RE: [802.3af] RE: Startup and PD input cap

[Yair Darshan <YairD@xxxxxxxxxxxxxx>]

All,
See my comments below.
Yair.


> -----Original Message-----
> From:	Huynh, Thong [SMTP:Thong_Huynh@xxxxxxxxxxx]
> Sent:	ω, ιεπι 16, 2001 5:39 AM
> To:	'Yair Darshan'; 'R karam'; 'Dave Dwelley'; 'Karl Nakamura'; 'Donald
> S. Stewart'; 'Rick Brooks'; 'Lynch, Brian'; 'Peter Schwartz';
> 'scott_burton@xxxxxxxxx'; 'Steve Carlson'; Koo, Ronald;
> 'mike_s_mccormack@xxxxxxxxxxx'; 'bruce.inn@xxxxxxxxxx';
> 'henryhinrichs@xxxxxxxxxxxx'; 'Jetzt, John J'
> Cc:	'stds-802-3-pwrviamdi@xxxxxxxx'
> Subject:	RE: [802.3af] RE: Startup and PD input cap 
> 
> All,
> 
> I realized that many of us have already shared the opinion about this
> issue.
> For whatever it worth, I would like to pitch in my 2c here, too:
> 
> 	1.  If we can make all the ports turn on simultaneously, why
> necessitate the 8 sec wait?
	[Yair Darshan]  We don't need to turn on simultaneously many ports
since from system point of view we don't have such a need.
	                        We want to control the peak power taken from
the power supply during startup and we want to control the stress
	                        on power components during startup.
				      If we don't need it why bother and
increase system cost?

> 	2.  Integrate 8 ports into a single IC seems like a good idea, but
> why stop there?  How about 16 ports?
	[Yair Darshan]  Check the cost for each device. 1port chip , 8 port
chip, 16port chip, 24port chip.
		                  Than compare it to market needs, 1port ,
8ports , 16ports, 24ports system.
	                        Than find the cost effective solution in
terms of package size, power dissipation on the package at normal operating
mode, and additional pins required 
	                        to manage all ports and send all status
signals etc. 
			            You will see that package size will
increase and along with it the ability to dissipate power. The limiting
factor will be the total power dissipation
	                        at normal operating mode as the number of
ports increased.

					I am strongly suggesting checking
the above. You will be surprised by the results.

					

> 	3.  Putting a single 1J pulse  on an SO-8 package is marginal
> (considering worst case ambient temperature).  Putting 8J on an IC limits
> the package choice to only a few.  Putting 16J on an IC is out of the
> question.
	[Yair Darshan] According to Dave info, 1Joule on 8pin is O.K. and
the actual energy is half of this, hence 100% margin.
	                      Again as I have mention above, no reason for
more than one port at startup mode out of a group of 8 ports.

> 	4.  System with multiple ports already has to struggle with size and
> heat removal.  Adding 48J onto a 48 ports system is indeed quite a burden.
	[Yair Darshan]  Yes, but if we don't need to dissipate 48J or
96Joule etc. due to the fact that we don't need to activate them
simultaneously,
	                        it is not relevant to system needs.

> 	5.  Having inrush limit PD ensures that the voltage at the terminal
> of the PD will be well behave during power up. 
	[Yair Darshan]  I disagree with you.
					The same results in terms of "well
defined" you will get if the inrush current limiting will be in the PSE with
out having inrush limiting in the PD.
	 				a- If the inrush current is in the
PD, the port voltage will jump quickly to around 30V or more.
	                            dV/dT on the cable can be
(30V-40V)/less than 1mSec.
	                        b- If the inrush is in the PSE, the port
voltage will start from zero and ramping up nicely and smoothly.
	                            dV/dT on the cable can be  (1V-2V)/mSec.

					From noise and transient point of
view I prefer option b. In both cases the voltage on the PD port is well
defined.
			            It is even better in option b. 

	                        

>  It will simplify the isolation switch's UVLO design.  I will also prevent
> the classification
> current source from turning back on because the PD voltage will remain
> above
> 30V.
	[Yair Darshan]  It is really ammeter of implementation. Isolating
switch with out inrush current limiter in the PD is much simpler than one
with inrush current limit.
	Have you build and test one to see the clear differences? when
working with PSE. There are system issues that should be address to have
complete and reliable
	solution.

> 	I would vote for having both inrush limit on PSE and PD with the
> inrush on the PSE is set higher than that in the PD.  This approach offers
> a
> more predictable behavior in the PSE/PD interface and add to the
> robustness
> of the standard.
> 
> BR's
> 
> Thong Huynh
> 
> 
> > -----Original Message-----
> > From:	Yair Darshan [SMTP:YairD@xxxxxxxxxxxxxx]
> > Sent:	Friday, June 15, 2001 6:53 PM
> > To:	'R karam'; Yair Darshan; 'Dave Dwelley'; 'Yair Darshan'; 'Karl
> > Nakamura'; 'Donald S. Stewart'; 'Rick Brooks'; 'Lynch, Brian'; 'Peter
> > Schwartz'; 'scott_burton@xxxxxxxxx'; 'Steve Carlson';
> > 'rk@xxxxxxxxxxxxxxx'; 'mike_s_mccormack@xxxxxxxxxxx';
> > 'bruce.inn@xxxxxxxxxx'; 'henryhinrichs@xxxxxxxxxxxx'; 'Jetzt, John J'
> > Cc:	'stds-802-3-pwrviamdi@xxxxxxxx'
> > Subject:	RE: [802.3af] RE: Startup and PD input cap 
> > 
> > 
> > I was out for 4 hours and partying when I arrived again home I have
> > noticed
> > that you still around, so here it is..
> > 
> > Roger the story here is very simple. No need to complicate it further
> and
> > I
> > believe that we can close it if we stick to engineering facts.
> > 
> > In St Louis we have agreed as a compromise that the PSE will contain
> > inrush
> > current limiter and will be responsible to limit the current for PD
> input
> > cap up to 50uF.
> > And from 50uF and up the PD will be responsible for limiting the
> current.
> > Right?
> > 
> > I have told you and the others back than that 50uF is too low and all of
> > us
> > agreed that Dave and I will check again the number if we can increase it
> > pending that integration in a chip is not impaired.
> > 
> > I did it and the numbers confirmed by Dave and others.
> > 
> > Now we can move the number from 50uF to 350uF(and still we have a lot of
> > margin). Simple right? So what is the problem now? 
> > 
> > As you told me in St Louis. If somebody will bring new and proven data
> we
> > will change it right?
> > 
> > 
> > In addition, see my comments below.
> > 
> > Thanks 
> > 
> > Yair.
> > 
> > 
> > 
> > 
> > > -----Original Message-----
> > > From:	R karam [SMTP:rkaram@xxxxxxxxx]
> > > Sent:	ω, ιεπι 16, 2001 2:03 AM
> > > To:	Yair Darshan; 'Dave Dwelley'; 'Yair Darshan'; 'Karl
> Nakamura';
> > > 'Donald S. Stewart'; 'Rick Brooks'; 'Lynch, Brian'; 'Peter Schwartz';
> > > 'scott_burton@xxxxxxxxx'; 'Steve Carlson'; 'rk@xxxxxxxxxxxxxxx';
> > > 'mike_s_mccormack@xxxxxxxxxxx'; 'bruce.inn@xxxxxxxxxx';
> > > 'henryhinrichs@xxxxxxxxxxxx'; 'Jetzt, John J'
> > > Cc:	'stds-802-3-pwrviamdi@xxxxxxxx'
> > > Subject:	Re: [802.3af] RE: Startup and PD input cap 
> > > 
> > > Yair
> > > 
> > > Now I thought you would be home partying by now, what are 
> > > you still doing at work (laughs)
> > > 
> > > all right, I think the right answer may lie somewhere between Yair and
> a
> > > bunch
> > > of us who beleive in our own story, that is to Inrush in the PD or not
> > to
> > > inrush in the PD.
> > 	[Yair Darshan]  I agree with the group that support inrush current
> > limit in the PD for an input cap greater than TBD uF. Now we are arguing
> > the
> > number. We have calculated it and now the number is clear. 600uF is O.K.
> > 350uF is proposed.
> > > and so at the time I thought the 50uf was a compromise of live and let
> > > live.
> > 	[Yair Darshan]  With 350uF you let live for more PD vendors and I am
> > not a PD vendor.......
> > 
> > > So Yair wants that increased now (the capacitance) .  I think it would
> > be
> > > nice to find 
> > > a way to  prove that we could setup a PD designer to fail without
> > inrush-
> > > or cause grief
> > > of any kind.  
> > 	[Yair Darshan]  I can give you a list of PD's without inrush
> > limiting that was connected to a PSE with inrush limiting and they are
> > working perfectly.
> > 	I can give you a shorter list of PD's that was connected to PSE
> > without inrush current limiter in the PSE and the results.
> > 
> > > we can not test every possibility obviously (if this was a case of
> > Yair's
> > > Box or
> > > mine -system and pd ) it would be easy, this is Yair's box and Mine vs
> > the
> > > rest of the world's PD's.
> > > the big question is :
> > 	[Yair Darshan]  Same as above.
> > 
> > > can we think of a single case where we cause grief and grief is
> > > interoperability problems like
> > > Dave  was saying, failure to power up, oscillating) .... without
> inrush
> > in
> > > the PD, if we can set
> > > up such a case prove it in the lab, then we may be done.
> > 	[Yair Darshan]  As long as the PD designer will design according to
> > his spec. Why he will face problems?
> > 	If the PD designer will design PD with inrush current in it and will
> > not obey the PD spec, he will face problem too right?
> > 
> > > if I have enough time, I shall do some of this.  
> > > so far I managed to repeat Rick's noise's measurements and we agree on
> > > most of it,
> > > there are some that require measuring (uv- common mode case) so the
> fun
> > > never ends.
> > > 
> > > roger
> > > 
> > >