AVR Freaks

The simulator in AVR Studio 4 Rev 4.14 still has no support for the USI port.

Basically you make it into an SPI by writing a loop to shift out the bits or unrolling the loop into a linear stream to shift out the bits. The code is in the datasheets for the ATtiny24 and the ATtiny261. Check it out.

The reason they did this is so the same hardwar can do I2C as well. Pretty clever actually.

I don't know if a USI implementing a slave device can get its data into the appropriate register before the shift clocks come in from the master. I have a TEAC transport where the master waits for an acknowledge to SS*(Slave Select) before sending the first byte. I think this is kind of essential for implementing a slave device.

App note AVR319 says:

So it looks like it's up to you to implement it - which presumably means keeping the pins tristated with USI disabled until the line you have picked to be _SS is seen to go low.

Cliff

Devil's advocate: It sounds like an awful lot of trouble and complexity to go through, versus just having the "main" processor read a port.

Also, from your description you will have an extra line to tell the "master" that new data is ready. Why not make the '8535 the SPI slave in that case?

How fast/often are these inputs changing? (are these user inputs such as switches?) How fast does the reaction time need to be?

I'm leading up to polling & debouncing on the '8535 if the objection is that there are no pin-change interrupts on that model. Another obvious step is to use a Mega164 instead which will have pin-change interrupts.

The described system is not going to have an "instant" response, anyway.

If at all critical, I'd feel a lot better about reading the data directly. Or at least send the data byte, and then a checksum or inverted byte to check for integrity.

Lee