"Richard Harrison" wrote
First difference from Richard Fry`s table is the loss of 3.55 dB
as the result of circular polarization (mostly) as half of the power
which a linearly polarized reference dipole would use is
cross-polarized.

That, and the fact that the radiation pattern from each element is not the
pure cosine function assumed for a 1/2-wave dipole. It has slightly less
gain peak gain.

The steps between doubling the number of elements
in Richard Fry`s table are all nearly 3 dB.

"Nearly" is right, but the difference is not uniform for successive doubling
of apertures.

A small variation in the bay-bay spacing (departing from 1 wavelength) is
needed as a function of the number of bays, to maximize the peak gain from
this type of an array. The arrays in my table all have exactly 1-wavelength
element spacing, and the peak gain from arrays of them is lower than
expected for lower numbers of elements, and higher than expected for higher
numbers of elements -- which stretches/compresses that nominal 3 dB delta.

Fine points, to be sure.

RF