ABSTRACT The detection of redshifted hyperfine line of neutral hydrogen (H i) is the most promising probe of the epoch of reionization (EoR). We report an analysis of 55 h of Murchison Widefield Array (MWA) phase II drift scan EoR data. The data correspond to a central frequency $\nu _0 = 154.24 \, \rm MHz$ (z ≃ 8.2 for the redshifted H i hyperfine line) and bandwidth $B = 10.24 \, \rm MHz$. As one expects greater system stability in a drift scan, we test the system stability by comparing the extracted power spectra from data with noise simulations and show that the power spectra for the cleanest data behave as thermal noise. We compute the H i power spectrum as a function of time in one and two dimensions. The best upper limit on the 1D power spectrum are: $\Delta ^2(k) \simeq (1000~\rm mK)^2$ at k ≃ 0.2h Mpc−1 and at k ≃ 1h Mpc−1. The cleanest modes, which might be the most suited for obtaining the optimal signal to noise, correspond to k ≳ 1h Mpc−1. We also study the time-dependence of the foreground-dominated modes in a drift scan and compare with the expected behaviour.