| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Jean, You’re right, the At-midspan will
probably take over from the Af-endspan eventually. It could take minutes,
hours, or weeks. But when it does happen, it will be an interruption of
service for that particular PD. In fact, if the power transient is
shorter than T_LIM then the Af-endspan will NOT turn off. So you may get
periodic resets of the PD without ever switching over to the At-midspan. This is why we have to make sure the At-midspan,
not the Af-endspan, powers up the PD initially. I agree it would be great if somehow we
could solve this problem without adding any components to the midspan, but I
don’t see how. Steve From:
owner-stds-802-3-poep@xxxxxxxx [mailto:owner-stds-802-3-poep@xxxxxxxx] On Behalf Of Picard, Jean You are right Steve that at initial power
up it is not guaranteed that the midspan will take control of PD power. However, this guarantees that the midspan
will take control of PD power when it is needed. It is not perfect, but simple. I mean that whatever solution will be
adopted, it should be simple and should not involve adding costly
components in the midspan. Jean From:
owner-stds-802-3-poep@xxxxxxxx [mailto:owner-stds-802-3-poep@xxxxxxxx] On Behalf Of Steve Robbins Jean, I don’t think your idea would
work. A PD that needs 20W may only need that much power occasionally,
while drawing much less power most of the time. So it probably
wouldn’t cause an Af-endspan to go into fold-back on initial power-up. Steve From:
owner-stds-802-3-poep@xxxxxxxx [mailto:owner-stds-802-3-poep@xxxxxxxx] On Behalf Of Picard, Jean All I think we should try to keep it simple. One simple way would be the following: If the af PSE tries to power the 20W PD, if
its current goes higher than Icut, it will go into overload fault and will turn
off for at least 750 msec. One way to make sure the midspan will then
power the high power PD would be to change the spec so that for a PoE+ PSE
Alternative B, the backoff time becomes 500 msec instead of 2 seconds. So, during the fault off time of af PSE,
the midspan will take control of the PD. Regards Jean Power Interface Systems Engineer Phone : 603-222-8683 jean_picard@xxxxxx From:
owner-stds-802-3-poep@xxxxxxxx [mailto:owner-stds-802-3-poep@xxxxxxxx] On Behalf Of Steve Robbins Yair, I think you missed the point. The solution is NOT to
“…mandate separate signatures for each 2P…” because we
also want “medium power” on 2P. The Task Force already
agreed to this. So there will be some PDs in the 15W to
30W(?) range that have an input structure like an Af-PD (SS with diode
bridges). The obvious upgrade path for a user would
be to add a medium-power At-midspan in-line with his existing Af-endspan.
These two PSEs can’t share the load since the PD is SS. So the
At-midspan has to supplant (take the place of) the Af-endspan. I showed a method to assure the midspan
will power the MP PD every time. I think this will be essential to the
802.3at standard. Steve From: Steve and all, The problem shown in this presentation can
be easily solved by mandating detection and classification signature for every
2P in the PD. Specifically separate signatures for each
2P and not a common one as measured at the RJ45. See attached presentation in page 13 for
the proposed scope of work. This is the idea of how standard can be
simplified. Doing the above is not requiring any
special specifications from the PSE. The Endspan PSE will detect and classify
pair A. The Midspan PSE will detect and classify
pair B. Both channels will turned on within the current
802.3af timings or similar to it. Since each channel A and B are not tied
together at the diode bridge output, the PD can be operated easily by using the
PD circuitry which is TBD (implementation independent). Yair From:
owner-stds-802-3-poep@xxxxxxxx [mailto:owner-stds-802-3-poep@xxxxxxxx] On Behalf Of Guys, I think I’ve solved another problem that’s inherent
with DS. See the attached pdf. (Only about 10 more tough problems
to go!) I’d sure appreciate some comments. Steve |