Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Yes, I agree. From statistical point of view and
from what we know from the field, you are correct. Yair From: Patoka, Martin
[mailto:mpatoka@ti.com] Yair, In our separate email chain, you proposed that
the reason to power was; "Yair: Yes this is true but it is possible the PD is equipped
with defective classification circuit and still will get power due to the fact
that he pass detection and he is truly PD and not dumb PD. The chances are low
but it is possible." It would seem more likely that a device that fooled
detection (and we know they are out there), but showed a bad classification is
far more likely than a PD that failed in such a way that only the
classification was bad but nothing else. That has to be a
one-in-a-billion failure. Martin
From: Statement #1: "802.3af PD that takes 50mA during classification is still a
compliant device." Yair, you are mixing up the ends here. You
are correct that it is perfectly legal for an AF PSE to power a PD that draws
<50mA during classification. I think you are overlooking the fact that
it is illegal for the PD to actually draw more than 44mA (see 33.3.4 of
802.3af: In addition to a valid detection signature, PDs shall provide the
characteristics of a classification signature as specified in Table
33-11). A PD that draws more than 44mA during
classification is non-compliant. Statement #2 "It is
allowed by the 802.3af and is treated by the 802.3af PSE." I think I proved above that it is not allowed for
the PD. Statement #3 "If it was a mistake to allow it or not in the 802.3af is not
relevant now due to the fact that it was allowed." It is only relevant for Type 1 PSEs -- and only
if anyone else cares to continue to allow power to non-compliant devices.
I have no problem closing a loophole in the first standard to disallow bad
behavior. Statement #4 "In addition, it was not a mistake to allow 802.3af PSE to power
802.3af PD with bad classification due to the fact that in 802.3af the whole
classification issue was optional and especially in 802.3af PSE it was optional
so in order to give the PD the same treatment when it is connecting to a PSE
that do classification to a PSE that is not doing classification you had to
power the PD in case of classification error of any kink." Classification was only optional in the PSE, the
PD is strictly required to conform to some sort of classification. It was
considered optional because you can get Class 0 for free with the detection
resistor. I don't really care if I don't get equal treatment between PSEs
that perform classification and PSEs that don't when we are talking
about a non-compliant PD. Maybe classification should have been
mandatory for the PSE. That is certainly a better solution than allowing
power to a non-compliant device. To clarify my statement that you didn't understand,
let me place some sentences here: from 33.2.8.1: "A Type 2 PSE that
has failed to complete mutual identification may provide Class 0 power." from July 2007 Plenary: "Move that Type 2 PSEs may optionally
power Type 1 PDs with Type 2 current limits." Figure 33-14 PI operating current templates, this
applies to both Type 1 and Type 2 PSEs and makes no distinction between T1 and
T2 PDs. It is not explicitly stated but via the
statements above I can rationally make a PD that fails to complete mutual ID
and gets Class 0 power. But according to the agreement in SF in July 2007
(moved by Schindler, Seconded by Darshan), the Type 2 PSE is free to power the
Type 1 PD with Type 2 limits. So now I can be a non-compliant PD and I
can reasonably expect that I will get 30W from a Type 2 PSE. And the PSE
is not required to police and only has to conform to the operating current
template which will allow 600mA for a Type 2 PSE. That was the point of
my statement. Statement #5 "The fact is The PD you mention is impossible to design.
If it draws no current during classification (disables the detection resistor),
it is still Class 0. Any other current falls into some category, Class 1,
2, 3, 4 or fail. I know we can't prevent all non-compliant behavior but,
in my opinion, it is bad form to make it SO EASY to misbehave and still get
powered. Statement #6 "Now what will cause PSE vendors or PD
vendors more problems and noise from the field? Type 1 PD that always working
with Type 1 PSE but not working with Type 2 PSE OR fooling our selves that we
have the ultimate solution how to prevent using dumb PDs? You can see to you
don’t have to be smart to create dumb PDs." My assumption here is that people have been
making compliant PDs and there will be no noise from the field. As I said
above, I am not fooling myself into thinking this can be made bulletproof, but
this is a glaring hole in the spec. And I disagree, you have to be very
smart to make the dumb PDs and you have to do it on purpose because you
thoroughly read the standard and understand all the intertwined rules that
allow misbehavior. - From: Hi 802.3af PD that takes 50mA during classification
is still a compliant device. It is allowed by the 802.3af and is treated by
the 802.3af PSE. If it was a mistake to allow it or not in the
802.3af is not relevant now due to the fact that it was allowed. In addition, it was not a mistake to allow
802.3af PSE to power 802.3af PD with bad classification due to the fact that in
802.3af the whole classification issue was optional and especially in 802.3af
PSE it was optional so in order to give the PD the same treatment when it is
connecting to a PSE that do classification to a PSE that is not doing
classification you had to power the PD in case of classification error of any
kink. This was exactly an interoperability issue only because PSE
classification function was optional. Regarding the argument " The other problem with your remedy is that there is a two layer
slight of hand going on here. If you allow a Type 2 PSE to assign this
noncompliant PD class 0 there is another statement that says the Type 2 PSE can
treat the class 0 PD as Class 4 and now we are back to the start. You are
allowing a noncompliant device to draw up to 30W, therefore enabling dumb
PDs. My opinion is this is completely unacceptable. If I were a
devious designer, I would just draw 60mA during class and then move on to
drawing 30W." I dot understand the point you are trying to make
since I am not sure that the facts are correct or I didn’t understand
you: You said: " there is another statement that
says the Type 2 PSE can treat the class 0 PD as Class 4 and now we are back to
the start. " The specification says that in 802.3af Class 4 PD
is treated as class 0 and not as you mentioned above. So I don’t
understand the argument? You
said: . "You are allowing a
noncompliant device to draw up to 30W, therefore enabling dumb PDs. My
opinion is this is completely unacceptable. If I were a devious designer,
I would just draw 60mA during class and then move on to drawing 30W." How I allow non compliant PD to get 30w? Case 1: PD is drawing 60mA during classification. What could be the options? 1. It is Type 1 PD which according to 802.3af it
will get power from the PSE. 2. It is Type 2 PD with bad classification circuit or a non compliant PD
or dumb PD. The fact is He design a PD with resistor signature and without
classification at all. So when voltage is applied it will take full power
without any issue. It is not compliant behavior but you can not prevent
it… Now what will cause PSE vendors or PD vendors more
problems and noise from the field? Type 1 PD that always working with Type 1 PSE but not
working with Type 2 PSE OR fooling our selves that we have the ultimate
solution how to prevent using dumb PDs? You can see to you don’t have to
be smart to create dumb PDs. Yair From: We only have to guarantee interoperability for
compliant devices. A device that draw more than 51mA during
classification is noncompliant. As I pointed out in my comment it was a mistake
in AF to allow a PSE to power a device that has failed classification and that
we should fix that in AT. I'm not concerned that a device that is
noncompliant will not get power from a Type 2 PSE. The other problem with your remedy is that there
is a two layer slight of hand going on here. If you allow a Type 2 PSE to
assign this noncompliant PD class 0 there is another statement that says the
Type 2 PSE can treat the class 0 PD as Class 4 and now we are back to the start.
You are allowing a noncompliant device to draw up to 30W, therefore enabling
dumb PDs. My opinion is this is completely unacceptable. If I were
a devious designer, I would just draw 60mA during class and then move on to
drawing 30W. the only thing we have to be careful of is that
we are not making all legacy PSEs noncompliant by removing their ability to
power devices that fail classification as Class 0. I would support your
statement if it said: If
a PSE detects Iclass_lim, it may power the PD strictly at Class 0 levels
or it may choose to enter the IDLE mode. A Type 2 PSE that decides to
power a PD that violates Iclass_lim shall limit Pport to 15.4W. In this
case, there is no allowance for the Type 2 PSE to power at Type 2 levels. or some other equivalent but better
wording. I would want a shall statement that generates a PICS that would
check that a PSE will police at 15.4W if it powers a noncompliant device.
Of course we would also have to figure out how to work this into the state
diagram. To me the easier solution is to make the decision
to not power these devices. - From: Hi
guys, I
have found a problem with the remedy on the above subject. We
have 3 cases in which we agree last meeting to return to IDLE state. Case
1: If
PSE type 2 reads class The
rational was that we don’t want to encourage someone using different
coding so we can reserve it for future use. So
far it makes sense. Case
2: If
PSE Type 2 pass detection successfully and fails to complete classification, it
will return to IDLE state. Rational: Classification
in Type 2 PSE and PD is mandatory. If it is not working then probably one of
the two parts PSE or PD is non compliant or defective. Case
3: If
PSE detects Iclass_lim, it will return to IDLE. The
problem is in Case 3. Rational: a)
If PSE type 2 connected to PD type 1, and PSE reads Iclass_lim then according
to last week change, PSE will return to IDLE state. b)
But If PSE type 1 connected to PD type 1, and PSE reads Iclass_lim, PSE is
assign Class 0 which mean the PD will be operated. Which
means that in (a), PD type 1 will not work ever never with PSE type 2 under
Iclass_lim and may work with PSE type 1. This
is violates backwards compatibility and creates interoperability problems
between PSE type 1 and PSE type 2 connected to PD type 1 exhibiting Iclass_lim. The
solution for this problem is: In
Case 3 (and maybe also in case 2) , The PSE type 2 shall assign Class 0 as well
as Type 1 PSE, if it detects Iclass_lim. What
is your opinion on these guys? Yair
Analog Mixed Signal Group Microsemi Corporation
Cell: +972-54-4893019 E-mail: <mailto:ydarshan@microsemi.com>. |