The number and relative geometry of collected measurements significantly affect the reliability of massive antenna array diagnosis. In this paper, we investigate and compare two deterministic algorithms for taking measurements based on a compressive-sensing-based approach for rapid and reliable detection of faulty elements in massive multi-input multi-output antenna arrays. We exploit the fact that the measurement matrix for a uniform linear antenna array reduces to a partial discrete Fourier transform matrix with rows corresponding to the measurements' locations. With the aid of the investigated algorithms, the measurements can be wisely taken to reduce the measurement matrix's worst-case coherence, which affects the detection probability of the defective antenna elements. In particular, one of the algorithms aims at constructing a measurement matrix with fewer distinct inner product values to reduce the worst-case coherence. The second algorithm is based on bounding the inner product between any pair of measurement matrix columns. Our study shows that uniform measurements can adversely affect the detection probability of the defective antenna elements, while using either one of the investigated deterministic algorithms leads to remarkable performance improvement.