Featured Articles
Announcing Pi
Technical Specifications
Raspberry Pi FAQs
Win a Pi
Raspberry Pi Wishlist
Actions
Forum Thread Details
  • Replies 5 replies
  • Subscribers 665 subscribers
  • Views 1116 views
  • Users 0 members are here
Related

Expansion board level shifter -  TXS0180E - how do we put it in push-pull mode?

romilly

Some of you may know that my company (Quick2Wire)  is working on a number of add-on boards for the Raspberry Pi.

 

We've hit an issue with the main interface board, and I'm hoping one of the engineers here can help.

 

We'd like to use a TXS0180E to level-shift the eight GPIO pins of an ARM SoC. We need stronger outputs than the TXB0108PWR..TXB0108PWR.. provides. The data rate of the TSX0108E is fine in push-pull mode but would be too slow in Open-collector mode.

I'm confused, though, about what determines the mode in which the chip operates. How do we control whether it's in Open-collector or push-pull mode? Is it just a matter of whether we have pull-ups attached to the lines?

 

It's not a show-stopper, as we know that we can use a TXB0108PWR..TXB0108PWR.. with a ULN2803, but that adds to the cost. I'd like to avoid that option if we can.

 

Thanks,  Romilly

Parents

  • Hello Romilly,

     

    You need to study that diagram on page 10 of the data sheet some more. image

     

    The TXB0108PWR..TXB0108PWR.. does not have OC ouputs - it has high impedance (or in TI speak - weak) outputs which are assisted during switching transitions by pull up or down fets. It's a neat way of getting bidirectional buffering and speed. It is inevitable that a bidriectional buffer with no direction control pins must have weak drivers so that the direction can be forced.

     

    If this is for a general purpose RPi board I suggest that it is a bad choice - this chip is very difficult to debug with a DMM because it is fundamentally dynamic in operation. I would expect RPi add ons to be easy to debug with simple test gear.

     

    To buffer the RPi pins for output you can use ULN series parts in DIL sockets for ease of replacement (and they are dead cheap if a bit slow).

    For input buffering use series R and clamp diodes. Now you can replicate the bidirectional level shifting virtues of the TXB108 with high power output and RPi input protection with cheap bits debuggable by volt meter.

    (The price you pay is more parts, the buffered ins and outs are on separate pins (possibly not a bad thing) and quite slow.)

     

     

    Michael Kellett

  • in reply to michaelkellett

    Apologies to all: the note should refer to TXS0108E, not TXS0180E. For someone who spent much of his life as a sofware developer, I type like a Platypus.

     

    I know the TXB0108PWR..TXB0108PWR.. does not have OC ouputs. The weak outputs from the TXB0108PWR..TXB0108PWR.. will not be enough for some purposes, hence the idea of using the TXS0108E. I was not sure if a lack of pull-ups would be enough to ensure that the TXS0108E would be in push-pull mode. I'm now clear about that.

     

    But...

     

    Having thought about it some more, I've been over-engineering. For the GPIO pins we can get by with diode clamps and a ULN2803. As you point out, Michael, that means that we can use easily testable, inexpensive though-hole components. That's a much better solution that either of the ones we were considering.

     

    Thanks to everyone - your ideas and comments have led us to a cheaper, better solution, which is what Element14 is all about.

     

    Best, Romilly

Reply
  • in reply to michaelkellett

    Apologies to all: the note should refer to TXS0108E, not TXS0180E. For someone who spent much of his life as a sofware developer, I type like a Platypus.

     

    I know the TXB0108PWR..TXB0108PWR.. does not have OC ouputs. The weak outputs from the TXB0108PWR..TXB0108PWR.. will not be enough for some purposes, hence the idea of using the TXS0108E. I was not sure if a lack of pull-ups would be enough to ensure that the TXS0108E would be in push-pull mode. I'm now clear about that.

     

    But...

     

    Having thought about it some more, I've been over-engineering. For the GPIO pins we can get by with diode clamps and a ULN2803. As you point out, Michael, that means that we can use easily testable, inexpensive though-hole components. That's a much better solution that either of the ones we were considering.

     

    Thanks to everyone - your ideas and comments have led us to a cheaper, better solution, which is what Element14 is all about.

     

    Best, Romilly

Children