Rafter question

Hi folks,

In teh Charles McRaven book on building Hewn log houses he makes a claim that on std roof rafters, a typical install with a knee wall would cause the middle section to bow out (of the wall) due to roof loading. One could use a collar tie of sorts to keep the rafters & walls from spreading.

So he suggests an alternative that you can add a timber from B to A that would transfer the load which spreads the walls to the ends. The ends being knee walls are tied together. Resulting in the red triangle in the plane of the roof.

Thoughts?

It looks to me that this would both stiffen the roof and help direct the load to the corners. how would the timber be positioned?
under the rafters? half lapped through the rafters?

He had the timbers positioned as small short timbers between each rafter as well as half lapped at times. What got me thinking about this more was to consider putting the roof decking (1" boards) on the same angle across the roof for the same effect.

would be good to hear from one of the engineering type on this question.... it could be like engineering on pegged braces, which in my experience engineers do not count as support for a span. But if you stand back and look at a pegged brace in a frame, there is definitely the impression that the brace would carry weight from the beam to the post. Looks aren't everything?

My concern about using many short timbers between the rafters is fit and precision. How far would the plate bow outward before all of the sections of the A-B compression beam come into full, load-bearing contact after shrinkage of the rafters? Other than that, as long as the A-B timber is in, or close to, the same plane as the rafters, I don't see why this wouldn't work.

I'm not a PE, just a mech eng and TF ameture.

Dear Bruce: As an engineering type, I agree that this will work if executed properly. On a small shed with plywood roof sheathing and no collar ties, the plywood is acting as your A-B and B-C members. In a larger building, your member C-D and the corresponding member at the opposite side would have to be designed as collar ties resisting a significant amount of tension. Also member A-C would experience a significant amount of tension. This scheme is not typically done for two reasons:
1. The connections must be engineered to accomodate the large loads and given geometry.
2. There is a lack of redundancy. If there is a failure of either tie, the roof will soon follow.
Given these drawbacks, it can be done if the situation warrants it. It may be cheaper though to install a monster ridge beam.