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

[802.3af] Some comments from an outside engineer trying to use 802.3af

Forgive me if my questions and comments are not wanted, but as an
outsider to this project and with a real need to engineer in
powered ethernet, I may be able to provide a perspective of some

I read the draft proposed standard and have the following
questions and comments:

1. It appears that since mid span requires alternative B (unused
   pair power) and the PSE can choose either A (signal pair
   power) or B (unused pair power), the most common fielded
   system in a 4 pair environment will be powering the unused
   pairs (alternative B) at the PSE.  Was that the intention?

2. Why is it a violation for a PD to use both A (signal pairs)
   and B (unused pairs) for power?  What harm is there in
   allowing for the use of all the copper?  See later my question
   on power limits...

3. I don't see a detailed description of how you would implement
   class 1-5 detection.  Specifically, what does the circuit look
   like on the PD end?  I see that the PD end has to operate in 3
   regions: class 0 detection, class 1-5 detection, and powered.
   I can see leaving class 0 circuits in place at all times, but
   class 1-5 circuits consume some power and you may not wish to
   leave them across the power rails.  Do we have an example
   circuit for class 1-5 PD?

4. Why limit PD power to 13 watts?  That seems like a poor choice
   since many devices would likely fall in the 15-30 watt class.
   I'm specifically thinking of video devices that capture video,
   compress it, then send it over ethernet.  These devices would
   benefit greatly from power freedom since they have physically
   specific locations (they have to be placed to pick up the
   image).  13 watts is not enough.  My computation suggests the
   cable resistance would support higher currents.  In
   particular, my needs suggest that a 4 pair powered system
   would work for me at 25-30 watts.

5. The paragraphs on isolation are terse.  It seems to imply that
   every port on a PSE must be isolated from all others and the
   PSE ground.  This implies isolated DC-DC converters for every
   port.  The PD is also required to be isolated.  This means a
   DC-DC on every end of the cable.  Is this what was intended?
   It appears all the effort to save costs on the detection
   system (single diode and resistor) are for naught when
   compared to the costs of the DC-DC converters.  Is there not a
   way to isolate only one end of the cable and save half the
   cost (and half the loss)?  It almost seems like an AC based
   system, say running at 25 KHz, would be cheaper since it
   allows simple isolation through cheap transformers, but I'm
   sure this has other problems.

All in all, the proposed standard is almost but not quite what we
need.  We need higher power delivery (30 watts) and we could use
a simpler power system (fewer DC-DC converters).  We may end up
building a system that would support 802.3af PD devices, but our
PSE is non compliant since it knows how to power our higher power
PD device, probablyusing all 4 pairs.  I wonder if I should make
this my own "class 4" or "class 5" device since those are
undefined by seem an obvious way to indicate higher power.

I apologize if my comments are unwanted.

Mike Ciholas                            (812) 858-1355 voice
CIHOLAS Enterprises                     (812) 858-1360 fax
5855 Fiesta Drive                       mikec@xxxxxxxxxxx
Newburgh, IN 47630