Network Isolation

[tboooe]

On a temporary basis, why don't you take the NAS downstairs and try it directly connected to the PC via the built-in switch. Use the other NIC for connection to the wireless bridge. If that gives you a SQ improvement, then you will know the minimum level of SQ obtainable from a fiber connection.

 

Unfortunately I cannot try your suggestion. My control pc has 2 network ports. 1 is used to connect to the audio pc and the other to the bridge. My old piece of junk nas does not have the ability to act as a virtual switch like the Qnap does.

 

I had thought about getting a music only NAS as well but I am not sure if locating it in my music room wired either to bridge or control pc via fiber and fmc is better then keeping the NAS upstairs and getting the music wirelessly.

[jabbr]

Jabbr, I think we have the same approach, you are just ahead of me with your fiber experience. I'll get there soon enough.

 

My next purchase will be a Skylake build to run Roon with HQplayer. The NAS will be retired, the Asrock q1900 itx will replace it, and both machines will be directly connected with Intel 520 fiber nics, hopefully without a switch.

 

I'll leave the isolation transformer in place on the internet side, as it carries no music signal, it should be good enough for the control information coming wirelessly from the tablet.

 

And yea, the Skylake build will use a mini-atx board.

 

Does this make sense to you?

 

Larry,

 

You clearly have enough experience to get direct point to point network connections working, so I have no doubt you'd be able to optimize your network.

 

In general I would avoid either multihomed networks and/or direct computer to computer connections because these can be more difficult for people to get working if they are not networking experts.

 

Using an SFP switch with LPS will not result in signal degradation, indeed the extra serialization/deserialization step is much like the USB Regen in that it can actually reclock and clean up the Ethernet signal potentially improving SI.

 

I also know that you are particular enough that you would prefer using HQPlayer or your Skylake connected over fiber optic to a low powered, electrically quiet NAA. So at the very least all of the NAS, Skylake and NAA connected by fiber -- this is where I use a small 8 port LPS powered SFP switch. I also LPS my NAA. The advantage is that the Skylake can churn CPU cycles doing HQPlayer up sampling to DSD256 or DSD512, and take advantage of a CUDA coprocessor -- but this is all noisy -- and the NAA, which is electrically connected to the DAC, is optically isolated from the noise.

 

Jon

[lmitche]

Thanks Jon,

 

Thanks, I appreciate the validation. We are indeed aligned in our approach.

 

Larry

[aps]

Really interesting thread. What are the benefits (other than cost) of the model of fibre / SFP vs an Isolation transformer for RJ45 (such as the EMO systems units)? The reason for the question is that migration to the fibre setup will be quite an exercise in my situation.

[tboooe]

Really interesting thread. What are the benefits (other than cost) of the model of fibre / SFP vs an Isolation transformer for RJ45 (such as the EMO systems units)? The reason for the question is that migration to the fibre setup will be quite an exercise in my situation.

 

I have the exact same question as I am trying to figure out which approach to try out (fiber w/FMC or isolation transformer). The idea of adding multiple power supplies required for the FMCs into my system has me leaning towards the passive transformer solution (i.e. EMO systems). However, I am not educated enough to understand if the fiber and transformer solutions are theoretically equivalent or if one is superior.

[tranz]

I found optical LAN better than EMO. Just my experience. My EMO is now used for the DVR.

[lmitche]

Over here, the EMO sounds better then two FMCs powered by two LPSes between the music server and AudioPc. My guess is that the FMCs/LPSes somehow couple with the PCs or add a ground loop.

 

I have yet to try two fiber nics powered by their respective motherboards in a direct connection between the two. One of my two PCs has no PCIE slot so a new build is a perquisite to trying fiber nics. That is likely to be my next upgrade after I purchase a new DAC.

 

The EMOsystems transformers cost $200. I have three more, one on each of the two cable egress devices, my cable modem and cable TV tuner (ceton ETH 6) and one between my wireless NIC and the MusicServer. Isolating the network in the house from the street has added better network performance and improved the quality of the picture on all three HDTVs.

[Animaniac]

On the topic of ground loop "remediation"

Battery Ground Tweak (BGT) - audiodirection

 

Please keep in mind after reading this that the owner is ill and hasn't answered my e-mails.

[aps]

I have the exact same question as I am trying to figure out which approach to try out (fiber w/FMC or isolation transformer). The idea of adding multiple power supplies required for the FMCs into my system has me leaning towards the passive transformer solution (i.e. EMO systems). However, I am not educated enough to understand if the fiber and transformer solutions are theoretically equivalent or if one is superior.

 

Yes, that's pretty much my situation. One specific question is what is the (theoretical) benefit in isolating a NAS if the audio PC is isolated via fibre? I'm struggling to see what additional benefit this implies.

 

Over here, the EMO sounds better then two FMCs powered by two LPSes between the music server and AudioPc. My guess is that the FMCs/LPSes somehow couple with the PCs or add a ground loop.

 

I have yet to try two fiber nics powered by their respective motherboards in a direct connection between the two. One of my two PCs has no PCIE slot so a new build is a perquisite to trying fiber nics. That is likely to be my next upgrade after I purchase a new DAC.

 

The EMOsystems transformers cost $200. I have three more, one on each of the two cable egress devices, my cable modem and cable TV tuner (ceton ETH 6) and one between my wireless NIC and the MusicServer. Isolating the network in the house from the street has added better network performance and improved the quality of the picture on all three HDTVs.

 

Ah - so you got cable from the street? And the biggest benefits is isolating the in-bound signal. Here we've got ADSL2 which sees a phone connection into a modem / router

[jabbr]

Yes, that's pretty much my situation. One specific question is what is the (theoretical) benefit in isolating a NAS if the audio PC is isolated via fibre? I'm struggling to see what additional benefit this implies.

 

 

 

Ah - so you got cable from the street? And the biggest benefits is isolating the in-bound signal. Here we've got ADSL2 which sees a phone connection into a modem / router

 

Not sure there is much theoretical benefit. From a practical POV noise seems to perniciously weave its way between systems connected by wires.

 

What you can do is put in a FMC/fiber link between components and see if that makes a SQ difference in your system. If not take the FMC/fiber out and then try another place.

 

I decided just to replace as much as possible with 10g/1g fiber and be done with it. When I had cable from the street, I isolated it. Now I have 1g fiber. Since that's my external speed, I use 10g optical internally.

[sig8]

Just trying to share something I experienced recently;

 

I have a couple of Mellanox and Intel x520 fiber NIC's, and I use an all fiber network (FMC's, fiber NIC's, fiber switch, etc.) for my 2-pc HQPlayer/NAA setup.

 

I have been using Intel NIC in my NAA machine, and a Mellanox NIC in my HQPlayer machine for no particular reason.

 

I read somewhere that the Intel fiber NIC's maybe of better quality, so I just replaced the fiber NIC in my HQPC from Mellanox to Intel x520. As soon as I put it in, I got little harshness in sound (very little). I thought may be it will go away so I let it burn for 3-4 weeks, but it did not change at all, so I finally gave up last night and put my Mellanox card back in HQPC, and things came back to where they were, and the upper end harshness disappeared.

 

It seems like everything matters, as we have experienced earlier how fiber switches can affect the sound.

 

I have another Mellanox fiber NIC, I will swap that for my Intel x520 in NAA next.

 

YMMV.

[lmitche]

Interesting, please let us know what happens to SQ with the Mellanox in the NAA box.

[ted_b]

Wow, Ashok, great feedback. I have the Mellanox in the HQP machine (thanks to jabbr) so I guess I lucked out. My current NAA has its lone PCIe slot filled with a JCAT card, so it's an FMC there.

[Crom]

Interesting! I've always wanted to replace the clocks and power supplies on the final network card in the chain (in my case the audio player mother board) to see if it has the same effect as doing the same to the usb output card. Never managed to find the time (yet)! I wonder if your find could be down to either the quality of the DC-DC convertors on the board or perhaps a less noisy controller chip.

[kevin1969]

Could someone explain the basic reasoning for all this distrust in playing music from a NAS over a copper Ethernet network? What exactly do people think is occurring that is such a huge problem?

[lmitche]

Ground loops!

[kennyb123]

A huge thanks to those of you who promoted the idea here of isolating a network using fiber optic. This was a big deal in my system.

[kevin1969]

Ground loops!

Exactly what is being affected? The digital signal, the analog, a component? How is it being affected?

[jabbr]

Could someone explain the basic reasoning for all this distrust in playing music from a NAS over a copper Ethernet network? What exactly do people think is occurring that is such a huge problem?

 

Its not that there is distrust. Its just that the fiberoptic sounds better and this isn't expensive. Fiberoptic gives complete galvanic isolation so one of the ways it likely helps is by inhibiting noise from travelling between network devices.

[jabbr]

Just trying to share something I experienced recently;

 

I have a couple of Mellanox and Intel x520 fiber NIC's, and I use an all fiber network (FMC's, fiber NIC's, fiber switch, etc.) for my 2-pc HQPlayer/NAA setup.

 

I have been using Intel NIC in my NAA machine, and a Mellanox NIC in my HQPlayer machine for no particular reason.

 

I read somewhere that the Intel fiber NIC's maybe of better quality, so I just replaced the fiber NIC in my HQPC from Mellanox to Intel x520. As soon as I put it in, I got little harshness in sound (very little). I thought may be it will go away so I let it burn for 3-4 weeks, but it did not change at all, so I finally gave up last night and put my Mellanox card back in HQPC, and things came back to where they were, and the upper end harshness disappeared.

 

It seems like everything matters, as we have experienced earlier how fiber switches can affect the sound.

 

I have another Mellanox fiber NIC, I will swap that for my Intel x520 in NAA next.

 

YMMV.

 

I've been meaning to mention for some time now that I've encountered some bogus/counterfeit Intel x520 cards on ebay. These cards may sort of work but often have problems with bios, and don't, for example, iSCSI boot etc. A good way to verify real Intel is to look for the 'Yottamark' sticker: Yottamark* Sticker on Intel® Ethernet Adapters for Network and I O

 

I wonder if the knockoffs have SQ issues?

 

I've had some problems with Mellanox cards also getting very hot in fanless machines ... YMMV.

[kevin1969]

Its not that there is distrust. Its just that the fiberoptic sounds better and this isn't expensive. Fiberoptic gives complete galvanic isolation so one of the ways it likely helps is by inhibiting noise from travelling between network devices.

 

Call me still confused but how exactly is this noise affecting the sound quality?

 

Noise is nothing other than unexpected frequencies traveling along the same medium. Are you suggesting that these frequencies are altering the digital bit stream on the wire or that these frequencies are somehow being picked up by equipment and then transfer through the equipment into the speakers or somehow confusing equipment from doing its job?

 

I would just like to know where exactly in the interconnect component path that this noise is causing a problem.

 

How does one measure this noise to prove its even there in the first place?

[jabbr]

Call me still confused but how exactly is this noise affecting the sound quality?

 

Noise is nothing other than unexpected frequencies traveling along the same medium. Are you suggesting that these frequencies are altering the digital bit stream on the wire or that these frequencies are somehow being picked up by equipment and then transfer through the equipment into the speakers or somehow confusing equipment from doing its job?

 

I would just like to know where exactly in the interconnect component path that this noise is causing a problem.

 

How does one measure this noise to prove its even there in the first place?

 

Firstly, all theory aside, you should try fiber vs copper Ethernet and use whichever you please. If you can hear a difference great. If any information here helps you, great.

 

There is much written about noise and sound quality -- too much to summarize but read back through the thread. The "optical network" thread also has practical information (not so much debates over theory)

 

I don't have equipment needed to adequately measure signals in the 1-10ghz band. Commonly "eye pattern" is used for networking equipment SI, and I've used published information, eg single mode vs multi mode vs copper.

 

Noise and jitter is always present-- how much difference this makes for your ears: you tell me.

 

What I did was to use published information and specs to give me enough info to spend $100 to give it a try, and liked what I heard, and then added from there.

[sig8]

I've been meaning to mention for some time now that I've encountered some bogus/counterfeit Intel x520 cards on ebay. These cards may sort of work but often have problems with bios, and don't, for example, iSCSI boot etc. A good way to verify real Intel is to look for the 'Yottamark' sticker: Yottamark* Sticker on Intel® Ethernet Adapters for Network and I O

 

I wonder if the knockoffs have SQ issues?

 

I've had some problems with Mellanox cards also getting very hot in fanless machines ... YMMV.

 

My card which came out of HQPC does not have any such mark, so it may be a counterfeit. I will pull out my other Intel NIC from NAA and let you guys know.

 

[jabbr]

Sigh ... what a strange world we live in ... I've heard that many of the counterfeit cards are very good copies but have different firmware, and skimp on things like crystal oscillators, power chips and discrete components.

[jabbr]

Ashok,

 

Could you copy that message along with pictures of the card to the "Optical" thread so we can keep a repository for other folks... trying to all learn from our collective mistakes.

 

Jon