Massive MIMO with Per-Antenna Digital Predistortion Size Optimization: Does it Help?

In this paper, we study the effect of optimizing the per-antenna digital predistortion (DPD) sizes on the performance of the downlink massive MIMO system with residual hardware impairment. In particular, we first quantify the capacity after applying some per-antenna DPD with different numbers of coefficients at each antenna and then maximize the capacity by finding optimum values of the per-antenna DPD sizes under a constraint on the total number of DPD coefficients in the system. Two closed-form sub-optimal solutions are derived and numerical examples illustrate that their performance is very close to the optimal solution. It is shown that when the channel large-scale gains have high variation over the transmitter array, which is the case in practical scenarios, using our proposed optimized DPD sizes can improve the system capacity significantly. This allows us to scale down the massive MIMO system but still maintain performance. We also study the asymptotic behavior of the derived capacity and show that by using the optimized DPD sizes, the system performance can approach the asymptotic bound with a significantly smaller number of antennas.