This is a repost from my studio construction thread. I don't think I am going to get any interesting discussion started in there so I hope you'll forgive the double post.

I did find a paper named bbc research department report from way back in 1990. It seems to approach the math considering diffusion from a slightly different angle then in the book I am currently using, but does not answer the following.

I wish to construct a 2-dimensional QRD, although not with the same number of colums as rows. One way that these are (or at least were back then) constructed is by taking two modulo series as vectors (one a column and one a row) and multiplying the elements to get a p1 x p2 matrix with p being the respective prime number. Another (mostly used for primitive root diffusers) is to take p-1 and dividing it into two numbers that give the dimension of the matrix. the elements are then set starting with the first diagonal and working from there.

I wish to create a diffuser in the neighborhood of p1=71 horizontally and p2=11 vertically. This is generally possible with both construction methods,correct?

here's a question to the more experienced acousticians out there:

i am aware that this construction provides me with a far better dffusion in one direction than the other. This is less of interest to me than the fact that I would reduce repetition of patterns on the horizontal level, providing a smoother diffusion. I don't know if this makes more sense than to just create pseudo-square diffusers and rotate them by 90° compared to their neighbors.

Can you think of any flaws in my thinking? Does my strategy make more sense? Or should I go for p1=p2 and rotate them?

I really really hope my post makes sense to you guys