Hi TomJones, I was just looking again at schematics but with 4 cable pairs (I can only speak about standard cables) and no GND near in sight there is not even a way to terminate. If you choose to use the unused pairs for GND then you indeed have created a new personal standard but I would then advise to invest time in real measuring if it really is better than standard ethernet.
Peufeu, I have solved RF issues in the past with the good old molded clip on ferrite cores. Good that you bring these up! If these bring as much improvement as the isolators in the setup of TomJones then we can guesstimate a little in which direction to look. Recently I had a nice issue where things were sometimes sounding good and sometimes they weren't. It turned out to be a smart phone that sometimes was very near the audio setup as eh ... it was used as ... quality source ...
Peufeu, I have solved RF issues in the past with the good old molded clip on ferrite cores. Good that you bring these up! If these bring as much improvement as the isolators in the setup of TomJones then we can guesstimate a little in which direction to look. Recently I had a nice issue where things were sometimes sounding good and sometimes they weren't. It turned out to be a smart phone that sometimes was very near the audio setup as eh ... it was used as ... quality source ...
Last edited:
Hi TomJones, I was just looking again at schematics but with 4 cable pairs (I can only speak about standard cables) and no GND near in sight there is not even a way to terminate. If you choose to use the unused pairs for GND then you indeed have created a new personal standard but I would then advise to invest time in real measuring if it really is better than standard ethernet.
Peufeu, I have solved RF issues in the past with the good old molded clip on ferrite cores. Good that you bring these up! If these bring as much improvement as the isolators in the setup of TomJones then we can guesstimate a little in which direction to look. Recently I had a nice issue where things were sometimes sounding good and sometimes they weren't. It turned out to be a smart phone that sometimes was very near the audio setup as eh ... it was used as ... quality source ...
The 66 ohms for CAT6 and 52.3 ohms for CAT5 cable in the article are for standard 4 pairs. So i have to measure it for my setup - puh! Or use standard cable....
But what about CAT7 or CAT8 cable resistance?
What the ability to ground the termination to the GND pins of a wall socket?
Sorry my friend but a wall socket (if you mean the 230V Schuko...) does not have GND but PE or "Erde". The termination does not ask for PE as that is meant for safety, it asks for GND of the circuit I guess. If you don't know the difference then don't bother at all.
Please note if you choose to design your own standard that you will couple GND to unused cable pairs only at one side of the TomJones standard network cables or you will have a galvanic connection (this is absolutely not done in ethernet and your creation may not be called ethernet anymore). It also defies the use of the solution for all problems, the Isolator. It also makes cables directional and nothing will be interchangeable in the bad sounding normal ethernet environments.
Also note that the PCB's of ethernet switches in general are designed quite OK and stringent at the well known brands. Random DIY attempts to improve things without any measuring certainly doesn't make things better although I meanwhile start to think the club that measures with ears probably has a way with that.
And what about the suggestion of peufeu? You ignore that suggestion?
Please note if you choose to design your own standard that you will couple GND to unused cable pairs only at one side of the TomJones standard network cables or you will have a galvanic connection (this is absolutely not done in ethernet and your creation may not be called ethernet anymore). It also defies the use of the solution for all problems, the Isolator. It also makes cables directional and nothing will be interchangeable in the bad sounding
normal ethernet environments. Also note that the PCB's of ethernet switches in general are designed quite OK and stringent at the well known brands. Random DIY attempts to improve things without any measuring certainly doesn't make things better although I meanwhile start to think the club that measures with ears probably has a way with that.
And what about the suggestion of peufeu? You ignore that suggestion?
Last edited:
But where to get GND in the middle of nowhere?
And what GND? Of the transmitter or the receiver?
Regards,
Tom
THAT are the questions you should have asked yourself when you started the design. Normally the cable is not interrupted and when it unfortunately is one uses a coupler. The coupler simply connects 2 RJ45 sockets and thus 4 twisted cable pairs (question: do you know all these cable pairs have a different number of twists?). Nothing less, nothing more. There is no GND in ethernet cabling/sockets. The GND stays in the devices. You want a GND in a GND-less environment The GND is not there as stuff is isolated. I hear myself thinking now
I just guess that when you start digging in ethernet magnetics you will find out the isolation voltage of the magnetics as used in normal equipment is more or less the same as the in the Delock isolator but I leave that one to you. The Delock lightning protectors you insert do need a connection to PE to be able to discharge to PE although I doubt they are fit for the task knowing lightning protectors in pro stuff. I suspect the part labelled "GT1" is a part specified for use till max. 6 kV and the device is further designed by Placebo Inc.
This operation is a failure from the start as it is not according the whole principle working of the protocol/devices/cables etc. I think I did my best to answer your questions albeit in a cynical way which is not to be avoided. Sorry.
The GND is not there as stuff is isolated. I hear myself thinking now I just guess that when you start digging in ethernet magnetics you will find out the isolation voltage of the magnetics as used in normal equipment is more or less the same as the in the Delock isolator but I leave that one to you. The Delock lightning protectors you insert do need a connection to PE to be able to discharge to PE although I doubt they are fit for the task knowing lightning protectors in pro stuff. I suspect the part labelled "GT1" is a part specified for use till max. 6 kV and the device is further designed by Placebo Inc.
This operation is a failure from the start as it is not according the whole principle working of the protocol/devices/cables etc. I think I did my best to answer your questions albeit in a cynical way which is not to be avoided. Sorry.
Last edited:
The termination resistors are about terminating the pairs relative to each other. That's what the article you posted explains, than in common mode, we consider each pair to be just one wire, which then forms another transmission line with the neighbor pairs, and that's what the resistor terminates. Now, I have no idea whether this will make it sound better, but at least there is an explanation, the effect is measurable, and it's not woo-woo.
Now, as Jean-paul explains, if you want to make a lowpass filter to get rid of HF noise, for example a LC or a RC filter, then you need to connect the bottom of the capacitor to ground. But in the common mode case, there is no ground. And at high frequency, the impedance of the Protection Earth conductor is high enough to make it irrelevant for filtering.
What the IEC filter in your appliances do is short the common mode noise to the chassis. It then becomes "ground", even if it is not. The important point is the chassis is a big chunk of metal, so it has low impedance, so even if there is common mode noise on it, since everything inside the box is at the same potential, the device inside the box does not experience the noise as a voltage between different parts of its circuit, so it is not bothered.
So what common mode chokes (like the ones I recommended) do is simply add inductive common mode impedance to your cable. With more impedance, less noise current flows. Being inductors, they're most effective at high frequency.
The isolation transformer also adds common mode impedance, it is almost infinite impedance at DC, but at HF there is quite a bit of interwinding capacitance, so it becomes less effective with increasing frequency. That's why ethernet magnetics have both a transformer and a choke. The transformer isolates at low frequency, and the choke adds impedance at HF.
That said, ethernet magnetics are really tiny and the common mode chokes are wound on a toroid core, so they still have some interwinding capacitance. A big fat clamp-on ferrite core won't have this problem. Try to pick one with a material that has high impedance at high frequency, for example this one. Well, okay, it isn't clamp-on, so you'll have to crimp a RJ45, but with some digging, you can find.
Note if you use the SPDIF output of the squeezebox, then the isolation, if it is needed, should be implemented with TOSLINK over optical fiber. If you put the boxes 6 feet away from each other and connect via optical fiber, there's not much doubt about the isolation
Now, as Jean-paul explains, if you want to make a lowpass filter to get rid of HF noise, for example a LC or a RC filter, then you need to connect the bottom of the capacitor to ground. But in the common mode case, there is no ground. And at high frequency, the impedance of the Protection Earth conductor is high enough to make it irrelevant for filtering.
What the IEC filter in your appliances do is short the common mode noise to the chassis. It then becomes "ground", even if it is not. The important point is the chassis is a big chunk of metal, so it has low impedance, so even if there is common mode noise on it, since everything inside the box is at the same potential, the device inside the box does not experience the noise as a voltage between different parts of its circuit, so it is not bothered.
So what common mode chokes (like the ones I recommended) do is simply add inductive common mode impedance to your cable. With more impedance, less noise current flows. Being inductors, they're most effective at high frequency.
The isolation transformer also adds common mode impedance, it is almost infinite impedance at DC, but at HF there is quite a bit of interwinding capacitance, so it becomes less effective with increasing frequency. That's why ethernet magnetics have both a transformer and a choke. The transformer isolates at low frequency, and the choke adds impedance at HF.
That said, ethernet magnetics are really tiny and the common mode chokes are wound on a toroid core, so they still have some interwinding capacitance. A big fat clamp-on ferrite core won't have this problem. Try to pick one with a material that has high impedance at high frequency, for example this one. Well, okay, it isn't clamp-on, so you'll have to crimp a RJ45, but with some digging, you can find.
Note if you use the SPDIF output of the squeezebox, then the isolation, if it is needed, should be implemented with TOSLINK over optical fiber. If you put the boxes 6 feet away from each other and connect via optical fiber, there's not much doubt about the isolation
There's an isolator for that! Signal is nicer from coax but as said, the Touch has no transformer and when the DAC does not have one the Touch or the DAC should be fitted with one. A nice chance to replace the RCA for a true 75 Ohm BNC. TomJones uses USB however...no doubt also with an isolator but at a spot where they make sense (if the +5V is separated).
Ahem, ... "6 feet"....
Ahem, ... "6 feet"....
Last edited:
Hi peufeu,
many thanks for your explanations.
So common mode choke without the problem of the transformer with termination will be the way to go.
I will test the ferrite rings you linked.
Should there be a great distance between the ring für RX and TX?
How many turns would you recommend?
There are also common mode chokes available for 100/1000base-t1 (single paire ethernet): https://www.we-online.de/katalog/datasheet/744232222.pdf
Will these have the same effect as the ferrite ring construction? But these are to tiny to solder by hand ;-(
Last edited:
The ferrite ring should be over all cable pairs and you can decide to have a few turns of the complete cable going through the ring in case of heavy assumed RF. The Common mode filter you link to is not specifically for ethernet and ethernet devices already have common mode filters built in. Again, adding superfluous stuff does not make signal integrity any better. Please have a device available that shows you visually what you do to the signals, this is NOT like analog where ears can decide what is better. I am sure it is even the worst approach to improve data transmission. Likely a "by ear" evaluation will lead to wrong decisions and mask issues at either sending or receiving side.
BTW when using analog signals you will have all problems solved. Then you can isolate as well but then with transformers. No RF issues.
BTW when using analog signals you will have all problems solved. Then you can isolate as well but then with transformers. No RF issues.
Last edited:
Hi Jean-Paul,
yes, the best would be to have a way to measure this approach - but I have only a multimeter ;-(
One way to do measurements is to use a network analyze software tool to measure package loss an - this is all i can do.
As the positive effect of the isolator is often discussed in different forums, i googled a while, if there are any measurements available to understand what happens - but there arent.
Regards,
Eric
Remember network analysers are for RF work and use 50R or 75R impedance. You'll need a specific analyser for ethernet cable applications.
Ferrite clamps come in all sorts of materials and you will have to decide what frequencies and attenuation is required. Only pass two loops of ethernet cable through the ferrite clamp any more will increase attenuation but the downside is it will narrow the effective frequency range and will result in less than ideal minimum cable radius bends.
Ferrite clamps come in all sorts of materials and you will have to decide what frequencies and attenuation is required. Only pass two loops of ethernet cable through the ferrite clamp any more will increase attenuation but the downside is it will narrow the effective frequency range and will result in less than ideal minimum cable radius bends.
It's audiophile stuff lol. There are never measurements.
I was simply suggesting to slip a ferrite core on the cable and see if it sounds better. Hey, even if it's placebo, it still sounds better!
Testing for common mode noise is rather complicated btw.
Peufeu,
one more question:
If one wraps one pair (TX or RX) multiple turns around the ferrite ring, would the result not be a kind of a common mode choke?
Regards,
Eric
Yes.
Clamping a ferrite core on the cable makes a common mode choke, it's a one turn inductor/transformer with the cable instead of a wire.
But if you use a core with a larger hole, like a toroid, you can make several turns, although Cat-5 is not exactly the most "bendable" cable... In fact scrunching it to bend it over a toroid to make turns would probably screw up the nice properties of the twisted pairs.
More turns increases both inductance and parasitic capacitance between windings, so you get better (higher) impedance at lower frequency and worse at HF, there is always a tradeoff.
Since Tom already has a toroid common mode choke and a transformer in his ethernet port, I guess the low frequency part is already handled, but these toroid common mode choke in ethernet magnetics have lots of turns on a tiny core, so they're not that good for parasitic capacitance. So I figured, why not complement it with a ferrite clamp on the cable.
But in order to know, you'd have to bring in the spectrum analyzer to see what sort of noise there is. Also there are tons of ferrite materials that are more or less effective at various frequencies.
Clamping a ferrite core on the cable makes a common mode choke, it's a one turn inductor/transformer with the cable instead of a wire.
But if you use a core with a larger hole, like a toroid, you can make several turns, although Cat-5 is not exactly the most "bendable" cable... In fact scrunching it to bend it over a toroid to make turns would probably screw up the nice properties of the twisted pairs.
More turns increases both inductance and parasitic capacitance between windings, so you get better (higher) impedance at lower frequency and worse at HF, there is always a tradeoff.
Since Tom already has a toroid common mode choke and a transformer in his ethernet port, I guess the low frequency part is already handled, but these toroid common mode choke in ethernet magnetics have lots of turns on a tiny core, so they're not that good for parasitic capacitance. So I figured, why not complement it with a ferrite clamp on the cable.
But in order to know, you'd have to bring in the spectrum analyzer to see what sort of noise there is. Also there are tons of ferrite materials that are more or less effective at various frequencies.
Last edited:
As thought... anything changed in the well proven ethernet standard should be measured for effect and proof what it does and what it doesn't. No serious manufacturer relies on designing a device without measuring its effect.
No friend in the network/datacenter field? These guys have everything to measure and see the patterns.
Hi I worked in industrial environments with IGBT high power rectifiers/inverters where the clamp on ferrite cores were the solution. This was with "slow" ethernet and a hostile environment with many frequencies and high voltage where odd things occur now and then.
Last edited: