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

Re: Specifying Discovery

Thanks for starting this discussion. (I was out of the office on Friday and
missed the initial flurry of discussion.)

I certainly agree with the view, that many have also expressed, that the
standard should not require particular implementations.

I think some relatively simple constraints on the PSE are appropriate.
Again, its implementation should be flexible. 

Regarding, "the actual method used to do discovery in the PSE should be
left to the individual implementor", I pause at that a bit. The "method" or
concept used can impact the reliability of detection, which we all have an
interest in. We have at least two paths we can follow: 

a.) We can specify enough about the detection "method" (e.g., shall be
based on at least two V/I measurements in such and such regions, with such
and such accuracy) so we have confidence a PSE will make a valid
measurement of slope. Then we primarily specify the slopes it accepts and
rejects. Or,...

b.) We can specify more rigorous test cases to try to verify that the PSE
measures slope with acceptable accuracy. Our hazard matrix experience and
examination of sensitivities led us to the "slope" being a very dependable
and specific signature; we do not want to lose those benefits.

I'm not locked into either approach at this time, but lean to the first
(a), probably just more familiarity. The right answer may be a hybrid of
the two.


Paul Moore wrote:
> I'm sure others have already thought of this, but having a dialog might help us get to completion sooner, so here goes.
> Now that there is a discovery method on the table I began thinking about how we specify it. Seems to me we only need to specify the identity network on the PD end, some sort of simple limitations on the PSE discovery signals, and then put an annex in the standard showing a proven implementation. The actual method used to do discovery in the PSE should be left to the individual implementor. Things like radiated and conducted noise, noise susceptibility, ESD susceptibility, etc. need not be in the spec., but an annex might point out the relevant documents governing them. I'm sure I've missed a few things here, but seems to me the actual spec. part is pretty short. Comments??
> \Paul
> Paul B. Moore
> Senior Manager, Hardware Engineering
> Small Business Solutions - Santa Clara
> Nortel Networks
> 4401 Great America Parkway
> PO Box 58185, MS SC05-02
> Santa Clara, CA 95052-8185
> Phone: 408-495-2466
> FAX: 408-495-5615

Donald (Don) S. Stewart				Phone: 732-817-5495, FAX x4666
Avaya Inc. 					e-mail: dsstewart@xxxxxxxxx
Cross-Product Architecture     			
101 Crawfords Corner Road
Holmdel, NJ 07733