[net2-wg] lqi distribution for parent links in MultiHopLQI

[Omprakash Gnawali]

> On Jul 13, 2007, at 11:34 AM, Omprakash Gnawali wrote:
> 
> >
> > I logged the adjust(LQI) values of incoming pkt that triggered parent
> > change in MultiHopLQI. Here is a graph that shows the distribution of
> > adjust(LQI) at tx power 8, 16, 24, and 30:
> >
> > http://enl.usc.edu/~om_p/net2/white-bit/lqi-distribution.gif
> >
> > For comparison, the graph also shows distribution of adjust(LQI) of
> > all the incoming packets (parents/non parents).
> >
> > Here is a table to convert between adjust(LQI) and LQI:
> >
> > LQI, adjust(LQI)
> > 93, 790
> > 94, 729
> > 95, 669
> > 96, 612
> > 97, 561
> > 98, 512
> > 99, 465
> > 100, 420
> > 101, 380
> > 102, 343
> > 103, 307
> > 104, 273
> > 105, 243
> > 106, 216
> > 107, 189
> > 108, 165
> > 109, 144
> > 110, 125
> > 111, 106
> >
> > The conclusion is most of the parent links have adjust(LQI) > 400 so
> > if we wanted to set a whitlisting or blacklisting threshold they might
> > be in a region near 400. On the fifth line, interesting to note though
> > is 75% of all the adjust(LQI) values are smaller than 400, which
> > indicates that with a blacklisting threshold of 400, we would
> > blacklist links 25% of the time.
> 
> The whitelist code I checked in for the CC2420 puts the LQI threshold  
> at 100. So, adjusted values < 400 (I assume your > is a typo above).  
> I picked this value by eyeballing Kannan's LQI/PRR curves.
> 
> Phil


That ">" was a typo. It should have been "<". My conclusion would
result in a value of 101 and your eyeballing resulted in a value of
100. So we are in agreement, at least for now.

- om_p