"How is that consistent with the RT framework? "
This is just my opinion.
I really do not think you can have a real serial based RT closed loop system!
The best you can do is come close and event then you have to be pretty loose in your definition of deterimism. Here is why I feel this way.
For simplicity, let us thow out the issues with recieve errors and re-transmision requests (how can be deterministic of the data transfer time can not be defined?) A serial transmision takes some mount of time to be transmitted. The higher the baud rate the shorter the time. So with a high baud rate you still have some amount of time between when the "correct value" is placed in the output buffer until the entire buffer is recieved and turned back into a series of byte
s. With serial this process will often (based on baud rate) be on the order of milliseconds. This introduces a delay in your output widget responding.
With analog or digital output control lines the "information propogation delay" is determined by the transmision media. These delays are typically on the order of microseconds.
Of course all of the above statements assume the rest of the system is ideal and reacts instantaneously.
I just thought that your question was a good one and wanted express my thoughts on the matter. Twenty years from now people will probably be making similar statements about analog control loops saying that if you cannot keep a loop closed with nano-second determinism then its not RT.
Just my thoughts,
Ben
