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

Re: Selection Criteria for choosing a Discovery Method




Rick,
Thanks for the fast comments. A few responses are below.
Don

> Rick Brooks wrote:
> 
> In general, I agree with your proposal list, and I state my responses
> below.
> 
> (I-2) Minimizes cross-talk interference potential to Gigabit Ethernet and
> 
> unidentified future services
> 
> I think that the "future" part is kind of nebulous. We should state the
> requirements that exist now.
[Don] Bob Love has not proposed we score these mathematically or that we
tell anyone how much importance they should place on any particular
criteria. If people do not think it is important to preserve capability for
future services, they can ignore this. Or the PRO for either technique can
spell out how far it goes in this dimension and people can judge for
themselves. An "objective" of 802.3af is support for Gigabit Ethernet, so
there is already a basis for including this.

> 
> Also, I'd rather say does not interfere with 1000BASE-T, rather than
> minimize.
> Otherwise, you could argue that a DC scheme will always minimize
> interference compared with an AC
> scheme. And you would be right.
[Don] I see the purpose of this is to put on the table information that
allows comparison and decision. Therefore it is useful to not just capture
"better than X" but to capture "how much better than X". People can judge
on their own if any extra margin one has over the other is significant.

> 
> (I-3) Minimizes EMI
> 
> Clearly, only a true "DC" discovery scheme can do this.
> Are we then done?
[Don] No. This is just one of several criteria. I understand the proposal
is that people will apply their own weights to each one.

> 
> (R-1) Avoids "false positive" detections (quiet and noisy environments)
> (R-2) Reliably delivers power to proper PEs (quiet and noisy
> environments)
> 
> We need to define "noisy" environment.
> How about passing the EFT (electrical fast transient) test and radiated
> immunity at 5 volts per meter.
> Also, we should use the Campbell clamp fixture to determine where the
> discovery process fails vs frequency,
> as it is stimulated with common mode fields.
> 
> Maybe we should add these cables to test: CAT-3, CAT-5, and CAT-6.
[Don] I'd like to see definition of the best configurations for simulating
real world noise injection. I'd encourage not testing just "better than x"
but showing the interference level that causes failure, to help comparison.
I realize this is a bit more work. In the interest of time, parties may
have to just test to the level they think is sufficient.
> 
> (R-3) Avoids giving power to paralleled devices
> 
> We need further definition on this one.
> No discovery method proposed so far can detect a short paralleled legacy
> device, depending on it's input configuration.
[Don] In the interest of time, parties may have to just address to the
configurations they think demonstrate a "benefit". As to "no discovery
scheme can detect...", I do not follow you. The "Resistive + diode drops"
sees a modified V-I curve for anything except a paralleled open circuit.

> 
> Also, I would add to the list:
> "Avoids failure to detect a valid powerable device".
[Don] This was covered under (R-2) "Reliably delivers power to proper PEs"

> "Automatic recovery from a cable discharge or ESD event".
> 
> (D-1) Low risk of equipment damage IF power is ever inappropriately
> provided
> 
> How is this an issue of the discovery method, that is not covered above
> under robustness?
[Don] I used "Robustness" to address the detect/no-detect performance, and
considered potential damage to devices as slightly different. I am not
sensitive to where we put this. This can be a PRO/COM of a discovery method
because one scheme may do more to safely measure the DC characteristics of
a device before applying full DC power.

> 
> (C-1) Leverages the necessary function of precisely monitoring current
> levels to inexpensively perform detection
> 
> Doesn't this assume a fact that is not yet in evidence?
> Are you saying that discovery and load current measurement must use the
> same method?
> To repeat: Are we then done?
[Don] Again, no single criteria says we are "done". People can integrate
over the body of information and draw their individual conclusion. This
does assume that people see a benefit in being able to predictably provide
up to the maximum safe current. For any given voltage limits we set, this
will ensure maximum power to the PE. If we have only a "crude" continuous
current measurement and cut-off, we effectively can only guarantee the
lowest power consistent with that. I think this is a valuable criteria.

> 
> Other survivability tests:
> Ability to survive RJ-45 Hipot test, whether DTE power is on or off, and
> meet the leakage resistance at the end of the test.
[Don] I have no issues with these suggestions.
> 
> comments?
> - Rick
> 
>      -----Original Message-----
>      From:   Donald S. Stewart [SMTP:dsstewart@xxxxxxxxx]
>      Sent:   Monday, September 25, 2000 12:25 PM
>      To:     802.3af
>      Cc:     RDLove; Donald Stewart
>      Subject:        Re: Selection Criteria for choosing a Discovery
>      Method
> 
>      I applaud Bob's suggestion and his offer to maintain this
>      information. I
>      think it will be helpful in reaching a decision. I include below my
>      proposals for criteria.
>      Regards
>      Don Stewart
>      ______________________
> 
>      Criteria Categories and Specific Elements:
> 
>      Non Interfering
>      (I-1) Does not interfere with 10/100 Ethernet via cross-talk
>      (I-2) Minimizes cross-talk interference potential to Gigabit
>      Ethernet and
>      unidentified future services
>      (I-3) Minimizes EMI
> 
>      Detection Robustness and Flexibility
>      (R-1) Avoids "false positive" detections (quiet and noisy
>      environments)
>      (R-2) Reliably delivers power to proper PEs (quiet and noisy
>      environments)
>      (R-3) Avoids giving power to paralleled devices
>      (R-4) Potential for detecting multiple classes of PE
> 
>      Avoids Damage to DTE
>      (D-1) Low risk of equipment damage IF power is ever inappropriately
>      provided
> 
>      Efficiency
>      (E-1) Ability to deliver maximum, usable power to the PE
> 
>      Cost
>      (C-1) Leverages the necessary function of precisely monitoring
>      current
>      levels to inexpensively perform detection
>      (C-2) Amenable to integrating into the PHY
>      (C-3) Compact, allowing high density PSE implementations
>      (C-4) Compact for PE implementation
>      (C-5) Amenable to low cost Mid-span implementations - Env A
>      (including main
>      necessary elements: sense over/under current, signature detection,
>      power
>      on/off control, on/off switch for current)
>      (C-6) Amenable to low cost LAN-switch integrated implementations -
>      Env A
>      (including main necessary elements: sense over/under current,
>      signature
>      detection, power on/off control, on/off switch for current)
>      (C-7) Amenable to Mid-span Env B implementations
>      (C-8) Amenable to LAN-switch integrated Env B implementations
> 
>      Survivability
>      (S-1) Ability to survive ESD
>      (S-2) Ability to survive Cable discharge
> 
>      --
>      Donald (Don) S. Stewart                         Phone: 732-817-5495,
>      FAX x4666
>      Avaya Inc.                                      e-mail:
>      dsstewart@xxxxxxxxx
>      (former Enterprise Networks Group of Lucent Technologies)
>      Cross-Product Architecture
>      101 Crawfords Corner Road
>      Holmdel, NJ 07733

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