|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
A thought to consider. If schottky diodes are used for this discussion, it should be remembered that schottky’s can be very leaky in the reverse mode. Sometimes several milliamps can flow in a back biased schottky.
So if leakage is a direct concern, this is an important parameter.
From a Motorola spec:
60V 3A MBRD360 Max Irev = 20mA @ 125°C
Analog Product Specialist
Texas Instruments Incorporated
HC66 Box 203
Mountainair, NM 87036
WEB SITE: <http://www.ti.com>
Office = 505-847-0576
Fax = 413-280-0812
1. If the spare pair have diode on each pair i.e. a diode on 4,5 and a diode on 7,8 than the problem I have described can not happen, however using a full bridge on the spare per had the additional advantage of being polarity insensitive and free us from being dependent on what type of cable is being connected.
2. The are few types of crossed cables:
Type 1: only pins 1,2,3,6 are crossed.
Type 2: all pins are crossed.
Therefore I find mandating diode bridge on the spare pair is useful too.
3. The wording of the current draft prevents injecting power backwards from the PD to the PSE however, we are discussing here very
low leakage current that may be enough to mass up the detection function.
I had a quick thought about Yairs point.
I agree, a PD should not send out power it receives on the other wires.
This may cause a problem in the switch. If the power sources for individual ports are connected in the switch, it may also cause problems in other devices attached to the switch.
However, I think if the data pair has a diode bridge, and the spare pairs have a diode in each pair ( a half bridge) ,
the problem you described cannot happen.
So there is no need to mandate polarity insensitivity on the spare pairs per se.
Such a polarity insensitivity would make sense only we consider a cabling that swaps the spare pins is considered as a scenario that needs to work with power over lan. ( May be using Gigabit crossover cables.... )
I´m not sure if the wording of current version of the standard prevents injecting power backwards from the PD to the PSE.
If its not in, it should be added of course.
I would like to discuss the benefits in mandating diode bridge at the input of both data pairs and spare pairs.
The PD is required to be ready to accept power from the spare pairs or from the data pairs.
Typical implementation of Oring the power from data pairs or spare pairs could be one of the following options:
1. Data pairs has diode bridge and spare pairs using single diode.
2. Data pairs has diode bridge and spare pairs has diode bridge.
3. Data pairs and spare pairs has has single series diode each, data pair should have diode bridge if the PD is auto-mdi-x.
Now lets consider the following case:
A multiport system activate port number x and send power to the PD.
The PD is configured per option 1 or 3.
Now, there is voltage present at the output of the oring diode, but, due to the fact that one of the leads of the spare pair is directly connected to one pair data pairs
There is a leakage current path from the data pairs to the spare pairs back to the PSE.
This leakage current will find its way to other ports in the PSE and may affect the detection function.
In some bob-smith termination configurations that was good for a switch without pse and are not suitable for switch with pse some ports may see voltages above 30V even if they are at OFF state.
In order to prevent such scenarios, option 2 is suggested that keep DC isolation from the spare pare to the data pairs and vice versa.
In addition, using diode bridge at the data pairs will fix the issue raised by Moti Goldish regarding the MDI-X/AUTO MDI-X issue.
Mandating diode bridge on both pairs will ensure powering of the PD in any PSE configuration and in any cable type straight or crossed cable
so we can eliminate the potential of interoperability problems regarding the ability to successfully powering the PD.
The data issue is solved by the definitions for the PSE and PD, by the pin assignment and polarity for the MDI/MDI-X/AUTO MDI-X configurations as described in tables 33-1and table 33-7.
Actually referring to Auto MDI-X in tables 33-1 and table 33-7 will not be required anymore.
The suggested remedy to support the above is:
Draft 4.2 page 60:
1. Delete the text at lines 50-51:
"If the interface is implemented as an MDI-X or Auto-MDI-X per Clause 14,the PD shall be polarity insensitive "
Replace it with the following text: "The interface in Mode A and in Mode B shall be polarity insensitive.
2. Consider to delete the reference for Auto-MDI-X from tables 33-1 and 33-7 as it is not required due to (1).
I believe that to mandate the above is required.
Please comment over the above issue as soon as possible.