Networking Logic - Intro

Cable Monkey

Moderator
Staff member
May 16, 2006
9,069
2,089
158
Birmingham, UK
AKA
Henry
HiFi Trade?
  1. No
As you imply a high quality Cat 5e cable (Connectix or Blue Jeans Cables) between switch and music renderer will galvanically isolate the hifi from the network. Switches regenerate traffic so a high quality one (e.g. HP, Cisco) will minimise noise. As long as the network is stable I don’t see or hear it as an issue.
Galvanic noise isolation is not absolute. It offers a degree of noise attenuation and what it offers can vary depending on the quality of the magnetics used amongst other things. Data integrity isn't the issue in these instances. Data can exist uncorrupted in very noisy environments. But your average £25 five or eight port SOHO switch was never designed to have something as sensitive as a streaming DAC with a noise floor of -130dB in the analogue domain. This sort of device is hugely disrupting because they simply didn't exist when the standards were drawn up.

I think, if the standards were re-drawn with things like the afore mentioned DAC or 4/8k UHD TV's it would look very different. But they won't be because for the very first time since it's inception we have WiFi fit for purpose. WiFi 6 fixes everything!
 
  • Like
Reactions: JustinTime

tuga

. . .
Wammer
Aug 17, 2007
14,342
7,001
173
Oxen's ford, UK
AKA
Ricardo
HiFi Trade?
  1. No
Galvanic noise isolation is not absolute. It offers a degree of noise attenuation and what it offers can vary depending on the quality of the magnetics used amongst other things. Data integrity isn't the issue in these instances. Data can exist uncorrupted in very noisy environments. But your average £25 five or eight port SOHO switch was never designed to have something as sensitive as a streaming DAC with a noise floor of -130dB in the analogue domain. This sort of device is hugely disrupting because they simply didn't exist when the standards were drawn up.

I think, if the standards were re-drawn with things like the afore mentioned DAC or 4/8k UHD TV's it would look very different. But they won't be because for the very first time since it's inception we have WiFi fit for purpose. WiFi 6 fixes everything!

The Wi-Fi interface will also generate noise. Ideally you'll wan't it oustide of the DAC and possibly filter the data coming out of it, which is why I have the Wi-Fi connected to a switch and a network bridge between before the DAC.

The ideal DAC would take the digital stream straight into the Switched Capacitor Filter (the D/A stage).
AKM's latest D/A chip allows you to do that but most chips will perform DSP before converting the signal, some (ESS) quite a lot of it, and processing generates noise.
Taking this into account, having the over- or up-sampling and modulation performed in a external box is actually a good idea, the problem is that you will need to use an interface to link that box to the DAC and interfaces can be troublesome, as evidenced by the MScaler's (jitter) measurements, particularly when dealing with high sample rates.

My view is that a dedicated network bridge (copper Ethernet or optical) buffering the digital stream and feeding it to the DAC is the most effective way to deal with this issue. I have used a Raspberry Pi, a CuBox-i and currently a microRendu for that purpose (with slight improvements between them).
 

Blackmorec

Newbie
Wammer
Feb 6, 2019
90
65
23
Thanks for confirming. Exactly the same clock then, but doing two different jobs in the two different machines operating in two different domains on two different kinds of data, only one of which is relevant to sound quality. 👍
Hi Justin,
Here’s a very specific definition from Innuos regarding the PhoenixNET vs the Statement. It’s exactly why I used both in my system.

The STATEMENT basically has all the PhoenixUSB technology built in, the PhoenixNET is actually an evolution beyond the Ethernet stage of the STATEMENT. The key differences would be as follows;

  • Isolation Transformers; a pair of these are found on the STATEMENT RJ45 ports, but the PhoenixNET uses much larger ones for better noise isolation.
  • Dedicated Clock; the STATEMENT shares its clock, whereas the PhoenixNET gets the benefit of the 3ppb OCXO being completely devoted, and is mounted straight to the mainboard for maximum accuracy.
  • Shared board vs bespoke PCB; the PhoenixNET uses a very high quality PCB board carrying the ports which we build ourselves specifically for this purpose, whereas the STATEMENT port is seated on the main overall motherboard.
  • Gigabit vs 100mb; Curiously a slower passthrough speed is actually better for audio which is why the PhoenixNET is specced at 100mb, whereas the STATEMENT Ethernet is Gigabit which, while faster, is a little noisier.
  • Socket gauge; the PhoenixNET uses very heavy duty RJ45 sockets with no disruptive LEDs at all, whilst the STATEMENT port is slightly more typical and does have LED indicators.

For all these reasons, the PhoenixNET does still represent a very decent upgrade to existing STATEMENTs, especially if frequently using streaming services or if you have another streamer in your system such as Linn or Naim etc.

Please do let us know if you require any further assistance."
 

tuga

. . .
Wammer
Aug 17, 2007
14,342
7,001
173
Oxen's ford, UK
AKA
Ricardo
HiFi Trade?
  1. No
Hi Justin,
Here’s a very specific definition from Innuos regarding the PhoenixNET vs the Statement. It’s exactly why I used both in my system.

The STATEMENT basically has all the PhoenixUSB technology built in, the PhoenixNET is actually an evolution beyond the Ethernet stage of the STATEMENT. The key differences would be as follows;

  • Isolation Transformers; a pair of these are found on the STATEMENT RJ45 ports, but the PhoenixNET uses much larger ones for better noise isolation.
  • Dedicated Clock; the STATEMENT shares its clock, whereas the PhoenixNET gets the benefit of the 3ppb OCXO being completely devoted, and is mounted straight to the mainboard for maximum accuracy.
  • Shared board vs bespoke PCB; the PhoenixNET uses a very high quality PCB board carrying the ports which we build ourselves specifically for this purpose, whereas the STATEMENT port is seated on the main overall motherboard.
  • Gigabit vs 100mb; Curiously a slower passthrough speed is actually better for audio which is why the PhoenixNET is specced at 100mb, whereas the STATEMENT Ethernet is Gigabit which, while faster, is a little noisier.
  • Socket gauge; the PhoenixNET uses very heavy duty RJ45 sockets with no disruptive LEDs at all, whilst the STATEMENT port is slightly more typical and does have LED indicators.

For all these reasons, the PhoenixNET does still represent a very decent upgrade to existing STATEMENTs, especially if frequently using streaming services or if you have another streamer in your system such as Linn or Naim etc.

Please do let us know if you require any further assistance."

Innuos' approach of dealing with all potential issues is laudable but that doesn't mean all or any of those issues will actually impact the digital playback system. This approach is known as over-engineering.
As is often the case in audio, their marketing blabber uses Fear, Uncertainty and Doubt to make their product seem special...
 

tuga

. . .
Wammer
Aug 17, 2007
14,342
7,001
173
Oxen's ford, UK
AKA
Ricardo
HiFi Trade?
  1. No
  • Isolation Transformers; a pair of these are found on the STATEMENT RJ45 ports, but the PhoenixNET uses much larger ones for better noise isolation.

Likely to help.
Alternatively you can use an Ethernet isolator in the last Ethernet leg.

  • Dedicated Clock; the STATEMENT shares its clock, whereas the PhoenixNET gets the benefit of the 3ppb OCXO being completely devoted, and is mounted straight to the mainboard for maximum accuracy.

Useless in this particular application (data transfer).

  • Shared board vs bespoke PCB; the PhoenixNET uses a very high quality PCB board carrying the ports which we build ourselves specifically for this purpose, whereas the STATEMENT port is seated on the main overall motherboard.

Useless in this particular application (data transfer).

  • Gigabit vs 100mb; Curiously a slower passthrough speed is actually better for audio which is why the PhoenixNET is specced at 100mb, whereas the STATEMENT Ethernet is Gigabit which, while faster, is a little noisier.

Bold claim, needs evidence for the particular application (data transfer).

  • Socket gauge; the PhoenixNET uses very heavy duty RJ45 sockets with no disruptive LEDs at all, whilst the STATEMENT port is slightly more typical and does have LED indicators.

Again a claim which needs evidence for the particular application (data transfer).

Same with vibration control.


As Mr. Ping famously said in Kung Fu Panda, "To make something special, you just have to believe it's special."
 
Last edited:

Cable Monkey

Moderator
Staff member
May 16, 2006
9,069
2,089
158
Birmingham, UK
AKA
Henry
HiFi Trade?
  1. No
The Wi-Fi interface will also generate noise. Ideally you'll wan't it oustide of the DAC and possibly filter the data coming out of it, which is why I have the Wi-Fi connected to a switch and a network bridge between before the DAC.
I use internal WiFi for my various AV devices but I do what you do for my HiFi setup.
 
  • Like
Reactions: tuga

Cable Monkey

Moderator
Staff member
May 16, 2006
9,069
2,089
158
Birmingham, UK
AKA
Henry
HiFi Trade?
  1. No
As Mr. Ping famously said in Kung Fu Panda, "To make something special, you just have to believe it's special."
If the device works in its totality then breaking matters down to LEDs and connector quality is irrelevant. Nothing is better because of the quality of connections used. It is better because you can connect and disconnect 100,000 times without failure instead of 10,000 times. That is what that buys you. However it looks great and generates confidence so it does serve a purpose though more marketing than overall performance.
 

JustinTime

Wammer Plus
Wammer Plus
Jan 18, 2023
699
1,397
148
HiFi Trade?
  1. No
@tuga hits the nail on the head.

As per @Blackmorec's Innuos quote, the PhoenixNET might indeed add something to what the Statement does but I'm not convinced it's anything to do with the attributes they mention. They operate (optimise?) in different domains.
Note in particular the clock description: "Dedicated Clock; the STATEMENT shares its clock, whereas the PhoenixNET gets the benefit of the 3ppb OCXO being completely devoted, and is mounted straight to the mainboard for maximum accuracy." It's maximally accurate, but there is no reference here to the impact on sound quality... because there isn't any.

LEDs can be noisy so avoiding their use, unless they tell us something really useful, is probably good practice.
 
  • Like
Reactions: tuga

Blackmorec

Newbie
Wammer
Feb 6, 2019
90
65
23
@tuga hits the nail on the head.

As per @Blackmorec's Innuos quote, the PhoenixNET might indeed add something to what the Statement does but I'm not convinced it's anything to do with the attributes they mention. They operate (optimise?) in different domains.
Note in particular the clock description: "Dedicated Clock; the STATEMENT shares its clock, whereas the PhoenixNET gets the benefit of the 3ppb OCXO being completely devoted, and is mounted straight to the mainboard for maximum accuracy." It's maximally accurate, but there is no reference here to the impact on sound quality... because there isn't any.

LEDs can be noisy so avoiding their use, unless they tell us something really useful, is probably good practice.
Hi Justin, one is reclocking the Ethernet in a switch, the other is reclocking the ethernet in a server. They are both reclocking an ethernet stream. Innuos did a lot of work to evaluate the impact of the PhoenixNET as the front end to a Statement in a variety of different systems. I was one of those who received a fully run-in pre production prototype to try. I was impressed enough to order one. It makes a very positive and substantial difference. And it’s really very simple why. Better in = better out. The PhoenixNET provides a superior input to the Statement, which in turn generates a superior output.
 

Blackmorec

Newbie
Wammer
Feb 6, 2019
90
65
23
If the device works in its totality then breaking matters down to LEDs and connector quality is irrelevant. Nothing is better because of the quality of connections used. It is better because you can connect and disconnect 100,000 times without failure instead of 10,000 times. That is what that buys you. However it looks great and generates confidence so it does serve a purpose though more marketing than overall performance.
No, its not better because you can connect and disconnect it 10,000 times without failure. It’s better because each time you connect it it makes a better connection. It’s why cleaning plugs and sockets has a positive impact on sound quality. It’s why superior contract pressure is important on good quality mains sockets. It’s why connectors are gold plated to stop oxidation. Taking out LEDs is simply removing a source of noise….in digital and analog, the less noise the better, just for different reasons. In analog you hear the noise directly. In digital the noise disturbs the signal processing, so you hear the noise indirectly
 
Last edited:

tuga

. . .
Wammer
Aug 17, 2007
14,342
7,001
173
Oxen's ford, UK
AKA
Ricardo
HiFi Trade?
  1. No
Hi Justin, one is reclocking the Ethernet in a switch, the other is reclocking the ethernet in a server. They are both reclocking an ethernet stream. Innuos did a lot of work to evaluate the impact of the PhoenixNET as the front end to a Statement in a variety of different systems. I was one of those who received a fully run-in pre production prototype to try. I was impressed enough to order one. It makes a very positive and substantial difference. And it’s really very simple why. Better in = better out. The PhoenixNET provides a superior input to the Statement, which in turn generates a superior output.

You cannot improve (the quality of) the data so the better 'Ethernet' clock makes absolutely no difference.
 

tuga

. . .
Wammer
Aug 17, 2007
14,342
7,001
173
Oxen's ford, UK
AKA
Ricardo
HiFi Trade?
  1. No
No, its not better because you can connect and disconnect it 10,000 times without failure. It’s better because each time you connect it it makes a better connection. It’s why cleaning plugs and sockets has a positive impact on sound quality. It’s why superior contract pressure is important on good quality mains sockets. It’s why connectors are gold plated to stop oxidation. Taking out LEDs is simply removing a source of noise….in digital and analog, the less noise the better, just for different reasons. In analog you hear the noise directly. In digital the noise disturbs the signal processing, so you hear the noise indirectly
Again the switch deals with data: if a run-of-the-mill switch connector does not produce lost packets a better connection is not needed.
It doesn't hurt to remove the LEDs but I'd still like to see evidence that it makes a difference.
 
  • Like
Reactions: Nestor Turton

JustinTime

Wammer Plus
Wammer Plus
Jan 18, 2023
699
1,397
148
HiFi Trade?
  1. No
Hi Justin, one is reclocking the Ethernet in a switch, the other is reclocking the ethernet in a server. They are both reclocking an ethernet stream. Innuos did a lot of work to evaluate the impact of the PhoenixNET as the front end to a Statement in a variety of different systems. I was one of those who received a fully run-in pre production prototype to try. I was impressed enough to order one. It makes a very positive and substantial difference. And it’s really very simple why. Better in = better out. The PhoenixNET provides a superior input to the Statement, which in turn generates a superior output.
No. The Statement clock is driving the digital stream, the PhoenixNET is clocking the ethernet stream. Unless you have evidence to the contrary: a link for example. Sorry to be persistent but it’s an important differentiation.
 

JustinTime

Wammer Plus
Wammer Plus
Jan 18, 2023
699
1,397
148
HiFi Trade?
  1. No
Again the switch deals with data: if a run-of-the-mill switch connector does not produce lost packets a better connection is not needed.
It doesn't hurt to remove the LEDs but I'd still like to see evidence that it makes a difference.
LEDs can, apparently, be noisy. You won’t find measurable evidence, I would wager. But why use LEDs anyway?
 
  • Like
Reactions: tuga

JustinTime

Wammer Plus
Wammer Plus
Jan 18, 2023
699
1,397
148
HiFi Trade?
  1. No
In analog you hear the noise directly. In digital the noise disturbs the signal processing, so you hear the noise indirectly
What do you mean, specifically, by “disturbs the signal processing”? You appear to suggest here that there is a thing called analog(ue) noise and a thing called digital noise. I think the first is otherwise called RFI and the only thing I can imagine might constitute the second is jitter. Is this what you mean? Jitter is of course not an issue with ethernet data.
 
  • Like
Reactions: Nestor Turton

Blackmorec

Newbie
Wammer
Feb 6, 2019
90
65
23
What do you mean, specifically, by “disturbs the signal processing”? You appear to suggest here that there is a thing called analog(ue) noise and a thing called digital noise. I think the first is otherwise called RFI and the only thing I can imagine might constitute the second is jitter. Is this what you mean? Jitter is of course not an issue with ethernet data.
Analogue noise is the noise added to an analogue signal. Hum is analogue noise, hiss is analogue noise, rumble is analogue noise. You can hear those examples of analogue noise because they’re in the audible frequency spectrum. Digital noise is noise you can’t hear because it’s far too high frequency. It’s in the frequency range in which digital electronics operate. Jitter is noise in the timing domain seen as variations from the correct timing. Jitter causes phase noise, noise in the frequency domain, which you can hear. EMI is radiated or conducted noise of any frequency, RFI is a subset of EMI and is noise in the radio frequency band, ie noise capable of being transmitted.
Sources of RFI and EMI include cell phones, microwaves, switch mode power supplies, computer circuits, Wi-Fi etc. You can’t hear RFI emitted by a microwave because its too high frequency but try to receive a wi-fi signal while sitting close to a running microwave and the whole thing either stops of just stutters along. That’s because the microwave RFI is causing huge problems with the digital processing such that error rates go up massively and data needs to be re-transmitted multiple times, causing huge amounts of processor noise, power supply noise and latency, all of which affect what you hear.

When you remove or better avoid jitter, EMI and RFI and when your digital bit square wave is fast, accurate and free of noise, when processors aren’t being interrupted, when error correction isn’t working overtime, when processors can execute without latency the resulting music sounds better. That means that when your clocking is more accurate, the level of network traffic is low, when there’s fewer demands on CPUs, when power supplies have very low noise and ripple and plenty of headroom, the music that results from a digital stream is easier for your brain to process
 
Last edited:

Cable Monkey

Moderator
Staff member
May 16, 2006
9,069
2,089
158
Birmingham, UK
AKA
Henry
HiFi Trade?
  1. No
No, its not better because you can connect and disconnect it 10,000 times without failure. It’s better because each time you connect it it makes a better connection. It’s why cleaning plugs and sockets has a positive impact on sound quality. It’s why superior contract pressure is important on good quality mains sockets. It’s why connectors are gold plated to stop oxidation. Taking out LEDs is simply removing a source of noise….in digital and analog, the less noise the better, just for different reasons. In analog you hear the noise directly. In digital the noise disturbs the signal processing, so you hear the noise indirectly
Sorry to disappoint but switches and connectors are graded by the number of times they can be operated reliably. A consequence of that may be better contact pressure when new but it was never the point of the exercise. Nothing used in electronics is designed to meet the needs of HiFi enthusiasts. It is designed for other uses and enterprising souls figure out their advantages and supply them to us at a premium! Often a very hefty premium. I was a Quality Assurance Officer testing for BT and the P.O. in a previous life, I knew this stuff backwards. Many of the tweaks that assist in our hobby are happy coincidence.
 

JustinTime

Wammer Plus
Wammer Plus
Jan 18, 2023
699
1,397
148
HiFi Trade?
  1. No
Analogue noise is the noise added to an analogue signal. Hum is analogue noise, hiss is analogue noise, rumble is analogue noise. You can hear those examples of analogue noise because they’re in the audible frequency spectrum. Digital noise is noise you can’t hear because it’s far too high frequency. It’s in the frequency range in which digital electronics operate. Jitter is noise in the timing domain seen as variations from the correct timing. Jitter causes phase noise, noise in the frequency domain, which you can hear. EMI is radiated or conducted noise of any frequency, RFI is a subset of EMI and is noise in the radio frequency band, ie noise capable of being transmitted.
Sources of RFI and EMI include cell phones, microwaves, switch mode power supplies, computer circuits, Wi-Fi etc. You can’t hear RFI emitted by a microwave because its too high frequency but try to receive a wi-fi signal while sitting close to a running microwave and the whole thing either stops of just stutters along. That’s because the microwave RFI is causing huge problems with the digital processing such that error rates go up massively and data needs to be re-transmitted multiple times, causing huge amounts of processor noise, power supply noise and latency, all of which affect what you hear.

When you remove or better avoid jitter, EMI and RFI and when your digital bit square wave is fast, accurate and free of noise, when processors aren’t being interrupted, when error correction isn’t working overtime, when processors can execute without latency the resulting music sounds better. That means that when your clocking is more accurate, the level of network traffic is low, when there’s fewer demands on CPUs, when power supplies have very low noise and ripple and plenty of headroom, the music that results from a digital stream is easier for your brain to process
I’m afraid I think most of this is untrue but will leave it there. Thanks for taking the time.
 
  • Like
Reactions: Nestor Turton

tuga

. . .
Wammer
Aug 17, 2007
14,342
7,001
173
Oxen's ford, UK
AKA
Ricardo
HiFi Trade?
  1. No
Analogue noise is the noise added to an analogue signal. Hum is analogue noise, hiss is analogue noise, rumble is analogue noise. You can hear those examples of analogue noise because they’re in the audible frequency spectrum. Digital noise is noise you can’t hear because it’s far too high frequency. It’s in the frequency range in which digital electronics operate. Jitter is noise in the timing domain seen as variations from the correct timing. Jitter causes phase noise, noise in the frequency domain, which you can hear. EMI is radiated or conducted noise of any frequency, RFI is a subset of EMI and is noise in the radio frequency band, ie noise capable of being transmitted.
Sources of RFI and EMI include cell phones, microwaves, switch mode power supplies, computer circuits, Wi-Fi etc. You can’t hear RFI emitted by a microwave because its too high frequency but try to receive a wi-fi signal while sitting close to a running microwave and the whole thing either stops of just stutters along. That’s because the microwave RFI is causing huge problems with the digital processing such that error rates go up massively and data needs to be re-transmitted multiple times, causing huge amounts of processor noise, power supply noise and latency, all of which affect what you hear.

When you remove or better avoid jitter, EMI and RFI and when your digital bit square wave is fast, accurate and free of noise, when processors aren’t being interrupted, when error correction isn’t working overtime, when processors can execute without latency the resulting music sounds better. That means that when your clocking is more accurate, the level of network traffic is low, when there’s fewer demands on CPUs, when power supplies have very low noise and ripple and plenty of headroom, the music that results from a digital stream is easier for your brain to process

Noise will affect the clock and D/A stage performance. Not many people are measuring this but Jussi Laako has been higlighting the issue for quite some time (I started a topic here about the subject but it was deleted without justification).
Sources are varied but mostly down to SMPS', DSP and packet noise.
According to Laako some DACs have very good filtering in their USB interfaces (e.g. Holo). He owns and has tested dozens of DACs.
He uses Intona USB filters and Meanwell medical grade SMPS' (with ferrites) and makes a clear stand against the use of audiophile switches and cables.
 
  • Haha
Reactions: JayC

Blackmorec

Newbie
Wammer
Feb 6, 2019
90
65
23
No. The Statement clock is driving the digital stream, the PhoenixNET is clocking the ethernet stream. Unless you have evidence to the contrary: a link for example. Sorry to be persistent but it’s an important differentiation.
You cannot improve (the quality of) the data so the better 'Ethernet' clock makes absolutely no difference.
I think you guys are missing my whole point here. Firstly there are 2 clocks in a Statement. One for the USB and the other for Ethernet. The PhoenixUSB is based on the former, the PhoenixNET on the latter. Take an Innuos Zenith which has no special reclocker, add a PhoenixNET to its input and a PhoenixUSB to its output and you get close to the sound quality from a Statement, so a nice upgrade path for Zenith owners.

To tuga’s point, I’m not claiming to improve the quality of the data. I’ve already stated that the stream is bit perfect, so the data quality is as good as it can get . The question is, how much noise is generated in order to make the bit stream perfect? If there are high error rates (because for example connectors are poor quality and vibrate a lot and it takes multiple transmissions to get that bit perfect data) the multiple transmissions are causing a lot more noise, which you can hear and occupying more bandwidth, causing latency, using more power and basically causing the music produced by the bit-perfect stream to be of a much lower sound quality.
My whole point is, a perfectly functioning network from an IT perspective is NOT the same as a perfectly functioning network from an audiophile’s perspective, where better sound quality is the goal. The former can have a ton of noise and still function perfectly. Heck its designed to function perfectly in a noisy environment. But in an Audio network, that same noise kills the sound quality. Its why audio networking is rich in noise and jitter reducing devices And IT networking isn’t.
 

Forum statistics

Threads
113,444
Messages
2,451,263
Members
70,783
Latest member
reg66

Latest Articles