TXB0304: Push pull does not work with CMOS logic from 1.8V to 3.3V

Please see below for the schematic. It is used in two places on the board, having troubles in both cases.

Case #1: level shifter for I2S signals from Qualcomm Snapdragon to TI TAS5754M. There is a nice clock signal of 1.8V coming from the Snapdragon, but no signal on the output towards TAS5754M.

Case #2: 

Left-hand side goes to a connector with a short cable to the board that uses 3.3V CMOS logic. Right-hand side goes directly to the same Qualcomm Snapdragon pins.

In one setup, adding a 10 kΩ pull-down on B1 helped solve an issue where it varied the signal on A1 between 3.3V high level and 1.8V low level. After adding the pull-down, low level was 0V and signal was read correctly. Could not be reproduced on a second board with a different cable.

Dear Dylan,

Thank you for looking into this! I checked the PCB, I confirm the OE pin is in fact routed properly on the board and pull up to the 1.8V rail, I don't know what the no connect symbol is doing there.

For Case #2:

Clock signal 1.8V before level shifter (pin A4)

Clock signal 3.3V after level shifter (pin B4) (presence of cables on the connector does not affect the shark fin shape):

Incoming data signal 3.3V side (pin A2) (note that the low level is not 0V as expected but 1.8V)

To solve the latter, a 10 kΩ pull-down was added on the 3.3V board, this pulled down the logical low level to 0V and things work properly. This same fix does not work, however, when another cable is used.

Thanks.

Hi Martijn,

Thank you for the schematic and the scopeshots.

On the schematic, you have noted that the OE pin needs to have pullup resistor and capacitor 100nF. I would not suggest using the capacitor on the OE pin. this RC causes slow rising edge on OE pin which is not good for the CMOS device.

on the post, you have mentioned that the A2 signal is 3.3V, whereas the Vcca port supply is at 1.8V, why is that? 

I would not suggest using a pullup or pulldown on the TXB010x or the TXB030x devices, and not such a strong resistor especially. The reason is that there is internal series 1K resistor on the TXB device which forms a voltage divider at the output. Please refer to the app note regarding this 

finally, there is another requirement for the TXB030x device to have the driver strength of at least 3mA to sink or source current in order to change(overdrive) the signal levels, as mentioned in the datasheet. Does the Qualcomm chip provide this drive strength?

Dear ShreyasRao,

Thanks for helping out.

[SR]>> On the schematic, you have noted that the OE pin needs to have pull-up resistor and capacitor 100nF. I would not suggest using the capacitor on the OE pin. this RC causes slow rising edge on OE pin which is not good for the CMOS device.

Thanks for this suggestion, I will try this out.

[SR]>> on the post, you have mentioned that the A2 signal is 3.3V, whereas the Vcca port supply is at 1.8V, why is that? 

Sorry, mistake on my part, I meant B2. I checked this also on the PCB to be sure. So Vcca is 1.8V and A side is 1.8V, B side is 3.3V (in both cases).

[SR]>> I would not suggest using a pullup or pulldown on the TXB010x or the TXB030x devices, and not such a strong resistor especially. The reason is that there is internal series 1K resistor on the TXB device which forms a voltage divider at the output. Please refer to the app note regarding this 

Thanks for sharing the application note. This led me to thinking about the other end of the clock signal for case #2, which also has a 100Ω series resistor. I don't know how the designer of the board came up with that value, but could it be that this is causing issues as well, and what would be a better value?

[SR]>> finally, there is another requirement for the TXB030x device to have the driver strength of at least 3mA to sink or source current in order to change(overdrive) the signal levels, as mentioned in the datasheet. Does the Qualcomm chip provide this drive strength?

Yes it is configurable between 2 and 16 mA, and currently it is set to 12 mA.

[SR]>> The shark fin shape also indicates to me that the output loading is large. do you know the scope probe capacitive loading. Are you using a passive or active probe? 
an SMB probe can have cap loading of at least 100pF + the board traces + input loading at the scope will be additional loading on the device.

I am using a passive probe. 

[SR]>> If its possible, please measure the output of the TXB device at the pin disconnecting the TI TAS5754 device and the series 100ohm resistors as well.

I will try as soon as possible.

I have removed the capacitor on the OE pin: no change.

I also removed the pull-downs and disconnected the cable to the other board.

What you see is this as input to the level shifter:

And the output is this, without anything attached at the output, e.g. no pull-ups/pull-downs, capacitors or series resistors:

The voltage rails on 1.8V and 3.3V show a straight line, no glitches at all.

Drive strength of the Snapdragon is set to 12 mA.

We don't have an active probe to test with.

Will try next week with the SN74AVC4T774 that we just got in, hopefully that will work better.

I created a measurement setup with the TXB0304 EVM and a home-made SN74AVC4T774 EVM. No pull-ups/pull-downs are present. Both devices are connected in the same way to the Qualcomm Snapdragon development kit. The 3.3V output is unloaded in both cases, just measuring with the oscilloscope.

These are the results:

Step function with TXB0304 (yellow = 1.8V input, blue = 3.3V output):

Step function with SN74AVC4T774 (yellow = 1.8V input, blue = 3.3V output):