Microphone arrays are commonly used to capture sound fields. As the number of sensors forming the array increases, the spatial sampling accuracy at high frequencies improves. Numerous prototypes of spherical arrays were developed over the last years. However, much less attention has been paid to the intrinsic performances of the sensors than to their number and arrangement. This study aims at evaluating the relative performances of two rigid spherical microphone arrays of the exact same size differing only in their capsules (pressure sensors). The two recording systems are based on higher order ambisonics and were used to acquire the exact same sound scene. Four short music excerpts were decoded as various types of audio content (mono, stereo and multichannel) and displayed through dedicated loudspeaker setups. The recordings issued by the two arrays were then to be compared by pairs, on a similarity basis and on a preference one, by twelve expert listeners (sound engineering students). The results showed that the perceived differences and preferences depended on the way stimuli were rendered. These assessments were consistent with those obtained from naïve listeners in a previous study, although experts perceived significantly better the differences and reported more pronounced preferences.