- Problem in DCI NDI : the implementation use 5 HARQ processes, but process 0 NDI never toggled because it was reused on subframe 5 that is not carrying format 1 DCI. Fix -> use 8 harq processes instead of 5.
 - SI scheduled on every subframe 5 (even and odd frames) instead of only on even frames
 - Add DLSCH scheduling on subframe 5 for odd frames
 - change default rballoc from 0x7FFF to 0x1FFFF to support maximum 10MHz throughput