So... here is the spreadsheet I used to find wall solutions (at least some of them!). Problem is that for most diameters you'll need approximations (a circle that is not really a circle). The method I use works only in the case where we are "close enough" of circle.
"Rounded Nseg" is the number of segments needed to approximate a given circle diameter. This number must be at least divisible by 3 (to close the circle), in orange, better by 6 (yellow) to have a full diameter 'on grid', even better (green) divisible by 12 to get two orthogonal diameter 'on grid'. "Effective R (studs)" gives the radius (distance from origin) for the first 1x2 brick. "EffectiveAngle" is the rotation around origin of the next 1x2 brick. "Diameter error" show precision of diameter. Closest to 0 (green) is best, but higher values are acceptable (orange), this is the case for D=23 circle attached.
"Rounded Nseg" is the number of segments needed to approximate a given circle diameter. This number must be at least divisible by 3 (to close the circle), in orange, better by 6 (yellow) to have a full diameter 'on grid', even better (green) divisible by 12 to get two orthogonal diameter 'on grid'. "Effective R (studs)" gives the radius (distance from origin) for the first 1x2 brick. "EffectiveAngle" is the rotation around origin of the next 1x2 brick. "Diameter error" show precision of diameter. Closest to 0 (green) is best, but higher values are acceptable (orange), this is the case for D=23 circle attached.