To be fair to Choueiri and his group, the Princeton measurement wasn't exactly +/- 3 dB (and due to the limitations of his anechoic chamber, the FR curves start at 300 Hz), see on the axis response curve (more like +/- 10 dB) .
fwiw, I contacted David JansZen back in 2012-2013 when these measurements were first posted and asked about the rising response on-axis. He responded that those measurements were of a prototype for the BACCH project that had no panel segmentation or internal response compensation. So higher directivity than the production zA2.1 which, evidently, was what they were after early in the project. I also recall there being a note in the comments section about the response being un-EQ'd (ie response of the raw panel) but I don't see it now. Flat unsegmented ESLs have a natural rising far-field response when driven with constant voltage. Dipole ESLs have the slope over their entire frequency range; monopole ESLs just above ka=1.
Last edited: