Cross Talk
Hi Yair
Please find my reply to the issues below.
I am glad Yair is looking into this, and more of us should, last meeting a lot
of data was presented but there is no way I can cover the infinite ammount of
details and do every measurement and examine every angle, also the impression
I got was that, it was not the expectation either (that I do all the work) !.
At 12:09 PM 8/13/00 +0200, Yair Darshan wrote:
>Hi Roger,
>Regarding cross talk issue, to my opinion the situation is more complicated
>since there is more than one port caring power
>to the media and it is not clear to me what will be the impact of crosstalk
>and other noise issues in the following cases:
>
>1. If a single power supply output is connected to the center tap of the
>data transformer (Phantom connection)
> and obviously we have more than one Data+Power output (12 or 24 etc..)
>than we can say that all center taps of all
> data transformers are "shorted" together at high frequency signals.
> (Each Receive-Transmitt center tap of the transformer is shorted by the
>power supply output capacitor. All the Receive
> transformer center taps are shorted together and the same applies for
>all the transmit center taps.)
Point well taken,
Before power came along, Center taps sat on AC ground via capacitors
So AC-wise they sat on the same reference or were connected together.
Again we are back to the implementation issue, if the center taps
are all tied together we still have to make sure that no high frequency
exits and possibly enters the system, why is it an issue if a capacitor to ground
is added on these center taps to make them sit at AC ground?
And there will almost always be a choke between the supply and the WIRE (or
the center tap) to insure EMI compliance thus high frequency attenuation!!)
again do we want to get into the design of the power supply/system at this stage?
unless of course I am missing the point??
>
>2. Same as 1, but all outputs has separate power supply output. In this
>case there is some parasitic capacitance between each
> DC output although they are isolated from each other.
The Data that I collected proves that high frequency noise on the center tap will
cause EMI headaches, and a FILTER will be needed of course unless the power
supply is designed to be Ultra quiet. Either way, that center tap must not see
any high frequency noise, and the noise at frequencies below 1MHZ did not seem
to affect us in anyway, please see the graph at the end of my presentation,
where the amplitude at different frequencies is quantified.
Like Bob Love said, 100TX does not have tons of margin, but again the noise current
that makes us fail EMI is on the order of (micro amps- common mode) and if care is
taken to pass emi chances are we are in good shape.
>
>3. Add to the above description the fact that from EMI reasons there is
>about 1 to 10nF between power supply output to
> power supply chassis ( = system gnd) thus the construction of the
>connections may not be balanced.
>
>Questions raised from the above cases are:
>
>1. What will be the impact on the requirements for cross talks between one
>port to other ports,
> is there a problem here?
By inspection I do not see an issue here, I am very sure we will have different
"2 cents" on this. Also we have built 48 porters and did not see any issues....
>2. Is there any restrictions in the 802.3 standards (or any other standard)
>for zero voltage isolation between each data path
> to an other port data path?
> (If there is, We will need isolated power supply for each port)
Yair, good question again, If I understand your ? right the spec is about 2250dc (from memory)
to ground or about 1500v ac to gound on the port, and 500v between ports. Some quad
magnetic may have affected the port-port isolation due to the proximity of the pins (adjacent
port pins) and the breakdown of the air-gap.
>
>What is your opinion regarding the above issue?
>
in summary, you ask valid and good questions, I will try to help analyze and measure, and I am hoping
that the work is shared among many of us present in the group.
thank's
roger