The plot in glob98sp_ps is imperfectness as a function of block length.
Optimal rate 1/2 convolutional codes of m=14, 19 and 25 have been used.
They was decoded by hierarchical decoder. The solid line denotes the case
which rate loss and Shannon limit shift due to the tail bits have not been
deducted. The reason to present the solid line is due to that the tail biting
convolutional codes does not need the tail bits. Now the work has been focused
on how to decode long tail-biting convolutional codes.
The dashed line denotes the case which the rate loss and the
Shannon limit left shift due to the tail bits have been considered.
For m=14 cases, if the M=256 algorithm is used, the imperfectness
is increased by 0.1 dB.
For 10^-4 packet error rate, the imperfectness is increased by 0.25dB for
the m=14 case, 0.2dB for the m=19 and 0.18dB for the m=25 cases.
The plot in glob98fad.ps is the packet error performance in independent
rayleigh fading channels. The rate loss due to the tail bits is about 0.45 dB
which has not been deducted.