Audiophiles giving too much credits on clock in network applications

[Windows X]

 

First of all, I agree that improving clock from stock industrial grade in network switch, network adapter devices can improve network performance and reliability to certain extent enough for audible improvements. I'm not against that at all. I also bought audiophile network adapter card and network switch for personal uses and upgrade clocks on motherboard's LAN port too.

 

But lately I saw many audiophiles focusing so much on networked audio stream to the extent that they put 10MHz on network switch or network adapter device. Sure that can make big changes and most consider it to be 'big improvement' but that's only when compare to stock grade clock and many privately admitted that some 10MHz input may need really good quality 10MHz clock to shine.

 

If they put this kind of effort into what really matters clock like putting 10MHz input into server as masterclock multiplying to 12-25MHz feeding motherboard chipset clock or storage device that keeps OS and music as source, yeah that'd be some really cool stuff befitting highend computer transport since those are actively used and can effectively use high quality 10MHz clock.

 

But for network protocol, it has 7 layers for basic infrastructure that has many underlying layer beneath making clock upgrade being less efficient due to many processes involved including error correction and packet fragmentation. Improvement from industrial grade network device to high precision/low phase noise clocks that can be swiftly replaced will also make big enough improvement too.

 

10MHz input clock in network will bring better improvement than smaller 25MHz clock in network adapter/switch under proper implementation that's for sure. But how much it'll be better than high quality clock with very low phase noise that can fit in there? I doubt there'll be audiophiles who'll try to upgrade clock on network adapter/switch with 10MHz input and compare with.

 

Let's break some theories here. Imagine you visit streaming service's data center and managed to request admin to upgrade clock on their servers. Do you think that will make streaming sounding so much better after running from another side of planet to your house? While clock upgrade can make noticeable improvement but that can be done to certain extent. That's what network is all about.

 

And the reason why I started this thread is to inform you that there's something more important than clock and something more dangerous, more sinister to your system than clock quality. It's everywhere in your house and in every networked device around you. Yes, it's router, device with full contamination of noise and pollution from other devices.

 

When you plug your server to your home router before using network switch, your router that has wifi spreading frequency and collect all the dust from other wired devices like TV, NAS, receivers that bounce back to your router and feed your server or streamer. You got those pollution from other devices to server via router too and that's the problem.

 

Some people made article before that using WiFi will get rid of this issue because it's not associated with other devices causing pollution. But they forget one thing, WiFi connection itself is also considered as big pollution and packet fragmentation in WiFi is far worse than wired network rendering 10MHz clock input to WiFi even less effective. There was time I got network phobia from discovering how network can really harm audio in ultra highend performance scale. I used to disable network completely and control from monitor for a while until I can get over it.

 

OK. Back to main story. As for these pollution hoping around wired network from devices to device, good network switch will have transformer placed behind LAN port to isolate noise and pollution from other devices. You can test this by using web remote feature to control your server so that network will have zero effect on audio itself and see how network switch and reduce pollution from router.

 

This feature that isolate pollution from router and other devices is the main reason why people should employ network switch. But not only that, it can also help in handling data transmission to be more reliable and more efficient with buffer in network and software vendor optimized for lower packet fragmentation than allowing router to handle everything manually.

 

So, no matter how much effort you put into clock upgrade, it won't reduce pollution from other devices in network. And another thing that most audiophiles overlooked is software implementation in network adapter/switch. As changing configuration in network adapter affect sound quality significantly, software configuration in network adapter/switch itself matters too no less than hardware circuit design. As we can't test and confirm about this part ourselves, there's not much we can do to verify though.

 

To summarize all these in few sentences, I want to share my research that clock isn't the most potent thing in networked audio and you should consider other things like handling pollution in network audio network and implementation in other aspects with higher priority. Don't just blindly go for clock upgrade and thing everything else comes second ignoring more pressing matters. And sometimes 10MHz clock input may perform worse than high quality clock upgrade on board if not done right too.

[Rexp]

Thanks, I'm suffering from network phobia, what minimum steps would you suggest to improve Tidal via wifi? 

[Windows X]

On 6/13/2019 at 8:09 PM, Rexp said:

Thanks, I'm suffering from network phobia, what minimum steps would you suggest to improve Tidal via wifi? 

 

With pleasure. Getting network switch will help a lot. I agree that industrial network switch can be improved and some audiophile mod or OEM one can help significantly. My point of this post is to explain that upgrading clock won't magically resolve every issues we have in network.

 

Regards,

Keetakawee

[Nikhil]

On 6/3/2019 at 7:34 AM, Windows X said:

This feature that isolate pollution from router and other devices is the main reason why people should employ network switch. But not only that, it can also help in handling data transmission to be more reliable and more efficient with buffer in network and software vendor optimized for lower packet fragmentation than allowing router to handle everything manually.

 

+1  

 

Great post KP.  I've found this to be true as well when I set up my audio network.  

A decent network switch + good clean power supply cleaned up the transients / leading edge of notes in my setup.

 

 

.

 

[jabbr]

On 6/14/2019 at 12:59 PM, Windows X said:

 

With pleasure. Getting network switch will help a lot. I agree that industrial network switch can be improved and some audiophile mod or OEM one can help significantly. My point of this post is to explain that upgrading clock won't magically resolve every issues we have in network.

 

Regards,

Keetakawee

 

Actually if you get a 10Gbe industrial network switch you can’t easily improve on by an audiophile mod because it already has an improved “femtosecond” clock and a good power supply. It’s also been tested to hit thst 10Gbe eye plot —if that’s not good enough for you try hitting a 100Gbe eye 😉 Go SFP(+). The pro stuff is quite good, much better than a small audiophile company can do. 

[Ralf11]

On 6/13/2019 at 6:09 AM, Rexp said:

Thanks, I'm suffering from network phobia, what minimum steps would you suggest to improve Tidal via wifi? 

 

I'm just gonna get rid of my router - maybe smash it with an axe.

[jabbr]

On 6/13/2019 at 9:09 AM, Rexp said:

Thanks, I'm suffering from network phobia, what minimum steps would you suggest to improve Tidal via wifi? 

 

Move to a rainforest? What else needs to be improved?

 

Seriously though Wi-Fi AC has great bandwidth, and get a NUC as an access point, and something like an iFi iDSD that’s got good input isolation.

[Windows X]

On 6/24/2019 at 2:45 PM, Nikhil said:

 

+1  

 

Great post KP.  I've found this to be true as well when I set up my audio network.  

A decent network switch + good clean power supply cleaned up the transients / leading edge of notes in my setup.

 

 

.

 

 

With pleasure. I hope this thread will help in providing more insights about how network works. Personally I recommend network switch with isolation transformer design from respectable companies. Volume amount of units, parts being exclusive to certain group, and technical engineering behind long history of professional can make really good results.

 

On 6/24/2019 at 7:39 PM, jabbr said:

 

Actually if you get a 10Gbe industrial network switch you can’t easily improve on by an audiophile mod because it already has an improved “femtosecond” clock and a good power supply. It’s also been tested to hit thst 10Gbe eye plot —if that’s not good enough for you try hitting a 100Gbe eye 😉 Go SFP(+). The pro stuff is quite good, much better than a small audiophile company can do. 

 

I did hear some highend network switch models from my friends. Some also sell certain models too. While I agree that better quality of clock matters but the way to implement those clock matters more. I used to be crazy clock modder until I realize 3rd-party clock module with shared regulation from other power source sending clock signal through stranding wires is worst idea to worsen clock performance. There's a good reason why they placed clock nearest to important modules and optimize circuit to be shortest possible. And the clock itself is less important than circuit design and improvements to provide more linear bitstream from network.

 

Regards,

Keetakawee

[jabbr]

40 minutes ago, Windows X said:

And the clock itself is less important than circuit design and improvements to provide more linear bitstream from network.

Linear bitstream as measured by eye pattern? Agreed, the 10G specs says that indeed requires end to end testing to meet spec. 

 

Older stuff like 1g and 100m didn’t require end to end linearity

[Windows X]

13 hours ago, jabbr said:

Linear bitstream as measured by eye pattern? Agreed, the 10G specs says that indeed requires end to end testing to meet spec. 

 

Older stuff like 1g and 100m didn’t require end to end linearity

 

Do you think streaming audio requires up to 10G? Well, maybe it's only me who has highend CD transport to compare with.

[jabbr]

5 hours ago, Windows X said:

 

Do you think streaming audio requires up to 10G? Well, maybe it's only me who has highend CD transport to compare with.

 

The point is that the 10Gbe Ethernet spec specifies an end to end jitter limit (as measured by an eye pattern) in order to meet the specification. This requires a really well made switch. When you use that same switch at 1G you get the benefits of it being really well made.

 

The lasers in a single mode SFP module are super high quality, such that the 10G bitstream can be sent very precisely kilometers over a fiberoptic cable. You use those same lasers when you are using the module at 1G in your home. Unbeatable precision. 

[Windows X]

3 hours ago, jabbr said:

 

The point is that the 10Gbe Ethernet spec specifies an end to end jitter limit (as measured by an eye pattern) in order to meet the specification. This requires a really well made switch. When you use that same switch at 1G you get the benefits of it being really well made.

 

The lasers in a single mode SFP module are super high quality, such that the 10G bitstream can be sent very precisely kilometers over a fiberoptic cable. You use those same lasers when you are using the module at 1G in your home. Unbeatable precision. 

 

Well, we used to think optical cable will sound a lot better than coaxial cable. We'll see in time. I'm also interested in testing SFP too. Already bought some stuff about it but don't have time to test with yet.

 

Regards,

Keetakawee

[jabbr]

1 minute ago, Windows X said:

 

Well, we used to think optical cable will sound a lot better than coaxial cable. We'll see in time. I'm also interested in testing SFP too. Already bought some stuff about it but don't have time to test with yet.

 

Regards,

Keetakawee

Fiber optic Ethernet is nothing like S/PDIF! Particularly the VSEL lasers and transducers are vastly better, but everything is different, particularly Ethernet is packet based, so, for example you could transmit an entire song during the silent moment between songs. The network doesn’t need to be active during playback. 

[Windows X]

7 minutes ago, jabbr said:

Fiber optic Ethernet is nothing like S/PDIF! Particularly the VSEL lasers and transducers are vastly better, but everything is different, particularly Ethernet is packet based, so, for example you could transmit an entire song during the silent moment between songs. The network doesn’t need to be active during playback. 

 

It's an analogy about people expecting new things to be better but we can't tell for sure until we try enough. Personally I have high hopes for 10G and researching on best 10G card for server at the moment to use with 10G switch I bought last year.

[jabbr]

28 minutes ago, Windows X said:

 

It's an analogy about people expecting new things to be better but we can't tell for sure until we try enough. Personally I have high hopes for 10G and researching on best 10G card for server at the moment to use with 10G switch I bought last year.

 

My listening comparisons have not shown what I consider a definitive difference between the Intel x520 vs very low latency Solarflare 7322F vs Mellanox Connectx-4 (100 Gbe) ... but that’s just me

 

there seem to be some crazy Solarflare bargains on eBay these days. 

[Windows X]

7 minutes ago, jabbr said:

 

My listening comparisons have not shown what I consider a definitive difference between the Intel x520 vs very low latency Solarflare 7322F vs Mellanox Connectx-4 (100 Gbe) ... there seem to be some crazy Solarflare bargains on eBay these days. 

 

How's the difference between 10-100Gbe against audiophile 1G switch? Since this thread is on active now, I'd like to share more insights about network switch modding that may not done right.

 

1. External clock upgrade module often use stranded wire to map from clock module to original clock circuit which is closest to main chip as much possible like few centimeters to main chip. Using external clock module, it'll need longer stranding wire to map from new clock output to older one. This will introduce more noise and interference during transmission to wire I don't see any solid core coaxial wire installed inside yet.

 

2. Some people put shielding layer on top of transformer block which has magnetic field generated. Putting those on will affect magnetic field and can potentially affect device's reliability rather than helping it. I used to discuss about this kind of tweak with my technician friend and he said we don't know the direction flow of generated magetic field around transformer and that's not sensitive parts that requires EMI/RFI shielding.

 

3. Removing isolation transformer from network switch design is not really a good idea. Ideally not having isolation transformer can have signal and data flow directly which should be better for audio servers. However, network switch will be connected to more polluted devices like main router with WiFi connecting to other things like TV/AVR/etc. You may also connect network switch to media players and NAS which also have noise and interference that can pass through network path to other devices. Isolation transformer will lessen those effects and I highly recommend to use network switch with this feature since network path isn't as clean as we believe.

 

That's what I found so far from my research with networked audio applications. Using WiFi extender with network switch is also a good idea to upgrade from pure WiFi connection too.

 

Regards,

Keetakawee

[jabbr]

32 minutes ago, Windows X said:

How's the difference between 10-100Gbe against audiophile 1G switch? Since this thread is on active now, I'd like to share more insights about network switch modding that may not done right.

 

I agree with your premise that mods not done correctly are likely to be worse. I also agree that we need to look at end to end performance rather that individual mods.

 

The reason I’ve heard to make audiophile mods to switches is to improve the power supply and clock. Both go hand in hand at reducing jitter. I won’t comment on whether reducing jitter in an Ethernet switch will improve SQ, lets take end to end jitter/phase noise as a parameter to be improved.

 

In the ancient days of 10Mb / 100 Mb and even 1Gb Ethernet, there wasn’t an end to end jitter standard — so circa 2002 when the 10Gbe standards came out, it was recognized that end-to-end jitter was important and these standards mandate minimum standards as measured in the eye pattern. 

 

This standard is already very very tight and because even tighter for 100Gbe such that only a handful of companies produce ASICS that can handle 100Gbe and only a handle of QSFP28 transducers for 100GBe ... and on and on to 200 and 400Gbe. 

 

So so what I am saying is that professional equipment from reputable companies has already been designed and tested to have extremely low jitter. 

 

I am not looking for an Ethernet switch to impart a SQ on my audio system, rather to provide clean, high quality bits (and not inject noise)

[Windows X]

11 minutes ago, jabbr said:

 

I agree with your premise that mods not done correctly are likely to be worse. I also agree that we need to look at end to end performance rather that individual mods.

 

The reason I’ve heard to make audiophile mods to switches is to improve the power supply and clock. Both go hand in hand at reducing jitter. I won’t comment on whether reducing jitter in an Ethernet switch will improve SQ, lets take end to end jitter/phase noise as a parameter to be improved.

 

In the ancient days of 10Mb / 100 Mb and even 1Gb Ethernet, there wasn’t an end to end jitter standard — so circa 2002 when the 10Gbe standards came out, it was recognized that end-to-end jitter was important and these standards mandate minimum standards as measured in the eye pattern. 

 

This standard is already very very tight and because even tighter for 100Gbe such that only a handful of companies produce ASICS that can handle 100Gbe and only a handle of QSFP28 transducers for 100GBe ... and on and on to 200 and 400Gbe. 

 

So so what I am saying is that professional equipment from reputable companies has already been designed and tested to have extremely low jitter. 

 

I am not looking for an Ethernet switch to impart a SQ on my audio system, rather to provide clean, high quality bits (and not inject noise)

 

That's interesting insights. Thanks for sharing. I used to be clock modder audiophile who put clock mod on many digital audio devices (and very good one like Superclock/Ultraclock). Those devices still sound nice to me but I later notice some added interference from external clock module mapping so I did some tests with replacing to best clock that fits in instead. I notice some quality from external clock modules are missing a bit but those added noise are all gone.

[TomJ]

Hi all,

 

I am a bit late to the party.

 

Has there ever been measurements of the ethernet signal from clock modded switches / NICs versus standard?

 

Maybe the clock modding worsen the signal and this will  cause a sonically change.

I made this experience with lan cable with wrong impedance.

 

regards,

 

Tom

[Miska]

On 6/3/2019 at 5:04 AM, Windows X said:

OK. Back to main story. As for these pollution hoping around wired network from devices to device, good network switch will have transformer placed behind LAN port to isolate noise and pollution from other devices. You can test this by using web remote feature to control your server so that network will have zero effect on audio itself and see how network switch and reduce pollution from router.

 

Just remember that it only works with unshielded UTP cables (plastic connector body). If you use shielded STP cables, it connects grounds of all involved devices together, through the shield. Thus totally spoiling the isolation you would otherwise have...

 

[TomJ]

On 6/3/2019 at 4:04 AM, Windows X said:

OK. Back to main story. As for these pollution hoping around wired network from devices to device, good network switch will have transformer placed behind LAN port to isolate noise and pollution from other devices. You can test this by using web remote feature to control your server so that network will have zero effect on audio itself and see how network switch and reduce pollution from router.

All switches and other ethernet devices have transformers, because this is part of the ethernet spec.

Transformers are integrated in the RJ45 or as a dedicated chip on board.

 

As the transformers have also an impact on the signal, I think it could also be a topic for modding, but as you mentioned before, audiophiles take to much care about clocks (with lousy implementation).......

Would be very interesting to see measurements and eye pattern of this home brewed devices and mods.

[yamamoto2002]

IMO quiet switch that does not emit coil whine is desirable for audio use

 

I read somewhere, OC-192 requires very low jitter clock that can reliably distinguish consecutive 511 zeros and 510 zeros, that is not happen on copper Ethernet.

[plissken]

8 hours ago, TomJ said:

As the transformers have also an impact on the signal

Transformers are part of the spec and design and they are taken into account on signal.

 

External, medical grade, isolation transformers can degrade throughput.

 

But all these transformers are passive and shouldn't exhibit any audible whine.  I'm not talking about the PSU transformer either. Plus with a generous 328ft to the spec you should be able to put wired network gear elsewhere.

[TomJ]

1 hour ago, plissken said:

Transformers are part of the spec and design and they are taken into account on signal.

 

External, medical grade, isolation transformers can degrade throughput.

 

But all these transformers are passive and shouldn't exhibit any audible whine.  I'm not talking about the PSU transformer either. Plus with a generous 328ft to the spec you should be able to put wired network gear elsewhere.

 

There are different transformers with different quality regarding return loss, crosstalk and insertion loss. Also there are transformers with only one transformer coil and others have common mode chokes. All this has an impact on the signal quality.

 

And if you think something is going to be improved on a very well developed product like a network card or switch, I am surprised (actually it doesn’t surprise me, since clocks is the new sport among audiophiles) that you don’t start here.

Most transformers are not made very precisely. 

Like this:

 

Just think that a signal produced by a $ 1,000 clock has to go through this common mode choke grave - throwing pearls to swine.

 

There are now machine-made transformers with a very high level of accuracy. 

Like this:

 

Regards,

 

Tom

[Miska]

On 8/28/2021 at 12:22 AM, TomJ said:

Just think that a signal produced by a $ 1,000 clock has to go through this common mode choke grave - throwing pearls to swine.

 

As long as the data is transmitted correctly... In any case, there's no relation between network clocks and audio clocks.