TCP/IP Networking, Cables and Numpties

[rdale]

Agree with Bob. Once you've worked in a data center you quickly realise how ridiculous the foo products aimed at computer audio are

I can't say I've worked in a data center, but I am working on a PCIE based appliance for use in data centers at the moment. One thing that surprised me about the PCIE specs was that they have the idea of 'spread spectrum clocking' where they intentionally introduce jitter on the bus to reduce the amount of EMI being radiated. See here for an explanation:

http://www.edn.com/design/analog/4363708/Spread-spectrum-clocking-reduces-EMI-in-embedded-systems-item-2

This is the exact opposite to what you would want to do with a computer audio signal, where we have very regular periodic signals sent via a clock which is as accurate as possible.

The problem with EMI from a PCIE bus actually doesn't apply to ethernet transfer in cables that we are actually talking about. But my point is that I think that high frequency square waves being sent down wires can cause of lot of EMI, and that EMI may have unexpected effects.

 

[Tony_J]

I can't say I've worked in a data center' date=' but I am working on a PCIE based appliance for use in data centers at the moment. One thing that surprised me about the PCIE specs was that they have the idea of 'spread spectrum clocking' where they intentionally introduce jitter on the bus to reduce the amount of EMI being radiated. See here for an explanation:

http://www.edn.com/design/analog/4363708/Spread-spectrum-clocking-reduces-EMI-in-embedded-systems-item-2

This is the exact opposite to what you would want to do with a computer audio signal, where we have very regular periodic signals sent via a clock which is as accurate as possible.

The problem with EMI from a PCIE bus actually doesn't apply to ethernet transfer in cables that we are actually talking about. But my point is that I think that high frequency square waves being sent down wires can cause of lot of EMI, and that EMI may have unexpected effects.

Clock jitter in PCIE (or on Ethernet or WiFi for that matter) is utterly irrelevant - the data packets are buffered by the receiver and the DAC will use a (hopefully stable) clock all of its own. Unless the designer is totally brain-dead, that is.

Yes, Ethernet cables do produce EMI, but so does WiFi - it is after all a radio-based technology - so not a lot of difference there. However, any decently designed network audio kit will be capable of filtering that out so that it doesn't manifest itself as noise on the analogue side. If it doesn't do an effective job of that filtering, toss it and use something that IS decently designed.

 

[kernow]

ahaha, only just noticed this. Hey how about a single switch with more ports on.

Jesus christ when it comes to audio all common sense goes out of the window sometimes.

 

[rdale]

Clock jitter in PCIE (or on Ethernet or WiFi for that matter) is utterly irrelevant - the data packets are buffered by the receiver and the DAC will use a (hopefully stable) clock all of its own. Unless the designer is totally brain-dead, that is.

You've completely missed my point. Introducing clock jitter (ie spread spectrum clocking) is part of the PCIE spec in order to reduce EMI. Encrypting the data being sent down the bus to reduce EMI is also part of the spec.

Imagine if this was an audio discussion and it was suggested that it was a good idea to introduce variable clock timing and data encryption just to reduce EMI, what reception do you think that would get? 'These guys are nutters', 'Bits are bits' etc etc.

 

[rabski]

You've completely missed my point. Introducing clock jitter (ie spread spectrum clocking) is part of the PCIE spec in order to reduce EMI. Encrypting the data being sent down the bus to reduce EMI is also part of the spec.Imagine if this was an audio discussion and it was suggested that it was a good idea to introduce variable clock timing and data encryption just to reduce EMI, what reception do you think that would get? 'These guys are nutters', 'Bits are bits' etc etc.

So how exactly can the connecting cable alter this?

 

[rdale]

So how exactly can the connecting cable alter this?

By offering better shielding. I am not claiming this is why ethernet cables might sound different because I haven't read any account of why audiophile ethernet cables are better than ordinary well designed cables (not many ethernet cables are 'well designed' in terms of meeting the CAT specs AFAIK).

My point is that in general square waves being sent down cables generate a lot of EMI, and you need good shielding and other techniques (eg PCIE spread spectrum clocking and data encryption) to avoid problems.

 

[Tony_J]

So how exactly can the connecting cable alter this?

It can't, unless the quality is so poor that frames get dropped. However, that would manifest itself as dropouts on the audio side, rather than in more subtle changes in tonal balance or whatever, so it would be very obvious. Ordinary off the shelf Cat5/Cat6 cables would not cause that kind of issue though.

Bear in mind, also, that 100megs and gigabit Ethernet are specified for cable lengths of up to 100 metres between nodes, so unless you live in a mansion, standard cables are actually way over spec for the kind of run lengths that you are likely to use. I live in a big old Victorian house and the longest cable run I have is about 20 metres.

 

[rabski]

By offering better shielding. I am not claiming this is why ethernet cables might sound different because I haven't read any account of why audiophile ethernet cables are better than ordinary well designed cables (not many ethernet cables are 'well designed' in terms of meeting the CAT specs AFAIK). My point is that in general square waves being sent down cables generate a lot of EMI, and you need good shielding and other techniques (eg PCIE spread spectrum clocking and data encryption) to avoid problems.

If the cables and transmission are generating EMI, then what they are carrying won't be in the least bit altered by shielding them will it? What possible problems are there?

Ethernet cables don't 'sound' different, because it simply isn't possible for them so to do. There is no 'sound' carried by an ethernet cable, and to assume so ignores the entire principles upon which tcp/ip and the like operate.

It's a system which inherently checks for packet loss and corrects accordingly. It really is fundamental. It either works or it doesn't. Drop outs or similar I can accept as an issue, but any even remotely compliant cable and connection will not cause this.

I've put together a fair few data networks over the years. These have had to deal with substantial information transfer and whether it's a high-definition audio file, a high-resolution image file, or a massive data transfer, the result is the same.

 

[Tony_J]

If the cables and transmission are generating EMI' date=' then what they are carrying won't be in the least bit altered by shielding them will it? What possible problems are there?

Ethernet cables don't 'sound' different, because it simply isn't possible for them so to do. There is no 'sound' carried by an ethernet cable, and to assume so ignores the entire principles upon which tcp/ip and the like operate.

It's a system which inherently checks for packet loss and corrects accordingly. It really is fundamental. It either works or it doesn't. Drop outs or similar I can accept as an issue, but any even remotely compliant cable and connection will not cause this.

I've put together a fair few data networks over the years. These have had to deal with substantial information transfer and whether it's a high-definition audio file, a high-resolution image file, or a massive data transfer, the result is the same.[/quote']

Yep.

 

[rdale]

If the cables and transmission are generating EMI, then what they are carrying won't be in the least bit altered by shielding them will it? What possible problems are there?

That the EMI and other forms of interference find their way to the DAC. I assume that if the EMI produced by cables is better shielded, then less of it will arrive at the DAC or USB to SPDI/F converter or whatever very sensitive kit you have.

But I have never claimed to actually understand exactly all aspects of computer audio. I just play music, that is what my system is for, it isn't a science research project. I might well buy stuff like cables just because it 'makes me happy' and doesn't cost a very high proportion of my pocket money. Audioquest cinnamon ethernet cable from a reputable brand, is mechanically solid and a pretty red colour and doesn't cost a great deal compared with what my CD player cost or whatever.

I don't mind saying that i don't understand everything about computer audio, and even looking a bit of a fool perhaps. But it does get up my nose when people say differences between USB to SPDI/F converters are 'physically impossible', or 'you need to study information theory' (as been said recently on the Computer based Hi Fi forum), as if they are the Albert Einstein of computer audio or something.

 

[Tony_J]

That the EMI and other forms of interference find their way to the DAC. I assume that if the EMI produced by cables is better shielded' date=' then less of it will arrive at the DAC or USB to SPDI/F converter or whatever very sensitive kit you have.

But I have never claimed to actually understand exactly all aspects of computer audio. I just play music, that is what my system is for, it isn't a science research project. I might well buy stuff like cables just because it 'makes me happy' and doesn't cost a very high proportion of my pocket money. Audioquest cinnamon ethernet cable from a reputable brand, is mechanically solid and a pretty red colour and doesn't cost a great deal compared with what my CD player cost or whatever.

I don't mind saying that i don't understand everything about computer audio, and even looking a bit of a fool perhaps. But it does get up my nose when people say differences between USB to SPDI/F converters are 'physically impossible', or 'you need to study information theory' (as been said recently on the Computer based Hi Fi forum), as if they are the Albert Einstein of computer audio or something.[/quote']

So the rough translation of the above is that you are banging on about a subject that you don't understand, and that you are happy to let the audio foo con-men help to spend your ££ on stuff that they know won't improve the SQ of your system?

 

[rabski]

That the EMI and other forms of interference find their way to the DAC. I assume that if the EMI produced by cables is better shielded, then less of it will arrive at the DAC or USB to SPDI/F converter or whatever very sensitive kit you have. But I have never claimed to actually understand exactly all aspects of computer audio. I just play music, that is what my system is for, it isn't a science research project. I might well buy stuff like cables just because it 'makes me happy' and doesn't cost a very high proportion of my pocket money. Audioquest cinnamon ethernet cable from a reputable brand, is mechanically solid and a pretty red colour and doesn't cost a great deal compared with what my CD player cost or whatever.

I don't mind saying that i don't understand everything about computer audio, and even looking a bit of a fool perhaps. But it does get up my nose when people say differences between USB to SPDI/F converters are 'physically impossible', or 'you need to study information theory' (as been said recently on the Computer based Hi Fi forum), as if they are the Albert Einstein of computer audio or something.

OK, I can buy EMI interference as possibly affecting the audio stages of a DAC, but the likely EMI output of any Ethernet cable is going to be vanishingly small. Plus, any half-competent DAC will be sufficiently shielded.

If that's the worry, I'd be far more concerned about mains-borne interference (measurable and substantive) from things like power supplies. This IS an issue.

Shielding the Ethernet cables isn't frankly going to make a cunt-hair's worth of difference. It doesn't matter of there's spurious radio-frequency stuff in the digital signal, it will be completely ignored and cannot affect the transferred data packets in any way whatsoever.

Sometimes, common sense is appropriate. We are talking about tcp/ip and similar transmission of data packets. When you print a digital picture from your computer, the eventual image will depend on many things, but not in any way whatsoever the connecting cable or the means of data transmission. You don't need a degree in IT technology to understand this.

Many things can/may/might make a difference. The jury is out on a lot. However, transmission of data packets simply physically cannot be affected in any way that will influence the eventual sound. It just is not possible. There aren't any comparisons in 'normal' like I can come up with, but it is fact.

 

[rdale]

So the rough translation of the above is that you are banging on about a subject that you don't understand, and that you are happy to let the audio foo con-men help to spend your ££ on stuff that they know won't improve the SQ of your system?

I am not convinced that random people like yourself posting to this forum understand computer audio any better than I do. You may well have a track record in designing state of the art digital audio equipment, but until if and when I find out you have some sort of track record and expertise, why should I think that you know more than I do?

I don't regard a reputable brand such as Audioquest as 'foo con-men'. I haven't spent a lot of money on my computer audio server and associated cables, and in fact I have spent much less than the average computer audiophile with a Windows tower or Mac Mini or whatever.

 

[rabski]

I am not convinced that random people like yourself posting to this forum understand computer audio any better than I do. You may well have a track record in designing state of the art digital audio equipment, but until if and when I find out you have some sort of track record and expertise, why should I think that you know more than I do?I don't regard a reputable brand such as Audioquest as 'foo con-men'. I haven't spent a lot of money on my computer audio server and associated cables, and in fact I have spent much less than the average computer audiophile with a Windows tower or Mac Mini or whatever.

If they're claiming that an Ethernet cable can have any effect on audio reproduction, then every bit of proven science regarding the way that data transfer over ip protocols works says they are foo con men.

I avidly await any proper proof otherwise.

As does everyone else with even the most basic knowledge of how data transfer over ethernet actually works.

 

[rabski]

From Audioquest's website and in relation to Ethernet cables:

"All audio cables are directional. The correct direction is determined by listening to every batch of metal conductors used in every AudioQuest audio cable. Arrows are clearly marked on the connectors to ensure superior sound quality. For best results have the arrow pointing in the direction of the flow of music. For example, NAS to Router, Router to Network Player."

To make things clear, any criticism and comments I make here are based purely on my own opinion. They in no way reflect the opinion or otherwise of HiFi Wigwam Limited and are made purely on that basis. If any person or corporate identity wishes to take issue with my comments they are free to do so on the basis that in all legal terms, my comments are made purely by myself. I am happy, however, to defend these comments in any way whatsoever.

Now that's out of the way, let me say, as the proud possessor of a doctorate in physics from a fairly well-known university, that is a complete pile of utter bullshit. Directionality in a cable passing internet protocol packets?

Further, "Solid High-Density Polyethylene Insulation ensures critical signal-pair geometry while minimizing insulation-induced phase distortion."

Really? Does it? Exactly what 'phase distortion' is in play here? Phase distortion for packet transfer. My oh my. Cisco and the rest who provide massive data transfer architecture should be mighty worried by this. Except they aren't. And guess why?

Bullshit foo con men you say? Well there's the proof as far as I'm concerned.

 

[rdale]

As does everyone else with even the most basic knowledge of how data transfer over ethernet actually works.

Didn't I just say that I didn't personally think that any differences were to do with data transfer? Clearly we all have a basic knowledge of how data transfer over ethernet works. As did the reviewer of the Audioquest ethernet cables on the Audiostream web site - he was very annoyed he heard a difference when he couldn't think of a reason why it might be so.

I can't say I've heard a difference either. I got back home after being away for nearly 5 months and I had a pile of ripped CDs to listen to in only a week, along with a Bel Canto USB to SPDI/F converter I had bought. But my CD player was certainly sounding better than it had ever done before. Maybe I'll try and do a comparison when I have more time in the future.

 

[rdale]

From Audioquest's website and in relation to Ethernet cables:"All audio cables are directional. The correct direction is determined by listening to every batch of metal conductors used in every AudioQuest audio cable. Arrows are clearly marked on the connectors to ensure superior sound quality. For best results have the arrow pointing in the direction of the flow of music. For example, NAS to Router, Router to Network Player."

To make things clear, any criticism and comments I make here are based purely on my own opinion. They in no way reflect the opinion or otherwise of HiFi Wigwam Limited and are made purely on that basis. If any person or corporate identity wishes to take issue with my comments they are free to do so on the basis that in all legal terms, my comments are made purely by myself. I am happy, however, to defend these comments in any way whatsoever.

Now that's out of the way, let me say, as the proud possessor of a doctorate in physics from a fairly well-known university, that is a complete pile of utter bullshit. Directionality in a cable passing internet protocol packets?

Further, "Solid High-Density Polyethylene Insulation ensures critical signal-pair geometry while minimizing insulation-induced phase distortion."

Really? Does it? Exactly what 'phase distortion' is in play here? Phase distortion for packet transfer. My oh my. Cisco and the rest who provide massive data transfer architecture should be mighty worried by this. Except they aren't. And guess why?

Bullshit foo con men you say? Well there's the proof as far as I'm concerned.

Yes, it all doesn't make any sense to me either. I could believe the arrows if the cable was earthed at one end and not the other.

 

[rabski]

Yes, it all doesn't make any sense to me either. I could believe the arrows if the cable was earthed at one end and not the other.

Which would make sense (one of van den Hull's better ideas) for audio interconnects to help the fight against ground loops, but makes absolutely no sense at all with Ethernet connections.

Anyway, time to bow out gracefully.

Point, I think, has been made.

 

[browellm]

He just goes on and on.

 

[Tony_J]

OK, I can buy EMI interference as possibly affecting the audio stages of a DAC, but the likely EMI output of any Ethernet cable is going to be vanishingly small. Plus, any half-competent DAC will be sufficiently shielded.If that's the worry, I'd be far more concerned about mains-borne interference (measurable and substantive) from things like power supplies. This IS an issue.

Shielding the Ethernet cables isn't frankly going to make a cunt-hair's worth of difference. It doesn't matter of there's spurious radio-frequency stuff in the digital signal, it will be completely ignored and cannot affect the transferred data packets in any way whatsoever.

Sometimes, common sense is appropriate. We are talking about tcp/ip and similar transmission of data packets. When you print a digital picture from your computer, the eventual image will depend on many things, but not in any way whatsoever the connecting cable or the means of data transmission. You don't need a degree in IT technology to understand this.

Many things can/may/might make a difference. The jury is out on a lot. However, transmission of data packets simply physically cannot be affected in any way that will influence the eventual sound. It just is not possible. There aren't any comparisons in 'normal' like I can come up with, but it is fact.

:^

- - - Updated - - -

Didn't I just say that I didn't personally think that any differences were to do with data transfer? Clearly we all have a basic knowledge of how data transfer over ethernet works. As did the reviewer of the Audioquest ethernet cables on the Audiostream web site - he was very annoyed he heard a difference when he couldn't think of a reason why it might be so.I can't say I've heard a difference either. I got back home after being away for nearly 5 months and I had a pile of ripped CDs to listen to in only a week, along with a Bel Canto USB to SPDI/F converter I had bought. But my CD player was certainly sounding better than it had ever done before. Maybe I'll try and do a comparison when I have more time in the future.

There seems to be no limit to the human capacity for self-delusion I'm afraid.