How to prevent short circuit between output pins

Question

I am trying to hack some signals coming from an old computer going to a Floppy drive using an ATtiny1614 microcontroller. Normally the computer has full control of the signal lines but at some window in time, the computer is decoupled and MCU takes over.

The signal lines (I am using 4, but for simplicity I have drawn only one) are cut off using a CBT3257 FET 4x 2:1 (de)multiplexer as not to disrupt the original timings, since it's high speed. Switching is controlled by the MCU pulling the output-enable (_OE) on the switch high which turns the switch output to high impedance, then making the SEL pin on the MCU an output.

• Vcc is 5V TTL everywhere
• No info on the current the computer can sink or source, or on the Floppy
• ATtiny can sink/source 45mA max per pin
• I cannot use the secondary input pins of the switch on 2 of the 4 ports since I am using those to switch 2 select lines around.

problem

If there is a bug in the MCU software it can occur that both the switch and MCU outputs are enabled, which then can short.

What can I do to protect the circuit in case I mess up the software?

Preferably using fairly inexpensive components, and not too many. I would like to avoid another IC.

what I've tried

• I need about 122 Ohm to limit 5.5V TTL to limit the sink/source current of the ATtiny's pins to a a safe 45mA. This can be implemented by putting down two 68 Ohm resistors on the line. The Floppy, however, fails to respond properly using anything much above 22 Ohm (as per the picture). Wether this is because the signal gets too weak, voltages out of spec or some timing issue maybe due to a capacitive effect I don't know.
• replacing the CBT with a 74LS257, which has proper TTL output, to boost the output and enable me to maybe use a higher resistor on the line, but this resulted in timing errors.
• AND-ing the lines from the switch and the MCU using diodes and pull-up resistor, but this also caused the computer-floppy communication to fail.

I have access to an 8 port logic analyser with digital (I assume CMOS level) inputs and an analog Scope (so no memory).

Any ideas? Thanks

Answer 1

The CBT3257 is a 2:1 multiplexer.

Connect one of the input signals to the Computer and the other to the MCU rather than as you have shown. Even if the software does not function correctly you can never have two outputs shorted together.

Answer 2

Use 74xx245s.

Bus transceivers like it are specifically designed for this exact purpose.

The downside is that they're half-duplex, so you can't use them in both directions at once, though you can mix the port-A and port-B connections to the microcontroller and use it with mixed directionality that way.

Answer 3

The Software Way

The easiest to overcome this is to do it in the software. You save time and effort to redesign the hardware and then rewire or fabricate new PCB. Here is your problem:

If there is a bug in the MCU software it can occur that both the switch and MCU outputs are enabled, which then can short.

To prevent software from destroying your hardware, I suggest you create a simple API to select the controller for the floppy drive. The following is a pseudo-code:

set_floppy_controller(device) {
  if device is MCU then {
    S  = 1                        // Disconnect CPU
    delay_a_bit()                 // To allow time for full disconnection
    set_pin(SEL_pin, PUSH_PULL)   // Let MCU in control
  } else if device is CPU then {
    set_pin(SEL_pin, INPUT)       // or Hi_Z
    delay_a_bit()                 // To allow time for MCU to disable output
    S  = 0                        // Connect to CPU        
  } 
}

// Example of usage
set_floppy_controller(MCU)
set_floppy_controller(CPU)

Test the API to ensure it works. This is done only once. Then if you want to switch the floppy controller, you MUST use the foolproof function and refrain yourself from doing otherwise.

If you do it otherwise, then you allow bugs to creep in.

The Hardware Way

Rewire your circuit like so

You can remove the resistor if you wish because the switch guarantees break-before-make switching. If you intend to damp the fast rise/fall signal edges, then let it be. This also means that you need not disable the CBT3257 (through OE) first to switch from CPU to MCU and vice versa, and then reenables it.