Switch from Slave to Master in I2C during execution.

[facosta.ie]

Hi, I'm trying to change a PIC18F25k80 from slave to master and vice versa during the program execution, is this posible?

My PIC works fine as master and as Slave individualy using the instructions:

#use i2c(SLAVE,ADDRESS=0x16,sda=PIN_C4,scl=PIN_C3,)
Or
#use i2c(MASTER,sda=PIN_C4,scl=PIN_C3)

But I need that the same IC2 works sometimes like master and sometimes like slave.

Thanks in advance.

[Ttelmah]

Setup the two uses with streams.
You will need to behave like in multi master mode, and not use the master, unless the clock line is high.
Use force_hw on the slave setup, and software for the master (you are already doing this, I2C defaults to software mode for the master unless you use force_hw).
Obviously your master should not try to talk to the address used by the slave.

[facosta.ie]

Slave works ok, but master is not working, I use the Slave Stream in the interrupt and the Master Stream when I want to read another Slave, any suggestion?

P.S when I quit the Slave USE the master works correctly.

[jeremiah]

perhaps there is a particular register bit that must be set master vs slave? If so, you may have to do that manually.

[Ttelmah]

You do understand that you have to switch off the slave peripheral to run the master. They can't both be enabled at the same time:

[facosta.ie]

Thanks for your answers, Ttelmah and jeremiah with the I2C_init() instruction was enough to enable and disable one or another, it works fine.

I was testing the clk pin(C3), for enable the master and it stucks, if I don't test this pin, the master enables and disables correctly.

For my application I'll be testing a control pin wich will tell me the moment to act like master or slave, with this pin I will not need check the clk pin right? Only will be one master enabled in the bus.

[Ttelmah]

A control pin is much better.
Remember though to not turn it off on whichever device operates it, until the transaction has completed.
You should be able to read C3 though. I suspect the problem here is TRIS. By default in standard I/O mode, reading the pin would change the TRIS. You read the pin in multi-master mode, and to avoid two devices trying to operate the clock at the same time, but when I've done this I've always been in fast_io mode (so controlling the TRIS). Even with this being read, you have the problem without a separate control, if two devices start at the same time. This is why you have to add bus collision detection when using multi-master. So a separate control line is very much better.

[facosta.ie]

Thank you again for your help Ttelmah, I think that the control line is what I need to sync correctly the devices and avoid collisions.

I'll test the complete system and will tell you the results.

Regards.

[jeremiah]

Side note: If you want to read a pin in standard IO mode without changing the input/output state, the input_state() function is an additional option. It'll read the pin without changing the TRIS settings. I prefer to use this rather than fast IO or fixed IO.