SN74TVC3306: Failing outputs

Part Number: SN74TVC3306

I am using the SN74TVC3306 part in a design for the past few years and all of a sudden I am seeing field failures where the pins no longer go to the rails. The part is used as a simple I2C buffer from a AM335x CPU to a LM75AD temperature sensing IC. I have the correct pullups and caps. I am trying to understand if TI has seen this issue on this part and what suggestions they might have for me. I saw another post back in 2015 talking about the same issue. Again, the parts do not fail initially but some do overtime. Not all tho fail. At least not so far...

12 Replies

  • Jeffery,

    We can try a couple of things to assist you.

    1) Does replacing the part solve the issue?

    2) Does the part removed from the board and placed on a working board cause a failure?

    3) Do you have any scope shots of the signals at the GATE, A1, B1, and B2 of a working and not working system?

    4) Do you have any schematics we can review?

    To initiate a customer return there is a customer return information form you can fill out following the Guidelines for Returns.  You can submit the form through the channel you purchased the devices, local TI sales or TI authorized distributor.   ( I believe this process will provide a report if there have been any other returns similar to your issue)

    Thank you,

    Adam  

  • In reply to Adam Torma:

    Adam;

    Thank you for the response. Answers to your questions are

    1. Yes every time

    2. Did not try that yet. I have several coming in from the field and I will do that next

    3. I have scope shots of a working board. I will get some from the parts coming in from the field

    4. Yes I can share the part of the schematic with the this part for sure.

    Attached is a partial schematic for you to review. I will get the scope shots next. Thanks, Jeff

  • In reply to Jeffrey Jacobson:

    Jeff,

    There is nothing that stands out immediately in the schematic. We are curious to see the voltages on those nodes when the controller is high and low to see how much current is flowing through the switch. It is possible that your pull ups are too weak and need more current in the system to drive the I2C communication.

    Adam
  • In reply to Adam Torma:

    Adam

    Are you suggesting that too weak of pullups could damage the part? 

    On other question is on the part, how do you determine that actual date code? Specifically how do you decode it?

    Thanks, Jeff

  • In reply to Jeffrey Jacobson:

    Jeff,

    The only clue you have provided of the failure mode is that the voltage is not raising to the desired level which is why I was curious if the pull ups were strong enough.  Without any scope captures it is hard to tell what you are seeing as "failure".  If you think the part is defective I would recommend submitting a return claim through the links provided in the first response.  This will initiate a process through our quality engineering team. 

    Does the part you remove from a failing board cause a good board to fail? 

    For the date code which package are you using?  Do you have a picture of the label on the box or a picture of the IC? 

    Thank you,

    Adam

  • In reply to Adam Torma:

    Below is a scope shot of pins 5 & 6 of your part. Channel 1 on my scope is the SCL line and it goes from 0v to 5v. Channel 2 on my scope is the data line and you can see only goes up to just over 2v. It is not always the same pin that looks loaded down. Some times its both. Also, the same goes for pins 3 & 4

    Yes I took a bad part and placed it on a known good board and it fails. I have also replaced the bad parts with new parts and the boards work and continue to work.

    I am using this part: 

    TI SN74TVC3306DCUR psop8_m50_w320x210

    I have seen now 12 failures of this part and am concerned of a higher fallout coming. None of the pins on this part are exposed to the human hand so it is not ESD. These parts are in the field for months and then fail.

    I will try to get some parts into your quality group but I know that will take weeks...

    Thanks for your help.

    Regards, Jeff

  • In reply to Jeffrey Jacobson:

    Jeff,

    Here is some information on how to decode the topside marking.  The two things to look at are the last character of the 4 digit alpha numeric code and dashes above and below the alphanumeric code.  The last character will tell you which factory it came from and for a DCU package it should be P or S.  The dashes above or below the alphanumeric code tell the date the material was made.  You can use this information to see if you can sort the failing material to a certain factory or date. 

    These dashes are hard to see some times but if you use a light and look at the package surface at a 45 degree angle it can help. 

    The scope shots you show are not what I was expecting.  I was expecting some distorted edges but it looks like you have very clean rising edges but it is clamped at 2V.  What are the voltages on pins 2, 7, and 8?  Are they able to provide enough current to drive up to the 3.3 V and 5V  you need on the failing board?  

    Adam

  • In reply to Adam Torma:

    Adam;

    Thanks for that info on the date code. With 4 bits, how would you represent a date of 2003 or 2015?

    Once the part fails, it is done and needs to be replaced. I have over 7000 units deployed so I know the design is solid. I am concerned of a lot code issue. Has TI seen any issues with this part? I was looking for errata but found none also. 

    Let me know. Thanks again for your help!

    Jeff 12/12

  • In reply to Jeffrey Jacobson:

    Jeff,

    The date code represents the last digit of the year. For example 2003 = 3 = 0011 and 2015 = 5 = 0101. The storage life of our products is 5 years so you shouldn't be getting new material that is more than a decade old.

    I think submitting these parts to our quality team is going to be the best way to go to find out what is going on with these units through failure analysis.

    Do you have a picture of the top side marking of a failing unit or a picture of the label on the box that would have the wafer lot information?

    Thank you,
    Adam