Hi Phil, Thanks for the explanation. That was quite helpful. I agree with your explanation. I agree that to have perfect reliability we have to implement end-to-end reliability (or at least network layer reliability). I do not have the exact numbers for the false acknowledgment rates, but at high data rates was starting to see around 10-20% of packets failing to make it through. Hence was trying to see if I can improve the reliability at the MAC layer. The only problem with implementing acknowledgments at the network layer is that it will introduce latency and also consume capacity (since the data packets are slightly larger). Hence was trying to take the shorter route by improving reliability at the MAC layer (without much success though :) ).
I also wasn't aware of the limitation caused by the DSN bits in the ACK's, in a dense deployment this can definitely cause problems. Thanks again for your explanation. regards, Aviansh On Thu, Feb 19, 2009 at 8:06 PM, Philip Levis <p...@cs.stanford.edu> wrote: > > On Feb 19, 2009, at 6:42 PM, Avinash Sridharan wrote: > >> Ah ok, >> Thanks a lot. The specs makes sense, I guess need to look at other >> places to understand the false acknowledgments. Any idea as to why >> this might be occurring ? > > What false positive acknowledgement rate are you seeing? > > 802.15.4 suffers from false acknowledgements in part because acknowledgement > packets do not have source addresses. Since the DSN is only 8 bits, it is > possible that a node hears an acknowledgement for another transmission and > thinks it is for its own. > > Also, it's possible that a packet is corrupted but passes CRC. The receiver > will issue an ACK. But, if, say, the AM type is different, your network > protocol may never receive the packet. > > Link-layer acknowledgements are not perfect -- no acknowledgements are. This > is the end-to-end argument. If you want to know that a packet has arrived at > the network layer, you need the network layer to send acknowledgements. > > The problem is a little more pronounced in 802.15.4 because CRCs are only 16 > bits and there aren't checks at many layers (layer 2, layer 3, layer 4, > etc.). > > Phil > > -- Phd Dept. of Electrical Engineering University of Southern California http://www-scf.usc.edu/~asridhar _______________________________________________ Tinyos-help mailing list Tinyos-help@millennium.berkeley.edu https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
