• State
  • Locked Locked
  • Replies 6 replies
  • Subscribers 63 subscribers
  • Views 376 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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: Random failures

Charlie McAndrew
Prodigy 75 points
Part Number: TPS23861
Other Parts Discussed in Thread: TPS23880,

We have been using the TPS23861 in our products for some years now with good results.  However we seem to be experiencing an increasing number of unexplained field failures.  The failed parts are unresponsive on I2C and there is no detection activity on any channel. They are not warm to the touch.  The vast majority of them have a characteristic mark, which can be seen in the somewhat fuzzy picture below.  There is a crater just below the P and the S in the TPS marking.  Most of our applications are outdoors, and some failures can be explained by lightning strikes, but this doesn't cover every case by any means.  When no other visible board damage has been found, replacing just the IC has restored normal operation in every case.  Does the location of this crater give any insight on what might have failed and why?  We are phasing in an updated board using the TPS23880 to add 802.3bt support, but we have an installed base using the 23861s.  Many thanks and best regards,     Charlie

  • 0
    TI__Mastermind 41760 points

    Hi Charlie,

    It's difficult to tell what exactly is getting damaged looking at the hold on the package. I've seen damage to the IC when the ABS Max ratings are exceeded. Do you have a schematic of the design for review? Thanks!

  • 0 in reply to Darwin Fernandez
    Prodigy 75 points

    Hi Darwin,

    Many thanks for your reply.  I'm attaching schematics.  This unit started out as a single board, but due to added functionality requirements the PoE function now spans 3 boards.  Schematic 01010 is the main board.  Plugged into it are a PoE board 01030 and a CPU board 01060.  PoE board J1 connects to main board J5.  It is +3.3V, /RESET and I2C.  PoE board J2 connects to main board J2.  The CPU board is actually a full-function Layer-3 managed network switch.  Currently it also manages the PoE, but this, too, is in the process of changing.  We are moving all the power functions down to the main board in a pending version.  From the PoE perspective, PoE power to the RJ45 connectors on the CPU board passes from J2 on the PoE board, through J2 on the main board to J4 on the main board, and then to J5 on the CPU board.  I2C comes from the CPU board via J107 to J3 on the main board, then to J5 on the main board and J1 on the PoE board.  The connectors are 0.1" pitch, 3A rated.  The boards are clamped together with screws and standoffs to ensure correct connector position and engagement.  Sorry this straightforward schematic is now so awkward to navigate.

    Many thanks and best regards,     Charlie01010SE.pdf01030SA.pdf01060SC.pdf

  • 0 in reply to Charlie McAndrew
    TI__Mastermind 41760 points

    Hi Charlie,

    What catches my eye is the shutoff circuitry (correct me if that's not what it is doing) around the FETs of the 861 controller. Can you check if there is spike with respect to GND on sense, drain, or gate pins of the IC when there is a fault (like a short) on the PSE port? Can you verify if it possibly goes over the abs max shown in the datasheet? Thanks!

  • 0 in reply to Darwin Fernandez
    Prodigy 75 points

    Hi Darwin.

    You are correct.  UL requested a secondary current limiting circuit in order to certify the product, so we added that.  We made some tests on Friday.  The port was powered at 60W on both pair sets.  We measured at Q6, which powers pair 1-2/3-6, and to overload we added a 50 ohm resistor across that pair to create an overload.  On the attached waveforms, Q6S is red, Q6D or Q6 G is blue.  Operation of the overload circuit is seen in the source voltage at the beginning, until the 23861 also sees the overload and turns off Q6.

    We also thought a disconnect could cause a transient, so with the same setup as above we unplugged the PD multiple times.  Waveforms also attached.  The doubling of the source voltage just before disconnect is probably due to pair 4-5/7-8 disconnecting first, resulting in double the current on 1-2/3-6.

    Sadly we don't seem to have found a smoking gun, and are still at a loss.

    Many thanks and best regards,     Charlie

  • 0 in reply to Charlie McAndrew
    TI__Mastermind 41760 points

    Hi Charlie, the transient will be pretty fast. I'm not sure the ms and s time scale will catch it. Is it possible to check on an oscope and zoomed in timescale? you can try triggering on abs max voltage values for these pins and see if the scope catches an overvoltage when the overload is applied. also can you try with a heavier load than 50 ohms? Since it's set to 60W, it is a slight overload but we want to hit the port hard with a high overload to force the PSE to protect itself.  thanks!

  • 0 in reply to Darwin Fernandez
    Prodigy 75 points

    Hi Darwin.  I take your point about possibly missing the transients with such a slow timebase.  However, I wonder if the limiting circuit would be the culprit.  First, in order for the optocoupler to turn off Q5, C8 has to be discharged by U5.  Also while this is happening, Q6 will still be firmly on.  I can see how the +54V might spike, and that the D5 (SMBL75CA) clamping voltage may be too high to protect the 23861.

    I have cut and pasted the schematics for 1 channel together on 1 page to make it easier to follow everything (attached).    Also I wonder if it might be wise to consider replacing the bidirectional parts with unidirectional parts, and lowering the clamp voltage from 75 to 58.PoE across all 3 boards.pdf

    I will continue tests and be back with results as soon as I can.  Many thanks,     Charlie