Hi Jeff and Ken,
Please see my inputs below.
From: Jeff Heath [mailto:jheath@xxxxxxxxxx]
Sent: Saturday, May 03, 2014 12:20 AM
Subject: Re: [802.3_4PPOE] Base Line proposal - 802.3bt End to End Channel P2P Resistance Unbalance ad hoc
I agree with nearly everything you said. I have the following comments:
When specifying a PI as a black box the specification needs to include all ‘current ranges of interest’ whatever they may be.
Yair: I am not sure that this statement is correct.
We need first to look at the operating states of the system e.g. Detection, Classification, Startup, Power ON and MPS and decided first what states we care?
And then to specify what we want.
For sure we need to specify the P2PRUNB requirements for POWER_ON state (PD is on AND WORKING).
Regarding other states, first we need to work to justify if P2PRUNB is relevant there. We are not there yet.
There is no point to specify Pair to Pair current unbalance in detection mode, work on it, spend time without clear justification. Same for classification etc.
If we use discrete points, interoperability is not guaranteed from a spec standpoint.
Yair: If you limit the discussion per operating state as I suggested above, automatically you specify the operating current that is relevant to this state, and then you just need to specify the worst
case P2PRUNB of YOUR PD and you don’t need to take more points.
More over 2, 3 or more points doesn’t guarantee interoperability. PD input impedance behavior is complex and can be almost any. It is not practical cost wise and any wise to base compliant on this
A text such "PD PI P2PRUNB at ON_STATE shall not exceed TBD% between any two pairs" (or equivalent) is sufficient for interoperability. (There is only one point where you will have the maximum P2PRUNB
and it need to be under spec limit.
This put the burden on PD vendor to guarantee it by design and test it under the following conditions as long as the following requirements are met:
-During Power_ON state (or other individual state that we decide)
-Over POWER_ON state (or other specified state) operating voltage current range.
-Other condition relevant to this stae.
This will help to guarantee interoperability. How many points is vendor issue not spec. issue.
Testing on the other hand cannot test every current in a range.
I am generally of the opinion that a specification needs to cover the range.
The test from a practical matter does not and there is always a leap of faith.
A hypothetical example of this might be the testing of DC disconnect. The spec is very clear. The PSE must not turn off a PD if the current is above 10mA for 60mS (in a given
period). One can test with 10mA but it is generally not necessary to test again at 11mA, 12mA …
Yair: agree. So if we specify the worst case in a given range, we don’t have to test the whole range, just define the worst case condition, and this requires some work. I don’t have the answer for
If however a PSE is mis-designed, and let’s say turns off a PD at 20mA, a tester may not catch it for the above reason, but a test could be done to show that the PSE is out of
specification and therefor at fault from an interoperability perspective.
Yair: Not clear?
However we end up specifying the PSE and PD PIs, I think we need clear specifications and operating conditions that cover the subject, and not a discrete set of points within
whatever range we choose to specify current imbalance.
Design Center Manager
402 East Carrillo Street, Suite D
Santa Barbara, California 93101
Here's some follow-up thoughts to consider regarding yesterday's Runbalance spec discussion:
1. In 802.3af/at, Runbalance was easily and necessarily defined, since it could focus only on resistance of wires and the points between the port and the power taps. That spec applies pretty universally to all devices.
2. An Runbalance spec exists primarily to address Iunbalance. Previously, Runbalance alone could be used to determine Iunbalance. This is not the case with 4-pair; other factors include:
2a) PSE Case: Voltage differences at the PSE output. Voltages are not inherently equal and can have a greater effect on Iunbalance than Runbalance does. Specifying a worst case Voltage unbalance at a specified current loading, may be more appropriate
because it would inherently include PSE Runbalance. Conversely, specifying PSE Runbalance alone does not remove the need to have a Voltage unbalance spec.
2b) PD Case: As in the PSE case, Runbalance alone at the PD input is not solely responsible for PD contribution to Iunbalance. Implementation flexibility imposes many potential influences. Other items mentioned below also factor in.
3. There are cases where an Runbalance spec at a PD PI is an unnecessary and perhaps burdensome requirement. Examples where PD PI Runbalance doesn't matter are:
3a) (As mentioned in the meeting) there may be PD's under a (TBD) power limit with inherent Iunbalance between pairs (ie a 200mA/100mA split)
3b) Some of the architectures under consideration (ie. x2 PD's)
3c) PD's implementing Iunbalance correction circuitry
4. PD PI Runbalance Testability is complicated. If a test method is suggested, it should focus on testing under a powered-on condition because 1) this is where it matters, and 2) the resistance at Powered-on Voltages can be different than resistances at unpowered
Voltages. A PD PI test recommendation ideally would be doable under normal operating conditions without having to probe circuits or develop special test modes.
There's a fundamental problem with testing PD powered-on input resistance: a PD can and typically does present a load which varies over time. If a test is proposed which involves sequential measurements, it will be prone to error due to varying loads. A simple
straightforward test would be to impose a Voltage on all four pairs and measure the currents. This would allow a resistance calculation free of errors. Note however that this is essentially doing a current unbalance measurement for the purpose of calculating
resistance, which in turn is used for determining current unbalance. Why not simply impose a maximum current unbalance spec at specified Voltage inputs?
1) Runbalance spec of a cable... no problem.
2) PSE's: it may be more appropriate to specify a Voltage unbalance at specified currents
3) PD's: it may be more appropriate to specify current unbalance at specified Voltages, with allowances for PD's under TBD watts.
4) PI Runbalance work done thus far can certainly be used as a basis for specifying the test currents and Voltages, but it doesn't need to be an additional spec.
On 5/1/2014 8:23 AM, Darshan, Yair wrote:
Attached please find the meeting material for today in addition to what was send already today. Total 3 files.
802.3bt End to End Channel P2P Resistance Unbalance ad hocAd hoc chair.
Power over HDBaseT Subcommittee
Chief R&D Engineer
Analog Mixed Signal Group
1 Hanagar St., P.O. Box 7220
Neve Ne'eman Industrial Zone
Hod Hasharon 45421, Israel