Ethernet switches for audio - Part B: why a regular switch will suffice.

[bob_m_54]

On 31/03/2022 at 3:49 PM, meotomvn said:

I don't need to prove anything to you or anyone. I never listen to music based on the measurements. 

The only difference I was able to distinguish from the two recordings was also that one was louder than the other, which could be due to a multitude of reasons, totally unrelated to using the isolator.

 

I tend to agree with @MLXXX that you need to use a much more precise method, if you really want to demonstrate to others the difference your changes make.

 

But if it makes an improvement for you, that's all that really counts, at the end of the day eh?

[MLXXX]

8 hours ago, bob_m_54 said:

The only difference I was able to distinguish from the two recordings was also that one was louder than the other, which could be due to a multitude of reasons, totally unrelated to using the isolator.

I've just listened after equalizing the levels. The recordings do sound rather different to my ears. The "no isolator" version seems to have a more pronounced stereo effect. The "with isolator" version sounds a little muffled. These differences I'd assume to be attributable to different microphone distance and location. However our friend meotomvn hasn't actually described how the recordings were made.

[MattyW]

This looks interesting for the price:

 

AU $681.74  30％ Off | 8 Bit Gigabit HIFI Audio Ethernet Switch Full Linear DC Power Supply SC Cut OCXO Constant Temperature Crystal Oscillator Upgrade
https://a.aliexpress.com/_mqYed0e

[Volunteer]

7 hours ago, MattyW said:

This looks interesting for the price:

 

AU $681.74  30％ Off | 8 Bit Gigabit HIFI Audio Ethernet Switch Full Linear DC Power Supply SC Cut OCXO Constant Temperature Crystal Oscillator Upgrade
https://a.aliexpress.com/_mqYed0e

 

I still don't get the fixation with the clocks. You could run a Rubidium or Caesium clock if you want, but it doesn't make any difference. Ethernet clocking requirement is way lower than even ocxo and when the timing that matters in the DAC is actually the other side of the data buffer, then clock accuracy in switches is demonstrably unimportant.

 

Still, whatever floats your boat.

 

Nice case though.

[dbastin]

7 hours ago, El Tel said:

I still don't get the fixation with the clocks.

This product seems to be a budget version of the approach taken by Innuos PhoenixNet.

 

I am still not compelled by the idea of a better clock.  Many people report considerable improvement using a better clock, but I wonder if it is the clock or something else that coincidentally occurs when a better clock is used.

 

I am very much enjoying my combo that has no fancy clock, except the ER clock and Mikrotik CRS 305 SFP+ seems to have a better than usually clock (ie. perhaps $2 rather than 5c). ...

 

nbn > ER (as FMC) > fibre > Mikrotik CRS 305 (as router) > server > wifi > endpoint

 

Thanks to your comments elsewhere I have/am reviewing use of cabling with shields. which is well worthwhile before contemplating clock upgrade (if at all).  I note other small things are important too, such as Isotek System Enhancer.

[Assisi]

18 minutes ago, dbastin said:

reviewing use of cabling with shields

What have you done so far?  As I said in a recent post somewhere "to shield or not to shield that is the question?"

John

 

[dbastin]

7 hours ago, Assisi said:

What have you done so far?  As I said in a recent post somewhere "to shield or not to shield that is the question?"

John

 

This is an aside to this topic, however, the short answer seems to be "it depends what it is connected to".

 

Firstly, I am starting to be more fussy about the grade/quality of the cables.  I have not ventured into industrial grade.  But certainly not using $5 garden variety.

 

I am quite confident that from nbn > ER side B, Cat 6a UTP is far better than Cat 6a STP on cables that appear to of similarly good grade/quality (not $5 garden variety), and than a far more expensive cable that is shielded and shield connected to ground at both ends.

 

However, the Cat 6a UTP was relatively dull, flat and 2 dimensional further downstream.  I will revisit that though.

 

From router > server I am quite enjoying a Cat 7 STP that has shield connected to ground at only one end, but it also has carbon fibre sheath.  This is competing with cables many times its cost in this part of my network.

 

I have also investigated how the shields are connected on my cables and discovered the designers made deliberate decisions after trial and error to chose a method that sounds best.  More often than not the shields are connected to ground at both ends.

 

I am not settled on anything.  Changes usually teach a lesson and prompt review of another aspect.

 

I am thinking, as a hypothesis, in a relatively high noise situation (eg. from nbn box or router), no or disconnected shields may perform better, whereas in portions of a network that are very low noise and very sensitive to noise (eg. from an audio server to streamer) there are sonic advantages using shields and connecting to ground depending on other aspects of the cable design and the gear they are connected to. 

 

Fibre fits somewhere in that spectrum.  But the Cat 6a UTP and Cat 7 STP impart qualities quite similar to fibre.  Further downstream, higher end cables (shielded) have previously performed better than fibre.

 

None of these are anywhere near the magnitude of benefit I get from wifi in my situation.  However, I may revisit that too.

[Assisi]

14 hours ago, dbastin said:

From router > server I am quite enjoying a Cat 7 STP that has shield connected to ground at only one end,

With the Cat 7 STP cable with shield connected at one end, have you tried it connected either way and if so, is there a difference? 

 

I have until recently used a 15 mt CAT 6 UTP cable to go from my upstairs router adjacent to the NBN box to the Router downstairs that is the beginning of the Audio network.  The result when streaming was fine.  Maybe not as good as files from the NAS but not critical. 

 

A few days ago, I received a 14 mt Cat 7 with Telegartner plugs. I understand it is S/FTP.  This cable goes from one router to the other and replaces the CAT6.  The multiple shields on the new cable are detached or not connected at one end.  The end with the shields not connected is designated as the receiving end.  So, it is directional. 

https://avantiaudio.com/

 

The cable has been connected for five days.  Over that time there seems to be a beneficial change as the cable settles.  Whether the change is due in any way to the shields being disconnected I cannot say definitively.  You would need an identical cable with the shields connected at both ends to compare.  If I change back to the CAT 6 cable that was replaced, there is a small but perceptible benefit with the new cable.  So regardless I am happy with the outcome.

I have also had the shield on one end of the cable that goes from the Melco 10 switch to the Waversa Filter removed.  Probable benefit.  Hard to tell but definitely not a disbenefit.  I am tempted to have more shields disconnected at one end of some cables.  Once it is done though it is not easy to go back.

John

[dbastin]

On 09/04/2022 at 11:42 AM, Assisi said:

With the Cat 7 STP cable with shield connected at one end, have you tried it connected either way and if so, is there a difference? 

Replied in this thread to move the subject into a more suitable thread.

 

Not yet, I need to work out which end has shield connected.
 

 

[MLXXX]

5 hours ago, dbastin said:

Replied in this thread to move the subject into a more suitable thread.

Thanks, dbastin.  The current thread is not well suited to claims bereft of technical argument or objective corroboration.  "True believers" can post to their heart's content about their uncorroborated subjective experiences, in the other thread.

 

I have respected the "sanctity" and special immunity  of that other, complementary, thread by avoiding challenging any of the subjective claims in it.

Edited April 10, 2022 by MLXXX

[MLXXX]

On 07/04/2022 at 11:12 PM, MattyW said:

This looks interesting for the price:

 

AU $681.74  30％ Off | 8 Bit Gigabit HIFI Audio Ethernet Switch Full Linear DC Power Supply SC Cut OCXO Constant Temperature Crystal Oscillator Upgrade
https://a.aliexpress.com/_mqYed0e

 

Given that incoming streams of data from an ethernet switch are processed and reclocked by the receiving device there is no apparent technical argument to support any need for a  highly stable clock rate in the switch.  

 

I note that manufacturers of audiophile devices tend to respond to and cater for consumer demand. In Australia, there is no clear legal obligation on the manufacturer to establish whether what the consumer demands actually makes an audible difference.  Nor is there a clear legal obligation on retailers to sell only those audiophile tweaks or add-ons that make an audible difference.  It is the old law of the commercial jungle: caveat emptor.  The consumer needs to do his or her own research. 

 

Of course marketing departments do their best to promote the product, regardless. We see gushing language such as in the case of this product:  "The power noise is as low as uV, which greatly reduces the various noise and interference on the board, which is expressed in the amount of sound information. Without reduction, [sic] the background is purer, and the high frequency is more rounded and shiny."

 

The above marketing language makes little technical sense to me.  Also, the last sentence appears to say the reverse of what was probably meant!

 

Edit: That last sentence could instead begin, "With such reductions in noise and interference, ... ".

Edited April 10, 2022 by MLXXX

[dbastin]

6 hours ago, MLXXX said:

Edit: That last sentence could instead begin, "With such reductions in noise and interference, ... ".

I suspect the 'marketting' is a less than accurate translation to English. Perhaps that is intentional to further baffle the reader.

 

However, I will say my experience is that power supply to ethernet devices can make an audible difference, anything from barely discernible to quite obvious.  Of course this depends on the context - the device it is powering, where in the network that is, the ethernet cabling/connections, overall how revealing the network/audio system is.

[Ittaku]

6 hours ago, MLXXX said:

Given that incoming streams of data from an ethernet switch are processed and reclocked by the receiving device there is no apparent technical argument to support any need for a  highly stable clock rate in the switch. 

Pretty much, and in some regards, assuming incoming data packets could somehow impart noise downstream (which is questionable enough as it is), perfectly periodic packets are more likely to impart a noise signature at one exact frequency than ones designed to have a random amount of acceptable jitter and be stochastic in nature whilst still meeting timing targets to spread out the noise and make them lower amplitude.

[Volunteer]

1 hour ago, Ittaku said:

assuming incoming data packets could somehow impart noise downstream (which is questionable enough as it is)

 

Yeah, totally. It's more than questionable. It's actually not possible for an Ethernet frame to impart noise to the other side of a transformer-coupled interface. Any level of RFI that impacts upon the cable is effectively rejected by the twisted pairs (their very nature); any larger issues result in payload checksum failing and the TCP packet being resent. I know I am preaching to the choir on this one 

[Ittaku]

19 minutes ago, El Tel said:

 

Yeah, totally. It's more than questionable. It's actually not possible for an Ethernet frame to impart noise to the other side of a transformer-coupled interface. Any level of RFI that impacts upon the cable is effectively rejected by the twisted pairs (their very nature); any larger issues result in payload checksum failing and the TCP packet being resent. I know I am preaching to the choir on this one 

Not to support the concept, but presumably even if it's not noise that's travelling with the data, I was talking about the arrival of a packet causing an interrupt to fire at the receiving end. Thus there'd be periodic interrupts within the receiving end device, and how making them perfectly periodic is more likely to create receiving end generated device noise of a certain frequency. None of this is "transmitted noise" from the sending device, and my point being that making the data periodicity more accurate could theoretically make things worse, not better.

Edited April 11, 2022 by Ittaku

[Volunteer]

10 minutes ago, Ittaku said:

Not to support the concept, but presumably even if it's not noise that's travelling with the data, I was talking about the arrival of a packet causing an interrupt to fire at the receiving end. Thus there'd be periodic interrupts within the receiving end device, and how making them perfectly periodic is more likely to create receiving end generated device noise of a certain frequency. None of this is "transmitted noise" from the sending device, and my point being that making the data periodicity more accurate could theoretically make things worse, not better.

 

Well, it's a theory...

[Ittaku]

2 minutes ago, El Tel said:

Well, it's a theory...

We need to market a stochastic packet transmitting ethernet switch. Genius.

[Volunteer]

15 minutes ago, Ittaku said:

We need to market a stochastic packet transmitting ethernet switch. Genius.

 

The irony.... CSMA/CD was a mechanism used in Ethernet bus networks prior to switching (it's still baked into 1Gbps Ethernet even though a full duplex negotiated link warrants it redundant). It's purpose was to detect other nodes transmitting or collisions with ongoing frame transmission on a shared wire and to, get this... "back-off for a random period" before attempting to send again. You could not make this up

[aussievintage]

1 hour ago, El Tel said:

It's purpose was to detect other nodes transmitting or collisions with ongoing frame transmission on a shared wire and to, get this... "back-off for a random period" before attempting to send again. You could not make this up

 

Don't knock it.  It worked,  and was more reliable than some coax token ring shyte I had to keep going before we got it. 

[Volunteer]

14 minutes ago, aussievintage said:

 

Don't knock it.  It worked,  and was more reliable than some coax token ring shyte I had to keep going before we got it. 

 

802.5 for the win!

 

A favourite pastime was to go find the beaconing NIC that was set to 4Mbps instead of 16Mbps. Although there was nothing favourite about it whatsoever.

 

I'm old.

[Godot]

If I understood this at all, while offering a product in this segment, Alvin’s also saying buy one if it will make you feel better. Don’t if it won’t.