just curious, ; why do you think the belly edge needs to be sharper? Or do you mean that the lower two edges each need to be as flat as possible, not rounded?
The lower two edges probably need to be thinned in the photo. Earlier, what looks flat here had been rounded -- just as one would radius a normal bows corner to reduce edge stresses. But I found that in the hollow design, the rounded edge became a stress riser in compression -- I got some small chrysals there, which I removed by flattening with a block plane (what you see now in the photo).
Cottonwood was a good teacher because it is so weak and chrysalled easily -- pointing out errors in what I was doing. I think it also demonstrated that hollow tillering definitely compensates for a weak tension wood -- the back never broke, even with the limb bent double. Cottonwood would normally be expected to to break in tension.
The actual corner with the cambered back was sharp after planing back, although it looks flat in the pic. The wall thickness at the edge was increased because of the planing, and I would probably have gouged more of it out if tillering had gone further, but I would not have radiused it.
just a thought, but if the belly is not taking strain in the conventional manner, then is there some side bending, or controlled "flapping" outwards that is taking the load?
Yes definitely, In fact I believe that was why the back did not break despite the extreme bend, and why the bow "reassembled" itself afterward. Simson has said that in his HLB designs, the sidewalls flex outwards a small amount relieving stress.
I think a less extreme camber to start with probably would have produced a bow. A nearly half circle section is asking a lot of the sidewalls.
