Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [802.3af] Late comment

Rich and all,

If those switches are using the same polarity and the same common and their
main power supply output is isolated from the system ground/chassis than I
don't see a problem.
If the common point is different than it might change the noise performance
which is something that hard to predict with out having the exact circuits. 
In case of two ports connected together in opposite polarity, the port
current limit will protect the port.


-----Original Message-----
From: Graham, Richard [mailto:grahamri@xxxxxxxxxxxxx]
Sent: Tuesday, January 14, 2003 7:03 PM
To: Yair Darshan; Dave Dwelley; Geoff Thompson
Cc: stds-802-3-pwrviamdi@xxxxxxxx
Subject: RE: [802.3af] Late comment


I have products already that run two connections to separate switches
for redundancy purposes.
In the future both of those switches will support 802.3af.
Either connection will provide power or data.

IF they come from two differnet vendors (or an inconsistant single
vendor... i.e. multile product lines) wouldn't I end up with your
"unlikely" case?

Already on the wireless product roadmap,
Rich Graham
Enterasys Networks


-----Original Message-----
From: Yair Darshan [mailto:YairD@xxxxxxxxxxxxxx] 
Sent: Tuesday, January 14, 2003 3:22 AM
To: 'Dave Dwelley'; Geoff Thompson
Cc: stds-802-3-pwrviamdi@xxxxxxxx
Subject: RE: [802.3af] Late comment


I agree that if all ports are floating than you may have 2*Vport between
any two ports however these ports would never touch each other under
normal operating conditions and installation.

In any case the preferred lead to be ground is the positive lead from
practical point of view (easy to design when the + is the common for a
single power supply) and from system point of view (positive lead to
system earth).

Currently in our spec we didn't defined if:
a. we stay floating in both leads or
b. ground the positive or 
c. use the positive lead as a common point without grounding it to the
system ground/chassis. d. ground the negative or 
e. using mixture of possibilities.

It is clear that option e is disaster to the standard.

option d is bad from practical design aspects and cost issues. (P
channel mosfets etc)

Option c : the best option by having a positive common point which is
floating too.

Option b is proffered if we find good reasons why we should ground the

Option a is very good due to max completely floating outputs.


I suggest the following text to the standard:

1.	The PSE port outputs shall be isolated from system ground, frame
ground and data circuits.
2.	In multi port system all ports common point shall be the
output (+Vport).


-----Original Message-----
From: Dave Dwelley [mailto:ddwelley@xxxxxxxxxx]
Sent: Tuesday, January 14, 2003 3:47 AM
To: Geoff Thompson
Cc: stds-802-3-pwrviamdi@xxxxxxxx
Subject: Re: [802.3af] Late comment

All -

Geoff is right. There's nothing to stop someone from building a system
a +/-48V supply, and wiring half the ports to -48 and half the ports to 
+48. The fact that the supply is floating relative to earth ground 
matter - there are at least two terminals that are 96V apart, and
could short them together.

Technically, because we're isolated, it's completely arbitrary which end
treat as "ground". Practically, there are arguments both ways:

- most non-telecom engineers (read: networking engineers) are most 
comfortable with ground on the bottom, i.e., +48
- most telecom engineers are used to -48
- most engineers of both stripes like buying 100V NFets in preference to

PFets, suggesting switching the negative rail is best.

I'd vote we specify that the more positive rail (environment A only) is 
"common to all ports" (don't call it "ground"), and the more negative
is switched. If we refer to that switched rail consistently as "isolated

-48V", it would be even more clear.

Now, what to call the rail that the logic on the isolated side runs
-44.7V? :-)


At 10:10 AM 1/13/2003 -0800, Geoff Thompson wrote:
>If you put a voltmeter between "hot" leads from 2 different PSEs, then 
>is the max voltage between the leads (fault conditions included):
>         VportMax
>         VportMax - VportMin
>         Zero
>         2 X VportMax
>I contend that any answer above except [2 X VportMax] is OK and
>We currently have no text to preclude  [2 X VportMax].