TXS0108E-Q1: Part failure: TXS0108EQPWRQ1 failure

Part Number: TXS0108E-Q1
Other Parts Discussed in Thread: TXU0204-Q1, , TXS0108E

Tool/software:

I use the TXS0108EQPWRQ1 in our product for the CAN Transceiver - Microcontroller communication line to act as a level shifter from 3.3V to 5V. The device is a telematics unit designed for Trucks and Buses.

We have received some devices back from the field due to failures in CAN communication, and during our troubleshooting, we discovered that the TXS0108EQPWRQ1 IC was defective, with slight heat being generated on the component.

Simply replacing the TXS0108EQPWRQ1 IC in the affected device resolved the problem. What could be causing this issue, and what steps can I take to prevent it?

Following is the circuit used 

                        

Adding some voltage graph comparison for more insights. 

Yellow: TX

Blue: RX

Port-B 5V Graph of CAN1 TX & RX line with NG IC.

Port-B 5V Graph of CAN1 TX & RX line with Okay IC.

Port-A 3.3V Graph of CAN1 TX & RX line with NG IC.

Port-A 3.3V Graph of CAN1 TX & RX line with Okay IC.

  • The TXS is a passive switch with edge accelerators. Noise or ringing on the data lines can falsely trigger the edge accelerators and cause oscillations. Long traces/cables increase the risk.

    For unidirectional signals, a more robust solution would be a unidirectional translator like the TXU0204-Q1.

  • Thank you for your response.

    Here are the trace lengths in our design from the level shifter ports to the ICs:

    Tx trace length from MCU to TXS port B: 29.5mm
    Rx trace length from MCU to TXS port B: 24.38mm
    Tx trace length from CAN to TXS port A: 4.7mm
    Rx trace length from CAN to TXS port B: 2.18mm
    Could you please comment on whether these trace lengths fall within the recommended range?

    Additionally, could you suggest any unidirectional drop-in replacements for the TXS0108E-Q1?

  • These lengths look OK.

    There is no P2P replacement.

    The schematic above is incomplete; are there any other components connected to these eight lines?

    The waveforms just show that there is oscillation. Please zoom in on some actual data transmission to see where the oscillation starts.

  • In the field devices, only 2 lines are connected for CAN 1.
    However, there is provision for 2 more lines, which are for CAN 2. 

    This is its schematic:  

     

    which is similar to CAN 1. And here is the zoomed version of the graph that was shared.

    Port-B 5V Graph of CAN1 TX & RX line with NG IC.

    Port-A 3.3V Graph of CAN1 TX & RX line with NG IC.

    Port-A 3.3V Graph of CAN1 TX & RX line with Okay IC.

  • Sorry, the sample rate is too slow to see the actual waveforms. Please use the largest sample rate that your oscilloscope can do.

  • I am adding the oscilloscope graphs taken with the damaged IC, sample rate from 1 second to 5 nanoseconds.

    Port-B 5V Graph of CAN1 TX & RX line with NG IC.

    1sec.

    10us.

    5ns.(This is our oscilloscope largest sample rate)

    Port-A 3.3V Graph of CAN1 TX & RX line with NG IC.

    1sec.

    10us.

    5ns.

  • Hello,

    The team is out on US holidays today and responses will be delayed. Thank you for your patience.

    Regards,

    Jack

  • All images show the same sample rate (100kSa/s). To change the sample rate, you have to adjust the horizontal scale while acquisition is running.

  • Our DSO has a highest sample rate of 2GSa/s. I am attaching the graphs for both OK IC and NG IC within this range.

    Yellow: TX

    Blue: RX

    Port-B 5V Graph of CAN1 TX & RX line with NG IC.

    Time Division: 20nS with sample rate 2GSa/s

    Port-B 5V Graph of CAN1 TX & RX line with OK IC.

    Time Division: 20nS with sample rate 2GSa/s

    Port-B 5V Graph of CAN1 TX & RX line with NG IC.

    Time Division: 5nS with sample rate 2GSa/s

    Port-B 5V Graph of CAN1 TX & RX line with OK IC.

    Time Division: 5nS with sample rate 2GSa/s

    Port-A 3.3V Graph of CAN1 TX & RX line with NG IC.

    Time Division: 20nS with sample rate 2GSa/s

    Port-A 3.3V Graph of CAN1 TX & RX line with OK IC.

    Time Division: 20nS with sample rate 2GSa/s

    Port-A 3.3V Graph of CAN1 TX & RX line with NG IC.

    Time Division: 5nS with sample rate 2GSa/s

    Port-A 3.3V Graph of CAN1 TX & RX line with OK IC.

    Time Division: 5nS with sample rate 2GSa/s

  • These are regular oscillations, which are typically caused by capacitances or inductances on the signal lines.

    Are there any components on the signal lines between the TXS and the MCU?

  • No, the MCU and TXS are directly connected; there are no components between them.

                        

  • Hello,

    Can you confirm if this is seen only on one board or multiple? It could be a possibility that this singular device was damaged via short or improper solder. Just want to rule out this possibility here before we analyze further. I did not see any immediate issues with the schematic of the TXS0108E.

    Regards,

    Jack

  • Hi Jack Guan, 

    We are encountering this problem on multiple devices, particularly those being used in the field, which are experiencing failures and are being sent back to us due to this issue.

  • Hi,

    The attached waveforms seem to indicate noise and not actual translation operation between the NG and OK devices (the waveform characteristics seem to indicate crosstalk). Do we have additional waveforms during level shifting to compare as well? 

    I found it strange that the RC time on this is very slow. Can we check if the inputs follow the input transition rate requirements per the datasheet? 



    Regards,

    Jack 

  • CAN Data Transition and Transmission of OK IC

    I am attaching a graph for your understanding. The lines in the graph represent:

    Yellow: Tx of Port A (3V side)
    Blue: Tx of Port B (5V side)
    Green: Rx of Port A (3V side)
    Red: Rx of Port B (5V side)


    I hope you find this graph easy to understand.

    And here, I am also attaching a graph of the rise and fall timings of waveform which is 340nS.

  • Hi,

    Thank you for providing the waveforms of the functioning IC. How do these compare against the NG units and what is different between the two boards (NG vs. OK ICs)? For fixed direction translation applications, it is generally recommended to use fixed direction translators. In your case, the TXU0204-Q1 would be a good fit (also features schmitt-trigger inputs so less susceptible to noise). 

    Regards,

    Jack