Landman Bester

Do you mean maximum or optimal number of directions? Maximum is easy. At least the maximum for a unique solution to exist. Optimal is a different story entirely... I don't...

Well, without a regulariser, you need fewer parameters than data points so roughly Ndir * (Nt/Delt) * (Nv/Delv) * Ncor * Na < Nt * Nv * Ncor * Na(Na-1)/2...

Ah ok. This is a bit of a leap of faith but might still give you a useful rule of thumb. Note that the m_{dpqt} term is really the source...