In this paper, we investigate pilot power allocation in multi-cell multi-user massive MIMO systems for enhancing channel estimation quality. it is known that, with good channel estimation quality, we can detect signals more accurately. To improve the throughput, instead of focusing directly on spectral efficiency (SE) of the system, we focus on improving the channel estimation quality by creating a closed-form formula for the normalized mean squared error (NMSE) between true and estimated channel. We consider well-known minimum mean squared error (MMSE) channel estimation method. After that, we propose a minimization problem to minimize the summation of estimation error in the system with constrain of a maximum pilot power per user. The optimization problem for is nondeterministic polynomial-time hard and it is very hard to find the global solution, so that we propose an algorithm to find the local optimum solution with polynomial time. Numerical results show the advantages of our proposed approach in channel estimation quality, sum SE of the system, and minimum SE in the system.