Kevin, fellow framers,
I don't work in the same tradition as most of the people here, therefore some of my comments may be of limited usefulness; all the same I thought I would offer them. Please excuse my presumption in that regard.
There is some use of housed dovetail joints in Japanese timber work- let's say it is one option at least - however, like most of the people posting ahead of me, I agree it is an inferior joint option in most if not all circumstances.
The dovetail promises something it really
can't deliver, most especially when green timber is used. You might say that the dovetail joint has excellent
branding, but like the New Coke, perhaps, doesn't quite deliver on what it appears to be selling.
The 'promise' is that the dovetail will act to keep together the girts or beams in between which it ties, that is, resist tension. It also offers
convenience in assembly.
It doesn't deliver on these promises for a variety of reasons:
- in green timber, one can expect shrinkage in service, especially in that first 1~2 years as the building dries out. The performance of the dovetail is absolutely dependent upon a tight fit of the tenon cheeks to the mortise side walls (which are end grain, largely). As the timbers shrink, the tenon will shrink disproportionately more across the width of the joist than the mortise will (in that same direction as the joist) due to their grain orientation. The
slightest shrinkage however, and the fit is loose, and with it the mechanical 'benefits' of the joint are largely gone. A joint with slop allows easy movement and is more likely to fail than a joint which maintains a tight on-going fit between surfaces.
- even
if one is using dry timber, the dovetail joint is problematic in terms of what it is
supposed to do. If in the life of the building there is ever some sort of spreading load moving the floor girts/beams in the frame apart from one another and the joists with their dovetail tenons are thereby loaded, what happens is that the end grain walls in the dovetail mortise, which are harder and less compressible than the side/face grain surfaces of the dovetail cheeks, will simply
crush the sides of the tenon as it tries to withdraw. Even if the load on the frame is not tending to fully separate the involved pieces, or is seasonal/occasional, the crushing of the side grain when load is applied causes the tenon cheeks to take a compression set, and thus the fit is worsened anyway.
Further, after a small amount of shrinkage in the tenon, most of which will be manifest at the fat end of the dovetail and less at the neck of the dovetail, the only portion of the mortise which will be doing anything is the sharp corner at its narrow end, and this small amount of material can fail in tension perpendicular to grain. Some woods are particularly vulnerable to such loads, Douglas Fir being one example.
Unfortunately, despite all appearances, the single dovetail tenon really offers limited resistance to withdrawal - worse than most other joinery options for that type of connection at least.
- a further negative about this form of joint is that it does not exploit the neutral axis of the receiving timber, but rather chops excessive material out of a critical compression surface (the top of the girt/receiving beam), and I say this especially when the dovetail tenon connection is compared to another relatively poor joint, the drop-in joist. A much better connection, in terms of the neutral axis issue, for cases where the joist and beams are flush with one another is the tusk tenon, though it is not so overtly convenient for frame assembly. That 'convenience' however really boils down to design, in terms of how the frame will be assembled, and I would say as well that in-service long term performance of the frame trumps a few extra hours here and there during the brief assembly process.
I imagine many people might choose a dovetail tenon for its apparent convenience when it comes to assembly, in that the joist can be installed after the surrounding frame members are installed.
However, given that fact that it offers no real structural performance advantage, is not really going to do much to resist tension loads, that it weakens the receiving timber excessively (or requires that the receiving timber be up-sized to compensate, which is wasteful), comparing the two connections the drop in simple housed connection is to be preferred.
Additionally, the drop-in joist may be strengthened to give it reasonable tension-resisting performance by the addition of an angled timber screw or two.
If I might offer some design points...The dovetail can be improved somewhat, in my view, by a few modifications, though I would, personally, still avoid it.
The drier the wood the better, the longer the tenon the better, and the narrower the tenon the better (to a point!). A wider dovetail, as is often illustrated in books, will shrink more across the width of the timber than a narrower dovetail will shrink. To give an example from outside the Western (er, North American) tradition, a Japanese dovetail on a floor joist, say on a 6" wide timber, might be 1/5 ~ 1/4 the width, or 1.2" ~1.5". A longer dovetail (I mean in the vertical orientation) offers more surface area and potential strength than a shorter one, however this approach would normally be paired with a joist that was deeper in section than width (as is structurally logical considering the vertical load that floor timbers are meant to resist).
A longer dovetail lends itself also to tapering, so that the fit tightens as the joist is lowered. This fact can be exploited, in some instances, by having the assembled joist sit slightly proud of the receiving beam, so that drying in service will cause it to shrink and drop down into a flush position with the surrounding beam, and still be reasonably snug in fit (one of course would need to consider the floorboard fastening situation carefully if this was the plan).
The wedged dovetail. This somewhat offsets the shrinkage problem described earlier, by pre-compressing the wood in the tenon, however, as an earlier poster pointed out, the wedge cannot be easily accessed after the floor goes on above it, so it is only useful for the period up to the floor going on, and in most cases that is a short time interval and the frame will have hardly done any drying. As far as mechanical performance goes, the wedging only adds very slightly, again by compressing the side grain of the tenon, to any hope that the joist's dovetail will resist withdrawal. Further, in order for the wedge to achieve an evenly-tight fit against the dovetail tenon cheeks, it would be better practice to slightly taper the opening (i.e., the tenon cheek and/or the mortise wall) into which the wedge is to be driven, along with tapering the wedge itself, and this adds time and cost to the work, and if the cut out is not really clean, it only increases the odds of the joint having less-than-ideal mechanical performance.
A
better reason to use the wedged dovetail tenon, in my view at least, is that it allows for a
horizontal assembly - the joist can be slid in along its length and then wedged, rather than dropped in. In some circumstances, this might offer a handy solution to a frame-raising difficulty (say some place where there was inadequate vertical clearance to drop any timber into place on a beam). However, that said, the dovetail joint is still better avoided and the design of the frame thought out so as to not require such joints.
As far as using a connection for floor framing which needs, for some reason, to resist tension, then a better option, in my opinion (and with the given that the situation will allow it, considering building/floor/ceiling height issues, etc), is to put the joists on top of the beams/girts (again, as another framer pointed out previously in this thread), and employ some form of partial cogged lap joint, with the halves secured to one another with a timber screw (or with a wooden connection if desired). That's likely to be the strongest connection, and is simple and quick to cut out, as an added benefit. If the floor joists must be flush with their receiving timbers, then the tusk tenon is the suggested choice in most instances. My $0.02
the Carpentry Way (my blog)