I am using a TXB0104DR chip to interface a ZynQ (ZC702) FPGA with a National Instruments PCI-6259 digital I/O card. The FPGA is outputting a series of 2.5v square waves, through the TXB0104, to the digital I/O card’s 5.0v digital inputs. When I use an oscilloscope to monitor the TXB0104’s 5.0v output, the square waves look fine. But as soon as I connect the DIO card connector to my circuit the square waves pick up a lot of noise. One can barely see the underlying square waves without difficulty.
The FPGA is outputting a 6 bit count value that increases by 1 every 128us, effectively creating 6 square wave outputs at varying frequencies (LSB has a 256us period, second bit has a 512us period, …, MSB has an 8192us period). These 6 FPGA output pins are connected directly to the TXB0104 VCCA side pins, and the VCCB side pins are connected directly to the digital I/O card pins through a shielded connector (supplied by National Instruments and designed specifically for the 6259 card).
My breadboard contains 2 TXB0104 chips. The 4 least significant bits (4 fastest square waves) are going through one TXB0104, and the 2 highest significant bits (2 slowest square waves) are going through the second. There is also a request bit and acknowledge bit using up the other 2 voltage translators on the second TXB0104.
I also have 0.1uF capacitors on my breadboard, filtering both the VCCA and VCCB pins to ground.
My questions are:
Are the TXB0104 chips designed to connect directly to the FPGA pins and digital I/O card pins without any components (resistors) between them? I’ve tried using different combinations of pull-down resistors and capacitors, but this did not seem to make any difference.
Are the TXB0104 chips rated to handle the speed of my LSB (a square wave of 256us period)? Relatively speaking, I think this speed is slow and the chip should be able to handle the switching times involved. But I just wanted to ask the question for my own edification.