Summation in Wideband vs Averaging in Subband in PMI selection

In 5G toolbox, in the hDLPMISelect.m file, we have two different approaches for wideband and subband to find optimum indices for PMI selection.

In the wideband case, we do sum of SINRs across the BWP for all layers for each PMI index. But in the subband case, we average the SINR per RE values across the subband for all the PMI indices.

I was wondering why this is the case? I couldn't find any related material for it.

Thanks

Answer 1

[Pratyush]

Hi Jake,

I understand that you want to know the reason behind different ways of selecting PMI for wideband and subband cases in MATLAB's 5g Toolbox.

This is done due to varying channel conditions:

• Wideband PMI Selection: Aims to maximize total throughput across the entire Bandwidth Part (BWP). It sums the SINRs across all Resource Blocks (RBs) for all layers to find the best PMI for the entire BWP.
• Subband PMI Selection: Accounts for channel variability across different subbands within the BWP. It averages the SINR per Resource Element (RE) within each subband to select the most suitable PMI for each specific subband.

The sum is used in the wideband case to optimize overall performance, while the average is used in the subband case for a more detailed adaptation to local channel conditions.I could not find the exact rationale explicitly documented, but it is consistent with the principles of link adaptation and precoding in the context of 5G communications. I think the choice of using sum or average is ultimately a design decision that reflects the different optimization goals for wideband and subband PMI selection.

[Jake]

Thanks Pratysuh. This makes sense. However, why not use another metric like effective SINR in this case? Something like MIESM (Mutual Information Effective SINR Mapping). Wouldn't it yield better result compared to taking average of SINR values?