You will recall that one of the fatal problems with BICEP2 I identified was that its imaging plane antenna elements were far closer together than the ultimate resolution of electromagnetic wave ( ~ one-third the wavelength – a basic physics limitation.) How did this situation come about?
There is now this paper which explains how the spacing was calculated.Here is the relevant portion:
This paper explains that the antenna spacing was determined inside a wafer of quartz above the antenna slots, with a dielectric constant of 11.8. The thickness of the wafer appears to be about one free space wavelength. The slots are much smaller than one wavelength.
By using the grating lobe criterion they find that the minimum spacing at 180 GHz should be 0.46 mm.
But this calculation is wrong. The dielectric constant that needed to be used in the above calculation is that of air (=1.0). Then the spacing would 1.6 mm – which would have avoided the spacing fatality issue I raised.