Reliable long 1-wire driver

[LukeW]

Maxim's application note here has much detail on reliable 1-Wire networks with relatively long cables - capacitive loading, reflections, etc.

https://www.maximintegrated.com/en/app-notes/index.mvp/id/148

Appendix A's "Improved CPU Bus Interface" shows a circuit for reliable operation, but it uses two microcontroller port pins.

Ideally, I want a circuit that can be a "drop in replacement" in series with a single data line, in a pull-up 1-Wire bitbanging arrangement - I don't want to change the code or use an extra GPIO pin. Is there any way to realistically get rid of it?

[rstofer]

Not with that circuit.  The FET does the signalling by pulling the signal line low but it's not possible to sense the line state.  So, the top pin is used to sense and the bottom pin is used to drive.

Just guessing...

[LukeW]

Does anybody have a code example (PIC, AVR, Arduino, whatever) that can support 1-wire in these multi-pin configurations in large networks?

[JVR]

LINBus not what you are looking for?

[kony]

Seconding the LIN suggestion. It is very resillient and well prooven. With bit of modification of the protocol, you can even add slaves IRQ. I have used it for modular system of actuators and sensors in past, and it made it very flexible for customisations, yet robust.

[SiliconWizard]

Does anybody have a code example (PIC, AVR, Arduino, whatever) that can support 1-wire in these multi-pin configurations in large networks?

You would just have to modify the existing code to use a separate input port and output port for writing and reading instead of using the same... doesn't seem that hard

Edit:

I don't want to change the code or use an extra GPIO pin

Those two statements seem a bit contradictory... 
Unless... you don't have access to the original code. In which case you'd have to rewrite the whole software anyway.

[David Hess]

The first thing I would do to improve reliability is replace the pull-up resistor with a current source.