Ethernet switch

[glevethan (Gregg)]

@Anupc

EMI/RFI is not magic, it’s easily measurable, and Media Converter manufacturer have to meet FCC compliance for EMI/RIFI. A typical Media Converter is many times simpler than any Ethernet Switch - it’s basically 2 Ethernet Ports/Magnetics, and 1 main chip - there’s no way it can generate more noise than the Switch itself. Coupled with a good LPSU, there’s no EMI/RFI generated that could cause any sonic degradation.

@TheFlash

The Ethernet Switch is an alternative solution. It is not the problem

$2000-4000 for an Audiophile Ethernet Switch, versus $200 for a Media Converter solution that permanently eliminates the Switch from sonic impact . You do the math

So am I spinning my wheels with my current solution?
Cisco switch w/Cisco SFP > Fibre cable > Cisco switch w/Cisco SFP > Ethernet cable to Rossini

Is the second Cisco switch introducing noise thus defeating the gains made by the preceding Fibre connection ?
Should the second Cisco switch be replaced with a Media Converter solution? Recommendations?
Thank you

Best
Gregg

[keiserrg]

Hi Bill,
What I meant was that for me, moving the discussion beyond the debate of whether the 0s and 1s are the same (now, finally, apparently settled in the affirmative) to one that focuses on possible sources and noise that can impact the equipment, and potential fixes, represents a step forward in this discussion.

Just my 2 cents, or 2 pence, if you prefer… ; )

[TheFlash (Nigel of Reiki Audio)]

@glevethan

So am I spinning my wheels with my current solution?
Cisco switch w/Cisco SFP > Fibre cable > Cisco switch w/Cisco SFP > Ethernet cable to Rossini

Is the second Cisco switch introducing noise thus defeating the gains made by the preceding Fibre connection ?
Should the second Cisco switch be replaced with a Media Converter solution? Recommendations?
Thank you

Best

For avoidance of doubt, how long is the ethernet cable to Rossini?

[jandersonhill (Jeremy)]

@Anupc

Data integrity is within the context of EMI/RFI, not in the absence of it

While EMI/RFI can (in extreme cases) cause data errors, an absence of data errors doesn’t mean an absence of noise. Noise which I believe can influence the sound, depending on the design of the DAC.

The designer of the Chord DACs and Uptone Audio equipment, John Swenson, appears to disagree with the assessment that noise from ethernet can’t enter the audio chain and he has written a white paper on this subject, which can be found here.

While it could be argued, as a designer of network switches (too), that he could have some bias, he has clearly thought this topic through.

[Urbanluthier]

For our purposes, I thought the geometry of properly certified Cat5e and Cat 6A cables ensure adequate common mode noise rejection. And isn’t ethernet galvanically isolated by design?

[jandersonhill (Jeremy)]

@Urbanluthier

I thought the geometry of properly certified Cat5e and Cat 6A cables ensure adequate common mode noise rejection

It does, but only for the purposes of making sure data can be received correctly at a specific data transmission speed, in a certain type of environment. It doesn’t make the cable immune to noise, however.

Because twisting the wires only reduces induced noise, as the data rate increases, so do the specified number of twists (per metre of cable) and you move from, say, CAT5e to CAT6 (and up).

[TheFlash (Nigel of Reiki Audio)]

@Urbanluthier

For our purposes, I thought the geometry of properly certified Cat5e and Cat 6A cables ensure adequate common mode noise rejection. And isn’t ethernet galvanically isolated by design?

Ethernet cables are not galvanically isolated by design and they don’t ensure noise rejection… but switches are; “galvanically isolated” may not equal total elimination of noise but it should do a good job of mitigating it. All an audiophile switch designer can do above and beyond this is to minimise the noise entering the case of the switch and minimise the noise generated by the switch circuitry. To use a phrase introduced here by @Anupc, it’s not magic.

[jandersonhill (Jeremy)]

@Urbanluthier

And isn’t ethernet galvanically isolated by design?

To add to my (and Nigel’s comments above), the transformers that provide galvanic isolation are used to reduce Common Mode noise, but it’s not a magic bullet that completely removes noise.

The way the isolation works is, if the induced noise is induced equally in each pair of wires transmitting a (balanced) signal, the noise should cancel itself out, but this doesn’t quite happen in practice, hence the need to increase the number of twists (in each pair of cables) as the speed goes up.

If I’ve understood Anup correctly, his argument is that this should be of no consequence as the data can still be received correctly (which is absolutely true, unless something goes very wrong such as a fault in the cable).

The counter argument is that this noise gets injected into the ground plane of a DAC (even with galvanisation), which affects the rest of the DAC circuitry (as its point of reference is no longer at the presumed 0V, but moving around).

[I would like to add that I honestly respect Anup’s deep knowledge of ethernet data transmission, we simply disagree on whether induced noise on ethernet signalling can influence a DAC. In addition, I believe dCS have put a lot of thought into this topic and their ethernet implementation is very, very impressive]

[TheFlash (Nigel of Reiki Audio)]

@jandersonhill

To add to my (and Nigel’s comments above), the transformers that provide galvanic isolation are used to reduce Common Mode noise, but it’s not a magic bullet that completely removes noise.

The way the isolation works is, if the induced noise is induced equally in each pair of wires transmitting a (balanced) signal, the noise should cancel itself out, but this doesn’t quite happen in practice, hence the need to increase the number of twists (in each pair of cables) as the speed goes up.

If I’ve understood Anup correctly, his argument is that this should be of no consequence as the data can still be received correctly (which is absolutely true, unless something goes very wrong such as a fault in the cable).

The counter argument is that this noise gets injected into the ground plane of a DAC (even with galvanisation), which affects the rest of the DAC circuitry (as its point of reference is no longer at the presumed 0V, but moving around).

[I would like to add that I honestly respect Anup’s deep knowledge of ethernet data transmission, we simply disagree on whether induced noise on ethernet signalling can influence a DAC. In addition, I believe dCS have put a lot of thought into this topic and their ethernet implementation is very, very impressive]

Basically, this ^

[glevethan (Gregg)]

@TheFlash

For avoidance of doubt, how long is the ethernet cable to Rossini?

I recall either .5m or 1m
The problem is both Cisco switches are located within .5 of a meter of the Rossini - no choice on that. Obviously the Rossini is on its own dedicated line wired directly back to where the power hits the house

[TheFlash (Nigel of Reiki Audio)]

@glevethan

So am I spinning my wheels with my current solution?
Cisco switch w/Cisco SFP > Fibre cable > Cisco switch w/Cisco SFP > Ethernet cable to Rossini

Is the second Cisco switch introducing noise thus defeating the gains made by the preceding Fibre connection ?
Should the second Cisco switch be replaced with a Media Converter solution? Recommendations?
Thank you

Best
Gregg

@glevethan

@TheFlash

For avoidance of doubt, how long is the ethernet cable to Rossini?

I recall either .5m or 1m
The problem is both Cisco switches are located within .5 of a meter of the Rossini - no choice on that. Obviously the Rossini is on its own dedicated line wired directly back to where the power hits the house

Thanks for this. So my guess is you might be spinning your wheels though (a) I’m not entirely familiar with the expression but I can guess! and (b) I/we would need more info to ascertain.

I’m not sure I’d describe the 0.5m as a problem, it sounds like more of an opportunity to simplify as I can’t help but wonder if you might have over-engineered this…

For avoidance of doubt, please can you provide some additional info on cable lengths and types, and switch connections?

Cable length/type:
Router > Cisco switch1 (copper?)
Cisco switch1 > Cisco switch2 (we know this is optical. Length)
Cisco switch2 > Rossini (we know this is a short cable and copper)

Connections
What do you have connected to your switch1 apart from the router and the optical out to switch2? Do you have anything other than the optical in and the Rossini connected to switch2?

Finally how far is it from your router to your Rossini (physically not in terms of cable length)?

Thanks

[Anupc (Anup)]

@TheFlash

You then talk about poorly designed cables as if, somehow, certified cables are immune to RFI/EMI. Spoiler: they’re not.

@jandersonhill

we simply disagree on whether induced noise on ethernet signalling can influence a DAC

The point is, they don’t have to be immune. Ethernet - both the physical layer and the networking stack as a whole - is designed to operate perfectly in the presence of all kinds of noise.

At the physical layer the differential-signalling and isolation transformer-coupling ensures noise is absolutely minimised to the point that very few bits are misinterpreted or flipped, and even when they do, there’s error checking that signals up the stack.

Yes, even certified cables are not perfect, and Ethernet ports are not perfect with parasitic capacitance all over the place, so there will be noise getting across copper, and it’s left to the Streamer/DAC designer to ensure that whatever noise that gets across does not pollute the platform enough to cause a sonic impact.

I highly recommend that you go take a very close look at how dCS have implemented their Ethernet interface; it’s port magnetics, grounding, de-coupling, network board, etc. You’ll come to realise that most of the conjecture in this thread have very little actual bearing on dCS DACs.

@TheFlash

Well here’s a thing. I talk about rational, and you talk about magic.

Here’s the thing; many forum audiophiles and audio journalists are somehow under the magical impression that only they know about noise and EMI/RFI, and that the Audio equipment manufacturers are dumb and don’t know how to mitigate it by design.

Granted not all audio equipment manufactures are smart, and some even imagine they can get Into the networking gaming without looking awfully stupid (case in point; TotalDAC’s Ethernet Switch). But I don’t count dCS among the dumb ones, and you shouldn’t too.

[Anupc (Anup)]

@glevethan

Is the second Cisco switch introducing noise thus defeating the gains made by the preceding Fibre connection ?

From an Engineering standpoint a Media Converter, inclusive of an SFP, will be significantly less noise prone than a typical Ethernet Switch, because of how simple the formers is, you just can’t cheat physics.

But, if you’re using something like a Cisco 2960-C - it’s a pretty very well designed Switch - you really have little to worry about when coupled to a dCS

[jandersonhill (Jeremy)]

@Anupc

The point is, they don’t have to be immune. Ethernet - both the physical layer and the networking stack as a whole - is designed to operate perfectly in the presence of all kinds of noise.

I’m not arguing ethernet doesn’t operate as designed - which is to transfer data with zero errors. Ethernet (the many iterations of it), however, was not designed to minimise the relatively small amounts of noise that might leak into the ground plane as, in a computer, this is rarely an issue. I believe, in audiophile applications, it can be.

@Anupc

At the physical layer the differential-signalling and isolation transformer-coupling ensures noise is absolutely minimised to the point that very few bits are misinterpreted or flipped

I have never argued on here or anywhere else that they are (in fact, I stated the opposite - the bits are intact, it’s about the ground noise injected elsewhere into the system)

@Anupc

You’ll come to realise that most of the conjecture in this thread have very little actual bearing on dCS DACs

I agree and I’ll quote from my own post you’re responding to:

@jandersonhill

…I believe dCS have put a lot of thought into this topic and their ethernet implementation is very, very impressive

I’ve been discussing whether ethernet could inject noise in general into an audiophile system. This is not the same as saying that is has a significant impact on dCS’s implementation, because I think they have done a very good job here.

@Anupc

I highly recommend that you go take a very close look at how dCS have implemented their Ethernet interface

I would love to study their circuits if someone will send them to me (the FPGA logic would be great too lol ).

[meltemi (Martin)]

It’s always the same.
Some manufacturers are capable of getting the job done extremely well, and others don’t.

[imprezap2]

Changing the power supply on my switch made significant SQ differences, switch is connected to my Rossini player. Tried four power suplies, all sounded different, dcs equipment might be designed well, but is not immune to outside influences IMO.

[TheFlash (Nigel of Reiki Audio)]

@Anupc

@TheFlash

You then talk about poorly designed cables as if, somehow, certified cables are immune to RFI/EMI. Spoiler: they’re not.

@jandersonhill

we simply disagree on whether induced noise on ethernet signalling can influence a DAC

The point is, they don’t have to be immune. Ethernet - both the physical layer and the networking stack as a whole - is designed to operate perfectly in the presence of all kinds of noise.

At the physical layer the differential-signalling and isolation transformer-coupling ensures noise is absolutely minimised to the point that very few bits are misinterpreted or flipped, and even when they do, there’s error checking that signals up the stack.

Yes, even certified cables are not perfect, and Ethernet ports are not perfect with parasitic capacitance all over the place, so there will be noise getting across copper, and it’s left to the Streamer/DAC designer to ensure that whatever noise that gets across does not pollute the platform enough to cause a sonic impact.

I highly recommend that you go take a very close look at how dCS have implemented their Ethernet interface; it’s port magnetics, grounding, de-coupling, network board, etc. You’ll come to realise that most of the conjecture in this thread have very little actual bearing on dCS DACs.

@TheFlash

Well here’s a thing. I talk about rational, and you talk about magic.

Here’s the thing; many forum audiophiles and audio journalists are somehow under the magical impression that only they know about noise and EMI/RFI, and that the Audio equipment manufacturers are dumb and don’t know how to mitigate it by design.

Granted not all audio equipment manufactures are smart, and some even imagine they can get Into the networking gaming without looking awfully stupid (case in point; TotalDAC’s Ethernet Switch ). But I don’t count dCS among the dumb ones, and you shouldn’t too.

I give up. Yet again, you’re talking about noise and its impact on “bits”. You’re the only person doing this. You appear unable or unwilling to think outside the digital domain. No-one has claimed, ever, that the noise of which we speak has any impact on the digital domain, on bits. The whole ethernet physical layer thing works as it has been designed to. You’re arguing with an imaginary opponent here.

Yet again, you’re talking about magic.

No-one is saying any manufacturer is dumb.

I own and use a dCS Puccini and U-Clock, fed until recently by a Bluesound Node 2i with high quality Sean Jacobs linear PSU and now fed by an Innuos PulseMini with pimped Innuos ZenMini linear PSU. I trust dCS implicitly but I also know what difference a switch makes in feeding my streamer. The issue switches address is not relevant only to those with cheap dodgy DACs connected via cheap non-standards-compliant cables.

[BillK]

@Anupc

From an Engineering standpoint a Media Converter, inclusive of an SFP, will be significantly less noise prone than a typical Ethernet Switch, because of how simple the formers is, you just can’t cheat physics.

But, if you’re using something like a Cisco 2960-C - it’s a pretty very well designed Switch - you really have little to worry about when coupled to a dCS

But I’m sure you realize most people are using a $20 D-:Link switch and media converters and SFPs purchased for the least amount possible on AliExpress or eBay, not equipment from name brand manufacturers.

[hifimckinney]

Without going to college to learn TCP/IP, I request all those who are still using stock power supply or some so-so LPSU with their switch/SFP to replace it with something like Farad or Sean Jacobs and hear the magic. My choice is SJ. Yes, it is like buying a $10 leash for $2 dog but something worth trying while you are in this hobby with these expensive equipment.

[glevethan (Gregg)]

Thanks for the confirmation