How to build a clavichord

How should I know? Kenneth Sparr took a course, kept a diary, and took pictures, which you can see here. There's also a diary listed on this page which I can't read because I don't have Japanese fonts loaded in my computer and also I neglected to learn how to read Japanese, but the pictures are interesting. Judy Conrad recognizes it as a kit from the Renaissance Workshop. I have put together a few harpsichord kits, but my talents, such as they are, don't lie in that area. When it came time for me to get a professionally made instrument, I contacted Owen Daly.

This isn't the place for introductory information on clavichords. If you don't know the difference between a belly rail and mousehole, Kenneth Sparr's site should help, or try here.

The instrument Owen built is modeled after a clavichord built in 1796 by the otherwise unknown maker Manuel de Sé do Carmo, Rua Direita de Santo Ildefonso, Porto (= Oporto), Portugal, currently in the collection of the Museu da música, Lisbon. This instrument has a 5-octave range and is fretted. There is a distinct tendency for large instruments such as this to be unfretted, but fretting has many advantages, chief among which for me is the ability to play earlier music idiomatically. A number of modern makers have built instruments based on Manuel's instrument; Owen made some interesting discoveries in building the instrument for me.

Owen kept an unintentional diary of sorts in the form of lengthy e-mails. Since he's a better writer than I, this "diary" appears below (with his permission), although severely edited for continuity. I have obscured some other names, since I didn't have everybody's permission to use them. Owen took all the pictures.

Pre-construction planning

First, Owen experiments with the historically appropriate material for the keyboard. A few modern harpsichords and organs have been made with chestnut keys, but we don't know of any modern clavichords.

I have always been intrigued but cautious about making a chestnut keyboard [as on Manuel's original], since the harder keys could conceivably be clattery and noisy. Well, I made a dummy chestnut key out of the one big plank of the wood I've been hoarding for years, and it is quiet and smooth as silk. Chestnut, even though it looks pretty much like oak or sassafras wood, is actually quite surprisingly light and easy to work. I found references to chestnut being tractable and easy to work puzzling until I actually cut into this stuff. But something even more interesting. The replacement dummy chestnut key, with a tangent installed, yielded a note almost twice as loud and rich in fundamental as the original basswood lever. All of this somehow related to the very fruitful discussion G***** D***** elicited regarding the role of mass in keylevers in discouraging the absorption of energy as heat by the keylever, and pushing it instead to the soundboard. Pretty damned cool....

Next, is the problem of Manuel's scaling, which has been finessed in the past, but maybe it's time to confront it.

There are some kinks to work out on the design. The scaling, as we discussed once before, is anomalous and puzzling. Was Manuel an idiot, or was there a practice we need to learn about? The scaling is quite short, really, up to well above middle c. But by c" the scaling is approaching about right or even a bit risky (on a clavichord; it would be fine on a harpsichord) for brass at A415, and at c" the scaling is the equivalent of a c" of 12.6", which is way off the charts for brass at a pitch even remotely conceivable for the rest of the clavichord. The two I built from this layout in the late '70s/early '80s, I learn from the drawings I just found, incorporated a rationalized scale, which, in itself, is fine. But another possibility is that there was a crossover to iron somewhere, say between c' and c". But the top half would still be short-ish for iron. The top-half stringing on the original is missing (!) I'm going to test some Voss iron on the monochord, and try an iron string or two on my big [unfretted] instrument at the length/pitch/gauge that would occur on the new instrument just to see what I find. Either we'll be at around 440, with a crossover to iron somewhere around c", or we'll rationalize the scale and string all in brass at around 415. And in passing, there are quite a few iron remnants on some of the other Lisbon clavichords, which Doderer ascribes to the muddle-headed meddling of 19th-century piano-tech repairers/"restorers," but one wonders.

... one simply cannot avoid the realities of Manuel's scaling. The really huge and odd #13 [in Doderer's list], expanded in 1855 or so (!), has a similar, though less dramatic, stretch in the treble with shortness in the mid-range. Several others show some extant evidence of brass up to about the middle with a crossover there into rather heavy iron. This pattern is the one thing which can be made to work convincingly with Manuel's set up. I have decided that for once one should take one's own advice and not try to outsmart the evidence. So my plan is to keep Manuel's scaling, string in brass to a little above middle c, and string, ahem, assertively. I've been using the big unfretted here [built several years ago] as a guinea pig. The other day I put .35mm (.014", which is huge) Voss iron on the choir whose length is the same as the c" on Manuel's instrument. WOW. Big, fat, strong sound. And that incredibly long 160mm c''', which raised the red flag in the first place, strung in .30mm iron just zings all over the place. And remember that with a pairwise-fretted instrument the net tensions will be not much greater than half of those on my own big [unfretted] clavichord. Now, by middle c the scaling gets quite short, and it's time for beefy brass. But at that point I started making comparisons with the relationship between pitch and length on [a smaller fretted clavichord], whose pitch I placed a tone lower than I normally might have done, and which works wonderfully well, and Manuel's lengths at 440 here are almost exactly the same (for the same absolute pitch) as on [the smaller clavichord] at 466.

Bottom - Sides - Interior frame members

sides on bottom [The camera got back from the repair shop], so finally took some pix. The bottom's douglas fir, with a thickness averaging around 23/24mm (hand-planed and it varies), and stiff as concrete. In a few days I hope to get the carcass rim glued on, the bottom edges trimmed and the keywell chunk stuck in. Then moldings around the bottom (dirty work, that) and the really dirty stuff's over.

The keywell chunk is the part of the front that goes in front of and beneath the fronts of the keys. I think the Portuguese call it the "regua da frente." Some instruments are made with a single front piece, with the keywell part cut out. More common, and more elegant, is for the soundboard front and toolbox front to be the termini of the front, with a separate piece wedged in in front of the keyboard. The case right now is plain, raw eastern white oak. Pretty pale. If we fume it, it will be browner, but it will be browner with a nice subtle oil/wax finish in any event. And will darken a lot with age. Case sides average 13mm thick. Moldings, as per usual, are the same wood as the carcass proper, so it all blends together. regua
de frente

I'll try to get a picture of those beautiful little trenails tomorrow. Also, the outer edge bottom moldings went on very nicely. Overall look of the thing really is Stickley-esque, what with the quarter-sawn oak. Oh: fuming. You put it into some kind of tent with a dog dish full of ammonia. STRONG ammonia, which is "aqua ammonia," which is 30% ammonium hydroxide solution (household NH3OH is 3%). Stuff will really wake you up in the morning. If you open the jug and hold your hand above the open neck, you can get blisters. Chemical reaction, not a stain, so there's no change in the texture or clarity of the wood.

bottom view What can I say? Trenails. Amazing 
how much pleasure these give.

The two little blocks against the spine went on today - and are actually three blocks, but you can't see the one further up by the bellyrail. All these are things for me to rest the backrail on (on which the key ends will rest) when I glue it down (next thing to make and glue in, making sure that its height above the bottom is consistent and correct). They won't be visible once the backrail goes on. Wrestplank is oak, from a humongous chunk we drove to Portland for, since I didn't gutswant to glue one up from smaller pieces. Liners are generic conifer, which is something I use a lot (in this case, of all things, some nice larch). The belly rail is tulip poplar because I had a conifer belly rail split on me once, and tulipifera doesn't do that. The liner that is inside the cheek overlaps the belly rail a tad. This spot is one where the soundboard can walk leftwards like a treble harpsichord soundboard, except for the part trapped in the cheek corner, and split with a shear crack. If the belly rail can lean or wobble leftwards this is exacerbated greatly, so the overlap cantilevers the support back behind the cheek and makes it very rigid.

The left hitchpin rail, made yesterday and glued in this morning, is basswood with an oak veneer, exactly as will be the rack/spine hitchpin rail, which is basswood where the slots are because tilia is really quiet ... with oak veneer to look good and secure the hitchpins. The latter piece of course can't go in until the keyboard is completely laid-out, so the rack slots can be laid out. left hitchpin rail

The backrail will be 20mm wider and parallel to the taper of the rack/hitchpin rail, which will sit down on top of it. A nice subtlety of the do Carmo design: the bass/left hitchpin rail is cutout (looking down) in a kind of slightly dovetail-like shape for 11 mm from the front face in the back left corner. The spine hitchpin rail/rack drops into this dovetail and this helps strengthen it against being pulled out of the instrument and flung into the faces of the audience.

go on to Clavichord, pg. 2

jump ahead to Clavichord, pg. 3