sure, let’s have polite discussion about this :slight_smile:

I translated the article you linked to (sent you a copy to verify that I’ve not missed anything), there are a number of issues mentioned in the article, but let’s start with the most critical one;

IMHO, there’s no argument that some Ethernet Switches will be noisier than others for whatever reasons be it Power Supply or poor Engineering. So, that’s a given. Resulting in a question of whether such noise affects the DAC and how (knowing full well that it doesn’t affect PCs to the extent that no applications are affected… more on this later).

The jitter problem they allude to is however inconsequential as the physical Ethernet stream is completely terminated on an Ethernet PHY chip (immediately after the Ethernet Pulse transformer at the port); with a Qualcomm Atheros AR8030 on dCS platforms. While data is assembled from IP Packets reconstructed by a separate CPU and networking stack; the TI Sitara ARM CPU/Linux on the dCS. Both the PHY chip and the ARM CPU/Linux are closely-coupled and isolated on the S800 board.

So, the issue is just one of noise, and the key question then becomes;

(a) Can such noise affect the sound of the DAC? (Enough to be heard)
(b) Can such noise be easily mitigated (without having to spend $2,000 on an Ethernet Switch)?

The answer to (b) is simple. There are any number of ways to eliminate noise from a noisy Ethernet Switch; fibre based isolation being one. The problem domain then reduces to just the Media Converter at the far-end where one can focus on cleaning things up. Most Media converters are built on a single chip coupled with a PSU, it’s trivial to produce noise-free Ethernet to attach to the DAC. The pictures in the linked article seem to suggest they measured both the Ethernet Switch directly and with a Fibre media converter, but they don’t show any of the measurement results. Maybe I’m missing it somewhere.

As for (a), let me put this thought experiment in your mind;

Take the cheapest/crappiest PC you can find. Attach it to the noisiest nastiest Ethernet Switch you can find. If you were to download an Excel spreadsheet, would you agree there is zero chance that the spreadsheet will have any error?

So, somehow, Ethernet noise has zero effect on the cheapest nastiest PC and application, but it affects the dCS platform & application (D-to-A)? Do you actually believe that the dCS Compute board is worse than the cheapest PC? :laughing:

Back to the article you linked to; there’s no explanation or demonstration of what they measured on the Metrum Streamer/DAC with the Wavecrest - unless I’m somehow missing it - they don’t show or explain the measurements, but they jump to this critical conclusion;

Does a switch affect the clock in a streamer or dac?

Yes: 100%. We have now proven that, we hope.

However: in terms of phase noise, we find 10 dBc/Hz quite significant. In addition, the audible influence is also quite significant between a router supplied free of charge compared to a decent Netgear or Dlink.

What have they proven exactly? They didn’t show any measurements.

Also, am I mis-reading this? They’re suggesting that the noise or jitter from the physical Ethernet stream causes phase-noise on the DAC’s CLOCK?? :astonished:

Within dCS DACs, the Ethernet stage is so far removed, in terms of stages, not physical distance, from the DAC’s Clock, that frankly, it’d be ludicrous to suggest that Ethernet noise or Ethernet physical layer inter-symbol jitter can affect it. Plus, with dCS’ hyper-focus on Clocking, do you actually believe they’d engineer the DACs to allow Ethernet to affects the D-to-A Clock?? Come on!!

My question/suggestion, did you truly honestly blind test yourself when you heard the difference? :slight_smile:


Thanks Anup, I have the same doubts as you as of their statements.

They do provide full measurements, but they are behind a pay wall. Unless you donate (Patreon), they are not shown. Their investigation took several months.

I did not donate. But from posted comments they suggest this is a time domain issue. Also they seem to have demonstrated that port isolation on a switch is not 100%, and also that galvanic isolation is not.

My old switch is just under the new Cisco. After listening on the new one I noticed a difference. It is not big, I switched back and forth, I still can hear it. I estimate it as big (or small) as half that of listening with or without a Rossini Clock.

P.S. Thanks for the translation you have sent. Maybe you want to post it here, so others can read it. Again, the measurements are only available after a Patreon donation.

Yes, with a managed switch you need to know how to configure it. If you have no clue, don’t get one.

However, it is not rocket science. Just to give you an idea, here is the Easy Setup Guide of my new switch. Now you can decide if you can do this:

Cisco_Catalyst_2960-L_Easy_Setup_Guide_20161105.pdf (2.6 MB)

I believe you, I was just trying to remember what was said in that kind of thread some time ago (on this forum, obviously).

I think this one:


Ahhh. Actually, I wouldn’t mind paying to see the measured results, unfortunately the open parts of the article doesn’t read with much credibility. For example, they totally lost me when they quoted Hans Beekhuyzen as a source of information on Jitter. That’s a joke. If you’re going to quote someone about understanding Jitter in the context of Audio systems, then there’s no one better than Julian Dunn.

I briefly look at the comments section. While it’s true that galvanic isolation is not 100%, the issue is parasitic-capacitance which is easily mitigated with a proper Pulse transformer. Jaap is suggesting that the Ethernet noise can “penetrate” the Pulse Transformer, the PHY chip, the ARM CPU (with it’s networking stack and the RAM buffers), all the way through to the D-to-A stage… Honestly, that’s quite laughable :laughing:

By the way, this one particular comment stood out.

Michael wrote:
11/02/2023 AT 12:38

What I wonder is if I remove the internet plug from the switch, the streamer continues to play and I hear no difference in sound change.

How about now exactly?

And Jaap’s response;

Jaap wrote:
11/02/2023 AT 12:50
Hi Michael,

It’s not that easy to compare. You empty a buffer that already includes the effects of the switch. You could compare directly in the router and via the switch where you restart the track. Then you could notice differences.

On the one hand he says data integrity is 100%, but on the other hand he’s saying that the data in the buffers are affected; that tells me he has no idea how Ethernet & TCP/IP works.

If what he says is true and Ethernet Switches have that kind of impact, then there’s absolutely no way the dCS Firmware update could possibly download successful! :rofl:

I would still recommend you do an honest blind test on yourself - you might be surprised that you can no longer tell the difference. The mind is a tricky thing, it’s easily fooled.

Translated version of the article attached (hopefully I’m not violating any Copyright in posting the translated piece, original article URL included :grin:);

How a Network Switch Affects Audio Playback - An Extreme Deep Dive (no pics).pdf (419.3 KB)


That’s a good post well worth reading, in particular this excerpt:


as always thanks for a great read and explanation.

Jaap measured the influence of a switch powersupply at a DAC’s clock, seems to me it’s not all “voodoo”

I have purchased another powersupply to give it a try on my Melco S100 switch, an Ifi elite smps this time, again the sq differences are there compared to the Plixir lineair power supply I was using.

Just give it a try if you like tweaking the streaming side of your system.