Maximum Throughput calculation for a specific TDD configuration with srsRAN

[Keilwerth11270]

Hi everyone,

I am estimating the maximum throughput in UL and DL for a specific TDD configuration based on the TS 38 306.

I am using the default TDD config of srsRAN, with

tdd_ul_dl_cfg: 
    dl_ul_tx_period: 10                     
    nof_dl_slots: 6                         
    nof_dl_symbols: 8                       
    nof_ul_slots: 3                         
    nof_ul_symbols: 0                       
    pattern2: 
      dl_ul_tx_period: 10                       
      nof_dl_slots: 6                           
      nof_dl_symbols: 8                         
      nof_ul_slots: 3                           
      nof_ul_symbols: 0  

The calculation requires the number of RBs in UL and DL, and a scaling factor.

With the default TDD config, we have 6 full slots and 8 OFDM symbols in the next slot in DL and 3 full UL slots in the TDD pattern periodicity of 10 slots. So, in 10 slots, DL used 6.57 slots, and UL used 3 slots.

For example, with 20 MHz bandwidth and 30 KHz subcarrier spacing, the number of PRBs is 51. The distribution of PRBs for DL is 33.507, and UL is 15.3. For this calculation, we have not considered additional factors such as DMRS, SLIV etc. and with this, the calculated throughput comes out to be 53.73 Mbps in DL and 26.24 Mbps in UL with scaling factor 1.

The measured OTA throughput under the best channel conditions is 42.5 Mbps in DL and 17.9 Mbps in UL.

What would be the best way to approximate the number of DL and UL RBs with the default srsRAN configuration?
And is a scaling factor of 1 correct for srsRAN?

[NicolasGresset]

Hello,

What would be the best way to approximate the number of DL and UL RBs with the default srsRAN configuration?

The way you mention the distribution of PRB for DL and for UL let me thinks that you consider that at a given time, 33.507 slots are allocated for DL and 15.3 are allocated for UL whereas the very principle of TDD is to allocate every slots to DL for a fraction of the frame then every slots to UP for another fraction of the frame (with a certain guard period between them to avoid collisions).
Therefore, I don't think this question makes any sense.

And is a scaling factor of 1 correct for srsRAN?

Consider scaling factor to be equal to 1. As far as I know, it is only used in specific scenarios with high modulations and multiple layers and I assume you do not use the latter.