??? 02/14/13 11:15 Read: times |
#189363 - more details Responding to: ???'s previous message |
Oliver Sedlacek said:
I can see that a ring is a really good topology for a small number of 8051 class nodes, but can you explain your twisted ring a bit more as it sounds really interesting. How many bits or bytes make up a packet? As many as you want: The slaves extract N bytes, following a Edge on the 9th bit, and flip Bit 9 as they do so. Thus the next slave sees the edge at an advanced packet-point, form the slave before it, and it removes the next block of data, and so on around the loop. Because address is implicit by location, all info is data. On the simple design we did, bytes-extracted was compile-time fixed, and the same for all identical slaves, but you could make the first byte after an edge a Message-Count, and trade off a little bandwidth for flexable messages. Oliver Sedlacek said:
Can any slave inject a packet? Yes. Usually all slaves remove N bytes after an edge on the 9th bit, and replace those with N replies. In all other slots (not next to the edge) they simply pass-on the Data stream. So the total bytes in circulation does not change, but the 9th bit moves phase, as you probe various points around the ring, and the content of the bytes changes from SendInfo to ReplyInfo, also as you move around the ring. Oliver Sedlacek said:
Where does the twist come in? It is not a physical twist, but based on a twisted ring counter (johnson counter to some) - which is a shift register + Inverter. The 9th bit in the loop, follows that same pattern, and you can think of the 9th bit's as a chain of D ff's |
Topic | Author | Date |
Multi-CPU designs | 01/01/70 00:00 | |
Commonly Done | 01/01/70 00:00 | |
Fun with network protocols | 01/01/70 00:00 | |
and also a twisted ring network | 01/01/70 00:00 | |
Please explain a bit more | 01/01/70 00:00 | |
more details | 01/01/70 00:00 | |
I'd use I²C | 01/01/70 00:00 |