- TBS calculation was performed 2 times. Once in "phy_procedures_nrUE_TX" and
another in "nr_ue_ulsch_procedures". The TBS calculation was wrong in "phy_procedures_nrUE_TX"
and correct in "nr_ue_ulsch_procedures".
- The data generation was performed in "phy_procedures_nrUE_TX" with a wrongly
calculated TBS because "ulsch_ue->nb_re_dmrs" wasn't set to a value, so the calculated
TBS would be more than it should be. This doesn't create a problem in the test because
the comparison between the TX and RX data is made using the correctly calculated TBS
which is smaller.
- Solution was to perform data generation in "nr_ue_ulsch_procedures", and remove TBS calculation
in "phy_procedures_nrUE_TX"
- removing "Nsymb_pusch" and "Nsc_pusch" from "NR_UE_ULSCH_t"
- using "number_of_symbols" from "NR_UL_UE_HARQ_t" instead
- removing "generate_ue_ulsch_params"
Added the following folder:
- ../openair1/PHY/NR_ESTIMATION/
Added the following file to CMakelists.txt:
- ../openair1/PHY/NR_ESTIMATION/nr_ul_channel_estimation.c
Added the following functions:
- nr_pusch_dmrs_rx
- nr_gold_pusch
- nr_ulsch_channel_level
- nr_ulsch_channel_compensation
- nr_ulsch_scale_channel is commented out (I think only used for downlink)
Changes from Khalid Ahmed Mahmoud Mohamed:
- nr_pusch_channel_estimation is used
- nr_ulsch_scale_channel is used
- nr_ulsch_channel_level is used
- nr_ulsch_channel_compensation is used
- multiple logging options for channel and signal magnitude
- removed gNB RT from "ru_procedures.c" to "nr_ru_procedures"
- added a new thread "nr_feprx_thread"
- added some functionalities for fep for rx in "nr_ru_procedures.c"
- changed Mdlharq to Mlimit when comparing with the round number
- Not incrementing the HARQ round in case of err_flag = 1, until HARQ is implemented
- Changed MAX_LDPC_ITERATIONS to 5 from 4 (provided better performance)
- Setting the payload to random data instead of all zeros at the Tx side
- Handling "nb_symb_sch" equal to 0 in "phy_procedures_nrUE_RX" which prevents
entering "nr_ue_pdsch_procedures"
- Handling symbol loop problem in "nr_ue_pdsch_procedures"
- Handling resource block loop problem in "nr_rx_pdsch"
This used to be done to ease power estimation in the UE,
but this is not correct. Let's put no energy in the REs
with no DCI data, which is what needs to be done.
Plus it consumes a lot of CPU because of calling taus()
way too much.
Some computation has been fixed in PUCCH reception.
SNR has been fixed, uses measurement instead of
a fixed value for the noise.
normalized_rx_power and target_rx_power have been removed,
replaced by snr and target_snr. All the files using those
variables have been updated.
Khalid Ahmed