| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Just to remind everyone of existing schematic implementations; look at:
http://www.ieee802.org/3/power_study/public/email_attach/Term_28oct99.pdf
This shows existing practices.
The common mode termination cap is usually around 1000pF rather than 0.1uF,
since it needs to be a 2kV cap in order to pass Hipot tests.
So common mode termination is AC coupled not DC coupled.
Due to Hipot, each wire of the RJ-45 must not be tied to ground.
Hipot is: tie all 8 pins together, ramp to 2250VDC for 60 seconds,
then ramp to 500 VDC, the leakage must be higher than 2 meg.
I hope this helps,
- Rick
-----Original Message-----
From: Sterling Vaden [SMTP:savaden@xxxxxxx]
Sent: Tuesday, November 21, 2000 2:27 PM
To: Dieter Knollman
Cc: tcobb@xxxxxxxxxx; a_flatman@xxxxxxxxxxxxxx; stds-802-3-pwrviamdi@xxxxxxxx
Subject: Re: Spare pair termination.
Dieter, I don't know definitively the answer to your questions, but will
state what I think is correct.
Dieter Knollman wrote:
>
> Terry, Sterling, Alan, et al.
>
> I have questions concerning high frequency termination, especially the
> spare pairs.
>
> How should the spare pairs be terminated in the PD and PSE?
If power is delivered on signal pairs, the best termination for the
spare pairs is differential plus common mode, i.e. 50, 50, 25 in a Tee
arrangement, (or equivalent delta) which is 100 Ohms differential (also
referred to as odd mode), and 50 Ohms common mode (also referred to as
even mode). The common mode leg should have 0.1 uF or thereabouts cap in
series to thwart ground loops.
This will result in a good impedance match to the cable at high
frequencies. The common mode of the transformer on active pairs should
also be terminated.
Now, if we are supplying power on whichever pairs, it follows that the
common of the PD must be isolated from the common of the PSE. Otherwise,
the resistance of the return leg of the power path will create a
substantial difference in potential through which current will flow....
It is apparent then that the PD ports must be isolated, but not
necessarily the PSE ports, although if you connected more than one port
of two PSE's together through different length paths this could
potentially release the smoke from the components... PSE signature must
be outside of 25K ohm range so they will not turn on. I defer to anyone
who can think up the different circuit configurations.
Now, if we hang our power supply, power sink on these pairs, the
souce/sink impedance will not match our ideal termination. Will this
cause a big problem? I don't know, probably not. It would be a source of
error for measuring crosstalk of the PSE ports, but probably not a huge
concern in terms of link performance.
>
> Or more specifically:
> What termination is needed for pairs 4&5 and 7&8?
> Differential:
> Is a differential termination required for these pairs?
Differential termination at least I think is necessary for 100 Base TX,
but doesn't matter for 10 Base T.
> Can they be shorted?
Shorting is not much better than an open, although you could argue that
a short is closer to 100 Ohms than an open and it does provide a path
for common mode currents. Think this way, a common mode signal signal
want's to return to it's source. to control EMI, we want to provide
return path with smallest loop area. An open termination forces some
current through large area ground path. A short will provide return path
through other cable pairs, a good thing. A load reduces reflected signal
due to impedance discontinuity. Ideally there would be no common mode
signal on pairs, but this cannot be assured.
> Common mode:
> Is a common mode termination to ground(?) required?
> Pair to pair mode:
> Is a pair to pair termination required?
The common mode terminations are to each pair. Actual ground termination
is not required or desirable.
>
> If a PD can accept power from signal pairs and spare pairs, is isolation
> required.
Yes
>
> If a PSE is designed for Environment A, is isolation required between
> ports?
Apparently not. That is why it is cheaper to implement environment A
PSE.
I have overheard some discussion that environment A is not do-able, but
I did not catch the reasoning.
>
> Please excuse my ignorance. If you think this is inappropriate, please
> state so.
I apologize if this is too much information.
Sterling Vaden
>
> Dieter Knollman