Thread Links  Date Links  

Thread Prev  Thread Next  Thread Index  Date Prev  Date Next  Date Index 
Hi Jeff, Thanks for asking. My approach to specifying a behavior of a system or subsystem is as follows: It has to be 100% mathematically proven. This is the only way known to science as bullet proof. Simulation is a tool to test your math but is not replacing it. With good and solid math, you can save tons of
work. There is no such thing to specify something and leave it loose by justifying it with "It is low probability to happen". Why? Because you/we have no clue what will be in the future, what we have in the past.
The ONLY WAY you can justify such "low probability" is to show mathematically and system performance wise why it is impossible.
 For your information such systems with very small resistance in the positive rail and larger in the negative rail do exists. But as I say ,to me it is not important if the exist now or tomorrow. I want a spec
that after standard ratification, all the parameters affecting system behavior, are well defined. It is also highly important during compliance test to not allow false interpretation by creative test houses.
Now addressing what I said at the meeting: The goal is specifying the PSE PI model In a way that its P2P unbalance parameter will limit the max current over the lowest pair resistance to a tolerable maximum value. This is the only goal of the PSE PI
model to me as a system vendor. If you did the math, you can see the I_max over a pair is not controlled only by Rdiff. See Ken presentation and see my work in the adhoc. To save you the time, here is the equation the controls I max: Imax=It*(1+PSE_P2PRUNB)/2. (Validated many time). It is the total PD current for all 4P.
PSE_P2PRUNB is a NUMBER between 0 to 1. PSE_P2PRUN=P2P_Current_UNB=(RmaxRmin)/(Rmax+Rmin). (Ignoring voltage difference effect for simplicity) PSE P2P RUNB = PSE P2P Current unbalance by definition. Digging further in this equation you will see:
Imax=It*(1+(RmaxRmin)/(Rmax+Rmin))/2. Now you can see that we are not using the P2PRUNB number but only its specified components.
And one last step for fully understand the problem: Imax=It*(1+Rdiff/(Rmax+Rmin))/2. Now I show you the equation with Rdiff or Delta Resistance as you call it. You can see that Imax is function of Rdif BUT also a function Rmax+Rmin. You need two parameters to close the spec:
Imax is given by system constrains such transformers etc. And you must have the flowing set of numbers: a)
Rmax, Rmin. (you can compute Rdiff.)
b)
Rdiff, Rmin. (you can compute Rmax) c)
Rdiff, Rmax. (you can compute Rmin) d)
P2PRUNB and Rmin ((you can compute Rmax) Having only Rdiff is not sufficient and leaves holes in the spec such major interoperability issue for example: You can have: Rdiff=RmaxRmin=10.9=0.1 ohms. Rdiff=109.9=0.1 ohm Rdiff=03.0.1=0.2 ohms. All the same Rdiff, but totally different Rmax, Rmin resulting with totally different Current Unbalance behavior that cannot be MATCHED WITH THE TOTAL END TO END REQUIREMENT. There is no other option per the math. Now after having knowing this what is the relevancy of low probability of this and that? And have bad specification WITH HIGH PROBABLITY FOR INTEROPERABILITY ISSUES?
Yes I suggest that the PSE will include the following parameters: 
Voltage difference spec. 
P2PRUNB spec. 
Rmin (tbd for now ) but I saw that we have a guarantee of at least 0.1 ohm. Other sets of numbers are possible as shown above. This is the best set in my opinion as it the closest to "implementation independent" requirement due to the fact the P2PRUN is a ration and not absolute number
as Rdiff only. If you agree , very good. If you don’t please use math to show why. This is the only way to prove your argument. Yair From: Jeff Heath [mailto:jheath@ here is teLINEAR.COM]
From: Jeff Heath [mailto:jheath@xxxxxxxxxx]
Yair, I believe you had an objection regarding doing a PI specification using maximum delta resistances on slide 8 for instance. Your objection was that if Rps1 was 0.95 ohms and Rps2
was 0.0 ohms the percentage was too large. Given that the goal is maximum pair current, can you explain why this highly improbable PSE case is an issue? Are you suggesting that the PSE PI will include a minimum resistance? If so, what would it be? Regards,
