- Organize into a main title and sub-headings
- Put var/func as inline code
- Add anchor links for the referenced step
29 lines
1.6 KiB
Markdown
29 lines
1.6 KiB
Markdown
# The gNB RACH Processing
|
|
|
|
## 1. Collect I/Q Samples
|
|
|
|
The gNB/ru main loop, in `ru_thread()`
|
|
If there is rach occasions in this slot
|
|
this main loop calls `rx_nr_prach_ru()` that stores the I/Q samples in the `prach_item_t` context.
|
|
|
|
## 2. Process the Samples
|
|
|
|
`L1_rx_thread()` has received slot indication by msg on queue `resp_L1`, it calls `rx_func() => L1_nr_prach_procedures()`
|
|
`L1_nr_prach_procedures()` fills `nfapi_nr_rach_indication_t` structure from result of `rx_nr_prach()` that process the content of `rxsigF[prachOccasion][aa]`
|
|
`rx_func()` sends the decoded rach to upper processing by call to `NR_UL_indication()` that calls `handle_nr_rach() => nr_initiate_ra_proc()`.
|
|
|
|
`nr_initiate_ra_proc()` creates a UE context if it needs, then sets `UE.ra.ra_state` to what it should do.
|
|
|
|
## 3. Send the RACH DL
|
|
|
|
`tx_func()` that is launched by `L1_tx_thread()` as it gets `L1_tx_out` msg (from main loop in [1.](#1-collect-iq-samples))
|
|
`tx_func()=>run_scheduler_monolithic()=>gNB_dlsch_ulsch_scheduler()=>nr_scheduler_RA()`
|
|
`nr_scheduler_RA()` process all existing UE contexts, to see it it has to schedule a RACH DL.
|
|
|
|
## 4. Global Synchronisation
|
|
|
|
`resp_L1` msg for ([2.](#2-process-the-samples)) is created in `tx_func()` that is launched by `L1_tx_thread()` (see [3.](#3-send-the-rach-dl)) when it receives slot indication as it gets `L1_tx_out` msg
|
|
this msg is sent by the main loop (see [1.](#1-collect-iq-samples))
|
|
`tx_func()=>run_scheduler_monolithic()=>nr_schedule_response ()=>nr_schedule_ul_tti_req ()=> nr_schedule_rx_prach()`
|
|
writes in the triggering structure to ([1.](#1-collect-iq-samples)) that it will have to run `rx_nr_prach_ru()` in a given future slot.
|