ISO Regen performance Improvement Cheap!

[mansr]

22 minutes ago, Quadman said:

So my RJ45 magic jack breakout boards showed up today with a RJ45 plug with transformers installed in the jack.  Continuity check with MM from pins 1-8 on the back of the board to the RJ45 pins showed that pins 1-3 connected together, Pins 4 & 5 isolated to pin 4 and pin 5 and pins 6-8 connected together.  Searching further I managed to find a data sheet for this jack.  Pins 1-3 connect to one transformer pins 6-8 connect to another transformer on the RJ45 plug side all pins seem connected to a transformer.  Based on my PoE teardown pins 4 & 5 were connected to DC +, and pins 7&8 connected to DC - (there was no transformer in that PoE).  So what would you guys think the best way to wire this PoE (wish I was an EE).  Seems to me if you want to use the full transformer effect then DC+ would go to pins 1&2 and DC- would go to pins 7&8.  Anyone with deeper understanding of this your help would be appreciated.  Direct wired Cat 8 certainly be easier, but I have to try all options.

 

Exactly what did you buy? Unless I'm misreading the diagram, there is no DC connection between the contacts and the solder pins.

[mansr]

16 minutes ago, Quadman said:

This, its a breakout board with a RJ45 soldered into it. the RJ45 has transformers inside it.  Solder pads are just visible on the rear of the unit.  Pads for pins 1-8, Y LED + and -, Grn LED + and - and shield.  I also ordered bare boards and RJ45 with no transformers so I can see the effect of transformer coupled RJ45 vs non transformer coupled RJ45.  This may all be moot now with the direct cat 8 connection but these parts were already on order so try I will.

 

 

 

DC can't pass through a transformer (that's why they are there). I don't think these jacks can be used for power delivery.

[mansr]

11 minutes ago, Quadman said:

BTW, these RJ45 do NOT have transformers inside.  I may get to those in the future once I confirm how DC goes thru these

It doesn't. That's the purpose of the transformers.

[mansr]

2 minutes ago, Quadman said:

There are articles on the web discussing sending DC over PoE magnetic transformers they are designed to do so.  I just have not had the time to research more and figure out are these PoE certified magnetic transformers.  If not then you are correct, if they are then it should be doable off the center taps on the network side.

To send DC power you need access to the cable side of the transformer. The connectors you mentioned don't seem to provide this.

[mansr]

Just now, lmitche said:

But yeah, over $10 a piece seems OTT to me.

You can get 100 regular plugs for that price. If durability isn't a concern, I see no reason to pay more.

[mansr]

3 minutes ago, wushuliu said:

Re: LT3045, it's worth using two in series for improved performance.

You can also put two (or more) in parallel to both reduce noise and increase current.

[mansr]

52 minutes ago, look&listen said:

Already parallel! 

Alexeys = 2 LT3045s for 1A

HPULN = 6 parallel for 3A

HC-HPULN = 10 LT3045s parallel for 5A

Look on LT3045 data sheet, see SET pin on chip used as reference to parallel chips to work together.

And if none of those suit your needs, you can always make your own PCB.

[mansr]

5 hours ago, austinpop said:

Recent postings have only reinforced the need for good grounding in audio.  

And low-noise regulators.

[mansr]

21 minutes ago, marce said:

I cannot see how dielectric absorption can have any effect at DC... Even low frequencies it has minimal effect at DC it will have none? Puzzled.

Same thing for inductance.

[mansr]

1 minute ago, lmitche said:

Do either of you have another explanations or ideas for the impact of these various DC and network cables on sound quality observed here?

Sure, I can think of at least two:

• Poor internal regulators and power distribution in the affected devices
• Imagination

[mansr]

24 minutes ago, marce said:

Well it definitely is NOT dielectric absorption, it is a high frequency ac issue, the main thing being ac not dc as we are discussing with these cables. Sorry but we can't ignore physics. 

Yeah, I don't see how dielectric absorption could ever be relevant for a DC cable. With a poor downstream regulator, I can imagine a spiky current draw being influenced by the inductance, though.

[mansr]

4 minutes ago, totoxio said:

Whichever the cause is, it seems that it's not blocked by galvanic isolation. 

Where do you have galvanic isolation?

[mansr]

15 minutes ago, totoxio said:

I use a Singxer F1 to feed my Chord Mojo. It has isolation after the conversion to I2S  and a DC-DC isolator.

Isolation won't help if the problem is jittery signals. DC power isolators do not necessarily block high-frequency noise. They are mainly intended to protect against dangerous voltages making their way in from the far side.

 

15 minutes ago, totoxio said:

I use a Jitterbug, a LT3045 board and a ferrite bead before the Singxer to filter some high freq noise.

A decent regulator with adequate capacitors on either side should render the nature of the input cable irrelevant. The LT3045 has impressive specs, but if not used correctly, it won't do any good.

[mansr]

If noise blocking has anything to do with it, a purpose-built filter would be preferable over a cable that randomly happens to have suitable properties.

[mansr]

6 minutes ago, lmitche said:

My expectation is that the benefit comes from the shielding which reduces the leakage of EMI/RFI generated from quickly switching current draws of digital devices.

A good power supply shouldn't present such a variable load.

 

7 minutes ago, lmitche said:

Sorry, I don't have equipment to measure and prove it.

I do, but I don't have the expensive cables, nor any intention of every buying any.

[mansr]

1 hour ago, Abtr said:

An (extra) active filter may throw back more noise to its input (and from there into other audio gear) than it clears at the output. A passive cable that simply rejects common mode noise by virtue of its geometry might be preferable..

Who said anything about an active filter?

[mansr]

Just now, Abtr said:

Could a passive filter with capacitors and inductors selectively reject common mode noise?

Since a cable is merely a distributed network of resistance, capacitance, and inductance, it follows that a network of resistors, capacitors, and inductors in their usual component form can do whatever a cable can and more since the component values can be chosen freely.

[mansr]

12 minutes ago, Abtr said:

Hm.. I think the geometrical action of a cable (e.g., a noise canceling effect of the twisted pairs of an ethernet cable) can't be modeled in a passive filter consisting of resistors, capacitors and inductors, which can only be a network of high- and low-pass filters.

Now you're talking about interaction with external electromagnetic fields. That's something else.

[mansr]

1 hour ago, Bricki said:

Is anyone else finding there are further improvements from longer lengths?? 

 

I am now running 3 × 1.5 meter lengths of audioquest pearl chained together (with PoE adapters in between) and 3 lengths sounds better than 2...worth a try if you can afford the voltage drop.

 

I'd be interested to know what others find in their systems... But so far for me - the more Ethernet cable the better - so it's looking like common mode noise rejection on twisted pair is the mechanism for the improvement... 

I think you may have misunderstood how twisted pair wiring works and the meaning of common-mode noise rejection.

[mansr]

9 hours ago, Bricki said:

? I may very well have misunderstood ... I'm not going to pretend to know what I'm talking about and I'm sure that you have infinitely more knowledge than I do... I'm just looking for a reason for why longer runs of Ethernet cables sound better than shorter ones when used as dc cables in my system. ?

Let me explain a bit. When a twisted pair is subjected to external electromagnetic interference, both wires pick up the same amount of noise. That's what common-mode means; it is common to both wires in the pair. The rejection of this noise happens in the receiver, not the cable. This works because the signal is the difference voltage between the two wires. Adding an equal noise voltage (relative to ground) to both wires doesn't alter the difference. A longer cable will pick up more noise, but the differential voltage (the signal) still remains unchanged and can be just as easily detected by the receiver.

 

That's how differential signalling works. However, what you guys are doing is something else. An Ethernet cable consists of four differential pairs. Power over Ethernet connects both wires in one pair to the power supply negative side and both wires in another pair to the positive side. The benefits of differential signalling over twisted pairs don't really come into play here.

[mansr]

Just now, marce said:

Slapping a scope on the end of the various wire configurations would be a better bet.

The intersection if people with scopes and people with fancy cables is rather small.

[mansr]

6 hours ago, fas42 said:

a scope is far too low in resolution

Sounds like you know what needs to be measured. Why don't you tell us so we can try it?

[mansr]

1 hour ago, Abtr said:

Well, I think a single twisted pair of wires for DC positive and negative conduction could function exactly as a common mode choke.

You might think so, but you'd be wrong. Sorry, that's just how it is.

 

1 hour ago, Abtr said:

But PoE doesn't do this because both wires in one twisted pair are connected to the negative side of the DC power supply and both wires in another twisted pair are connected to the positive side.

 

In this configuration a possible mechanism for PoE to improve SQ is that each twisted pair cancels out the magnetic field that a single wire would generate around itself, thus preventing it from inducing noise into the other twisted pair (or in other audio gear).

No, it wouldn't do that. The magnetic fields from two roughly parallel conductors carrying current in the same direction do not cancel. At a distance of an inch or more, the magnetic field will be very close to that from a single conductor.

[mansr]

14 minutes ago, tims said:

Looking for a Ugreen Ethernet cable on ebay and came across this one here:https://www.ebay.com/itm/UGREEN-Ethernet-Cable-Cat7-RJ45-Network-Patch-Cable-10-Gigabit-For-Laptop-PC-Mac/152108723581?hash=item236a62ed7d:m:mTy5KCpsGONMDL2q5i_RZXg

I don't know if it's the same Ugreen cable recommended in this thread but there's a nice close-up of the cable:

Each pair has a ground wire/line - what function would this have? 

That would be a drain wire for the shield. This reduces the lengthwise resistance without making the shield foil thicker and stiffer. The crosswise resistance from any point on the shield to the drain wire is negligible.

[mansr]

2 hours ago, Abtr said:

I was thinking that because in a twisted pair both wires are tightly coiled around each other in opposite directions, their respective magnetic fields may be (partially) canceled

If the current flowed in opposite directions, sure. In this case (PoE), the current flows the same direction through both wires of a pair, the return being a different pair.