Calibration of Sionna SYS

[rbachu21]

Is the system level simulator calibrated? If yes, how can we recreate those results? What were the reference results used?

I see that the the throughput curve for proportional fair scheduler is concave near origin instead of convex (which would provide some minimum throughput for low SINR users).

[lollodealma]

Hi @rbachu21,

Effective SINR values are computed via the EESM method, with multiplicative coefficients stored in sionna/sys/esm_params/eesm_beta_table.json.
These coefficients are taken from [1].

It is possible to try replicating the coefficients from [1] using Sionna.
If you do, please let us know—it would be interesting to compare the results.

FYI, new BLER tables can be generated as in the tutorial https://nvlabs.github.io/sionna/sys/tutorials/PHY_Abstraction.html .

Regarding your concern about throughput, I'm not sure I understand the issue.
Users are allocated wrt the classic proportional fairness metric, which allocates to each user approximately the same n. resources in the long-term.
Upon resource allocation, in the downlink, power is decided according to a second fairness criterion:
https://nvlabs.github.io/sionna/_modules/sionna/sys/power_control.html#downlink_fair_power_control

Make sure to set guaranteed_power_ratio>0 (default: 0.5) if you want to guarantee a minimum service to all users.

Best regards

[1] S. Lagen, K. Wanuga, H. Elkotby, S. Goyal, N. Patriciello, L. Giupponi. New radio physical layer abstraction for system-level simulations of 5G networks. IEEE International Conference on Communications (ICC), 2020

[rbachu21]

Hi @lollodealma

I was actually referring to calibration as shown in 3GPP TR 38.901. Basically, we need to get the same pathloss CDF and geometry SINR CDF. Was wondering if this was done

Also, for the throughput, I see the following curve

It is showing 0 throughput for several users with low SINR. The default value of guaranteed_power_ratio was not changed.

How do I get hold of a copy of the reference in your response? I do not have IEEE access.

Regards.

[lollodealma]

Hello,

1. Calibration results as in 3GPP TR 38.901 are not currently available, but they can be reproduced. Sionna includes 3GPP-compliant channel models, user drops in a hexagonal grid, and wraparound support.

2. In the example you mentioned (https://nvlabs.github.io/sionna/sys/tutorials/End-to-End_Example.html), users with zero throughput experience poor channel quality, for which even the lowest MCS results in a TBLER of 1 (cfr. the TBLER histogram).

3. Here's the public arXiv version of the paper you request: https://arxiv.org/pdf/2001.10309

Best regards