TXS0108E: TXS0108E

The schematic is incomplete; it does not show anything connected to the TXS.

1. TXS outputs have a drive strength of 20 µA. Apparently, your load is higher.

2. What are the voltages? What are the loads on both sides of the TXS?

Hi Rajan,

Do you have any waveforms showing the anomaly? Is VCCB tested and biased to 3.3V? What is the state of the IOs when nothing is driving the line? (It should be 3.3V, as the device has internal pullups at the outputs to VCC).

Please also note that external pullups/ pulldowns are not recommended as the device has internal pullups. 

Regards,

Jack

Clemens,

I have set VCCA = 1.8V and VCCB = 3.3V, but the B-side outputs have a magnitude of 2.5V - I was expecting 3.3V!

the 1V8 side is connected to an FPGA and I have confirmed the expected logic states (and voltages) on the A side.

the 3V3 side is connected to the static control inputs of either a RF switch or a digitally-controlled bandpass filter, both of which have a min VIH of 2V. I am not measuring 3V3 and in some cases, the opposite logic state to the corresponding 1v8 input is being measured, which is clearly wrong.

Not sure how to proceed, need to fix this ...

Jack,

I have set VCCA = 1.8V and VCCB = 3.3V, but the B-side outputs have a magnitude of 2.5V - I was expecting 3.3V! The 1v8 and 3V3 supplies have been confirmed and measured at the power pins of the TXS0108E. I am NOT using any external pull-ups/downs as shown by my schematic above!

the 1V8 side is connected to an FPGA and I have confirmed the expected logic states (and voltages) on the A side.

the 3V3 side is connected to the static control inputs of either a RF switch or a digitally-controlled bandpass filter, both of which have a min VIH of 2V. I am not measuring 3V3 and in some cases, the opposite logic state to the corresponding 1v8 input is being measured, which is clearly wrong.

The waveforms will be static d.c. levels and I have used a multi-meter to confirm the logic level (and voltage)

Not sure how to proceed, need to fix this urgently!

Jack,

If I drive the A (FPGA) side of the TXS0108E to high impedance 'Z', I observe the same behaviour on the B-side as described above, i.e. either 2.5V max or in some cases, either 0.5 or 1.9V. 

What should be the expected behaviour if the A-side inputs are high impedance?

Thank you,

Rajan

The TXS is a passive switch; it cannot pull the line low. The line is weakly pulled high by the internal pull-up resistors, so if you have a voltage lower than 3.3 V, then some other device must be pulling low. According to your measurements, the FPGA outputs 1.8 V, so it must be somewhere on the 3.3 V side.

Please show everything that is connected to the 3.3 V I/Os.

I have one bus called, ch-select, that receives three 1v8 signals from the fpga and the intent is to generate three equivalent 3v3 signals to connect to the select inputs of two AMUXs, 74hc4051 as well as the select inputs of two RF switches, i.e. a fanout of 4 devices in total as illustrated below, U6 and U7 are the the TXS0108Es 

I also have one bus called, filter_select, that receives ten 1v8 signals from the fpga and the intent is to generate ten equivalent 3v3 signals to connect to the select inputs of bandpass filters, i.e. a fanout of 6 devices as illustrated below:

What do you think?

is there a higher output drive, pin-to-pin compatable replacement for the TXS0108E?

ch_select(2:0) = 1.8V on fpga side, ch_select(2:0) = 2v5 on B side

when filter_select(9:0) = 'ZZZZZZZZZZ' on fpga (A) side, i.e. high-impedance, on the B-side:

 filter_select(9) = 2V

 filter_select(8) = 2V

 filter_select(7) = 2V

 filter_select(6) = 1.9V

filter_select(5) = 0V / 91 mV

filter_select(4) = 0V / 91 mV

filter_select(3) = 1.89V

filter_select(2) = 1.89V

filter_select(1) = 0.88V

filter_select(0) = 0.44V

I was not expecting the above values on the B-side!

In theory, all these inputs connected to the 3.3 V side should have a high impedance. Anyway, please show the schematic.

Auto-bidirectional translators cannot have drive strength. Why are you using the TXS when you do not have bidirectional signals? The TXU0104 would be easier to use and more robust.

ok, I should have used the TXU0104 as my use case is unidirectional, however, the my existing circuit should still function!

If the A-side inputs are high impedance (Z), should I expect 3.3V on the B-side? As I measured above, this is not the case!

No, I mean the control inputs of the devices connected to the B-side I/Os have a high impedance, and therefore should be 3.3 V.

Please show the schematic.

pdfjoiner (4).pdf

I attach the schematic, the first page shows the TSX0108E devices connected to fpga bank 2 and the other sheets show the two busses, CH_SELECT(2:0) and FILTER_SELECT(9:0) connected to the d.c. digital select inputs of an RF switch and/or filters as illustrated above in the layout diagrams

What should be the expected behaviour if the A-side inputs are high impedance (Z), should I expect 3.3V on the B-side? As I measured above, this is not the case!

Clemens,

I have made some progress and when the fpga outputs logic low, I now correctly measure logic low, i.e. 25 mV at the output from the TXU0108: there was an error in my fpga assignment of the lower two bits of filter_select

There is still an issue when I output logic high as listed below:

ch_select(2:0) = 1.8V on fpga side, ch_select(2:0) = 2v5 on B side

when filter_select(9:0) = '1111110000' on fpga (A) side, i.e. high-impedance, on the B-side:

 filter_select(9) = 2V

 filter_select(8) = 2V

 filter_select(7) = 2V

 filter_select(6) = 1.9V

filter_select(5) = 0.7V

filter_select(4) = 0.7V

filter_select(3) = 25mV

filter_select(2) = 25mV

filter_select(1) = 25mV

filter_select(0) = 25mV

filter_select(9) to filter_select(6) are wrong!

For a high output, the internal pull-up resistor has a value of about 4 kΩ. So the ch_select lines sink 200 µA, the filter_select(9…6) lines sink 325 µA, and filter_select(5…4) sink 650 µA.

Are there any pull-down resistor anywhere? Do the RF devices have internal pull-downs? The schematic does not show device names. What exactly are those RF switches and bandpass filters?

The 3V3 outputs from the TSX0108E connect to the select inputs of a multi-channel RF switch and to the bandwidth select inputs of the filters: both the RF switch and filters have a VIH (min) of 2V.

For the RF switch and AMUX, the TSX0108E is currently outputting 2.5V, so I am compliant, but for the bandpass filters, it's very uncertain!

I don't know if the select inputs of the RF switch or bandpass filters contain internal pull-downs, but have asked the question: I attach their datasheets!

how did you calculate the above currents?

(3.3 V − 2.5 V) / 4 kΩ

I have just been informed:

"All three of these parts use the same process, so they have the same digital circuitry. There is ~45kohm pulldown which would require less than 200 uA of drive current required at logic high."

Comments .....

that is not a viable solution, one thing I considered was lifting the 3v3 power pin from its pad on the pcb and soldering this pin directly to a 5V supply, but this might only work for the RF switch, but the filters are powered using 3V3!

Why does ch-select(2:0) output 2V5, is it because its fan-out is 4 devices? 2.5V and 0V are just about fine! This means I can still select between different RF channels, however, I need some way to select different passbands. At the moment, I can use the bandwidth setting when filter-select(9:0) = 0000000000.

Clemens, you have been very helpful, but we need to find a solution for the filter-select(9:0): the board is complex, uses many TI parts and we simply can't just redesign, we have to get the existing hardware working as we have to share results with our customer ...

the fpga side of the TSX0108E has the expected, correct amplitude of 1.8V: why have you recommended external pull-up resistors on the A-side?

Compared to filter-select(9:0) with a fan-out of 6 devices and logic high voltage levels around 2V at the B-side of the TXU0108, why does ch-select(2:0) output a slightly higher voltage of 2V5? Is it because its fan-out is 4 devices? 2.5V and 0V are just about fine and means I should still be able to select between different RF channels, however, I need some way to select different passbands. At the moment, I can use the bandwidth setting when filter-select(9:0) = 0000000000.

Clemens,

I am confused by your last response, are you suggesting I can add external pull-ups to the B-side to raise the logic high potential from 2 to 3.3V? If so, what value of resistor?

It might be easier to lift the B-side 3V3 power pin and connect this to 5V, but I need to check with the filter provider.

Compared to filter-select(9:0) with a fan-out of 6 devices and logic high voltage levels around 2V at the B-side of the TXU0108, why does ch-select(2:0) output a slightly higher voltage of 2V5? Is it because its fan-out is 4 devices?

Thank you,

Rajan.