The documentation of the Link Quality Indicator leaves something to be desired, as the little bit that is documented is clearly not correct: [quote]
The LQI gives an estimate of how easily a received signal can be demodulated by accumulating
the magnitude of the error between ideal constellations and the received signal over the
64 symbols immediately following the sync word. [/quote]
The above appears to have been copied word for word from an early datasheet describing a chip that has a fixed 64 byte payload, which in the Wixel is of variable size. But in any case, what might “ideal constellations” mean? Does this assume Forward Error Correction is on?
Has anyone experimentally determined how the LQI value depends on the size of the packet buffer, or if it is even useful? ((I’m hoping to save myself the trouble).
I think you’re quoting section 13.10.8 of the CC2511 datasheet, which reads:
I don’t immediately see anything wrong with that documentation. The CC2511 does allow variable length packets, but it still can be valid to talk about the “64 symbols immediately following the sync word” and use to measure the link’s quality. Similarly, books have a variable number of pages, but it still can be valid to talk about the 64 pages immediately following the table of contents, and use them to measure the book’s quality.
I don’t think there is any assumption about Forward Error Correction.
I don’t know how the LQI depends on the length of the buffer. I think the most important number to look at is the packet error rate, and in some recent tests we did using the test_radio_signal_tx and test_radio_signal_rx apps there seemed to be no correlation between the packet error rate and the average LQI of the packets that were received successfully. However there did appear to be some correlation between RSSI and packet error rate.
Thanks for providing those Wikipedia links. The one to “constellations” was helpful, and suggests that what is measured is the average Euclidean distance between the idealized points (“symbols”) in the phase diagram encoding the bits and the actual positions of the received points on the phase diagram. I have found other definitions, though and some just rescale the RSSI to be positive.
Since one symbol usually encodes several bits, a two-byte packet would not likely be composed of 64 symbols. The question remains as to how the LQI is actually calculated and how it might depend on packet length. However, further literature research revealed that the usefulness of the LQI is debatable, consistent with your experimental findings, so it is probably not worth the bother. The results of some testing with both LQI and RSSI indicators is at sing.stanford.edu/talks/emnets2006.ppt
Incidentally, PRR on the vertical axis of the graphs stands for “Packet Received Ratio” (I assume without content errors).