Ebrahim Vakilpour wrote:

It was 500 bytes/second (at 15 HZ), do you still think fifos are not acceptable ?

I am not saying FIFO's are not acceptable. I am just not sure they are needed. You say 500 "Bytes/second (at 15Hz)" What exactly do you mean? If you send at 15Hz, you will not send 500 Bytes per second... Where does the 15Hz fit in?

Besides, even 500 Bytes/second is _not_ a lot. It would take some 10-15 instructions to have an interrupt routine read the byte and put it in a buffer. That would take say 20 useconds, and this 500 times per second. You would lose 10,000 useconds per second, or 1% of your available CPU time. I still believe that if you cannot spare 1% of your cpu-time for this task, you will probably run into problems anyway, as your processor is probably to slow to handle all the tasks assigned to it.
People with more experience in this field than myself are welcome to agree/disagree with this remark of mine.




And Patrick can I ask your opinion about cross checking feature, please (http://www.8052.com/forum/read.phtml?id=64218 #4 in the case you forgot)


Didn't forget. Maybe I got no opinion on this. Or my opinion is in line with Erik's: "Why (ab)use your precious CPU-time watching your other CPU, if you can install a (cheap) supervisor/watchdog chip?"



As to the development of the protocol, I can add some general remarks:

If you want to specify a protocol, then make sure your specification
1) is complete.
2) leaves no room for interpretation
3) covers all possibilities
4) defines the "boundaries" where the protocol is applicable
5) uses consistent terminology.
6) has a glossary.
7) has whatever is needed to prevent mis-interpretation....

Look at other protocol specifications and see how these are written. (Personnaly I like the IETF RFC's as they are clear and readable, but then maybe I am biased).


As we have seen in this thread, a lot of confusion has been introduced due to lack of a "common" terminology. The same applies to the incomplete description you gave initially.

Furthermore, as I have said before: don't let me stop you from developping your protocol. It's just that with the data you have provided, I see no justification for a FIFO and/or complex protocol. I would just stick to something existing and proven, making sure I do not re-invent the wheel where it is not needed.

regards