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

RE: [802.3af] inadequate power PSEs

But this is exactly the argument FOR allowance of Mike’s underpowered PSE. 


The spec already allows for a designer to make a multiport PSE and underpower it for full power on every port.  It is wholly legal to make a 48 port PSE and put one 15.4W power supply in for POE.  Of course this is a ridiculous theoretical product, but it is allow in the text. 


It would also require a label to explain that only one port is able to provide full power and/or a page full of text in the product docs to cover it fully for the end user.  The only difference between what Mike is asking for and a multiport port PSE with only 15.4W available is the spec allows the latter.  This was added to the spec to ease the cost requirement on the power supply for a multiport PSE.  Mike is just asking for some leeway for a single port PSE.


My half-a-cent.


Chad Jones


-----Original Message-----
From: owner-stds-802-3-pwrviamdi@xxxxxxxxxxxxxxxxxx [mailto:owner-stds-802-3-pwrviamdi@xxxxxxxxxxxxxxxxxx]On Behalf Of Roger Karam
Sent: Thursday, January 23, 2003 3:20 PM
To: Geoff Thompson
Cc: Steve Jackson; Scott_Burton@xxxxxxxxx; stds-802-3-pwrviamdi@xxxxxxxxxxxxx
Subject: Re: [802.3af] inadequate power PSEs



I am with Goeff on the plug and play issue if we open this up to labeling
then everyone will get 'creative' about going around the power supply issue
but not by doing the right design to account for power - but by shipping
1/2 the wattage needed and claiming compliance ....  Then again chances
are Mike will come up with argument that tempts me to reconsider.

to me that fact this is NOT too clear, and is debatable, proves that
the risk of it making a mess are great.... my 2c anyway....

also we run hard to 'graduate' and we seem to go back and open up subjects that
were sweated for months before.. just because we can not recall the details:
classification and class zero though not too sophisticated gave us a start
for hooks into the phy later, a back up plan for discovery that was plagued with
low current suseptibility to noise, and leackage of diodes and power FETs.
these were real concerns that we tackled for months.... I shall invite
Rick to refresh our memory ....

also we have increased the detection signature to a much wider value than
the one the hazard matrix was based on... so can we stop fixing 'unbroken things'
and move on....

at the end of the day, it may be hard to tell the exact right thing to do.
but the is what voting is for....

At 10:49 AM 1/23/2003, Geoff Thompson wrote:


I think I will stick to my position. Steve's arguments for the other side below only (in my opinion) prove my point.

20 AMPs in the guest bathroom is just fine. It corresponds to our full power situation. You can run an iron or an electric heater off a 20 AMP circuit. If you try to do both then you deserve to have the breaker trip. The breaker BTW, these days is a GFI in the bathroom. You don't have to go down to the basement to reset.

The cigarette lighter socket in Steve'car is the 2nd socket. There is still a high amperage socket for the lighter within reach of the driver and the car came with a lighter in that socket. When additional socket started appearing in cars it was in a mature socket situation and nobody (well, there was probably some fool somewhere) was looking to add more cigarette lighters, they were looking for a place to plug in their cell phones. The bulk of the market understood the situation quite well.

Our situation, on the other hand is somewhat different (Opinion piece).
        1) The market we are approaching is used to wall warts. The market expectation for those is that they are absolutely not interchangeable between products. Our goal is that DTE Power just works everywhere. Kill wall warts!
        2) I am of the opinion that if we had not had Auto-Negotiation for 10/100 Ethernet then 100BASE-TX would have been a bust, or at least a slug. This would have been true even if product had Auto-Negotiation but what was sold was like to not be able to find a common operating mode between the 2 ends. One of the reasons (albeit probably a minor one) that 100VG-AnyLAN failed was that you had to manually configure it to match the two ends of the wire. It wasn't plug and play.

Therefore, I am still of the opinion that we should position the front end of the market so that everything just works. When folks understand it a little better, a couple of years down the road, then we can back down.


At 12:09 PM 1/23/2003 -0500, Steve Jackson wrote:

... pardon the tongue-in-cheek title ...

This issue really has me torn.

Mike's rationale is great, and Geoff's concerns are also 110% valid.

I'm going to speak up, since I haven't for a long time. I support the low-power-warning label "deal" for the following reasons:

1. Power-limit-at-the-socket paradigms are a fact of life in our consumer society. Having a label stating this is a bonus that isn't always offered; my car's cigarette lighter socket, which cannot source enough power to perform that function, is so labeled. The AC outlet in the guest bathroom can't source 20A, and isn't labeled. The circuit breaker is. Mike cites some .3 precedent but that isn't a concern to me. A 4 watt port is OK as long as you tell me about it, and besides, that's why they make polyfuses.

2. I didn't like the classification idea from the get-go, which, as you'll recall, started out as an optional feature. Having this 'feature' opens up (encourages) the possibility of deployment of poorly-architected .af systems, defined by me as those not capable of delivering battery-backed-up full power on all ports. Now that someone WANTS to sell such a system ON PURPOSE (even worse than the power allocation scheme) I say, go ahead. As long as it's labeled. Make sure the warning-label color is specified as bright yellow; caveat emptor. I also think that power-allocated ports ought to be so indicated by a blinking yellow port-status LED, but I never thought I'd get anywhere with such a proposal.

3. Oh, yeah: I mentioned power-allocated systems in #2. If we allow them, we have explicitly already allowed what Mike wants. I admit ignorance of the discussion-goings-on since I no longer can attend the .af meetings, but, heck. Why the fuss?

Nomex suit on.