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

Re: [802.3af] voltage offset proposal (was 1-point detection,monotonicity)




Out of this discussion, I believe that I now have a better way
to spec the PD signature.  It has been misleading to say that
the conditions for the required slope were that:
    v is between 2.8 and 10.0 volts, and
    i is between 0 and 500 ua.
This leads some to believe that one could use any current
between 0 and 500 to probe for the signature.
Whereas one could probe with any voltage between 2.8 and 10.0
and this will work for any valid signature,
not every current between 0 and 500 will work all the time.
Rather it depends on the offsets in the signature.
Therefore, I propose to modify the PD signature requirements
to be as below.  The new wording has a required resistance over
a voltage range, a required max voltage offset,
and a required max current offset.  When these are plotted out,
we get the working current range of 0 to 500 ua:

    Signature resistance: same min and max as before, under the
    condition that the voltage is between 2.7 and 10.1 volts.
    (This has the slight voltage change to provide some tolerance
    overlap for the PSE.)

    Voltage Offset for the signature resistance: 0 to 2.7 volts.

    Current Offset for the signature resistance: 0 to 123 ua.

This gives us the same overall possible current range as before,
i.e., 0 to 500 ua.  However, I believe that "500" number was
somewhat arbitrary, but round.  Now that we are specifying the
current offset, let's make it round instead.  Call it 100 ua,
and the new possible current range would be 0 to 477 ua.

John Jetzt wrote:

> Roger,
>
> The 500 ua was chosen to allow the PD to have some amount of
> current offset.  Here is a way to visualize the slope requirement
> of the PD:  In the V-I space we define a rectangular area
> where the sides are
>     top:     v = 10.0 volts
>     bottom:  v = 2.8 volts
>     left:    I = 0.0 ma
>     right:   I = 0.5 ma
> The requirement on the PD is to have a V-I characteristic that
> enters the space through the bottom and leaves through the top,
> and inside the rectangle the characteristic must have the required
> constant slope.
>
> In order to achieve this characteristic, the PD could theoretically
> have a pure voltage offset of up to 2.8 volts.  However, once the
> characteristic enters the space it must have the constant slope.
>
> Similarly, the PD could theoretically have a pure current offset
> of up to 0.123 ma.
>
> The above is the requirement on the PD.  It does not matter if the
> PD has an offset, and it does not matter how the PD achieves an
> offset.  It only matters that the slope in the area has the
> required constant value.
>
> The requirement on the PSE is to look within the space for the
> characteristic.  The PSE is told what the must-detect slopes are,
> and the PSE is warned that there might be an offset.
>
> Therefore, the must-detect characteristics for the PSE are:
>
> a) 19K to 26.5K ohm resistance between pairs,
> b) no more than 110 nF of capacitance,
> c) a voltage offset of at most 2.8 volts DC in the signature,
> c') a current offset of at most 0.123 ma in the signature.
>
> John
>
> R karam wrote:
>
> > Hi John
> >
> > I think the 10.1v Max conflicts with a 500ua max out of the PSE
> > during detection for 500ua* max Rsig=26.5k (if i got this right)
> > that puts is at 13.25v max.  do you know where the 500ua spec
> > came from?
> >
> > also are we happy with the 2.8v being one VBE (diode drop away)
> > say at cold we allow 0.8v per diode that is 1.6v is this too tight?
> > thank you for taking the time to do this - there is no shortage of
> > confusion on the numbers...
> >
> > regards
> > roger
> >
> > >I have a proposal for the signature voltage offset requirement
> > >as seen by the PSE.  However, before I generate an official
> > >comment to the editor, I am sending it to the reflector
> > >for comment.  I believe the proposal below is in
> > >the spirit of our agreements, and it also eliminates
> > >the one-point detection "loophole" in our former discussion.
> > >
> > >The Proposal (in two parts):
> > >Part 1:  Change the valid detection signature maximum
> > >voltage offset that the PSE might see to 2.8 volts,
> > >instead of 1.5 volts (see p.8, l.11).
> > >Part 2:  Change the valid detection signature
> > >voltage range for the PD V-I characteristic
> > >to 2.7 - 10.1 volts, instead of 2.8 - 10 (see p.16, l.22).
> > >
> > >Rationale for Part 1: the 2.8 volt offset:
> > >Several meetings ago we agreed that the detection
> > >test voltage would be 2.8 to 10 volts.
> > >The 2.8 number was selected to allow for
> > >voltage offsets.
> > >
> > >Where did the 1.5 number come from?
> > >I think this is what happened.
> > >The PSE Detection Source Material gave its "must pass"
> > >requirements in the form of circuits that could include
> > >the "typical" two diodes.  D1.2 put this into words.
> > >
> > >Rationale for Part 2: the 2.7 - 10.1 PD voltage range:
> > >This leaves a 100mV band for qualification testing,
> > >and ensures that a slightly-out-of-spec PSE
> > >will still detect a slightly-out-of-spec PD.
> > >(Thanks, Dave D.)
> > >
> > >John Jetzt
> > >