TXB0108: Driving TXB0108 Outputs with a Spartan-6 FPGA

Hi Ian,

The TXB is a weak driver as compared to the LVC device which can provide up to 32mA of drive strength.
Why is the input signal for the TXB degraded?
If the output of the TXB has large cap/resistive loading, then the TXB would not be able to support it.
furthermore, since you already mentioned that the steady high or low input at the TXB results in the correct output, it points to me that the signal frequency may not be supported at this 2.5Mhz frequency.
If you could reduce the frequency of the signal and measure the output, it should get better.
What is the probe setting? Is it high impedance or 50ohm? What is the capacitance contribution from the probe?

Hi Shreyas,

What is the probe setting? Is it high impedance or 50ohm? What is the capacitance contribution from the probe?

I had not considered this. The probe is indeed high impedance (1 Mohm, 11.5 pF from the oscilloscope + probe impedance). I have added a 50 ohm load to the output and measured the following:

With the output of the TXB0108 shown in yellow, and the SN74LVC4245A left in green for comparison. The duty cycle looks like but the amplitudes are obviously not.

Why is the input signal for the TXB degraded?

It should not be, since the input is coming directly from the FPGA in an identical fashion to the inputs of the SN74LVC4245A. Here is the measured input with 50 ohm load at the output:

The blue is the input to the TXB while the yellow is once again the output. Finally, removing the 50 ohm load for open (high impedance) load conditions:

The input, shown again in blue, exhibits reasonable signal integrity (it should be 3.3 V pk-pk) with an open output.

If the output of the TXB has large cap/resistive loading, then the TXB would not be able to support it.

This is possible, but currently there is only a load resistor and an alligator clip... and a 52 ohm resistor on the output line supplying the load.

furthermore, since you already mentioned that the steady high or low input at the TXB results in the correct output, it points to me that the signal frequency may not be supported at this 2.5Mhz frequency.

Looking at the datasheet for the TXB0108, it claims to support up to 100 Mbps. 2.5 Mhz shouldn't be an issue then, should it?

Thanks a lot for your time and the timely response!

Ian

Hi Ian,

As I had mentioned earlier, TXB is a weak driver. It cannot support more than 20uA of current drive as compared to 32mA for the LVC device.
In that sense, TXB definately cannot support the 50ohm setting on the scope. In this case, you are trying to ideally have 3.3V/ 50ohm = 66mA from the device. The TXB nor LVC can support up to 66mA per channel.

Do you also have an external pullup or a pulldown resistor? Please remove them, if not required. The effects of external pullup or pulldown resistors are mentioned in this app note below:
www.ti.com/.../scea054a.pdf

The datasheet mentions 100Mbps under the rated loading of 15pF and 1Mohm. I would imagine that the same R/ C loading is not currently present in your setup.
I would suggest you to use the SN74LVC8T245 or the SN74LVC4245a device as you probably require greater drive strength.

Hi Shreyas,

Thanks for your response. You are right about the driving current, and we only have a single pull-down on the output-enable pin as specified in the recommended configuration from the datasheet, and no pull-ups.

To summarize my confusion about the line/load impedance the TXB0108 has at its outputs:

-As was shown in the previous post, with an open output the signal integrity at the TXB0108's input is reasonable, but it will not output the correct wave, while it will output a steady high state (5 V DC). 

-The LVC is receiving an identical 2.5 MHz squarewave at its input, and successfully outputting a level-shifted and buffered version at 5 V.

-My oscilloscope has an input impedance of 1 M, 11.5 pF, and given that I am able to read the output of the LVC, it is reasonable to conclude that my leads are not adding too much parasitic effect to the system.

Given that the datasheet mentions 100 Mbps with a load of 15 pF and 1M, should I not then be able to connect my oscilloscope to the outputs of the TXB? That results in the original distorted waveform I shared... Is it possible there is reflection on the line? Or a parasitic reactance? In any case the LVC device is driving this fine, so it's almost certainly related to a different in chip spec.

Thanks again for the prompt response!

Ian

Ian,

Yes. TXB and LVC are from different family of translators. Their internal architecture is significantly different.
So the performance of these will be different.

Hi Shreyas,

Before I declare this issue resolved: given that the PCB output lines of the TXB and LVC are nearly identical (a 6 mil trace to a jumper), what might be the cause of this difference in performance be? Inductance on the line?

I assume it is something on the PCB since I cannot measure a good output signal with my high-impedance oscilloscope.

Thanks again,
Ian

Hi Ian,

If you have the high impedance probe line with 11pF cap load, it should be fine. I would suspect there could be other resistive loading on the line which may be causing this behavior.

I had measured the TXB0108 on our board and didnt see issues with it.

Hi Ian,

I also dont think reflections are causing it. With a pullup and pulldown resistor on the line, the voltage divider formed can create issues such as this.

Hi Shreyas,

I appreciate your correspondence as we seek to integrate this TI device into our commercial product...

As I mentioned, there are no pull-up or pull-down resistors. There is however a 50 ohm resistor in series at the end of the output trace before the jumper. Could that be the the additional load on the line? Here is a screenshot of the measured voltage at the end of the PCB trace, but before the 50 ohm surface mount resistor and jumper wire (shown in blue, with the desired wave again in green):

What makes you believe the issue is not related to reflections? Is this because this issue would be independent of the difference in current capacity between the two chips?

I assume it must be some AC phenomenon since a steady high state is driven properly by the TXB on one of its other outputs.

Thanks for your support,

Ian

Hi Ian,

Thank you for considering the TXB product into the system.
Since you have confirmed that the steady state waveform is working, it points to loading on the output.
Have you tried reducing the frequency to see at which point it starts to degrade?
I am not sure what could be going on without a schematic or looking into the board.

You can initiate an FA with the supplier so that they can test it or can send it to me for testing it here.