gluing up the frame

Well, now I think I understand all the things I have done wrong so far, in building this frame.  The *next* one, I'll be able to do a much better job on!  But what about this one?  The big problem is that I made the angled cuts on the ends, before the lumber was fully dry.  As it dried, the lumber changed shape slightly, and so the precision of my cuts suffered.  I felt some time-pressure due to only having access to the chop-saw for a limited time, and this in itself was probably a problem.  This stuff just takes time.  The trade-off for saving money by buying cheap, non-kiln-dried framing lumber and trying to use it for a musical instrument, is that it takes an extra six months for the lumber to be ready for use.  During the drying period, the lumber should be restrained to keep it flat and straight.  I made some efforts in this direction, but clearly not enough; I think I will need to build a special restraining frame for drying lumber in the future.  My longest piece, the spine piece, the full 8' long, developed a pronounced twist: maybe as bad as 10 degrees or so.  Well, rather than throw away all this time and money investment, I will try to make it work anyway, partially as a way to develop techniques for dealing with these problems; and even if it's not perfect, I really want to get something built up to the point that I can string it and verify whether my basic design can withstand the tension and produce good tone.

So here's my attempt to fix the twist in the spine-piece (the long one on the left).  I have been gluing the frame members together one by one.  First, I attached the horizontal member, not visible in this first photo, at right angle to the spine, which locked the front end of the spine piece into its correct orientation, i.e., at right angle to the plywood bottom.  The spine member twists CCW, going towards the tail end, so by the tail, its top edge is leaning outwards by the 10 degrees or whatever.  Here, I have attached the short crosspiece to the spine at the tail end, and then the longer of the two "bentside" segments, on the right side.  I connected these joints flat together, directly as the original cuts dictated -- as opposed to "fixing" (compensating for) the spine twist at the first joint in the tail.  Thus, the twisted spine causes the bentside member to angle upward, rather than lying flat on the plywood.  As you can see, I propped up this member during gluing, so that it would retain the "wrong", twisted position.  This now gives me a big lever (assuming the joints hold -- which is also under-test in this operation), which can be used to force the spine to twist back.  (Maybe.)

Next I glued all the other joints, of the smaller triangle close to the keyboard.  In this photo the final joint of the long bentside piece to the intersection in the middle of the bentside, has not been made yet.  You can see that I have wrapped an area of the spine timber with plastic, and inside I have paper towels saturated with water.  I will keep this wrap on for a few days or a week as I finish the gluing and put the spine under torsion, hoping that extra humidity will help bend the wood and get it "used" to its new orientation.  I don't know how wise this is, the opposing concern is that I don't really want to return the wood to its fully-wet condition like when I bought it (or worse), because drying from that was the cause of all the problems in the first place.  But anyway, we'll see, it's all just an experiment.

As the wood dimensions changed during drying, you can see how far off this formerly tight joint ended up!  As I have closed up all the other joints already, the error effectively is all transferred to this one.  I fit the two pieces of basswood, well-saturated with Titebond II, with the originally-intended single 2+1/2" #8 screw going through the whole "sandwich".  Hopefully this joint, primarily under compression and shear, will be strong enough.  Pretty, it ain't!  This will be fully enclosed and hidden -- but I still wouldn't sell this to a customer.  This is the life of a prototype...

It is exciting to see the frame, now able to stand as a solid unit by itself, rather than just the pieces laid out.

Completing the bentside 3-way joint has put considerable twist on the spine, and the twist error is reduced partially.  But still, left unrestrained, the short tail piece (naturally) does not lie flat, so there is more counter-twist to be obtained.  I plan to attach the bottom plywood, glued-and-screwed, and then weight the tail down flat while it dries.  (I have a lovely Sun Microsystems 19" monochrome CRT monitor that volunteers for this task.)  And by then I will have removed the wet-pack.  Then, I'll let the lumber, as well as the glue, dry out for a good number of days before removing the weight.  I'll work on other stuff like the soundboard while this is going on.  And... we'll just see how it all works out.

Even once the tail frame members are all flat, there will still be some twist-error; I guess I should have added a little overcompensation or "leading" into the spine-tail joint, so that it would all settle out to the right amount of twist.  But I didn't have the confidence to guess right on that, so I went with the straight, best-fitting position of the joint, i.e., as originally cut, knowing that with no overshoot I would not be able to completely iron out the spine twist.  It helps a lot; but, what about the remaining error?  And in general, the outer faces of these frame members are pretty rough and warpy.  The issue is, I want to attach the outer layer of nicer-looking lumber, my poplar 1x6s, by gluing them flat to the frame timbers.  Thus, they will contribute significantly to the top-to-bottom flex resistance of the frame, as well as adding a smaller amount of sideways stiffness.  But the twisted spine piece especially, is not suitable for such attachment, and probably other frame members would be problematic as well.  Any slight deviation from the vertical plane will mess up my beveled joints, and will just look crappy.

So the obvious solution is to plane (or otherwise machine) the sides.  That's what "real" woodworkers do, as a matter of course, in building stuff like this.  I am attempting to use the existing faces of the lumber, as much as possible, to avoid the need for "real" woodworking tools and a shop.  But this means compensating for and anticipating this reduced precision, through the design.  In this case, though, especially with the twisted spine piece, I do need to find a way to machine the wood.  Possibilities include, building a special saw-guide (or modifying/adding to the one I have) to trim a vertical plane; or maybe, I have a handheld vibrating power sander, which I could maybe build some kind of carriage for, to hold it in a vertical orientation.  Stay tuned...

Not necessarily as an alternative to machining the wood, but perhaps to complement it, in certain places where I inevitably or by-design have space in a joint, a gap, I think I'm going to have to develop a Titebond-and-sawdust mixture.  I hoped originally that, at least with small gaps, I could count on the glue to fill these.  The Elmer's type glues tend to have this filling ability; though I had noticed that their "new" wood glue formulation, which seems to be similar to Titebond II, was significantly different in consistency and behaviour, and did not seem to fill as well.  In any case, Titebond II itself has virtually no gap-filling capability.  Partially, this is perhaps simply because the glue in its liquid form is not very viscous, and can't be made to even "seem" like it's going to fill many cracks.  But even if it appears to be filling a void in the short term, I find that it usually wicks away into the wood and causes voids to be open, by the time it dries.  They're not kidding when they say that joints need to fit well and be well-clamped, for this glue to work.  The low viscosity, and/or high "wickability" (perhaps this is the same property?), clearly contribute to the strong bond capability of this glue.  I have read that it was standard practice for ancient instrument builders to mix (hide) glue and sawdust, to make a filling compound.  I think I will need to do the same, for some areas in this project.

Also: in several places in the design of my action mechanism, I rely on a bead of viscous glue to lock-in the heads of nails, fitting through holes in a rail.  This is sort of one of my design idioms.  But Titebond II will clearly not work in this application, any more than superglue -- which I tried in my dulce-melos zither project.  I guess I will still need to "stock" Elmer's Glue for these applications.  (I have this other silicone glue, which I used to attach the sharp tops in the pedalboard project, but I was not happy with it, it never seems to fully dry; it would be better as a sealant perhaps, than as a glue, at least by my definition of "glue".)