Do I really need an " Audio Grade Network Switch "?

[cjf]

Good post but I think that having IT experience here could be a partial handicap in a way if it prevents you from trying these network based tweaks. Noise is a huge deal in a system and many things can generate it. But I also think you make a good point about wall warts. They can contribute a lot of noise in my experience.

The great thing about your background is that it could help us focus in on more likely culprits.

I continue to have new experiences that show that everything in the chain matters.

I can think of only a few things that might offer better SQ in terms of the network infrastructure used to deliver content to ones audio system.

1. The use of an optical break (ie..Fiber, Fiber media converter) immediately after the ISP modem. Basically sitting between the main WAN Port on your Firewall/Router and the ISP modem. I do this today, but not really for any SQ related reasons. My main reason is to shield the Internal network equipment and Internal power grid from surges that can enter over the coax cable leading into the house (I use cable internet but DSL would be the same story). As a secondary reason, this can provide an an electrical break from outside world noise found on this line. There are obviously plenty of other ways for outside world noise to enter ones home but eliminating even one of those many paths for little cash outlay and hassle is hard to deny or ignore.

2. If you can get away with it, depending on the audio equipment being used, I think not using any switch at all between your music server and the DAC is a better choice than using an "Audiophile Switch". This would assume the DAC offers an Ethernet Input of course. I'm talking about having a "Home Run" Ethernet connection from a music server having more than one NIC card to the DAC directly. I do this today in my own audio system.

NIC #1 connects to your main switch/home network
NIC #2 connects directly to the DAC via the "Home Run" Ethernet connection

NIC #2 would utilize APIPA (Automatic Private IP Addressing) to supply an IP Address for its own NIC and also it would provide the IP Address for the DAC (assuming the endpoint is set for DHCP). Both would end up with a 169.254.x.x IP address and can speak to each other directly.

The way I see it, at the end of the day, it comes down to how well implemented the Ethernet Input is on the DAC side. If its sensitive to electrical noise propagation over the Ethernet cable then the DAC has other problems these tweaks are compensating for but in terms of the data stream itself (ie..the 1's & 0's) its impossible (as far as I'm aware) for this feared noise/grunge to inject itself into the data stream without also corrupting or changing the 1's & 0's that make up that original content. In that case, the stream just wouldn't work or it would be painfully obvious that something is wrong.

So IMO, unless there is a pre-existing issue in the system, the use of an "Audiophile Switch" or high-end Ethernet cable offers little benefit.

 

[Blackmorec]

The final 4 steps in a computer audio chain are:
conversion of the bit stream into a continuous voltage to represent the music (DAC)
conversion of the continuous voltage into sound pressure waves (amp and speakers)
conversion of the sound pressure waves to nerve impulses (ears)
conversion of nerve impulses to a conscious awareness of a rhythmic set of frequencies we recognise as music.

The problem with this audio networking discussions is the way in which we measure the results. In networking we have a number of different ways to characterise/measure our success in moving a bit stream from one place to another, but in audio, the final two steps of the process can‘t be measured in the conventional sense, but seem to be exquisitely sensitive to what happened across the rest of the network.

On its way to the DAC the bitstream passes through a number of devices which retime and resynthesise the bitstream. The bitstream may be converted to radio signals or light pulses then back to voltage. All these conversions require clocks (oscillators) and DC power supplies and each of these steps creates an indelible imprint on the music. Each of those conversions has a specification that defines ‘perfect’ and the further from perfect those conversions are, the greater the magnitude of the imprint It leaves. Deviations from ‘perfect’ include
Any noise or ripple of the DC power used
high impedance on the DC power supply
Any vibration or noise that affect the timing accuracy of the oscillator
the overall quality and repeatability of the oscillator
Any HF noise that finds its way into the data stream
CPU and IC noise
interruptions from non-music related network traffic

Despite our best efforts to preserve that music signal across the whole chain, deviations from ‘perfect’ in all the conversions leave an imprint that we can hear. The closer to perfect those conversions become the bigger the impact of any deviations.
The great thing about digital audio is that we can ‘repair’ the data stream, so deviations from ’perfect’ early in the stream can be corrected later in the stream and the sonic imprints left by those early flawed conversions removed. That’s exactly what switches do. By providing more accurate oscillators, emi filtration, superior isolation from vibration and upstream electrical noise, superior DC power with less noise and lower impedance etc. we can produce a bit stream that when converted by the DAC, ears and brain sounds more like what we recognise as music happening in a venue. The more you clean and refine the data stream, the higher the perceived quality of the music. Network experts may report that they are unable to detect any difference in the actual bitstream therefore no changes to the final sound are possible but thousands of audiophiles report hearing major differences so the obvious conclusion is that we are, as yet either unable to measure the changes or to correlate certain changes with what we hear. quite possibly, i our hearing is a lot more sensitive than our measurements.

Bottom line? In highly resolving systems, switches used to refine bitstream quality can have a majority impact on our interpretation of music.

 

[John T]

Great explanation of what switches actually do and what they attempt to accomplish. Well done. Just because something cannot be measured doesn't always translate to no "sonic differences".

 

[Blackmorec]

Great explanation of what switches actually do and what they attempt to accomplish. Well done. Just because something cannot be measured doesn't always translate to no "sonic differences".

There’s logic to stand behind that statement. The last 2 processes in the audio chain happen in your ears and the final process, the delivery of the results so to speak, happens in your head. We are able to understand the process in the ears but we’re a long way off fully understanding how nerve impulses are converted into concious awareness and appreciation of music, so we’re a long way from measuring anything about the process. We simply can‘t measure the final conversion and delivery, so any changes we sense there are, by definition unmeasurable and subjective is the only type of measure possible. As we dont fully understand and cant measure anything, its very difficult to know exactly what affects the process and why. What we can do however is evaluate any changes in the sound subjectively then look at what we changed for that subjective shift to happen and record that. When enough people report the exact same change and effect, you basically have something that is ’indicated by concensus’ rather than measured. That’s as good as it gets for any changes related to the last two steps in audio

 

[John T]

It really is a fascinating aspect of sonic interpretation. One could break it down further. Brain neurons collect, decipher, in a variety of pattern. Individual responses, retrieve and collect differently. Hence as you said 'indicated concensus' will have to do.

 

[Phantom-Audio]

On a few Dacs iv tried to play around with the Network Audio Switches made a difference, However on the Playback Design MPD Series it never so i think it system dependent too or how the DACS/Streamers are handling the incoming data.

 

[Mark38]

It's been my experience that adding a switch, or multiple switches offers a very nice improvement. They don't have to be audiophile switches. The Netgear gs105 is a solid piece with a very wide soundstage. The Netgear gs110mx is a more solid unit with a copper shielding. The Bonn is a solid unit to begin with, add two of the 105 or the 110 and you have seriously elevated the experience. Each additional unit brings increased presence and weight. LPS is a must. Three is the sweet spot for my system.

It's difficult to overstate how much of a steal the gs105 is.

Meraki very disappointing.

 

[antigrunge]

@Blackmorec,

This is the most elucidating commentary I have yet read about engineering measurements and audiophile hearing in digital audio. I believe that timing accuracy and noise elimination at every step of the chain are the prerequisites for good performance. One would though think that with a bit of engineering genius ther‘d be a better way than just daisy chaining a bunch of switches…and as far as understanding what‘s really going on is concerned: we are only,at the beginning of understanding what it takes to transmit the bitstream in audio quality via digital media, be they cables or converters

 

[Blackmorec]

@Blackmorec,

This is the most elucidating commentary I have yet read about engineering measurements and audiophile hearing in digital audio. I believe that timing accuracy and noise elimination at every step of the chain are the prerequisites for good performance. One would though think that with a bit of engineering genius ther‘d be a better way than just daisy chaining a bunch of switches…and as far as understanding what‘s really going on is concerned: we are only,at the beginning of understanding what it takes to transmit the bitstream in audio quality via digital media, be they cables or converters

Hi there antigrunge,
Thanks for the kind comments….appreciated. I completely agree and in fact I believe we do have at least one piece of electrical, audio and SW engineering genius that requires very little enhancement of the Network to sound great and that’s the Taiko Extreme server with TAS SW (soon to be superseded). But that’s not to say that it can‘t get better with some network optimization. To this end Taiko is working on a switch and router to supercharge the SQ.

 

[antigrunge]

Well, InnuOS is on a similar track; why one though should need to buy separate clocks ( PhoenixNET, PhoenixUSB, Server, Dac) however escapes me. There seems to be a more logical approach in using one clock for every step including the switch, but so far only DCS and MSB seem to think along those lines…At a minimum high end should provide inputs for a master clock

I also don‘t understand why you‘d want a separate preamp if your DAC‘s analogue section is state of the art and only requires an attenuator to act as such. As this industry matures surely the number of boxes (and required cable connections) will drastically reduce.

 

[Blackmorec]

Well, InnuOS is on a similar track; why one though should need to buy separate clocks ( PhoenixNET, PhoenixUSB, Server, Dac) however escapes me. There seems to be a more logical approach in using one clock for every step including the switch, but so far only DCS and MSB seem to think along those lines…At a minimum high end should provide inputs for a master clock

I also don‘t understand why you‘d want a separate preamp if your DAC‘s analogue section is state of the art and only requires an attenuator to act as such. As this industry matures surely the number of boxes (and required cable connections) will drastically reduce.

Hi there,
Innuos uses built-in clocks and power supplies in order to achieve full DC power/ clock/ switch chip ‘intimacy’. This removes all DC and clock cables and connectors and several boxes. In terms of the HW you mentioned, this means 4 clock cables, 4 separate LPSs and ca 9 DC cables.

As far as the rest of my system goes, I follow the same ‘intimacy’ strategy with a Devialet 440, which places DAC, phono amp, pre-amp, and 2 amps in a similar highly intimate arrangement.

At absolute maximum gain, my system is still perfectly silent, so all the volume control does is to increase the perceived signal to noise ratio

 

[Birdwatcher]

Even if you cannot tweak every little part of the system on your own every week/month again and again, system integration is in my opinion highly welcome and I like when the system is minimalized, it is really sexy. I think this very sophisticated audiophile network tweaking is quite new, many small tweaks are done in the DIY way, but not many complete one-box solutions are commercially available.
Innuos Phoenix Net is such a box.

 

[skids929]

Do you "need" it no, you can survive just fine without it..Will it benefit your system? I am in the camp of absolutely it will (assuming the gear can reveal it). I bought a switch after agonizing over the prices of the switches users on this forum spoke highly of. Like literally $4-5k on a switch! No thanks..I found this fella in Vietnam who makes switches based off the Buffalo (same as what Melco uses) and a DC4 power supply design. These were no where near the cost of the Pink Faun clocked switches, of the Telgartner m12 gold (which is the Accuphase of network switches, totally over priced for what it is).

I'd say if you are considering network switches the Innuous is a good candidate they pop up used quite a bit. Otherwise I'd say consider HNE Labs, build quality is incredible and they really clean up the sound coming off the network. I like my switch so much I bought a second one to layer them. His new design is based off Tesla coils (like Ansuz). 1st Picture attached shows a preliminary design, mine will have the coils all on one PCB. Second picture is the existing switch I am currently using that HNE sells, only difference is mine was built with Toroidy in lieu of Talema.

 

[Kingrex]

Yes you want an audio grade switch. You do not need any sort of optical isolator unless you have a ground loop issue which I have never run into with a DAC.

 

[rDin]

You do not need any sort of optical isolator unless you have a ground loop

My experience, and that of others here, is that optical (SFP), even without ground loops, is very beneficial to sound quality. And so, worth investigating.

 

[skids929]

My experience, and that of others here, is that optical (SFP), even without ground loops, is very beneficial to sound quality. And so, worth investigating.

I couldn’t agree with this more. My system is as quiet as can be and the fiber made a noticeable difference.

 

[Kingrex]

I could not agree less. I have a Gigafoil and an EN30. Both create a quiet. A "Black" background. Both add a subtle compression or loss of life. I have taken the Gigafoil to customers houses and they find the sound to be a bit more Hifi. I would agree.

If you want them, I am happy to sell them. I find no use for them.

Set your electrical power up correct. Set your network infrastructure up correct. Get a Farad or some such LPS to feed the modem and separate router. Dont use an all in one modem router. Get separates. Get an Audiophile grade switch like an Etherregen. Use cat 6 and terminate with Belden Revconect jacks.

Spend time setting up your speakers and room. This is more impactful than you might imagine. If you can not take the time to set up your speakers, don't throw money at your system. Enjoy it. If you have latitude to dial in your speakers, get the Jim Smith "Get Better Sound" DVD and dial them in. After you understand what "Tunning" your stereo is about, start messing with additional components to shape the overall system sound.

Thats my recommendation.
Rex

 

[skids929]

ENO is passive, and the implementaion of the GF is different than what I have set up. GF sounds great and never once heard it sounds “hifi” from any end users, thats a new one. Your recommendation has him doing something other than what he asked and buying a bunch of power supplies and random boxes.

Get a one and done switch, it matters, if it offer fiber decoupling built in go for it. No need to go down a rabbit hole with this watching DVDs and setting up speakers doing things completely unrelated to what he asked about.

 

[Kingrex]

I am simply making the point, if you want to improve system performance, speaker setup is critical. The OP has very nice equipment. Magico/Pass/Everest. A switch is a tweek. Speaker setup is transformative.

I would get a good audio grade switch as the OP requested. But I would also look at the modem/router and associated power supplies to get the most from streaming. I have experienced streaming as a system. Its a bunch of separate parts that make up a whole. Its not just a switch or DAC. Its many parts working together. Like a TT. A TT needs a good cart, arm, table, preamp, clean records. Its not a one and done. And a TT will perform much better if the speakers are set up correct in the room.

One of the best digital systems I heard has the switch on a dedicated platform with stillpoint under it. I believe he pays as much attention to the router and modem. And he does not skimp on the power that feeds all that equipment.
I know you would hear an AQ conditioner feeding the PS's that feed said modem, router and switch. I have an isolation transforner feeding the 3 x PS that feed my data utility infrastructure. And it all better all be on the same electrical phase with as much attention paid to the circuit feeding that gear as you put into feeding the rest of your system.

I have seen quite a few isolators come and go between a good switch and server. I don't find them a benefit. There are a lot of other details to pay attention too. IMO.

I do like the idea of an isolator between a coax utility cable and the modem. It's a great piece of lightning protection. It has no sonic benefit and will drag your download speed down. But it may save your equipment if your in a high lightining zone.

 

[Alrainbow]

For starters I’m not a fan of optical conversion for data steaming unless it’s made for the purpose who makes the server or streamer
data transfer is very complex and I’ll admit above my pay grade in some ways.
but any optical used must use a conversion method
this means an Input device and transmit device per side
this uses power and lends itself to many types of noise
then the same procedure on the other end
if the goal is electrical isolation use an emo systems medical grade
not much money and true galvonic iso rated at 5kv or more
very low loss and it’s non software intensive
any sw , router not only must covert but also be part of the sampling reconstruction process with error correction.
a made by Maker I feel is the only way an optical can be better as all of the chain is on them to be best