Did you try the interrupt version? - I've had it running all day (I was out) - and it has not missed a beat. Also - being interrupt driven means that the slave doesn't have to busy wait on the i2c status. The master controls the data transfer (send and receive!) - and inter-mcu communication is good!

A bit more playing and I almost got it to work by tweaking the CAL_I2C file - but it would work for one time and then stop. After a bit more fiddling - no joy. So - I rewrote the slave side using the TWI interrupt to handle receipt and sending of data. This worked remarkably easily - so here is a ve...

Another step forward

The RTC is acting as a slave device. Could maybe look at the output for this and see what/how it varies?

Another 'oddity' - when writing m->s there is an ~18us delay between bytes. On receiving s->m there is a much shorter 'gap' (~8us) and adding a delay after each byte is transmitted doesn't affect this as expected - adding a 1ms delay gives 2 bytes with first value (and a 8us delay between them) then...

Looking at your screenshot - after the 'receive' SDA stays low which holds the i2c bus as 'busy' - and prevents further transmission. If you look at the trace I posted - you can see that SDA goes high again after the 'read' It might be a slightly different board - could you try a pull-up resistor on...

It is very odd = what hardware are you testing on? As you can see from my screenshot it had counted up quite a few iterations when I took it? Can you 'disable' the 'Read' in the sender (master - will perhaps have to think again about the naming!) Iain - can you remember what you got returned by the ...

I've had one step forwards and one 'to the side' So - having ascertained that transmit works AOK (for me :-) ) I rewrote the code to use the I2C CAL component and transactions )notice how this makes the code for the master much simpler) I also just use the scope here (and as smartscope has decided t...

Cool - glad it's working for you...

It's odd that it just worked one time for you - repeated every second for me. Strange. Initialise is needed before using the i2c (once only), the transaction routines allow a simpler interface (it handles start and stop conditions. Transactions are also the only available interface on some MCUs but ...