Mobile signal booster

[michael adams]

To whom it may concern, and as a matter of record, I did not post the above message.

michael adams

...

now posting exclusively via GigaNews

[michael adams]

To whom it may concern, and as a matter of record, I did not post the above message.

michael adams

...

now posting exclusively via GigaNews

[The Natural Philosopher]

You can't supply DC via a transformer

[Andy Burns]

You don't need to; PoE uses a phantom* power arrangement so the DC is applied across different twisted pairs, not within each twisted pair.

Equipment that recovers the power will pass the DC through a diode bridge, not because there is any AC which needs rectifying, but just to ensure correct polarity in case it has passed through a odd number of crossover cables.

[*] ignoring the "dumb" PoE method that simply applies DC to the unused pairs of 10/100 ethernet, but that doesn't use transformers either.

[Mark Carver]

Voice over LTE (Long Term Evolution, aka 4G)

[Michael Chare]

You just have to use diodes to rectify it. On the router I looked at there is a big transformer chip just before the RJ45 sockets.

I would think there must also be some electronics after the transformer to make sure that the signal voltage is square rather than rounded.

[rick]

I have both Vodafone and EE signal boosters .... what they do is give you in effect a local cell, your phone connects to it by WiFi and it routes your calls over Broadband. Probably be discontinued in near future, as WiFi calling now supported.

[Andy Burns]

It's neither, it's flat.

[Michael Chare]

So how does it contain any information? For 100base-TX, on each pair, the voltage varies between +1, 0 and -1.

[Andy Burns]

We're talking about the power not the data ... the DC power is "imposed" not directly on any given data pair, but between the data pairs.

[John Walliker]

No, they use a 3G GSM transceiver, not WiFi.

John

[Paul]

A GbE has eight wires, arranged as four differential pairs, with transformer isolation at the source hardware, and transformer isolation at the destination. This is a big big improvement, over the shared coax cable scheme and vampire tap used in the first networks.

___ ___________________ ___ \||/ \||/ /||\ /||\ \||/ \||/ ___/||\___________________/||\___

___ ___________________ ___ \||/ \||/ /||\ /||\ \||/ \||/ ___/||\___________________/||\___

___ ___________________ ___ \||/ \||/ /||\ /||\ \||/ \||/ ___/||\___________________/||\___ | | (+)---------+ +-------------- (+) phantom power received ___ ___________________ ___ \||/ \||/ /||\ /||\ \||/ \||/ ___/||\___________________/||\___ | | (-)---------+ +-------------- (-) phantom power received

You can apply a DC offset to the floating cable portion, because it floats and has no effect on the data transmission going on differentially on the pair, at relatively low amplitude. The DC only appears between the two transformers, and neither end, the data pads on the chips, don't see that DC voltage. The transformer does not pass DC.

The transformers are also present, to reject common mode noise. The wires in the Ethernet cable are twisted pairs. The pairs are twisted, so that if an interference source is present, the same noise signal is imposed on both wires of the pair. The transformer "nulls" out the noise, because A-B = 0. Both the A noise and the B noise are the same amplitude, the transformer takes the difference, and zero noise ends up at the destination.

If you use an oscilloscope and measure the "floating" noise on a pair, I can easily see a 10V 60Hz signal on a wire in the cable, in a lab situation. The transformer makes that disappear, by taking the difference. At the time I first looked at that in the lab, I was shocked by the sheer amplitude of the signals I was seeing. The signals are not "real" in the sense that you cannot extract a free lunch from the induced signal and light up a lightbulb with it :-)

Now, if you run DC over the cable, like in the diagram, there's a chance some noise could be induced on the wires. Maybe the DC that is delivered, is not clean enough to use directly for anything, but after being run through a DC regulation solution, might be made a bit cleaner. It won't have my 10V AC signal imposed on it either. The POE DC source will be stiff, and the induced noise signal will have reduced amplitude.

The specs tell a story. They use high DC voltages at low currents, and those can be converted, via a buck converter, to a usable digital supply.

Ethernet uses a Bob Smith grounding scheme, a ground plane near the media end. It has 2kV capacitors, offering some sort of ESD or induced lightning protection. (A "bit of protection", not belt and suspenders. Your NIC could be blown to hell, in spite of this scheme.)

(Page 86) SMSC LAN83C175

The transformer has some sort of breakdown voltage. The intention is, to give a measure of induced HV protection. If there is a lightning strike 100 feet from the house, your hope is, the floating bits don't blow through and transfer an insult to the terminating equipment. Maybe the portion around the transformer is damaged, and "your NIC needs to be changed". If the onboard NIC on a desktop is ruined, you can try installing a PCI Express x1 NIC in its place. Laptops, not so much.

When I did an Ethernet at work on a computer board, one of the other engineers mentioned the keyword I needed. I'd never heard of Bob Smith before. (Ethernet hadn't been invented when I was in school, and the very first Ethernet I did, used vampire taps and wasn't nearly the same as the second one with the eight wire cable. Bob Smith goes with the eight wire version.)

Paul

[The Natural Philosopher]

there are no spare pairs in gigabit ethernet

Like I said., POE doesnt use transformers

I think you missed the point. Ethernet is inherently isolated because it uses transformers to couple input and output pairs.

POE inevitably breaks that isolation

[The Natural Philosopher]

transformers do not come in chips

Christ on a bike....

[Andy Burns]

We're not talking about the spare pairs, that was the "dumb" scheme I mentioned, "proper" PoE doesn't need separate pairs for data and power.

Not sure we're saying the same thing.

depends whether the PSU itself is isolated?

[Andy Burns]

[The Natural Philosopher]

[Paul]

The voltage on POE (~40VDC or more), implies an additional level of SMPS is used at the destination. Nobody wants to build logic running at 40VDC.

It is at that point, that isolation can be added if desired.

A buck converter doesn't offer isolation. It has one magnetic for energy storage, and is ground referenced.

Whereas an SMPS with transformer isolation, and classical rectification on the secondary, that has a withstanding rating provided by the transformer. On an ATX supply, that withstanding is somewhere on the order of 1100V or so. And is also why the ATX supply has a sticker on it announcing "HIPOT tested". Which means each unit receives a test to prove it can withstand a voltage near that value, for some number of seconds or minutes. And that the primary and secondary really are isolated.

You can put a smaller version of that, inside a POE powered item.

Even the original Ethernet had power isolation. For the cable feeding the brick transceiver from 3COM. I don't know if much remains on the web about that stuff.

Paul