Shielded vs. unshielded Ethernet and Grounding

[jabbr]

1 hour ago, Superdad said:

And they don’t do anything to block AC leakage currents—which John has measured traveling over Ethernet lines.  That’s an important factor to address for audio—especially when one purpose of using small end-point Ethernet renderers is to provide a truly isolated music source.

 

Medical/hospital Ethernet isolators absolutely are exactly designed to block leakage currents. In fact that's their entire purpose.

 

The standard Ethernet PHY transformers are fine to block high level voltages/currents.

 

Imagine an ICU setting where there is a catheter in the heart which is measuring (electrical) cardiac signals -- imagine what a very low level (AC) -- perhaps microamp --  leakage current might do! Leakage currents are literally a life and death issue.

 

 

[jabbr]

9 hours ago, Em2016 said:

 

Is this stuff unshielded? There's no mention. But can you tell by the specs mentioned?

 

 

There are different types of shielding "S" refers to a screen shield, "F" refers to a foil shield and "U" refers to no shield.

The shield can be around the entire cable, or around each twisted pair (TP) hence the nomenclature:

 

S/STP -- the first "S" is the outer screen, the second "S" are screen shields around each twisted pair (TP).

U/STP -- no outer shield, each twisted pair has a screen shield

S/FTP -- there is a screen around the cable, and foil around each twisted pair.

etc...

 

https://www.belden.com/blog/digital-building/stp-utp-ftp-cable-more-7-types-when-to-use-them

[jabbr]

9 hours ago, Miska said:

Do NOT use shielded ethernet cables with audio equipment! Only use standard inexpensive CAT6 UTP patch cables - those are the best.

 

Cat 6 U/UTP is best for audio (10Mb -> 1GbE)

The shields around each twisted pair reduce crosstalk between the twisted pairs and are used with 10 Gbe... e.g.

U/FTP

For the vast majority of home applications there is no need for outer shielding -- particularly with bonded twisted pair there is not significant EMI emission ... for example your Ghz Wifi works great right next to a Gb Ethernet cable -- nor is there significant RF pickup.

 

[jabbr]

3 minutes ago, mansr said:

Ethernet jacks/transformers are not generally rated according to IEC 60601, the standard specifying safety requirements for medical electrical equipment. A random TE Connectivity datasheet I checked doesn't even mention the relevant parameters.

 

For such purposes, you'll need a barrier much closer to the probe, or else it will be disturbed by noise generated within the equipment itself.

 

Yeah in fact, the standard specifies allowable leakage currents as well as methods of testing: https://www.mddionline.com/leakage-current-standards-simplified

 

In the USA, the equipment itself is regulated by the FDA.

 

Those old heavy "ultra isolator" low leakage current transformers were not made with audio in mind

 

[jabbr]

29 minutes ago, Superdad said:

 

Sorry, I misspoke. Of course the medical EN isolators block low-impedance (touch current) leakage.

John tested a popular one that someone sent him and found that it did not block high-impedance leakage.  Can't seem to find his post right now, but he did use his leakage test setup on it--during the time he was comparing switches--and reported that the medical isolators were only effective with the low-impedance leakage.  Not that it is difficult to design a switch that shunts the high-impedance leakage.

 

As I said before, there is more than one issue with leakage current. "Touch current" is the easiest to understand, and most widely applicable, and ultimately you don't want your 2 year old kid to get shocked either by a low impedance path, but there are more stringent situations as well, i.e. the cardiac compromised ICU patient with an electrode actually inside the heart. In the most sensitive situations e.g. brain monitors, actual fiberoptic sensors are preferred due to their blockage of even higher impedance leakages!

 

What are you defining as "high impedance"? IEC 60601 defines the test mechanism which requires a minimum >1 M ohm input impedance on the meter. If you look at the literature which if course exists and is fairly extensive, issues with even higher input impedance e.g. Gigaohm are discussed, as well as issues arising from insulator breakdown (can be in the Teraohm range). That can indeed be hard to measure!

 

I can't speak for the Ethernet isolators that you tested, but Topaz ultra-isolators have certainly been used in healthcare environments and are known to have low parasitic capacitance e.g. < 0.0005pF etc.

 

In defense of the IEC 60601 it does really make sense to measure the net leakage at the end of the day rather than merely the individual components because the circuits can get complicated.

[jabbr]

Let me expand a bit:

 

Every current path is governed by the equation V = IZ (Z=R at DC). Nothing magic about leakage current, they just travel along a certain path, typically one that involves parasitic capacitance/inductance -- if the current didn't travel along the parasitics then it would be a "normal" part of the circuit.

 

So the higher the path impedance, the lower the current at a specific voltage.

Now consider the measurement equipment itself -- the probe etc have their own parastics. Youve discussed providing a low impedance alternate path to shunt away parasitic leakage currents -- when the measuring device can be this!!!

 

So ... in order to be able to measure low level currents (high impedance leakage circuits), the input resistance of the measuring equipment needs to be substantially higher than that of the circuit impedance. That's why (duh!) IEC 60601 requires a high impedance measurement -- and if you want to get even more sensitive then you need even higher impedance.

 

So, to flip this around: what is the minimum current that you want to measure? 1 nA, 1pA? --- really now what is your background noise folks??? 

[jabbr]

3 hours ago, octaviars said:

Look here @jabbr I think this is the one @Superdad was looking for were John measured the Baaske isolator.

 

https://www.computeraudiophile.com/forums/topic/35129-a-novel-way-to-massively-improve-the-sq-of-the-microrendu-ultrarendu/?do=findComment&amp;comment=723196

 

Ah good. FWIW there are “medical grade”/rated SMPS that have lowered leakage. I tend not to use SMPS when possible although some of the old HP test equipment have internal SMPS which are quite good. Similarly a SMPS is used in the LPS1 and also mitigated so there are indeed different strategies!. I think it’s the brand/quality that’s a big issue and John has shown that they can be modified to greatly reduce. Also as we all know noise from cheap SMPS can travel into other devices so tackling the source of the problem is indeed better than mitigating.

 

BTW: what are the levels (Y axis) here? Are we talking microAmps? nano? pico?

 

I dont use Baaske isolator myself — I have a fiberoptic network which isn’t an issue. 

[jabbr]

For the final connection between your wall socket to your DAC or streamer by all means use Cat 6 U/UTP patch cables and you won't have a ground loop.

 

For a new installation in your walls, I recommend Cat 6a either F/FTP or U/FTP ... the reason the shields are needed around individual twisted pairs is to reduce cross talk which is helpful at 10 Gbe ... now if the receiving device has a socket which doesn't connect ground then not an issue for ground loops. Audio doesn't need 10 Gbe but a short segment of Cat 6 is fine etc.

 

Personally I have Belden U/UTP 5e in my walls but have added LC-LC fiber.

 

If you are having a need for heavy duty shielding around your twisted pair cables there is another problem going on that you should first solve. Perhaps your listening room is sitting atop a high powered radar array?

[jabbr]

17 hours ago, Superdad said:

 

4 Gigabit copper Ethernet ports, 1 Gigabit SFP cage, 1 10/100Mbps ultra-clean and isolated port (expensive GMR isolators), low-phase-noise master clock driving clock synthesizer for various clocks, LT3042/45 LDO regs, specially chosen magnetics for best leakage isolation, and a few other details.  

Price to be much lower than competitive units that are simply modified off-the-shelf switches—versus our from-the-ground-up approach.  Targeting $500 but it depends upon final BoM cost.

https://www.computeraudiophile.com/forums/topic/38968-etherregen-early-general-details-please-dont-ask-too-many-questions-yet/

 

 

Not bad!

 

There is a market for a consumer audiophile type switch. Hopefully one that can meet professional switch specs.

 

17 hours ago, Superdad said:

 

Your understanding is wrong.  Except for the very last part, where the money rolls in—due to happy customers enjoying audible benefits and a solid value. Same as with our other products—all sold with a 30-day, money-back guarantee. 

 

Your, the UpTone, products are very reasonably priced, particularly given the small audiophile market. 

 

I would like to see to see an end-to-end jitter/phase noise measurement the same as which is specified for the 10GBase-X Ethernet spec (2002) ie <5.5 picosec end to end jitter. Not required for 10/100 Mb Ethernet but give me a reason not to use a 10G switch at 1G, for example.

[jabbr]

17 minutes ago, Fitzcaraldo215 said:

When you say "end to end" that must necessarily include the DAC, unless you think that is of no importance in our setups.  

 

The good thing about IEEE specs is that the measurement details are specified. The standard measurement is what I’d like to see at a minimum for an Ethernet switch that claims to have improved functionality over consumer off the shelf $15 units. IEEE 2002 made a conscious decision not to specify the specs on individual components such as clocks, psus etc, rather looking at end-to-end specifications. That was >15 years ago.

[jabbr]

7 minutes ago, Emcee said:

What about something like BJC Cat 6a which have a ''floating'' shield (unattached at each end)?

 

I also have the Supra Cat 8 STP which introduced a nasty buzz in my system until I removed the copper jacket on one connector.

 

Is either of these theoretically better/worse than basic unshielded Monoprice Cat 6?

?

[jabbr]

1 hour ago, mansr said:

How is that relevant for audio? The DAC isn't running off the link clock.

It’s not relevant for you nor intended to be, however for people for whom upgrading Ethernet switch clocks or for whom cable shields, avoiding leakage currents is important, of for anyone interested in Ethernet signal integrity, I am pointing out that IEEE standards for 10Gbe fiber already have very stringent requirements and people should get some comfort in using equipment which meets such standards.

[jabbr]

3 hours ago, Fitzcaraldo215 said:

Ok, but if those efforts have no provable effect on the output from the DAC, other than giving people "comfort", of what relevance are they to our listening?

Everyone is free to do whatever they want, and use whatever metrics or advice they want, at least within the limits of their budget.

 

Who says there is no provable effect on the output of the DAC? Hasn't been proven one way or the other and unquestionably hasn't for every combination of PC, switch, renderer and cable ...

 

Could I construct a situation where there is no question about a measurable effect of a network cable on the DAC? Certainly. (simple case is ground loop with shielded cable -- seen that, heard that ...)

 

Does upgrade a clock within a switch matter? I'm saying just use a 10Gbe fiber switch if you are concerned and don't worry about it. I'm also asking audiophile companies to at least show us the measurements that professional networking companies do all the time.

 

Does it make a difference in the DAC output? Not inconceivable but I'm not going to measure someone else's substandard switch with someone elses DAC. Here's what I can tell you from my own experience:

 

There are sometimes subtle factors that affect power plane and ground plane noise. One of which that hasn't been discussed enough, but one that unquestionably makes a surprisingly big difference is rise time. Too short a rise time results in overshoot and ringing at the receiver. Lots of the cheaper logic ICs generate too fast rise times. Ringing at the receiver can result in power and ground plane bounce which can result in additional phase errors at the receiver. Phase errors at the transmitter combines with issues at the receiver can thus result in increase phase errors at the receiver -- this is a fairly well accepted issue and one that IEEE 802.11ae acknowledges which is why the end-to-end phase error/jitter be measured and not just the receiver's.

 

In any case whether this makes its way to the DAC output is speculation and if I were ever to commercialize my DAC I assure you that I would publish my own data but right now we have what we have. I try to make recommendations taking reasonable price into account and at the moment, fiberoptic ethernet is so inexpensive that I choose to use it and not worry further about it.

[jabbr]

@Fitzcaraldo215 do you have an answer for @wanta911 that will satisfy him/her?

 

@wanta911 you got me. I can't conceive of an electrical explanation because you've got complete galvanic isolation from the wifi router to the aries and to the DAC. I can't conceive of a way that cables connecting your NAS and wifi router could make an electrical difference.

[jabbr]

5 hours ago, Darryl R said:

You guys are fun, really.  And I truly appreciate all the participation in my thread.  I've nothing to prove, just merely posting my findings, and I wouldn't necessarily classify the difference as subtle.  Let me also reiterate that I wanted to put the money elsewhere, and I had the cable on contingency deal with a return authorization in my inbox* as I listened.  I did NOT want to hear a positive difference.

 

??? Your OP stated that you hear a hum ... is that an improvement?

 

If you like one cable better than the other why are you asking? 

 

To be clear your “findings” are that the shielded cable has a hum ... why not just go with @AMP‘s recs since that’s your equipment?

[jabbr]

31 minutes ago, plissken said:

 

Archimago for starters: 

 

http://archimago.blogspot.com/2016/11/measurements-on-value-for-ethernet.html Also, WiFi, as is to be expected and is my choice connectivity, shows to be superior because it can't pass on 60Hz AC hum. 

 

http://archimago.blogspot.com/2015/02/measurements-ethernet-cables-and-audio.html

 

http://archimago.blogspot.com/2015/02/measurements-intercontinental-internet.html

 

 

 

You don’t get it.

 

I said I could construct a situation where a particular Ethernet cable has a measurable difference on a selected PC/DAC combo. 

 

Want to bet? (your favorite way to settle these issues) — let’s not piddle with $1K — let’s do $10K

 

I pick the cable, PC & DAC/AMP & oscilloscope.

 

Measurable — wanna bet  

 

also don't misquote me

[jabbr]

25 minutes ago, plissken said:

 

 

You aren't being misquoted. 

 

IMO you are being disingenuous wrt to what is the main topic at hand: $340 cables vs $13 where a prudent person could assume that both have like construction, reputable manufacturer, passes spec.

 

Sure if you want to talk about constructing a cable, as @Mansr did with his USB cable experiment to exaggerate a scenario to finally show error and the inability to recover that's all well and good. 

 

Yep I like shutting up yahoo's that have zero idea about how packet networks and NRT-OS's function but continue to tell others about the sighted gospel of their experience. I would like, no LOVE, to take not only their $1000 but watch tail get tucked. I've always stated it would have to be a spec passing cable and I'll construct on the spot a like cable. Ears only evaluation. I'm not interested in seeing microscopic variability between even cables from the same manufacturer but different production runs. 

 

Now if you're up for that let me know. 

No I’d demonstrate a difference in “SQ” at the DAC output between a shielded and unshielded cable of my own construction. Also measurable difference. ? 

 

Thats the topic at hand here. 

 

Regardless of what Archimago posted. 

 

Take home point: just because someone doesn’t measure a difference doesn’t mean a difference can’t be measured. (That’s no excuse for smoke and mirrors nor so called “audiophile” Ethernet cables)

[jabbr]

1 hour ago, Fitzcaraldo215 said:

So, have you published those scope traces, even photos of them, as well as details of construction and connection details? If not, why let an as yet unaccepted bet for a measly $10k stand in your way?  The rest of us are all dying to know.

 

@mansr got it right: a nasty ground loop that runs through the shield connected at both ends... do you really need to see a scope trace? 

 

There's a spectrum of noise currents which share a ground return (loop). Ground loops should be uncontroversial. Anyone who's done enough connections has encountered them. They are well discussed. 

 

So-called "leakage currents" are really the same thing as ground loops but frequently flow through parasitic capacitance. Depending on the degree of parasitic capacitance as well as frequency, the leakage currents will have higher or lower magnitude based on well known equations -- the complex form of Ohm.s Law V = I.Z

 

In the case of Ethernet cables as we are discussing here, leakage currents can flow either through the shield (aka ground loops) or through the higher impedance of the PHY (parasitic capacitance of the ethernet transformer).

 

 

1 hour ago, Fitzcaraldo215 said:

 

But, if you are going to publish, could you also possibly do simple frequency response both ways at the output of the DAC?  It might be interesting to determine if, as mansr suggests, it might be a simple grounding issue with spikes at 50/60 Hz and harmonics thereof.  

 

This electronics is all well known and I'm not going to publish oscilloscope traces the "prove" Ohm's Law. Do you really want to pay me $1000/hr to prove this? You don't need me to... it is a simple ground loop/leakage current issue at the very least.

 

Same thing for USB ... break the ground at perhaps 90% of the issue goes away. That should be the take home message of this thread.

[jabbr]

35 minutes ago, mansr said:

Right, so if the capacitance is small, only very high frequencies will get through. Noise in the megahertz region is inaudible by itself. However, if it is modulated at a low frequency, it might cause audible disturbances. That's still a lot easier to mitigate than noise entering through a direct connection, ground or otherwise.

 

Certainly. I'm not trying to be cute. I believe that most cases where cables make a difference have to do with grounding and noise and antenna effects and shielding. Like lifting the USB ground ... common theme.

[jabbr]

1 hour ago, Em2016 said:

 

Hi jabbr. John Swenson has stated that he's had to build very customised testing gear to show the latter. Does your testing gear show the same or are just going off what John has said?

 

I've asked what voltage levels the leakage noise under discussion are, or the currents ... if we are measuring picovolts or femtoamps, yes you need very specialized equipment (for example) ... but does this matter?

 

Quote

I have a lot of respect for John S btw, but just like to get an idea if anyone else has the means to easily see these things themselves.

 

Certainly measurement in the microvolt range is frequently done (active high impedance probes) [1], nanovolt measurements are more difficult etc. Lets not get the idea that you can simply fire up your 'scope and stick in a probe ... you can show that your cables can be effective antennas 

 

1. http://download.tek.com/document/48W_28061_0_HR.pdf

[jabbr]

30 minutes ago, Em2016 said:

 

Well yes, I would assume you've verified everything you say? Not that I need to see any measurements but I would have assumed you've verified with your own eyes, for yourself?

 

I mean this respectfully of course - not a personal attack. I've followed many of your posts and you are clearly a clever guy.

 

Verified what? What I say about Ohm's Law? Probably had to verify that in Physics class some decades ago...

 

I'm asking a genuine question and not giving a specific answer but food for thought. Goes along with a common theme I've been posting about. Let me pose the full question.

 

What levels of leakage currents might be significant in audio circuits?

 

(I haven't verified the answer to this question with measurements).

 

microamps? what is the background current noise level in the circuit? If the current noise is in the milliamp range then microamp noise is unlikely to matter.

 

But for me personally: I get the best isolation transformer I can find (Topaz 0.0001 pF) to sit behind my audio system, and then remove cheap SMPS and use LPS when feasible even if to power a dc/dc converter. I also use fiberoptic ethernet ... so these reasonably inexpensive techniques mitigate leakage currents

 

so I'm being lazy and practical but if you want to be precise  you or anyone could measure the baseline noise of your circuit

rule of thumb:  say SNR 120dB, 4V full range = 4 microvolt noise floor

So these >1GigaOhm impedance leakage currents are rather likely below the noise floor of the circuit. (willing to look at specific measured numbers though).

 

 

 

 

[jabbr]

30 minutes ago, Em2016 said:

 

No, you stated: "leakage currents can flow either through the shield (aka ground loops) or through the higher impedance of the PHY (parasitic capacitance of the ethernet transformer)."

 

Oh that ... no I haven't measured what the parasitic capacitance of the Ethernet PHY.

 

I do know that leakage currents across copper Ethernet are/can be a concern in healthcare. 

 

Quote

 

The part in bold is John's claim (which I believe of course). I was just wondering if you've verified this claim or just assuming it's correct.

 

Not just a blind assumption ... I've read about it. Makes sense to me. The claim about >1 Gigaohm leakage currents (personally hate that terminology but equates to a nanoamp current) hasn't been demonstrated to me to be relevant e.g. above the noise floor (but willing to learn).

[jabbr]

39 minutes ago, Em2016 said:

 

What was the old equivalent term/s before that?

 

Every electronic circuit has a schematic which has the major electronic components but also parasitic capacitances and inductances that although not listen on the formal schematic, are present in the actual circuit.

 

Leakage currents are currents that flow to the ground (chassis typically): https://www.sunpower-uk.com/glossary/what-is-leakage-current/

 

If you connected a wire directly from the signal to ground that would be a "short circuit" and a relatively large current would flow. That's because the wire has low impedance (this is all the complex form of Ohm's Law).

 

Leakage currents typically flow through parasitic capacitances (easiest way to think about this). The impedance of the path depends on the amount of capacitance. The lower the parasitic capacitance the higher the impedance. https://www.dummies.com/education/science/science-electronics/generalize-impedance-to-expand-ohms-law-to-capacitors-and-inductors/

 

Every circuit consists of current loops, typically in multiple directions, and when dealing with high frequencies these loops are three dimensional. The amount of current flowing in each loop is inversely proportional to the impedance of the particular loop. Higher impedance means lower current. 

 

I'm not trying to be clever here, just trying to demystify this. These concepts are essential to electronics, so as I've said, I haven't verified since Physics class (my formal training isn't EE though I've done some stuff). In cases where its important to measure leakage current to meet standards, there are leakage current meters. These do measure leakage currents through Ethernet among many other things and I've seen them used by my engineers and physicists a few years ago. Among the more difficult measurements we made were using single photon diamond detectors in the 1990s... probably off the shelf these days...

[jabbr]

Let me add that I believe @JohnSwenson used the term “high impedance leakage current” because he had to use a high impedance probe to measure low current leakages.

 

Certainly. If the probe is low impedance then the probe shunts away the current it is trying to measure, so that’s why you want an active high impedance probe for sensitive measurements — that’s its own topic

[jabbr]

2 hours ago, Superdad said:

 

As far as I can recall, John never used “current” in his reference to high-impedance leakage. Others have somehow chosen to make that part of the phrase. 

Of course it is measurable and we have published some examples.

Ha ha ... so high impedance leakage ... could it be voltage? Nope ... gas?!