Ethernet System for Audio: putting it all together

[davewantsmoore]

It would be fascinating to perform a blind test at John (or anywhere) 's house.

 

Set things up so that all the networking equipment is moved out of the music room  (may be difficult to arrange, depending on the reality of things)

 

 .... and have it so that two (similar if possible) cables are entering the room.    Obviously, these might need to be long (5m+) and fairly standard network cables ... and hopefully that wouldn't "invalidate the test".

 

Path 1:    NBN > cheap "modem/router" > Audio Player

Path2:    NBN > cacophony of audio-grade boxes, cables, filters, etc. > Audio Player

 

You could then just swap cables into the audio player .... and write down the listening notes, etc.

 

 

 

I've done said tests.... but it's much less boring when others do it.

[Grizaudio]

1 minute ago, davewantsmoore said:

It would be fascinating to perform a blind test at John (or anywhere) 's house.

 

Set things up so that all the networking equipment is moved out of the music room  (may be difficult to arrange, depending on the reality of things)

 

 .... and have it so that two (similar if possible) cables are entering the room.    Obviously, these might need to be long (5m+) and fairly standard network cables ... and hopefully that wouldn't "invalidate the test".

 

Path 1:    NBN > cheap "modem/router" > Audio Player

Path2:    NBN > cacophony of audio-grade boxes, cables, filters, etc. > Audio Player

 

You could then just swap cables into the audio player .... and write down the listening notes, etc.

 

 

 

I've done said tests.... but it's much less boring when others do it.

 

100%. 

[davewantsmoore]

4 minutes ago, Grizaudio said:

 

I understand this, but if you galvanic isolate (end point) from network, the problem either never existed, and/or doesn't once isolated. 

 

What if noise is entering by the mains cord, or by the air.  <shrug>

 

 

[acg]

Just now, davewantsmoore said:

What if noise is entering by the mains cord, or by the air.  <shrug>

 

 

 

Galvanic isolation makes its own noise too...

[Nada]

17 hours ago, El Tel said:

 I guess we should ask who would care or notice noise at -50dB at 200 times beyond the range of human hearing? Higher frequencies can pass through smaller gaps, we can't block them all, but there comes a point where blocking is moot when it is this far beyond the range of our senses.

 

Its great to see some discriminate analysis applied to audio.

 

The idea that high frequency noise riding on data signal cant be audible is a limited hypothesis when that noise is going through a DAC conversion. 

 

Noise in the DA data or clock spectrum could conceivably modulate the DAC process and ride into the analogue output. The result may not be discernible with current ADC quantifiable measurements, but that does not preclude listening differences.

 

An informal trial where a lab signal generator injects high level high frequency noise would be fun to do initially. The noise freq could be played with to see if its the clock freq or bit clock freq that might generate audible changes.

 

Once the noise amplitude and freq is elucidated, then a DBT could be arranged with a bunch of audiophiles at a get together. A ten minute trial could generate sufficient data for p<0.5 validity.

 

Once noise is reliably found to be audible then different ethernet devices could be inserted and tested to find what blocks the noise.

 

Science is fun

[Assisi]

2 hours ago, acg said:

 

Crazy good cable... incredible what they do.   The first Vertere ethernet cable trialled in my system was a real "you've got to be kidding" moment. 

Yes incredible.  My one and only Vertere will stay that way.  To replace all the other Ethernet cables I have is not financially viable.  I do wonder though what the outcome would be if the cabling was all Vertere?  I will never know.

John

[davewantsmoore]

1 minute ago, Nada said:

Noise in the DA data or clock spectrum could conceivably modulate the DAC process and ride into the analogue output.

Very much so.... but to know if that is happening a listening tests is a very poor choice of methods.

 

You can measure it happening at levels which are much smaller than what is audible.

 

So you can know that it is (or is not) happening.....before you listen for it.

 

The listening doesn't tell you if it is happening or not .... the listening tells you if you can hear it (if it is happening).

[Nada]

11 minutes ago, Grizaudio said:

 

I understand this, but if you galvanic isolate (end point) from network, the problem either never existed, and/or doesn't once isolated. 

 

 

You must be thinking using USB with the Intona isolation was a smart move. Ethernet is a quagmire.

[Grizaudio]

1 minute ago, davewantsmoore said:

What if noise is entering by the mains cord, or by the air.  <shrug>

 

 I guess everyone needs to draw the line in the sand somewhere. 

If differences require very close listening and/or are difficult to detect, my opinion has always been they don't matter to me. 

I'm more inclined to focus on source quality, amps, speakers, room, equalisation.  

[Nada]

1 minute ago, davewantsmoore said:

Very much so.... but to know if that is happening a listening tests is a very poor choice of methods.

 

That's arguably incorrect.

 

The most scientifically valid measurement for music is human perception. Its people that listen to music. Not machines.

 

Lab measurements are interesting, have very high reliability, are useful for design, build and hypothesis testing but not of high validity

 

 

1 minute ago, davewantsmoore said:

You can measure it happening at levels which are much smaller than what is audible.

 

Audible differences may not be measurable. You know that.

 

 

 

[Assisi]

7 minutes ago, davewantsmoore said:

may be difficult to arrange, depending on the reality of things

May be difficult! That is an absolute total understatement.  There is no way that will happen.  I have taken specific switches out in the past and tried them in other persons systems.  The last time I did it I declared to my self never again.  It is as if the network is a living thing.  It takes time to settle again.  Even power failures are disruptive sometimes for days after.

 

John

[Assisi]

15 minutes ago, Nada said:

Science is fun

Listening is more fun.

John

[Nada]

1 minute ago, Assisi said:

  It is as if the network is a living thing.  It takes time to settle again.  Even power failures are disruptive sometimes for days after.

 

Even these subtle observations are testable with a DBT to see if they are repeatable.  It is possible to clarify if they are associated with physical changes or better explained by the complexities of attribution.

 

Just need two of the same switch.  Put them in the test system for a week, both  running. Do a DBT to ensure they are no different. Then turn one off. Turn it back on. Retest. Easy.

[Nada]

2 minutes ago, Assisi said:

Listening is more fun.

John

 

Doing both together with company is even more fun

[Grizaudio]

Golly Gosh..... 

15 minutes ago, Assisi said:

The last time I did it I declared to my self never again.  It is as if the network is a living thing.  It takes time to settle again.  Even power failures are disruptive sometimes for days after.

 

Refocus.... 

[dbastin]

14 hours ago, Assisi said:

As an aside I am trying a just one Vertere cable as the last link in the network chain at the moment.  Seriously incredible.  The best I have tried/heard.  Unfortunately, relatively quite expensive. 

 

5 hours ago, acg said:

Crazy good cable... incredible what they do.   The first Vertere ethernet cable trialled in my system was a real "you've got to be kidding" moment. 

 

It would want to seriously incredibly crazy good for its cost ...

https://soundgallery.com.au/product/vertere-pulse-hb-ethernet-rj45-cable/

 

From the Vertere website:

• High Purity Copper
• bespoke internal conductor designs
• proprietary shielded data conductor
• terminated with highest quality RJ45 connectors with our unique internal shielding configuration.

Pulse-HB’s proprietary conductors are all constructed from scratch using high purity copper with various diameters with each strand silver or tin plated as required. These bare conductors are then made into the individual insulated internal conductors. Finally, the internal conductors are all combined with a unique configuration and twist, shield wrapped, shield braided and coated with a special outer covering.

 

3 hours ago, Grizaudio said:

I wonder if quality galvanic isolation makes all this concern redundant. 

I.e. Using Opto, transformer, wireless or other. 

Maybe its time you stop speculating and give something a try, for example:

• router > Cat 6 > FMC > fibre > FMC > 
• router > Cat 6 > switch > fibre > switch > 
• router > Cat 6 > switch > fibre > FMC > 
• router > Cat 6 > EtherRegen > fibre > Optical Module > 

There are tonnes of possibilities given that each device is likely to be affected by its PSU and power cord, each device imprints its signature (ie. shape of noise), and there is a wide variety of SFPs which imprint their own signature too.  Many people have already been on their own version of this journey.  It works but is not this  nirvana you are hoping for ... 

3 hours ago, Grizaudio said:

I understand this, but if you galvanic isolate (end point) from network, the problem either never existed, and/or doesn't once isolated. 

 

3 hours ago, davewantsmoore said:

So, there is no universal solution (something that will work to improve, or affect, all audio equipment)  ..... and there is also no universal problem (some things may not be affected by the network).

That's it in a nutshell, because there is such a wide variety of devices, possibilities and listeners.

[davewantsmoore]

3 hours ago, Grizaudio said:

I guess everyone needs to draw the line in the sand somewhere. 

If differences require very close listening and/or are difficult to detect, my opinion has always been they don't matter to me. 

I think any improvement in the signal matters.....  even if it is just academic.

 

... but you don't listen to a signal to determine whether or not it has changed (from what is it supposed to be).   'cos you will neither know if it really changed (maybe you couldn't hear it, or maybe you got it wrong and though it changed when it didn't) ... and you also won't know which was was the more correct way (ie. when the signal looked more like it was supposed to).

 

3 hours ago, Assisi said:

Listening is more fun.

John

It is... but science can help you get better listening results..... not body here is for here for anything else but good sounding audio.

 

21 minutes ago, dbastin said:

• router > Cat 6 > FMC > fibre > FMC > 
• router > Cat 6 > switch > fibre > switch > 
• router > Cat 6 > switch > fibre > FMC > 
• router > Cat 6 > EtherRegen > fibre > Optical Module > 

There are tonnes of possibilities given that each device is likely to be affected by....

 

Indeed.   It's actually quite easy to imagine a situation where all of those combinations make the audio device(s) perform worse than if they weren't there.  ie.   router -> audio device.

[dbastin]

3 hours ago, Grizaudio said:

I'm more inclined to focus on source quality, amps, speakers, room, equalisation.  

This is a key point I have stated before but I will repeat to highlight it.

 

If you are streaming via a network (server, NAS, nbn), or connecting your audio gear to a network with a wire for any other reason, you need to redefine 'source'.  The network is a source of noise (like the AC grid, and RFI).

 

BUT, even in my system where the Devialet streamer is connected only via a dedicated lithium battery powered wifi access point, changes to the network can still be detected.  Those changes could impact the Antipodes server, but it is isolated from the network via fibre!.  Actually, I can fairly easily detect changes from doing things to improve that server (footers, caps, grounding, etc).  So, in my case, isolation alone is not the ultimate solution, and nether is the noise rejection capabilities of the Devialet because it has no wired ethernet connection to inject noise in the first place.

[davewantsmoore]

3 hours ago, Assisi said:

May be difficult! That is an absolute total understatement.  There is no way that will happen.  I have taken specific switches out in the past and tried them in other persons systems.  The last time I did it I declared to my self never again.  It is as if the network is a living thing.  It takes time to settle again.  Even power failures are disruptive sometimes for days after.

 

Then you could do like this.

 

Leave you network as it is.... don't touch anything, unplug anything, or power off anything.

 

Get a 20m (or whatever it takes) network cable, and run that from your NBN router to your audio device.

 

Then for the test, the only think you touch it the ethernet cable connecting to your audio player.

 

Cable 1:   Your stuff

Cable 2:  Long cable coming direct from your NBN router.

 

 

Or... in the case where your NBN router only has one ethernet socket, and/or you thought touching that might invalidate the test.

 

Cable2 could alternatively be connected to a little battery powered router, that uses a mobile phone for its internet connection....  so in that way NONE of your network was being touched, and the ONLY thing that was changing was the cable (1 or 2) plugging into the back of your audio device.

 

 

 

 

[davewantsmoore]

3 minutes ago, dbastin said:

The network is a source of noise (like the AC grid, and RFI).

I think that's actually very unlikely the be true.

 

However... noise is relative.   Nothing is completely "noise free".

 

Also the susceptibility of various audio devices to any "noise" may vary.

[dbastin]

3 minutes ago, davewantsmoore said:

Indeed.   It's actually quite easy to imagine a situation where all of those combinations make the audio device(s) perform worse than if they weren't there.  ie.   router -> audio device.

Correct, it depends on one's gear and particular set up.  So we need to be careful making broad assumptions and then trying to convince people to agree.

 

It would be better to share "this worked really well for me with this set up" in case it helps someone narrow down the possibilities of what to try in their set up.  And that is the intent of this thread.

[dbastin]

6 minutes ago, davewantsmoore said:

I think that's actually very unlikely the be true.

I enjoy you objectivity and accept that is a possibility which you have shared enough times now to make your point clear.  This thread is not to debate that subject but instead for those many people whose experience listening indicates otherwise.

 

Rather than objectively speculate, please feel free to share any magical endpoint that is immune to any changes to the network it is connected to.   Grimm Audio MU 1 could be one such thing, I gather Chord M Scaler has a similar benefit, and both these rely on code to filter out noise (wherever it may come from).

[frednork]

5 minutes ago, davewantsmoore said:

It would be fascinating to perform a blind test at John (or anywhere) 's house.

 

Set things up so that all the networking equipment is moved out of the music room  (may be difficult to arrange, depending on the reality of things)

 

 .... and have it so that two (similar if possible) cables are entering the room.    Obviously, these might need to be long (5m+) and fairly standard network cables ... and hopefully that wouldn't "invalidate the test".

 

Path 1:    NBN > cheap "modem/router" > Audio Player

Path2:    NBN > cacophony of audio-grade boxes, cables, filters, etc. > Audio Player

 

You could then just swap cables into the audio player .... and write down the listening notes, etc.

 

 

 

I've done said tests.... but it's much less boring when others do it.

I would caution anyone doing these kind of tests, whether with others or by themselves , as they may well be drawing the  incorrect conclusion from what they have done. 

 

These kind of tests are not just about blinding although that seems to be all that is ever discussed. The test is about using subjective sensory assessment (which is typically unreliable) and to turn that into a reliable objective instrument. For large obvious differences , the test may not need to be very robust (but the difference would be bleeding obvious so why bother) but as the level of difficulty of detection rises then so too must the level of effort put into the test. 

 

The methodology only really allows for a definitive answer in the positive. Ie a difference can be reliably detected. If it cannot be detected there is no way of saying definitively that it is the correct result as the test is not designed that way. However we might be able to say we can detect a, b, and c and we think that is a simlar level of difficulty to detecting e so is more likely that it is not detectable. We just cannot say it cannot be detected.

 

What is never done in these home tests is some tests to show that the testing methodology is at least capable of detecting known differences at a small level which are measurable, repeatable and well understood .  usually, people focus only on the really difficult unknown differences. It is done as a one off exercise and then ridiculed by people like me.

 

To anyone doing Blind testing at home please first test your testing protocol and try detecting small known differences and see how you go. If you are unable to do that then there is no point continuing with unknowns. you will need to improve your protocol. This does not mean that you wont hear a difference in casual listening, its just that your test has unintended problems which need sorting.

 

I would also reiterate what I have said elsewhere that if you dont think there is a difference then it is highly unlikely that you will be willing to go to the effort required to show a difference so keep that in mind.

 

If you do think there is a difference and do show it under agreed conditions dont expect people to believe you! As shown previously in the audiophile style ethernet cable challenge.

 

 

[Grizaudio]

1 hour ago, dbastin said:

Maybe its time you stop speculating and give something a try, for example:

• router > Cat 6 > FMC > fibre > FMC > 
• router > Cat 6 > switch > fibre > switch > 
• router > Cat 6 > switch > fibre > FMC > 
• router > Cat 6 > EtherRegen > fibre > Optical Module > 

 

Happy to try, just not happy to pay without listening first.

 

I think this deep dive is getting a bit much for me.

I might need to wave a white flag and tap out. 

 

I also disagree with what Dave said about "I think any improvement in the signal matters.....  even if it is just academic.".

IMO, Improvements are all relative to what one can achieve elsewhere in the system with the same investment. 

 

If a listener needs to concentrate intently, compare A to B 20-30 times to hear a benefit - I would argue the benefit doesn't exist or it simply doesn't matter. 

For instance, if I could spend $3000 on network improvements v's $3000 on an new amplifier or subwoofer addition, I know which I would prefer. 

If differences are more easily heard, well then a value judgement needs to be made. 

 

ATM, my current streamer uses Wi-Fi and I'm happy with that.

I not against your suggestions, but my next step is to upgrade my streamer and subs.
 

 

 

 

 

 

 

 

Edited November 30, 2022 by Grizaudio

[davewantsmoore]

 

1 hour ago, frednork said:

The methodology only really allows for a definitive answer in the positive. Ie a difference can be reliably detected.

That's correct.

 

1 hour ago, frednork said:

For large obvious differences , the test may not need to be very robust (but the difference would be bleeding obvious so why bother) but as the level of difficulty of detection rises then so too must the level of effort put into the test. 

For "large and obvious difference" .... it is often still very interesting to run the test.

 

As you say, if the difference is obvious... then the test will be a doddle.

 

1 hour ago, frednork said:

If it cannot be detected there is no way of saying definitively that it is the correct result as the test is not designed that way.

That's right.

 

All you know is that you couldn't reliably pick anything.   (ie. the difference probably wasn't obvious, even you thought it was)

 

1 hour ago, frednork said:

usually, people focus only on the really difficult unknown differences.

Yes.

Notice that I'm not really recommending such a test for the minute.   I'm recommending said test, for the "it's obvious".... ie. to test large, known (at least, perceived), differences.

 

I also recommend keeping the test as "casual listening" as possible...... listen for a long time (if you want).  Don't think too much about it....  as said, if "it's obvious" then the test should be easy.

 

Caveat -> some amount of appropriate 'blinding' is essential.