You have been asking the screen manufactuer instead of the lens manufacturer.
You radius the screen to the lens specification. Stewart has done their own experimenting and have a range of radiusses typically falling between 30 and 40 feet.
Amore precise formula was obtained from Isco assuming a 1.8 to 2.4 backup lens on the isco 3 2 to 1 TD, with 25 feet TD the radius is 60. I will pass this data to Stewart as they may want to revise their depth of curvatures.
"Our attachment III is designed for flat screens. But can also be used for slightly curved screens.
The attachment causes a small pincushion distortion. This is quite normal for anamorphic attachments.
Additionally the attachment amplifies the distortion of the projection lens.
Because of the 33% wider image when using the attachment the distortion is more visible on the screen.
If you use a cylindrical curved screen then the pincushion distortion will be reduced horizontally by the screen curvature.
The vertical distortion is not affected by the cylindrical screen curvature.
For optimal condition the radius of a 360 degree cylindric screen can be calculated with the formula r(cyl) = distance x 2.4;
For example with a distance of 4 m: r(cyl) = 4 m x 2.4 = 9,6 m.
For calculating the radius of a vertical curvature you have to use the slightly modified formula r(vert) - T = distance x 2.4;
Or calculating the distance: distance = ( r(vert) - T) / 2.4 ; with r(vert) is the radius of the vertical screen curvature and T is the depth of the vertical bending of the screen. Please see the attached drawing: "AnAtt formula vertical screen curvature.pdf".
Seems like the TORUS is dead in the water... It will be interesting to see what the new custom-radiussed DNP .8 all-format Supernova ISF