esp. for LEO NTN scenarios the timing drifts during initial sync, which is mostly compensated with --ntn-initial-time-drift.
But there might be a residual time offset, and by extending the peak search window we make this more tolerant.
The rotation compensating the shifted OFDM position _before_ the channel estimation makes the CIR peak move back to 0.
We have to take this into account on gNB side when calculating the timing advance command, and on UE side in the time adjustment.
Currently we cute the PUSCH and PDSCH OFDM symbols starting within the last 1/8th of the CP to avoid ISI.
This time shift leads to a rotation in frequency domain.
The current PRB based channel estimation does not compensate for the rotation _within_ a PRB, leading to bad performance.
This commit should _not_ make it into develop branch, it is just for testing!
Instead, we should do some compensating rotation before channel estimation.
- this was necessary because as soon as the gNB starts, it tries detecting
the prach preambles and it mesaures the background noise. However, there is
a delay of several dozens of frames before it considers the preables as
detected.
- the solution at the UE consists in avoiding sending preamble as soon as the
UE syncs by introducing an offset, this keeps the power measurements low at
gNB side as long as the preamble detection counter is under threshold
When a header is changed, all including files need to be recompiled,
too. Thus, including a header needlessly all over the place drives
compile times up without any reason.
- using nr_extract_dci_info instead of nr_generate_ue_ul_dlsch_params_from_dci as in phy_procedures_nr_ue. The latter was only a (poor) wrapper and not doing what the name suggests. also with FAPI things have to be done differently now. This will come in a next commit.
- other minor changes to remove warning
prepreatory changes
Thomas Schlichter