What a radius does is randomize the frequencies affected by the edge over a wider range and reduces the amplitude of the affect due to diffraction. If you leave one edge as a sharp corner then you will get some increase in ripple at frequencies whose wavelengths are associated with the distance to this edge. If the edge is 4" away, then there will be small hump at 1700 Hz, a small dip at 3400 Hz, and so on. This ripple will be reduced though compared to all sides being a hard break, so they will be on the order of 2-3 dB at 3400 Hz, decreasing with increasing frequency. Will it be measureable? Yes. Will it be audible, I really doubt it. The thing about diffraction is that off-axis it can fill in these peaks and dips fairly well, and what may look a little rough on one axis may look different on another axis, and overall sound very smooth. I really doubt that any affect will be heard by leaving a top edge sharp, without a radius. In other words you may not be getting much benefit for the cost, other than a slightly smoother measurement on a given axis.
Hey Jeff Let me grab you on one of your infrequent visits. Based on your real world experiences, and simulations with your software, can you say anything about how much benefit from roundover (let's say 3/4") you lose if you just radius the sides of the baffle around the mid and tweet and not the top? The former can be done much easier with veneer than the latter, so I'm curious what the cost-benefit is. Thanks.