Bugzilla – Bug 1628
Bug in LteMiErrorModel on b and c coefficients of BLER curves for MCS 11, 15, 18 and 19.
Last modified: 2013-12-10 05:02:03 UTC
As reported by Oleksandr in mailing list: https://groups.google.com/d/msg/ns-3-users/F0_ZbBnSgAI/tVA9zhGXDtQJ There is an incongruence between the BLER curves reported in the documentation and the "b" and "c" coefficients used for model the curves with Gaussian cumulative model. In detail, their behavior in the simulator is less conservative respect to the expected one from the curves (e.g., the knee of the curve to zero error probability is shifted to higher dB, in some cases even higher respect to adjacent less robust modulations). This might imply an optimistic evaluation of the error probability of such MCSs and strange behavior in the CQI evaluation in the LteAmc::AmcModel::Vienna model (i.e., the spectral efficiency of MCS from 12 to 19 are not considered in the CQI assignment).
The bug has been solved in lena repository and it will be integrated in ns3 official distribution with ns3.18 version. The b and c parameters have been updated for the MCSs affected. I would like to thanks Marco Mezzavilla for the support in solving this issue.
Marco's changeset e7367028d8a6 is in ns-3-dev