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.

TXS0102: Peaks on SDA Line

Part Number: TXS0102
Other Parts Discussed in Thread: TCA9517A, LSF0102, TCA9517-Q1, PCA9306, TCA9517

Hello together,

we have changed our level translator for i2c bus to the TXS0102.

With the new transloter we got the problem, if connect two level shifter parallel there are some peaks on the SDA Line.

These peaks are at the same time as the high level on the SCL line. With an unfavorable constellation, the SDA line goes to low and then the CLK and thus a new start condition on the bus results and a bit shift occurs.

Furthermore I have removed the additional PullUp resister, but we got also the same problem.

If we connect only one level shifter to the pcb we have no peaks on the bus and the communication is working.

Has there ever been such a problem before?

And do you have any idea how we can solve this problem?

Thank you so much!

Best regards

Simon

  • I guess the noise coupled from the SCL line is so strong that it looks like the beginning of a rising edge, which triggers the TXS's edge accelerator.

    Shield the lines from each other, or use a passive level translator like the LSF0102, or use an I²C translator like the TCA9517A.

  • Hey Simon,

    Why are you connecting two translators in parallel?

    Can you show a schematic of how everything is connected?

  • Hey Emrys,

    yes, of course.

    There is the block diagram how we connected the level shifter.

    Because the Slave are on sperates PCBs connected with cable. Therefore we did on every PCB the level shifter, that we shift the level directly in the near of the slave.

    Best regards

    Simon

  • Why is this type recommended for the I2C bus?

    If we will get there some trouble on the bus.

  • The TXS works with correctly implemented I²C systems. But is designed to also support higher speeds, and has strict requirements on the signal quality.

  • Thanks Simon, that helps.

    My team's voltage translators are more general purpose -- TXS can be used for I2C, but it's not its sole purpose.  The TXS translators are pretty complex internally. We recommend to keep the load on them under 70 pF, which is probably difficult to do in your described application.  The more items you add in parallel, the higher your load will be. In addition, you're adding more TXS devices, which all have the chance of having issues with large capacitive loads. This can cause a lot of strange behaviors from the device.

    In this case, you might consider either using one of our I2C team's dedicated I2C devices, or a simpler translator like LSF0102.

    We do have some dedicated I2C parts in TI, but I don't directly support them. I can add our I2C team to the thread to see if they can make any suggestions for how to solve that issue specifically with their parts.

  • Simon,

    Have you considered using the TCA9517-Q1 for your application?

    It is important to note that this device must be set up such that the A-sides are connected (shown above). You cannot connect TCA9517-Q1 B-sides together due to the device's buffered output characteristic.

    Best,

    Danny

  • Hello Danny,

    thank you for your answer.

    What do you think about the PCA9306 level shifter? 

    Which of them would be better suited for our purpose? The PCA9306 or the TCA9517?

    Furthermore, I think that explain why we get some issues with the i2c signal.

    This explains our problem, but I think that 70 pF is very little capacitive part, because at 100 kHz 400 pF is assumed.

    Thank you!

    Best regards

    Simon

  • The PCA9306 is identical with the LSF0102, and does support 5 V signals on both sides. (Only Vref_A needs to have a lower voltage. Alternatively, connect EN to the lowest supply, and leave Vref_A and Vref_B open.) Also, the PCA9306 does not have any A/B side connection restrictions.

    The PCA9306 is a passive switch, which means that the devices at the ends must be able to drive the entire bus. The TCA9517(-Q1) does buffering.

  • Hello Clemens,

    thank you for your answer. 

    We know that the PCA9306 supports 5V as this is also stated in the data sheet. Furthermore, in our older version of the ciurcuit we used the PCA903 and the i2c bus worked.

    I think we will use the PCA9306 for the our application, because we have on the 3V3 side also a microcontroller which can drive the bus.

    Thank you for your support.

    Best regards

    Simon

  • Simon, Clemens,

    Great catch! When I posted my prior reply, I had a completely different device in my brain. I've removed that response to minimize any confusion I may have caused.

    The [actual] PCA9306 should fit in this socket, as you both noted. And if buffering is needed - though it sounds like it is not - then you can use the TCA9517-Q1.

    Best,

    Danny