In this article, a real valued time delay neural networks (RVTDNN) based power amplifier predistorter is presented. A two hidden layers neural network, trained with measured data, is used as predistorter. Nonlinear distortion compensation is increased by implementing proposed approach. An achievement of the spectral regrowth improvement of 10 dB for a LTE 1.4 MHz 64 QAM signal at 880 MHz as center frequency is presented. The highly nonlinear MIMIX CFH2162-P3 power amplifier is used as a device under test (DUT). The effectiveness of the proposed approach is validated with experimental results.