|Thread Links||Date Links|
|Thread Prev||Thread Next||Thread Index||Date Prev||Date Next||Date Index|
Thanks Christian to start that discussion, I see the clear benefits both on costs and efficiency of using a single diode (or FET) instead of a full bridge, and it looks very interesting.
I’m wondering however about backwards compatibility, for example if the “bt” PD, using ALT-A, is connected to an existing switch which happens to use reverse polarity to provide PoE on ALT-A.
Could you explain how this would work?
Thanks and Regards
Thank you very much for brainstorming, and finding out pros and cons of this subject.
I would like to state again my position: My only goal is to have a defined polarity at the PSE output, which may allow well-defined systems to work without a full input bridge.
I don’t want to get rid of diodes at all, I like the plug&play feature, and as a customer I don’t like to bother about the right cable to use.
But, anyway, what we are facing today is a new power standard, for application which will need more power than ever.
For this new application there will be new PSEs and probably need for better cables.
In .at we already specified CAT5E as minimum requirement, so CAT3 cables might not work for 25W applications.
What is the difference between this requirement, and the new one for .bt, which might be: use CAT6 , straight, 37-max bundle-sized cables if you want to reach 99.9W?
Then I would like to address also other points from Jeff and Jean:
In general all the reverse-polarity issues can be solved with a single diode (FET) at PD input. To drive a low-side FET (on the lower voltage leg) is simpler than a FET bridge, and the losses are halved.
About backfeed voltage and auxiliary sources I see some issues with FET bridge also (it is currently an issue). It won’t stop the reverse current automatically. So the requirement in the standard must stay there. It will be a system implementation issue.
Very good point Jeff regarding the back-feeding that can mess detection.
This is a very good point Jean.
I would add we should also consider the back feeding from a 2 pair connected .bt PD to the unpowered pair (e.g. midspan case).
I think there is one important case that is overlooked in this discussion.
There are a LOT of PD equipments using a DC adapter input, which means it can be powered from PoE or the adapter.
You need blocking diode(s) to avoid the adapter being short-circuited by the PSE power supply through the port connection.
I would confirm what Keng-hua reports, that I know of many PHYs with the crossover correction, on both sets of pairs independently. I agree that this feature in the PHYs works data-only, and will not effect powering. However, because of this feature, we therefore do not see any trouble calls from mis-connected crossover cabling. Therefore we do not seem to have a good way to see whether the two sets of pairs are crossed over independently in the field.
One reason that the feature is popular is that the equipment vendor no longer has to worry about the configuration of the wiring (with regards to crossovers). Specifying a specific configuration for 4-pair POE applications seems to be a step backwards, and would lead us back to the days when the end-user would have to have multiple sets of patch cords handy when connecting a new link so that he could make sure he had the right crossover combination at the end point. Remember that there are several connections in the link topology where a crossover may or may not occur (either by mistake or by design).
Principal, CME Consulting
Experts in Advanced PHYsical Communications Technology
(PLEASE NOTE NEW EMAIL ADDRESS. THE OTHER WILL STILL WORK, BUT PLEASE USE THIS FOR CME BUSINESS)
Thanks Goran, Kenghua,
So, if we agree that Auto MDI-X is a data-only feature, and the PoE voltage will not be affected by this feature, I would propose not to mention Auto-MDI-X when talking about PoE since it creates confusion.
The only reason to mention auto MDI-X is to say that this feature makes crossover cables not needed anymore. And I know it is very widely-adopted feature.
For a brand new installation, for high power PoE application, we are specifying cabling performance anyway (cable category, bundle size etc..).
So, it seems to me that specifying to use straight cable would be fair as well. It would add no cost, and it would make lose no feature.
I’m not trying to get rid of active bridge in application where plug&play and backward compatibility are requested and appreciated feature!
I would like to specify a single voltage polarity at new 4PPOE PSE output, so that the new 4PPOE PD would not need diode bridges for high power application (e.g. > 51W)
Doing so, there will be no need for designing PSEs with voltage auto-selection as Goran is proposing.
I think this would add extra-cost and complexity to the PSE, and also could impact system efficiency since extra-switches are needed.
It is data-only feature. The transceiver will cross the data path internally but the physical connection of the power delivery path remain the same.
Hi Christian, all,
I would assume that it is a data feature only.
However, it would be really nice to remove those rectifying bridges from the PD, to both increase efficiency and reduce cost and complexity in the PD.
I have taken a look at what could be done at the PSE side.
If you could consider that a 4-pair powered device (PD) would accept power a bit differently from the old PoE standard, by instead powering the pairs 3,6 and 4,5 with the negative voltage and the pairs 1,2 and 7,8 with the positive voltage, the polarity at the PD side wouldn't change regardless of straight cable, "Val's Crossover Cord" or "Goran's Crossover cord".
The PSE would then need to be able to swap polarity between two of the pairs (for instance between pairs 4,5 and 7,8) to maintain compatibility with the old PoE standards.
When typing this down, I came to think about that it could maybe be solved even more elegantly by supplying the pair 4,5 with the negative voltage and the pairs 1,2 and 7,8 with the positive voltage regardless, and after identification of the PD type, the PSE would either power the pair 3,6 with positive voltage for an old (2-wire) PoE PD or with negative voltage for a 4-wire PoE PD. The 4-wire PD would then need to be able to connect the pairs 3,6 and 4,5 together in the powered state. It would also need to be able to rectify the voltage on the pairs 3,6 and 4,5 during detection and negotiation (classification). This reduces the total number of needed transistors at the PSE and PD but adds a rectifier (that only need to support detection and negotiation currents) and a switch (transistor) to the PD. This PSE solution would also only support 2-wire powered PoE PD that accepts power on the data pairs.
There should be other possible solutions to be able to get rid of those rectifying bridges, I would be happy to see some more brainstorming about this.
Thanks Peter, all,
Let’s consider a PSE switch where those transceivers with Auto-MDI feature are present.
Would it automatically adapt the port voltage polarity as well? Or is it a data-only feature?
It might be telling that there are 10/100/1000 PHY transceivers that automatically adapt to crossover (Auto-MDI) with BOTH the "data" pairs (1,2/3,6) and the "spare" pairs (4,5/7,8) independently.
No virus found in this message.