Fibre-optic ethernet connection

[davewantsmoore]

1 hour ago, frednork said:

The kinds of characteristics that might be found in specifications

It isn't the kind of thing which can be converted in a "single number" kind of spec.

 

Decent quality streamers should be immune to electrical noise carried over the ethernet cable.   If not, you could get a a streamer with digital outputs, and then a DAC which does aggressive reclocking.

[blybo]

5 hours ago, frednork said:

What about discussing more broadly the characteristics of a well setup streaming rig with a focus on the network side ie modem- characteristic a b c, cable abc  router, abc, server abc which gives me/us a better chance of finding something that is more likely to work well.  The kinds of characteristics that might be found in specifications would be best (or something else?). Is that doable? 

I'm no expert but 1 thing I did was turn off the wifi in my wifi modem/router (bridge mode?) and then put a separate wifi router in another central park of the house to deal with all our wifi devices. The streaming part of my network is all wired  with cat 5 and due to NBN recently being installed in a less than ideal location my cable runs have almost doubled and I had to add another switch. I have not found a noticeable difference with the increase in runs.

 

I can't say I could tell a big difference with isolating the wifi from my modem however the network has been more stable ever since and almost never requiring a re-boot. I use domestic linksys network gear.

Edited September 10, 2019 by blybo

[frednork]

4 hours ago, davewantsmoore said:

It isn't the kind of thing which can be converted in a "single number" kind of spec.

 

Decent quality streamers should be immune to electrical noise carried over the ethernet cable.   If not, you could get a a streamer with digital outputs, and then a DAC which does aggressive reclocking.

Not sure if you are but are you suggesting as long as your streamer is "immune" the rest upstream doesnt matter?

[Guest rmpfyf]

10 hours ago, davewantsmoore said:

The trick is that this doesn't require expensive parts, or particularly fancy things.... just good engineering.   It could be found in a device which costs hundreds of dollars, not thousands.   Although if you have good performance, then it's typical to charge as much as you can for it.

 

Agree all points but wary on magnitude here.

 

An iancanada isolator/FIFO/reclocker rig might cost $1k at DIY prices once you've got the clocks, power supplies, etc going.

 

Yes, parts might've less in production but development would be longer. It's not under $1k cost in production.

 

Plus markup, plus etc... Similar stuff is only going to appear in specialist or exy gear.

 

Whilst I don't agree with the rationale you can see why there's a thriving industry of sub $1k band aids.... They're actually affordable, and to many degrees accessible.

Edited September 10, 2019 by rmpfyf

[davewantsmoore]

3 hours ago, frednork said:

Not sure if you are but are you suggesting as long as your streamer is "immune" the rest upstream doesnt matter?

Not sure exactly what you mean.

 

If you had a "streamer" with an internal DAC, which accepted data over the network .... The streamer device might be susceptible to electrical noise entering via the ethernet cable ..... but otherwise, then yes, everything upstream will make no difference (or at least could be configured to, but doesn't need to).

 

[davewantsmoore]

55 minutes ago, rmpfyf said:

An iancanada isolator/FIFO/reclocker rig might cost $1k at DIY prices once you've got the clocks, power supplies, etc going.

Off the shelf products are much more "neatly" designed than this... and more cost effective.

 

I'm not saying they would/wouldn't outcompete what Ian did.... but you don't need to do what Ian did to get excellent isolation.

 

As an off the cuff example.... I have a firewire interface, which uses dice/jertpll .... it isn't the last word in performance, but it doesn't care what you plug into it.   It's got 8in/out analogue and heaps of digital mixing, mic preamps, and everything for <<$2000.

 

[Guest rmpfyf]

40 minutes ago, davewantsmoore said:

Off the shelf products are much more "neatly" designed than this... and more cost effective.

 

I'm not saying they would/wouldn't outcompete what Ian did.... but you don't need to do what Ian did to get excellent isolation.

 

As an off the cuff example.... I have a firewire interface, which uses dice/jertpll .... it isn't the last word in performance, but it doesn't care what you plug into it.   It's got 8in/out analogue and heaps of digital mixing, mic preamps, and everything for <<$2000.

 

$2k and it makes too much sense to be covered in audiophile sparkle dust.... You've proven my point

[Zardoz]

This stuff is my day job.

 

Fibre is useful in the following situations:

1. If we need greater performance that UTP cannot provide.  This includes 25GbE+ in most cases, or 10GbE if we're connecting older gear.

2. Power constraints. 10GbE's PHY's are power hungry; 10GbE multimode fibre SFP+ modules will dissipate less heat and use less power.  

3. Electrical isolation, including links between buildings.  This is A Very Good Idea.  If you're running ethernet out to the shed, it's probably a good idea to run fibre.

4. If we need to go long distances up to 1G, or not so long distances at >1G.

5. If we're running temporary high speed networking over a long distance at an event.  A 50m fibre cable weighs a lot less than 50m of cat6.

 

Some media converters have poor quality power supplies.  They supply sufficiently clean power to operate the conversion process (this is bit perfect or else the network packets would not be decoded on the other end) but the power supplies are typically cheap switch mode that can introduce interference into your power supply.  I'll leave the impact up to the experts, but this is measurable in some cases.  I believe that if you have audio gear that will ingest this noise and give you an audible result, you should look at better gear.  I'd prefer to stick to UTP coming out of a Cisco switch any day of the week.

In addition, adding fibre introduces both design and operational complexity in your system.  More parts to go wrong, more things to maintain and operate and more things to introduce potential noise into a system (putting aside the debate of whether this matters or not).

 

The optical SPDIF vs coax SPDIF argument holds a LOT more weight than the fibre ethernet argument ever will.

 

[Cruncher]

If someone at a hypothetical dinner party said to me that they replaced a copper based Ethernet system with a Optical based system and they felt Optical was better, but a friend "who knew a lot about computers" said it could not be because in Ethernet 'bits are bits'. I have no problem agreeing with them, particularly when I asked about the system and it was a home system and the budget was $200 -$500. I would not expect this with a $20,000 - $50,000 system.

 

Why do I think in theory they are the same but in practice Optical can be better than than Copper.

 

The way I think about things is the world is analogue. We create schemes (e.g ASCII code, Morse code, A/D) to turn the world into Binary ( 1 or 0) and then encoding / decoding schemes to overlay Binary onto a analogue world.

 

In Ethernet, the binary / analogue world bump into each other at ISO Layer 1. Another influence on my thinking was a design engineer who's principal I  adopted was to build  " Accurate transmitters, forgiving receivers". 

 

In textbooks for Ethernet on copper there is always diagrams showing lovely square waves. In the picture below, on the far left side is what a signal actually looks like on a copper cable at the receiving device.

 

The traffic is being encoded by the transmitter using a Tomlinson-Harashima Precoded (THP) version of pulse-amplitude modulation with 16 discrete levels (PAM-16), encoded in a two-dimensional checkerboard pattern ( 128-DSQ +LDPC).  I know what some you are thinking, what a co-incidence, you were only discussing the Tomlinson-Harashima Precoded variant of Pulse Code Modulation at the dinner table the other day and next thing it appears in SNA.

 

You can see the PAM-16 signals (the scheme uses 16 voltage levels) on the right hand side slowly emerging at each stage from the input on the left hand side.

 

When I look at the incoming signal on the left hand side I am amazed anyone or anything can recover a signal let alone 16 from the incoming mess. Then I understand the susceptibility to EMI etc. and poorly made cables.

Don't fret there is tons more of this stuff (including all the mathematics) in the Ethernet Standards WorkGroup meetings, IEEE standards, manufactures and Alliance documents. Figure 4 above is only the first of multiple steps. The signal processing required is immense and is usually done in a DSP which means more heat, more power, more cooling, more cost.  Ethernet on copper since it uses 8 conductors each end needs 4 transmitters and 4 receivers which feed into the DSP.

 

On the other hand it's twin, using Optical transport is pretty simple to explain.

 

Each end has one transmitter and one receiver. It uses a Binary e.g light / no light. It uses the simple 64/66 Byte sub-layer encoding scheme. Transmitters and Receivers are much simple. Above 10GE ( 40GE, 100GE, 400GE) there is no Ethernet over copper and there will never be. For short distances you can optimize cost by using much cheaper optics and even clear plastic instead of glass for the cable. Optical cables electrically and magnetically decouple the sender and receiver. The do not give off any emissions ( no cross talk between cables or effect on other devices) and and not effected by external signals / noise. This isolation works for all speeds.

 

In theory all Ethernet is the same, in practice and implementation they can be different.

 

 

 

 

 

 

Edited May 21, 2020 by Cruncher

[pwstereo]

12 hours ago, Cruncher said:

In theory all Ethernet is the same, in practice and implementation they can be different.

Excellent information and explanation, thank you.

[Guest rmpfyf]

I wouldn't focus on the data integrity, whether fibre or copper it's going to get there just fine in home.

 

On 20/09/2019 at 10:47 PM, Zardoz said:

This stuff is my day job.

Mine too, but in different ways. Shortest run I've ever seen is 2m. It's also common in industrial batteries (high EMI environment). 

 

There is nothing to lose in trying fibre, there are other things to consider other than data corruption.

 

Would agree the usual media converters have poor supplies.

[Stereophilus]

On 10/09/2019 at 12:47 PM, davewantsmoore said:

Decent quality streamers should be immune to electrical noise carried over the ethernet cable.   If not, you could get a a streamer with digital outputs, and then a DAC which does aggressive reclocking.

Im a bit comfused by this statement. 

 

Streamers are essentially network connected simple computers with digital outputs.  Whilst they might have ways to help reduce network noise entering (eg optical network input like the Sonore OpticalRendu) they are inherently internally noisy devices, like all computers.  Their product (the rendered sound file) is extremely jitter sensitive.  Or, put another way, highly sensitive to time domain interference of the data output.

 

Most design decisions relating to streamers are a series of compromises.  If you put an optical converter in your streamer, it can inflict a penalty of generating its own noise within the streamer.  Same with adding wifi modules within streamers.  Put a streamer right next to your DAC, it has the potential to negatively affect the DAC through the noise it generates.  Put it further away and you have greater susceptibility to jitter again.

 

So when i say im confused by your statement, im really asking how you percieve the design of streamers is so simple as to just eliminate all this potential compromise and complexity?

[Stereophilus]

11 hours ago, rmpfyf said:

There is nothing to lose in trying fibre, there are other things to consider other than data corruption.

 

Would agree the usual media converters have poor supplies.

It is bitterly ironic to me that when i advanced the same conclusions in my earlier thread, following my simple experiment, the thread was closed due to a few angry trolls.

[Assisi]

On 20/05/2020 at 9:08 PM, Cruncher said:

Each end has one transmitter and one receiver. It uses a Binary e.g light / no light. It uses the simple 64/66 Byte sub-layer encoding scheme. Transmitters and Receivers are much simple. Above 10GE ( 40GE, 100GE, 400GE) there is no Ethernet over copper and there will never be. For short distances you can optimize cost by using much cheaper optics and even clear plastic instead of glass for the cable. Optical cables electrically and magnetically decouple the sender and receiver. The do not give off any emissions ( no cross talk between cables or effect on other devices) and and not effected by external signals / noise. This isolation works for all speeds.

 

In theory all Ethernet is the same, in practice and implementation they can be different.

I play my music over a network. No USB. My understanding of how Ethernet works is virtually non existent. I am often surprised that my network functions as it does. I have an audio quality switch. I have tried two optical converters in the link from the NAS or the Roon core to the switch or from the switch to a network player. The inclusion of the converters does make a difference. I have not been convinced that the difference is beneficial.

 

 

Firstly a preamble. I obtained wonderful new speakers a few weeks ago. They were installed with no other changes to the system set up. Settling in of the speakers took time. Twelve days ago the entire system was pulled apart to try to tidy up cables and other things a bit. All the many accessories in the system were left out when the system was put back together again including the converters. The accessories were left out because the system will be pulled apart again at the week end for the installation of an isolated cable from the switch board to the system.

 

 

I was impatient so the other day I put the converters back in one morning. Because of the change in the speakers and the missing accessories I had been taking time to adjust to the new sound. The converters seemed fine. I took the converters out that night. No question. They are not beneficial in my system at all. They provide detail only. What is missing is weight, a solid fullness of sound, density. The new speakers have helped me realise that the converters are not beneficial for me. Without the important missing elements the overall SQ is less. The reason for the dis benefit to me is not obvious. It has been said to me that the reason is that at every point of conversion there can be a minimal loss of something in the audio signal. ( my words).  Too many steps. I do not know. It would be interesting as to what the situation would be if there were less steps in the links with just optical core to player for example.

 

John

 

[Guest rmpfyf]

1 hour ago, Stereophilus said:

It is bitterly ironic to me that when i advanced the same conclusions in my earlier thread, following my simple experiment, the thread was closed due to a few angry trolls.

Bloomin ell, that was a little harsh.

 

39 minutes ago, Assisi said:

 Too many steps. I do not know. It would be interesting as to what the situation would be if there were less steps in the links with just optical core to player for example.

This is my soon to be rig - optical from switch to PC, dedicated card, independent PSU.

 

But most consumers aren't rigged up for fibre - let alone wired Ethernet - so we adapt. Most of what's in the audiophile space is just adapting.

[Stereophilus]

4 minutes ago, rmpfyf said:

Bloomin ell, that was a little harsh.

Not a dig aimed at you in any way... Just sticks in my craw that I can put forward experience relating to fibre converters in my hifi many months back, only to have the thread closed for saying essentially the same thing as what you have stated in this thread.

[Cruncher]

I don't want my post to come across the Optical solves all problems.

[Guest rmpfyf]

8 hours ago, Stereophilus said:

Not a dig aimed at you in any way... Just sticks in my craw that I can put forward experience relating to fibre converters in my hifi many months back, only to have the thread closed for saying essentially the same thing as what you have stated in this thread.

Well.., that does suck. Sorry to read.

 

8 hours ago, Cruncher said:

I don't want my post to come across the Optical solves all problems.

Can be part of a better implementation, but nothing's a panacea.

[Cruncher]

 

20 hours ago, Assisi said:

My understanding of how Ethernet works is virtually non existent.

Nor should you really have to know how it works.

We supplied a finance application to a small company who were experiencing ‘random’ outages that was bringing the business to it’s knees. Troubleshooting went like this : Finance Application -> DB -> Storage -> Network ( which was dropping up to 30% traffic)

.

They found the culprit and it looked a bit like this ….

 

I am not really a cabling guy but even I can pick up fake Cat6 cable, uneven length wires, 8 cm untwist ( untwist should be in mm, will cause NEXT / FEXT), a nice big kink in the middle and the list goes on. Under instructions from management, the cable was cut in four and sent to the four corners of the earth.

It turned out a tech had decided to “save the company money” and not buy those expensive ( $30 ?) retail cables so he bought a spool of Cat6 cabling for a bargain price, a bag of Ethernet connectors and a terminating tool, all off the Internet.

I was amazed their system worked at all.

For most people, Ethernet just works. When there is problem there is usually logical explanation.

 

 

[pwstereo]

1 minute ago, Cruncher said:

It turned out a tech had decided to “save the company money” and not buy those expensive ( $30 ?) retail cables so he bought a spool of Cat6 cabling for a bargain price, a bag of Ethernet connectors and a terminating tool, all off the Internet.

I had a similar experience at a manufacturing company. A young sysadmin taught himself how to make Ethernet fly leads from a You Tube video.

He seemed to have completly missed the 586A/B pinout part of the tutorial and had random colour/pair assignments for his connectors

[Cruncher]

1 hour ago, pwstereo said:

I had a similar experience at a manufacturing company. A young sysadmin taught himself how to make Ethernet fly leads from a You Tube video.

He seemed to have completly missed the 586A/B pinout part of the tutorial and had random colour/pair assignments for his connectors

 "Oh, by the surprised expression on your face, you can not believe that I made this myself ! after only watching half of a YouTube video  "

 

 

[davewantsmoore]

On 21/05/2020 at 8:23 PM, Stereophilus said:

Im a bit comfused by this statement. 

 

Streamers are essentially network connected simple computers with digital outputs.

Some have analogue outputs (ie. a DAC inside), but let's talk about digital outputs....

 

On 21/05/2020 at 8:23 PM, Stereophilus said:

they are inherently internally noisy devices, like all computers.

The amount of noise they generate and/or are suceptible to (note: these are different things) is quite variable.

 

On 21/05/2020 at 8:23 PM, Stereophilus said:

Their product (the rendered sound file) is extremely jitter sensitive.

That depends...  on the DAC, and everything inbetween.

 

The digital signal output from the streamer, contains jitter ..... but it depends on how sucptible downstream stages are to it and/or if any downstream stages "reclock" the jitter away.

 

 

On 21/05/2020 at 8:23 PM, Stereophilus said:

Most design decisions relating to streamers are a series of compromises.  If you put an optical converter in your streamer

Converting ethernet to optical "inside" (or at the edge) of a streamer device would be a dumb way to make the device immune from electrical noise carried in on the ethernet cable.    

 

On 21/05/2020 at 8:23 PM, Stereophilus said:

it can inflict a penalty of generating its own noise within the streamer

It could.... it doesn't need to though.    I assume you're referring to "noise generated in the power supplying the optical>electrical ethernet converter" .... in which case, just supply it with a very good quality power supply.

 

On 21/05/2020 at 8:23 PM, Stereophilus said:

Put a streamer right next to your DAC, it has the potential to negatively affect the DAC through the noise it generates.

 It doesn't need to.  This is just all engineering. 

 

On 21/05/2020 at 8:23 PM, Stereophilus said:

So when i say im confused by your statement, im really asking how you percieve the design of streamers is so simple

I didn't say it was simple.

 

What I said was that a good quality streamer will (should) be eginered with mechanisms to make it immune from a sensible amout of "noise" carried in on the ethernet cable.     ie.   I should be able to inject varying amount of noise, and have the device continue to work without any change in it's (digital signal output) performance.

 

Audiophles often talk like better quality devices will be "more" suceptible.

[Guest rmpfyf]

10 hours ago, davewantsmoore said:

Converting ethernet to optical "inside" (or at the edge) of a streamer device would be a dumb way to make the device immune from electrical noise carried in on the ethernet cable.    

Thankfully someone said it.

[Stereophilus]

9 hours ago, davewantsmoore said:

Some have analogue outputs (ie. a DAC inside), but let's talk about digital outputs....

Let’s move beyond semantics.  We both understand that streamers can have DACs included and that there are “streaming DACs”.  The part that “streams“ is a network connected processor which sends a rendered sound file to a connected DAC.

9 hours ago, davewantsmoore said:

The amount of noise they generate and/or are suceptible to (note: these are different things) is quite variable.

But the amount of noise (EMI, RFI, or whatever) they generate is never zero.  And their susceptibility to noise is also not zero, although I understand this second point may be contentious.

9 hours ago, davewantsmoore said:

That depends...  on the DAC, and everything inbetween.

To some degree, but again, not absolutely.  I am of the opinion that there is no DAC that is completely immune to a “poor” digital feed.  The is no digital feed (the “inbetween” as you put it) that is completely immune to potential causes of jitter.

9 hours ago, davewantsmoore said:

Converting ethernet to optical "inside" (or at the edge) of a streamer device would be a dumb way to make the device immune from electrical noise carried in on the ethernet cable.    

And yet such a product exists with exactly that claim (Sonore OpticalRendu).  I don’t necessarily believe their claim, but these are the engineering decisions and compromises that some streamer manufacturers are adopting.  Some reports on these devices are very favourable... However “dumb” the principle may seem.

9 hours ago, davewantsmoore said:

It could.... it doesn't need to though.    I assume you're referring to "noise generated in the power supplying the optical>electrical ethernet converter" .... in which case, just supply it with a very good quality power supply.

There is also the electrical noise, EMI, RFI generated by the conversion as well.  But yes, a good power supply will help as well.

9 hours ago, davewantsmoore said:

It doesn't need to.  This is just all engineering.  
I didn't say it was simple.

Ok... so “complex” (opposite of simple) engineering can mitigate much of the noise and jitter issues within streamers and DACs.

My perception here, and without any particular prejudice, is that even at the highest echelon of consumer DAC and streamer technology (dCS, MSB, Esoteric, etc) there is still audible differences in sound quality resulting from the digital stream into the DAC.    Which is also to say, again contentiously, that even the most diligently engineered (re expensive) DACs are not immune to upstream jitter/noise.... Some manufacturers even go on record stating they prefer their USB over their I2S input, or vice-versa.  
 

The point being this perceived ideal DAC, immune to upstream noise/jitter, is something of a fallacy.  Hence why audiophiles look upstream and modify and notice differences when they do.  
 

Not wanting to derail the thread here, but if someone know of a DAC that is completely immune to what is upstream of it, then please share and we can all dispense with needless optical isolation of networks, USB reclockers, fancy streamers, shielded Ethernet cables and everything else that improves SQ when used on every other DAC.  

[davewantsmoore]

1 minute ago, Stereophilus said:

Let’s move beyond semantics.

I wasn't trying to get hung up on smantics.  I just would have said slighty different things depening on whether we were talking about box with a DAC inside.... or a box with just a digital audio outputs.

1 minute ago, Stereophilus said:

To some degree, but again, not absolutely.  I am of the opinion that there is no DAC that is completely immune to a “poor” digital feed.  The is no digital feed (the “inbetween” as you put it) that is completely immune to potential causes of jitter.

It really depends.

1 minute ago, Stereophilus said:

And yet such a product exists with exactly that claim (Sonore OpticalRendu).
I don’t necessarily believe their claim

That it dispenses with any electrical noise that would otherwise be carried in on an ethernet cable? .....   No, that's a fine claim for them to make.

1 minute ago, Stereophilus said:

Ok... so “complex” (opposite of simple) engineering can mitigate much of the noise and jitter issues within streamers and DACs.

I didn't say it was "complex" either.

 

1 minute ago, Stereophilus said:

that even the most diligently engineered (re expensive) DACs are not immune to upstream jitter/noise

Indeed....  but it all depends on what type of jitter.

 

When noise entering the system via ethernet is changing the analogue output .... then the streamer (or whaver) device is defective for passing that into the digital audio that it was so big that it materially affected the DAC .....   or (in the case the digital audio source was not terribly affected by the ethernet bourne noise) the DAC is defective for being affected by a (so small) error in the digital audio.

 

1 minute ago, Stereophilus said:

The point being this perceived ideal DAC, immune to upstream noise/jitter

The DAC will not be immune from any jitter (it depedns what).     However a streamer should be immune from passing noise entering via ethernet into the digital audio signal in any relevant magnitude.

 

1 minute ago, Stereophilus said:

Not wanting to derail the thread here, but if someone know of a DAC that is completely immune to what is upstream of it

I was talking about a steamer with (only) digital outputs .... and how it's digital audio output signal should be immune from noise carried in via ethernet.

 

Moving the context to  "a DAC that is completely immune to what is upstream of it" .... isn't the same.