• State TI Thinks Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 5 answers
  • Subscribers 64 subscribers
  • Views 3116 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TPS23861 4 pair high power implementation

Charlie McAndrew
Prodigy 75 points
Other Parts Discussed in Thread: TPS23861, TPS23861EVM-612, TPS2378EVM-602, TPS2378, TPS2379

I am in the process of designing a PoE injector that supports 802.3af, at and also higher power over 4 pairs.  I am currently using a TPS23861 in auto mode with no I2C interface connected.  Everything appears to work fine with the af and at PDs we have tested.  I realize the there are several high power implementations commercially, but in the absence of 802.3bt there is no real standard.  We currently have a Hikvision DS-2DF8836IV-AELW 60W PTZ camera as the PD.  I am told that it works with a Suplet LAS60-57CN-RJ45, but have been unable to obtain enough detailed information to go straight to a working design.

I have tried using 2 ports connecting one pair to each, and 1 port connecting both pairs to it.  The former does not seem to work even on af and at PDs.  Looking at "VPORT" it seems that detection fails.  The latter runs all af and at PDs we have tested.  With the 60W camera it succeeds at detection, does a 2 event classification, and turns on, only to turn off again presumably because the current exceeds ICUT.

Do you by any chance have some insight into what I might be able to do to resolve this.  Many thanks for your help.

  • 0
    TI__Genius 15655 points
    Hi Charlie,



    I am glad to see you get all .af and .at PDs working fine with TPS23861. Regarding the high power port, I think your first approach is better. Please refer to TPS23861EVM-612(www.ti.com/.../tps23861evm-612 ) which has two high power ports. How to operate TPS23861 for high power application really depends on how the PD is designed. I am not sure about the design of the Hikvision PD.

    Currently we have two nonstandard approaches of PD for high power:

    TPS2378EVM-602(www.ti.com/.../tps2378evm-602): dual TPS2378 devices on the board, with this PD solution, TPS23861 is able to operate in Auto Mode and output 60W per RJ45 port.

    TPS2379EVM(www.ti.com/.../TPS2379EVM-106): single TPS2379 device on the board, need MCU to control TPS23961.

    I guess the Hikvision PD might have similar design with TPS2379EVM which has single PD on the board. If two PSE ports operate in Auto Mode, there will be some problems during detection since there's only one PD interface on the other side of the cable while there are two PSE ports doing detection. Under this condition, we only allows one port to do the detection, classification,turn on and manually force to turn on the second PSE port to double the power if a high power is needed on the PD side. So, host control is needed to configure TPS23861 to output high power.

    Since TPS23861 has the highest ICUT set to 920mA, it doesn't surprise me if all the current goes through one PSE port.



    Best regards,

    Penny
  • 0 in reply to Penny Xu
    Prodigy 75 points

    Hi Penny,

    Many thanks for your reply.  In the absence of detailed information I eventually bit the bullet last Friday and opened up the PD.  The unit belongs to my customer's customer, and costs thousands of dollars!  It uses a TPS2378 with Rdet=24.9K and Rcls=63.4ohm.  It then seems to rely on the PSE to power all 4 pairs which go thru bridges straight to the PD power supply.  I can understand why it was done this way, because it should be back-compatible with at and af.  Armed with this information I modified my PSE as follows:

    Mod Schematic

    The philosophy is to use only one PSE channel because there is only one PD controller.  Channel 2 was used because it was easier to implement on the board.  Initially detection occurs with Q4 and Q2 off.  Then, for classification, only Q4 is used.  The drop across D1 means that Q2 is not involved in classification.  Following this, at power-on, Q2 is also turned on.  The current is split approximately 50/50 between Q4 and Q2.  This effectively provides 30W per pair set, 60W total.  I tested it on the PD and it works like a champ!  Very solid.

    However, now my new problem .....  as a sanity check we connected a lower power 802.3at PD.  It used to work fine.  Now it goes thru detection and classification and powers up.  About 600mS later it powers down, and repeats the process.  This clearly not an Icut issue now.  The only thing that came to mind given the timing is that MPS was not working.  As far as I know the TPS23861 uses only DC MPS.  Theoretically there should be plenty of headroom for it to be recognized.  I measured the steady state voltages across the current sense resistors to confirm this.  While the unit was turning on and off, the dv/dt was too much to get a good DMM reading.  There was so much ground trash on my scope reading that even going differential would have given trash - trash = meaningless.

    Just to see what happened, I removed R40 above. The unit powered up and stayed powered up (!), although this is obviously not the final answer.

    I checked the Vgs vs Id characteristics of the IRLR120N and even though Vgs-on is about 11.5V vs 9V, the difference in Rds should be very small.

    Currently the only thing I can think of is that my Kelvin measurement wiring may by now not be the best in the world.  I will change this in the morning to see if it helps.

    I fully realize that this is a non-standard situation, but any input you may have will be greatly appreciated.  With continually evolving high power PoE and the continuing lack of any officially recognized standard, this could also help others who may end up in similar situations.

    Many thanks and best regards,     Charlie

  • 0 in reply to Charlie McAndrew
    TI__Genius 15655 points

    Hi Charlie,

    Thanks for your feedback which helps me to have a better understanding of your application. 

    Before any technical discussion related to your application, I want to give some information about the standard. Before the new IEEE802.3bt standard comes out(will be available in 2018), any high power solutions(>30W) are considered as non-standard. 

    If there's only one TPS2378 in the camera, there should be some external circuit to achieve 60W, since TPS2378 itself has 1A(TYP) current limit. On the other hand, one TPS23861 port can not provide 60W either. You have the right idea that parallel two ports together. However, I wouldn't suggest you to connect them together before the RJ45. Additionally, removing R40 would impact port 2's current sensing:the sensed current is twice of the real current.  

    Please refer to TPS23861EVM-612(www.ti.com/.../tps23861evm-612) which has 2 high power ports. Basically, it connects port 5 to pair 12 36 and connects port 6 to pair 45 78. So the total current could  split to 4 pairs of the Ethernet cable. On the input side of the PD, there should a diode bridge, so the PD doesn't care which pair the current comes from. 

    During the start up, you can set first port in Semi-Auto mode and second port in manual mode.  Only enable first port's detection&classification. If everything is good, you can power on the first port and then turn on  the second port without detection and classification.

    Hope this helps. 

    Best regards,

    Penny

    Best regards,

    Penny

  • 0 in reply to Penny Xu
    Prodigy 75 points

    Hi Penny,

    Sorry, it has been a long time since we communicated.  I am pleased to report that we have succeeded to implement the TPS23861 for both the camera were were previously testing, for af/at PDs and for UPOE PDs.  We are putting this unit into production.  There are, however, two looming problems about which I would highly value your opinion.  Both of these would require running the TPS23861 in manual mode.

    Problem #1 is LTPOE.  I think this is a small obstacle.  We should be able to access the A/D at detection and decide for ourselves if the detection resistor is 25K or 12.5K, and take appropriate action.

    Problem #2 is a little more esoteric.  I have been reading a white paper from Microsemi concerning a possible multi-finger classification system being proposed as part of 802.3bt.  It seems to also be referred to as POH (Power over HD baseT).  I am think that it may be possible to implement this in manual mode by repeatedly performing classification as necessary, but this raises issues including timing, values obtained during classification, and I'm sure many others.

    Our goal is to be able to support all these implementations as they may appear in the future, by means of downloadable firmware upgrades to existing hardware.

    Your input on this would be most valued.  Many thanks and best regards,     Charlie McAndrew

  • 0 in reply to Charlie McAndrew
    TI__Genius 15655 points

    Hi Charlie,

    It's really great news that you got the TPS23861 working with your system. Please see my comments below:

    1. Can you confirm if my understanding is correct: you want to turn on TPS23861 port while a nonstandard PD is connected which has a 25K detection resistor? If so, yes, you can check the detected resistance register(recommend to access the LSByte and MSByte at the same time to avoid any mismatch) to make the decision.

    2. TPS23861 can't support multi-finger(>3) classification at this point but we are developing .bt PSE now which will support all IEEE802.3bt features. Please leave your email address and I will contact you offline to give you a clearer picture of the differences between the existing proprietary high power solutions(PoE++,HDbaseT) and the upcoming IEEE802.3bt standard.

    Thanks.

    Best regards,

    Penny

  • 0 in reply to Penny Xu
    Prodigy 70 points

    Hello Penny,

    Been using TPS23861 in two different scenarios. One scenario, it's 4 pair, both Type A and B in the same time, in parallel. Hence using only a single channel on TPS23861. And it's being pushed to the max 0.92A.
    Now, for the new generation of that product, I would like to push it to 1.2A per single TPS23861 port. Is it possible to achieve that, maybe by cheating a bit with the sense resistor? E.g. cut the sense resistor in half (0.125R), and set it at 0.6A Icut.
    If not, do you have some new device coming which will allow 1.2A per port (trying to save on the MOSFET, sense resistor, fuse,....)?

    Regards,
    MTD

  • 0 in reply to MTD
    TI__Genius 15655 points

    Hi MTD,

    I don't recommend you to change the sense resistor since the classification current measurement is also based on the sense resistor. If you change the sense resistor, you will not get correct classification results. 

    I am wondering why you connect one TPS23861 channel to both Alt A and Alt B pairs. Have you considered connecting a second TPS23861 channel to AltB pair, and then you can get 0.92A*2 current capability on that RJ45 port. Please take a look at TPS23861EVM-612. We have two 4 pair ports(UPoE) on the board which are connected as described above. Please let us know if that works for your system.

    We do have a new part under development which will be .bt compliant and has a higher current limit. But 1.2A per channel will be still non compliant applications. Usually, more than 30W they power needs to be delivered through 4 pair and 2 channels. Thanks.

    Best regards,

    Penny