• Resolved

TXB0108: Avoiding oscillations and ringing when buffering TXB0108 signals for off-board GPIO

Prodigy 20 points

Replies: 2

Views: 45

Part Number: TXB0108

Hi,  

I’m working with a 3rd party device which contains a TXB0108 bidirectional voltage level translator. It is used, likely inappropriately, to drive GPIO signals to components external to the 3rd party device and the outputs of the TXB0108 are directly connected to a 40 pin GPIO header. It seems like a non-ideal choice as data sheet section 8.3.3 (Output Load Considerations) specifically states that PCB traces need to be kept short, and a small cable can easily violate that. Still, we are stuck with it.

We are seeing lots of cases of signal instability and oscillation which we have traced back to reflected signals returning to the TXB0108 after > 10 ns propagation delay. This is constant with the warning in data sheet section 11.1 (Layout Guidelines.). Doing things as simple as probing with a logic analyzer can send the TXB0108 into oscillation. Since we can not change the 3rd party part that uses the TXB0108, our solution is to make a daughter board that fits as close to the 3rd party board connector as possible and to place a Schmitt-Trigger buffer (SN74HCS125-Q1) as close to the connector on our board as possible.  Traces are ~ 250 mm long, so the reflection should be within 2 ns of signal start and w/in the one-shot activation period.  Initial testing seems to work, but our limited probing ~ 50 mm from the TXB0108 is still showing ~30% ringing back at the TXB0108 (not sure how much of this is contributed by our test equipment) and after all the experiences we've been having I'm paranoid and want to make sure I get this right.  My questions are as follows please:

- Do you recommend any additional termination elements be used prior to the receiving buffer? With the limited drive strength it will not be possible to use parallel termination. I’m not sure I can get close enough to the TXB0108 to use series termination at the source. AC termination before the receiving buffer might work and if you concur I would appreciate guidance on capacitor and resistor sizing if this has been explored already. My first thought would be 25 pF and 300 kOhm. 

- Are there any buffers known to have already been tested with the TXB0108?  If so is there a reference design please?

- Is there another buffer you would recommend instead of the SN74HCS125-Q1?  We do not need to preserve the bi-directional functionality and it is fine if a line is permanently configured as an output.

Thanks in advance for your help

  • The most important measure is to minimize the trace length. You already did this.

    See AN-903 for how to design AC termination. This load might be too high for the TXB; no termination might be the best option.

    All CMOS input have high impedance; the logic family does not matter.

  • In reply to Clemens Ladisch:

    Thanks.  We'll give it a shot.