Question for Caelin Gabriel on digital cables

[Puma Cat]

Hi Caelin,
I've been doing some experimenting at home comparing various inexpensive Ethernet cables for my digital streaming setup with Mac Mini connecting via Ethernet-> optical fiber->Ethernet to a Sonore microRendu going to my Schiit Gugnir Multibit DAC (aka Gumby).

So far, I've been comparing cheap generic Ethernet Cat5e vs a well-made, commercially-available cable utilizing Belden 10GX32 Cat 6A cable and Sentinel RJ-45 connectors.

Do you know if its possible, based on construction methodology, strand layup, and materials or dielectrics, or any other factor, etc., for an Ethernet cable to cause an audible "ringing components" in the higher frequencies? Or, higher-order odd-order harmonics, e.g. 7th, 9th, and 11th order harmonics? If so, could these odd-order harmonics accurately be viewed a class of ringing components?

Lastly, could intermodulation distortion be a factor here, as well?

Just doing some experimentation with data cables for streaming music files from a Mac Mini server to streamer, and my scientist-brain is going full-tilt.

Any info or insights you may have would be most interesting. Thanks!

[CGabriel]

Hey Stephen,

As a scientist you appreciate a system analysis approach; specifically, regarding the importance of baseline controls and of variable discipline. It appears from your explanation that you have two different subsystems within the larger audio system (correct me if I am not correct about this). So, I am not going to layout the system block diagram - you are more than capable of doing that.

The sub-system that you are attempting to optimize is the [computer => DaC] interface. But in doing so you are actually directly comparing two completely different transmission methods: UsB and Ethernet. Each method has multiple variables including devices and cables. In the case of the USB method, you have achieved an exceptional high level of performance with the introduction of the Alpha USB cable (exotic and expensive). So while it may be fair to use the overall sonic performance you are obtaining as a reference, it would not be fair to compare that to a completely different subsystem that has not been worked out and optimized. The TP link and blue jeans cables are quite inexpensive and may be not be reprentative of what the Ethernet method is ultimately capable of achieving. Therefore, I would treat the Ethernet subsystem as a separate and isolated experiment in optimization. Then, once you are satisfied that you have achieved the maximum from that approach - then it would be fair to compare whole subsystems to one another.

My approach would be to establish a “simple” baseline control for the Ethernet subsystem with emphasis on the simplest possible number of variables to deal with. Ensure that your baseline is valid before moving on to the introduction of more variables (devices and cables).

I can talk more, later, about my opinion on your specific questions.

[Puma Cat]

Thanks, Caelin!

Yes, you are exactly right, I am working optimize two different data transmission subsystems within on overall system functional architecture from "Server to DAC". I do believe I've got, as you say, subsystem 2, Streamer to DAC, well optimized within my present budget; the streamer could be upgraded, but performance is currently "fit for purpose" at the present time.

It just occurred to me while writing this, that Installing the Shunyata Alpha USB cable, and running an excellent higher-gauge power cord, my Shunyata Black Mamba CX to my multibit DAC from the Triton, has effectively provided significant attenuation of subsystem 2 noise factors so that the components are capable of realizing their full engineering specification of performance (in fact, many Shunyata products do this throughout the audio system).

And you're correct, I've gone back to my "baseline" (datum) and as you recommend, minimizing system variables (control factors in systems engineering parlance) as a starting point: Mac Mini and Ethernet over Power adapter connected via generic Cat 5e cables to the ethernet ports on the back of the Pace router. This has restored the exceptional sound quality I've been enjoying since installing the 7M run of Tripp-Lite optical fiber from the dining area FMC to the FMC in the audio rack, using a quality ethernet cable (presently an AudioQuest Cinnamon) from the downstream FMC to the Sonore streamer, and installing the Alpha USB from the Sonore to Gumby.

I then started doing OFAT (One Factor AT a Time) experiments last night, the first experiments have already determined that connecting the Mac Mini and bedroom EoP adapter to the inexpensive TP-Link Gigabit Ethernet Switch is putting noticeable and nasty noise components into subsystem 1; this is something to definitely think about.

Hans Beekhuizen has a good video on YouTube about the AQVox Ethernet switch and notable improvements it brings. Link here: https://youtu.be/gwTATKd69FQ

So, from a systems engineering perspective, the Ethernet switch, if needed or used, is a control factor that needs to be considered, because, like many control factors, it can also produce not inconsiderable, and clearly audible, noise components.

I will put up a systems architecture diagram as point of departure for discusion later today.

Looking forward to continuing this very interesting discussion.

Cheers and many thanks for the reply.

[W9TR]

I’d add that along with the Ethernet switch the cables themselves may make a difference in terms of the noise they may conduct into the audio system.

Specifically, moving up the Cat 5e / Cat 6 / Cat 7 path may cause trouble as the higher spec’d cable may use RJ-45 connectors that employ metal shields connected to the cable shield drain wire.

If your Ethernet components also have metal shielded Ethernet jacks, you will create a direct noise path into your system.

By design Ethernet data interfaces are already transformer coupled or otherwise galvanically isolated. By adding shielded Ethernet cables, you may create a path for noise to enter your system’s ground.

So it is worth trying some high quality bonded Ethernet cables but also ensuring that if the connectors have metal shells they are not connected to the cable shield. You can verify this with an ohmmeter.
Tom

[CGabriel]

So far, I've been comparing cheap generic Ethernet Cat5e vs a well-made, commercially-available cable utilizing Belden 10GX32 Cat 6A cable and Sentinel RJ-45 connectors.

Don’t under estimate the Belden cable. We used certain industry cables as a baseline in our Ethernet cable research tests. Certain “common industry” cables outperformed many well known “audiophile” cables.

My recommendation would to find the absolute best performing Ethernet cable that you can obtain. This may take some time to test several candidates. Even a relatively expensive Ethernet may be less expensive than all the extra devices and multiple cables needed to do a copper-optical-copper conversion. It is quite possible that with the optical solution you could end up with worse performance than with a simple Ethernet cable.

Supposedly, going optical solves the galvanic isolation issue. But - is it really an issue in your system? How would you know if there is any correlation to galvanic isolation and perceived sonic performance? Even if it does contribute to a performamce loss, it introduces several tangential variables all of which are known to be possible problems. High on that list are the two wall warts you will plugging in to your power strip. Just the act of plugging those into the same circuit as your audio system may degrade performamce. Additionally, you now have two ethernet cables at each end of the chain to deal with and the quality of each of those does affect sound quality.

Added system complexity and the variables associated mathematically reduces the possibility of optimal performance without a lot of testing and attention to detail.

[For The Love of Music]

Sounds more complicated veering off course with the fundamentals we have been taught along with experience

[CGabriel]

What fundamentals?
What course?

[Puma Cat]

Originally Posted by W9TR

I’d add that along with the Ethernet switch the cables themselves may make a difference in terms of the noise they may conduct into the audio system.

Specifically, moving up the Cat 5e / Cat 6 / Cat 7 path may cause trouble as the higher spec’d cable may use RJ-45 connectors that employ metal shields connected to the cable shield drain wire.

If your Ethernet components also have metal shielded Ethernet jacks, you will create a direct noise path into your system.

By design Ethernet data interfaces are already transformer coupled or otherwise galvanically isolated. By adding shielded Ethernet cables, you may create a path for noise to enter your system’s ground.

So it is worth trying some high quality bonded Ethernet cables but also ensuring that if the connectors have metal shells they are not connected to the cable shield. You can verify this with an ohmmeter.
Tom

Thanks, Tom! This is very interesting information. Currently, the only cable I have that has a metal shield is the Audioquest Cinnamon cable that goes from my FMC to the Sonore microRendu. I'll see if I can determine if the metal shield is connect to the cable shield drain wire (somehow..). All the other cables have plastic RJ45 connectors.

[Puma Cat]

Originally Posted by CGabriel

Don’t under estimate the Belden cable. We used certain industry cables as a baseline in our Ethernet cable research tests. Certain “common industry” cables outperformed many well known “audiophile” cables.

My recommendation would to find the absolute best performing Ethernet cable that you can obtain. This may take some time to test several candidates. Even a relatively expensive Ethernet may be less expensive than all the extra devices and multiple cables needed to do a copper-optical-copper conversion. It is quite possible that with the optical solution you could end up with worse performance than with a simple Ethernet cable.

Supposedly, going optical solves the galvanic isolation issue. But - is it really an issue in your system? How would you know if there is any correlation to galvanic isolation and perceived sonic performance? Even if it does contribute to a performamce loss, it introduces several tangential variables all of which are known to be possible problems. High on that list are the two wall warts you will plugging in to your power strip. Just the act of plugging those into the same circuit as your audio system may degrade performamce. Additionally, you now have two ethernet cables at each end of the chain to deal with and the quality of each of those does affect sound quality.

Added system complexity and the variables associated mathematically reduces the possibility of optimal performance without a lot of testing and attention to detail.

Thanks, Caelin. I'll reply with more details, including the cable configuration, and rationale for the run of optical fiber tomorrow. What I will say for now is the two small FMCs and run of Tripp-Lite fiber only cost $65, and brought a notable reduction in the noise floor compared to the 7 meter run of cheap Cat 7 Ethernet cable I was using.

Using the level of improvement that Shunyata Alpha USB brought to the system as a reference, and setting its "level" of improvement at 100% I would say installing the run of optical fiber using the FMCs gave me another 15-20% of what the Alpha USB brought to the table.

As for the wall warts, the upstream FMC is using the TP-link supplied wall wart; I know its sub-optimal.

At the downstream end connecting from the FMC to the Sonore, I am using my iFi iPower 9V ultra-low noise LPS for the FMC. Its plugged into my Hydra 4, also. So, it should be pretty quiet.

The Cat 6A Belden-based cables I tried were Blue Jeans Cable, and I did not like the sound of them in my room and with my system and speakers at all. They were particularly bad when used in conjunction with the $15 TP-Link Ethernet switch.

Thanks so much for your comments; this is a very interesting discussion.

More background and info tomorrow.

[Puma Cat]

Originally Posted by CGabriel

What fundamentals?
What course?

LOL..



Yup. And, I should add that this is all an experiment...'cause that's what scientists do!

If I had a metrology system in place to measure S/N, the noise floor, and/or any distortion or noise components, I would be doing a full-on DOE using the JMP 14 Custom Designer. Most likely, four factors at 2 levels with quadratic terms in the model to examine curvature of the response surface(s).