• State
  • Locked Locked
  • Replies 9 replies
  • Subscribers 18 subscribers
  • Views 6841 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.

I/O problem with TXB0108 level translator

Joannes Laveyne
Prodigy 130 points
Other Parts Discussed in Thread: TXB0108

I'm designing a circuit which employs a couple of TXB0108PWR voltage level translators to interface a 3.3V microcontroller to 5V peripherals. I followed the datasheet and the application note, but somehow I am experiencing some strange results.

When I apply a 5V DC test signal to the high side of the TXB IC, the low side correctly translates this to 3.3V. However when I remove the 5V signal both the high and low side remain in their high state (5V and 3.3V respectively). When I remove my measurement probe from the high side pin and reapply it, the 5V is gone and the low side goes to zero also. It looks like the pins of the TXB remain clamped until a (minimal) change in impedance (the measurement circuit has 5 Mohm impedance).

When applying a 3.3V DC signal to the low side of the TXB, the high side correctly translates this to 5V. However, when applying 0V again, the high side hoovers around a noisy 2.5V. This only happens when a 3.3V signal has previously been applied, just after power-up the high side is at 0V.

Included is the schematic of how I hooked up the TXB. I'm not using pull-up/down resistors, I'm not driving any (capacitive) loads beside my multimeter or oscilloscope, my inputs can source sufficient current and my trace length is not too long (not more then 2 inches). My power rails are clean and grounding is good.

Any help would be appreciated!

  • 0
    TI__Expert 6430 points

    Hi,Joannes

    Could you post some pictures to show input and output signal when this case happen. It will provide a lot help to analyze this case.

    Thanks.

    Lawrence

     

  • 0 in reply to Wei Lang
    Prodigy 130 points

    Of course, see below.

    Signal 1 is the output of the 3.3V microcontroller connected to the low side of the TXB. Signal 2 is the 5V high side of the TXB.

    In this scope image I let the pin of the microcontroller output a square wave with 1s period. The output of the microcontroller seems clamped by the TXB low side at around 2.3V, the high side output clamped at 2.4V.

    When I disconnect the probe connected to the low side, the high side output is offset by around 2.5V.

    At first sight it would look like a grounding problem, but the grounding is solid (I even added an extra flywire to the GND pin of the TXB). It also can't be caused by oscilloscope ground loops because the TXB and microcontroller are battery supplied.

    When configuring the microcontroller pin as input and applying a 5V to the high side of the TXB, the low side jumps to 1.6V. When removing the 5V (TXB high side floating), the high side goes to 2.5V, low side remains at 1.6V.

    Only when connecting the high side to GND do both lines go low.

    When disconnecting the probe at the low side of the TXB, the high side of the TXB correctly goes to 5V when applying a 5V, but remains at 5V until connecting it to ground again. There is no noise on the high side when the probe on the low side is disconnected.

    When removing the micro and just testing on the TXB, results are alike. When applying a 3.3V to the low side, the high side goes to 2.5V.

    When applying a 5V to the high side, the low side goes to a (noisy) 2V. When removing the 5V, the low side goes to a (clean) 3.3V and the high side remains at 5V. Only after connecting the high side to GND do both sides go low.

    Notice how most of the signals are quite noisy, while my power rails are crisp clean. It's like the TXB adds noise to its in/outputs, even when they're connected directly to the rails.

    Thanks for you assistance.

  • 0 in reply to Joannes Laveyne
    TI__Expert 6430 points

    Hi,

    Thank you for providing detailed tesing images. I have some questions and want to double check with you. 

    First, do all channels have same testing results?

    Could you check PCB board again to avoiding weld defect?

    There is a 4k  serial resistor on input and output signal(as described on page 10 in TXB0108 datasheet). Please check the equivalent resistor of probe and signal source if there is no any external resistors on PCB board.

    Thanks

    Lawrence

  • 0 in reply to Wei Lang
    Prodigy 130 points

    Thanks for your response.

    Yes, all channels behave the same way. I've also tested on a TXB0108PWR on another PCB with similar results.

    The PCB has no defects. I've checked for appropriate VCCA, VCCB and GND by measuring on the pins directly. I've even added flywires to these pins and connected them directly to the power rails just to be sure.

    My probe has multiple megaohms impedance. There are no series resistors on the board or any external resistors. When used as input, the pin is connected directly to the VCCA or VCCB power rail (depending if low or high side is used as input).

    My I/O lines are packed together quite tight, with only 0.5mm between traces. However, that should not be a problem for DC or low frequency signals I suppose.

    Thanks for you assistance.

  • 0 in reply to Joannes Laveyne
    TI__Expert 6430 points

    Could you provide some information about power-up sequence?

    Thanks

    Lawrence

  • 0 in reply to Wei Lang
    TI__Expert 6430 points

    Hi,

    Could you mearsue the resistance between OE and GND when power is zero.

    Thanks

    Lawrence

  • 0 in reply to Wei Lang
    Prodigy 130 points

    Hello

    Below I have a scope image of the 5V (signal 1) and 3.3V (signal 3) rails powering up. 2V/div

    An image with the supply rails and (unconnected) high side (signal 3) and low side (signal 4) powering up. Note: signal 1 and 4 are set to 5V/div.

    An image with the rails powering up while the low side of the TXB has been connected to the 3.3V rail.

    An image with the rails powering up while the high side of the TXB has been connected to the 5V rail.

    Now below I have an image with the rails powering up while the low side has been connected to the 3.3V rail, then connected to ground, then connected to 3.3V again and finally connected to ground once more. Behavior seems ok, except that I have to connect the low side to ground explicitly to get the high side to 0V. Just removing the 3.3V signal from the low side input has no change in the high side output.

    Below I did the same thing for the high side. Rails are powering up with the high side of the TXB connected to 5V, then connected to ground, then to 5V again, then to ground again.
    Notice the initially correct behavior, but then after grounding the input and reasserting it to 5V, noise appears on both rails and the 3.3V output (green signal 4) remains clamped at around 2V. When removing the 5V input, the high side (purple signal 3) hovers around 2.5V and only drops to 0V when grounded again.

    I measured the resistance between OE en GND. It's 80 kOhm, and 54kOhm when I switch the probes.

    Thanks for your assistance.

  • 0 in reply to Joannes Laveyne
    TI__Expert 6430 points

    Hi,

    First thanks for your providing information.

    The TXB series parts of directional auto sensing voltage-level translator have some special thing need to be care when using it.

    First, the external driver will be more 2 mA of current to drive TXB0108.Please check microcontroller output current.

    If driver current is not enough, it would be result in : (In this scope image I let the pin of the microcontroller output a square wave with 1s period. The output of the microcontroller seems clamped by the TXB low side at around 2.3V, the high side output clamped at 2.4V.) (In your comment on 9-19)

     Second, please make sure the input pin is not at floating state when you measured the input or output signal. The chip needs determined state at input (the note is on the page 4 at its datasheet). For the input floating, the output signal state is not well defined.

     It would be result in :(When I disconnect the probe connected to the low side, the high side output is offset by around 2.5V. When configuring the microcontroller pin as input and applying a 5V to the high side of the TXB, the low side jumps to 1.6V. When removing the 5V (TXB high side floating), the high side goes to 2.5V, low side remains at 1.6V.)(In your comment on 9-19)

    As for noise on signal, there are two possible reasons.

    First, there is requirement on input transition rise or fall rate. The maximum of its value is 40ns/V (on page 3 at its datasheet). Second, noise problem is also can be caused by input floating. Could you retest it when enhancing this parameter and giving determined input signal?

    If you redo the test, any update, please let us know.

    Thanks

    Lawrence

  • 0 in reply to Wei Lang
    Prodigy 130 points

    Hello,

    The microcontroller can deliver 25mA per pin, so this should be enough to drive the TXB IC.

    I've done some more testing, and I've noticed that one of the TXB IC's on my PCB works much better (almost perfect) than the other. I'm going to redesign my PCB a bit to rule out possible design flaws and do some additional testing. The only big difference between the two IC's is that the 'bad' one has a 100nF cap on the 3.3V line close to the OE pin.

    Thanks for your advice.