L1/L2 low-layer signaling issues
This is a collection of issues pertaining to disfunctions in the L1/L2 procedures (mostly eNB).
Corrected issues (ready for merge with develop)
- the PUCCH handling (power control, resource mapping, L1 receiver bugs) for scheduling request and ACK/NAK. This has been mostly fixed, PUCCH1 and PUCCH1a/1b power control, PUCCH1 thresholds.
- uplink and downlink MAC scheduling (subframe 0,5 for DL, full UL scheduling procedures SR/BSR/CRNTI-CE). Subframe 0/5 scheduling has been added to eNB scheduler. DL throughputs on Cat 4 has been measured at 16 Mbit/s for 5 MHz channels and 33 Mbit/s for 10 MHz channels.
- the "nCCE" bug. This has been resolved, but not fully tested (i.e. with 2 or more UEs). This involved moving some code from the PHY PDCCH procedures to the MAC scheduler which removes impossible CCE allocations.
- included an activity timer and a connection timer (instead of the consecutive error count). only when that is reached, UE is remove from MAC.
Remaining (major) issues:
- UE reporting otherFailure via RRCConnectionRestablishmentRequest. This happens when a UE is connected and other UEs in the vicinity try to connect but fail. This is easily triggered by establishing a connection with a COTS UE with an OAI sim. Take any random smartphone, do a scan of all networks and select the OAI cell. The connection will fail and result in the other UE abandoning the connection and trying to reconnect with "otherFailure". This case is probably due to a confusion of the SRBs between the 2 UEs (to be verified)
- Handling of RRCConnectionRestablishmentRequest. This currently results in a RRCConnectionRestablishmentReject, which triggers a reattach by NAS. This procedure should be implemented to recover after radio-link failure or "otherFailure"
- Normal UL scheduling (i.e. based on SR/BSR reports) still has issues which limit UL throughput. SR seems to be working properly and the UL connection is fine. eNodeB doesn't allocate enough resources to maintain a high enough throughput.
- SRS is not functional yet. This should be reactivated and debugged.