The finished instrument

Photo is by my colleague, photographer and professional wind player Colin Savage:



Sorry I blurred the name, but I'm not giving the game away at this late stage :-)

Audio examples

Here, at last, are three sample tracks demonstrating the three possible stop combinations of this harpsichord. Files are high-bitrate MP3s; you may be prompted by your web browser to permit your default media player to play these tracks. The examples need a moment before they're ready to play.

Dietrich Buxtehude (c.1637-1707)
Excerpts from Suite in C Major, BuxWV 230

Allemande (back 8' register):


Courante (both registers):


Sarabande (front 8' register):



Notice that there isn't a huge difference in the sound between the two individual 8's: this is because their plucking points are quite similar. Instruments that have a 4' stop usually place the 4' jacks between the two 8's, and this extra separation of the plucking points alters the sound of the 8's more noticeably.

Slideshow

Here's a slideshow of every picture posted on this blog since the beginning. You'll need Adobe Flash to see these pictures.



To see any slide as a still picture, just click the slideshow. You'll see the current picture on its own in a new window; an album of all pictures can be seen by clicking View All at upper left in that window.

Final thoughts

In no particular order:

• Although the woodworking part of this project was complete back in late August, I already had a hunch that there would be no definitive conclusion to my labours, in the sense of putting down the final tool and proclaiming "Done!" Instead, the inevitable tweaking that was necessary to coax the best out of the sound and action continued on until late November, gradually tapering off until there seemed to be nothing left to do. So, after more than two years of work, things ended not with a triumphant bang, but a whimper of sorts—an enjoyable one, of course. In any case, I understand that professional makers also go through a period of tweaking after the major part of the work is done, until such time as they at last release the finished instrument to its new owner.
• From a woodworking standpoint, I learned (and here I can imagine the seasoned woodworker saying "Well, duh!") never to cut anything to its final dimension until as late in the game as possible. Eventually I clued in to the fact that I could cut something a whisker on the long side and sneak up on the correct fit by planing or sanding. But until then, I was forced to discard a couple of pieces of otherwise-good lumber because I had cut them a bit short during an attempt to nail the correct final dimension with just one cut. I did actually manage a correct cut in several crucial circumstances, and now I'm amazed I pulled that off, given that failure would have severely inconvenienced me. Beginner's luck, I suppose.
• The previous point has taught me great respect for the hand plane, which can easily nick off a thousandth of an inch when it's well sharpened and properly tuned up. I expect to spend time improving my hand-planing skills.
• It should be clear that this instrument is by no means a faithful copy of the original Trasuntino, for several reasons. Firstly, it doesn't use exactly the same materials, which should have been cypress for the case, soundboard and bridge, to name just one departure, or the same building techniques; secondly, most of the decorative details are different on my version; and thirdly, the process by which I laid out my instrument does not entirely replicate the methods used by the old builders, who designed their instruments using geometric constructions, deriving the dimensions from simple ratios of some basic quantity. The width of the keyboard, for example, was often used to determine the length of the cheek and the overall length of the instrument. String spacing between consecutive notes was typically 1/2" (the old Italian inch was a little larger than today's inch, being a bit over 26 mm, and varied slightly from town to town). But one aspect of this project that is in accord with the old way of working is the idea that whatever materials are easy to obtain are the ones used in making the instrument. In Italy, woods like cypress, pine, maple, chestnut and beech were widely available, so that's what they used. I went with cedar, poplar and walnut because I could get those easily: they have a track record in various schools of harpsichord-building, even if some of them weren't used in making the original Trasuntino.
• I've been keeping track of how much I've spent on tools and supplies, but out of respect for professional makers, I won't release any figures for the material costs of the supplies that go into making a harpsichord. The professionals deserve compensation for their time and experience, and the difference between the materials and the final price might look too much like a straight mark-up to an unsympathetic reader.

FINISHED

The instrument works. It's playable, the action is reliable, and it even sounds pretty good for a first attempt.

So, with that said, I declare this blog finished, apart from a few final posts to summarize things I've learned, present a few concluding thoughts, display a slideshow of all photos of the construction process, plus (soon) a few audio tracks of the instrument being played.

Final adjustments

In the past two months, I've continued along the lines suggested by the previous post, partly restringing the back 8' register in the lower regions and voicing the whole instrument down a bit. As was the case for the front 8', this has changed things for the better. Both rows of jacks now play well and repeat reliably.

With the overall strength of voicing altered, I tweaked the separation between both rows of jacks by revisiting the jack end screws. I ended up giving each another half-turn, so the back 8' now sits beneath the strings by the equivalent of 4 half-turns, and the front 8' by 8 half-turns. This gives a tiny bit more slack to the mechanism.

It's important to recognize that the primary means of controlling the interaction of the two registers when both are turned on is by means of consistent voicing, and not indiscriminate cranking of these end screws. The screws are meant to adjust how long it takes for the plectrum to rise up and contact the string from beneath. There may be small differences here and there due to the quill angle in the tongue, which are a result of the quill mortise punching process, and an extra half-turn or two of the end screws can correct for this, but that's the extent of their usefulness.

More debugging

I've been working on and off with the Trasuntino for the past two months, trying to get a handle on the troublesome repetition in the tenor and bass regions. What stymies me is that the jacks work perfectly well in the upper two-thirds of the instrument, yet they become troublesome about an octave below middle c'. No harpsichord will repeat perfectly right down to the bottom of the keyboard, but my problem areas are much too high up for me to feel completely comfortable playing the instrument.

After soliciting suggestions from the harpsichord list, I learned something very interesting that pertains to my difficulties. In a nutshell, the longer a string is, the thicker it must be in order to maintain a reasonable amplitude of vibration. "Reasonable" in this case means an amplitude that is not difficult to damp out when the jack comes back down. If the amplitude can't be controlled, repetition will be compromised.

The list experts reviewed my stringing schedule and a consensus developed that the schedule was too light in the lower regions by at least one, if not two, gauges. Several suggestions were offered that advised me to end with gauge #1 in the lowest regions (0.020"). One maker said that anything less than this would behave like a "loose clothesline". Since heavier wire must be under greater tension to reach the correct pitch, the resulting increase in stiffness means the wire won't flap around as much when vibrating.

Part of my stringing problems stemmed from a slight misunderstanding of mine regarding how stringing by numeric progression works. The original instruments on which this was practiced had keyboards extending up to d''', and starting higher by those extra two notes would allow heavier gauges to be used by the time one reached E in the bass. The Trasuntino only goes up to c''', which means I finished with gauge #3. I should have started with 9 strings of #9, and not 10, in order to get the correct results.

The new schedule has the crossover points between gauges occurring on higher notes than the current schedule. In adjusting the stringing to conform to the new schedule, there will be minor changes in the alto and treble, while the lower part of the instrument will require extensive restringing.

The changes higher up are so minor that I decided not to bother with them: besides, this isn't the problem area. So I applied the new schedule only from c' downwards. Because the crossovers occur higher up in this region, I was able to recycle a lot of existing wire by removing strings and tuning pins together, shortening each string slightly, and reinstalling it about a fourth higher. By reusing wire, I saved myself the trouble of depleting my wire supply, and also gained the advantage of using wire that has already stretched out over many months and developed its tone.

From F# downwards, new wire in gauges #2 and #1 was required. Since the numeric progression rule dictates that 2 wires of #2 and 1 of #1 are called for, there wouldn't be enough to reach C unless I wanted to use even thicker gauges , and excessively thick wire has its own problems (it can sound dull or otherwise strange). So I bent the rule by using 3 of #2 and 4 of #1. The short-octave note GG/BB was left unaltered, as it was originally strung quite heavily (.022" red brass).

I tried this schedule out on the front 8' at the beginning of October, and found that the stiffer wire did indeed improve repetition noticeably, without completely curing the problem. I decided to give the new wire several weeks to settle down and develop its tone.

Now, at the end of October, I've concluded the new wire sounds good, so I'll keep the changed schedule. Recently I've been revoicing the plectra in the restrung region, as they are not so well matched to the new wire. My sense of the register as a whole is that my initial voicing back in August was too heavy, so I went ahead and thinned all the quills in the entire register. With a lighter touch, the repetition has improved even more, to the point that I think the problem is almost cured. Next, I'll do the same to the back 8'.

Amplitude and string vibration have a more crucial effect on the playability of a harpsichord than I initially suspected. Originally I thought that, provided the scale was correct for the intended pitch level, one could string a harpsichord with purely sonic considerations in mind, but I now see that the mechanical reliability of the action is impacted by the string schedule too.

Debugging phase

Construction of the harpsichord is essentially done. Only the music desk remains to be made at some point.

The harpsichord has been playing since the end of August. I'm now in a kind of debugging phase in which I play the instrument and track down things that need improvement. So far I have

• increased the depth of touch about 1/32" by substituting a thinner cloth for one of the two layers under the jackrail
• shimmed up the keys with more punchings at the balance pin to help increase the depth of touch
• moved both of my gap spacers which, despite my best efforts, had not been located exactly under the strings and were slightly rubbing the nearest jacks
• tweaked the off positions of the registers to make sure the plectra cleanly miss the strings
• glued little cloth squares between the registers so they don't touch themselves or the edges of the gap
• done a little remedial voicing to quills that feel stiffer or sound louder than their neighbours

The main outstanding issue to be resolved is the troublesome repetition of notes in the bass. In this region the amplitude of the plucked strings is large. Damping can be problematic as a result, and the tongue is often flung backwards quite energetically when the jack descends and the plectrum touches the string. So I am experimenting with damper shapes, stiffer springs and various other factors in an effort to resolve this problem.

Dampers

Dampers are installed after the preliminary voicing, as it is easier to see the quills without them.

Dampers come in the form of a strip of felt about 1 cm wide and several feet long. A portion is pushed into the damper kerf of the jack and cut off from the back, leaving the damper a bit long. To determine the correct length, the damper is trimmed from the front while the register is in the off position (i.e. further from the string). If the damper sits on the string in this position, it will also work in the on position (closer to the string).

It's important that the dampers not be excessively long in the off position; otherwise, when the register is turned on, the damper may reach past its own string and damp a neighbouring one. This also means the register's off position cannot have the plectra too far away from the strings, otherwise by the time the register reaches the on position, the dampers may be too close to a neighbouring string.

I'm using so-called flag dampers, which are more or less square in shape and flat along the bottom edge. These dampers allow the jacks to hang slightly from the strings, which makes transposing the keyboard easy. If the jacks actually sat on the keys, the resulting weight and friction might interfere with transposition.

The bottom edge of these dampers should be just slightly above the tip of the plectrum. This little bit of slack keeps the action reliable during changes of humidity.

Voicing

Voicing, in the context of harpsichord-making, refers to the cutting of plectra (also known as quills) in a manner that draws forth a musically enjoyable sound from the instrument and feels controllable to the performer's fingers.

The same term is also used to describe the adjustment of organ pipes after manufacture. Unlike organ voicing, however, the voicing of harpsichord plectra cannot fundamentally transform the sound of the instrument. The tone is largely determined by various construction factors and the stringing materials, not by how plectra are cut. Volume and touch can be affected, along with the degree of shrillness, but not much more.

Historically, raven and goose feathers were the preferred materials for quill. Today, real quills are still used by some makers, while many others use Delrin or Celcon plectra. Both of these are hard, slippery plastics that have a long working life. They don't sound exactly like real quill, though they can get close if worked on by a knowledgeable voicer.

The process begins with ghosting, or setting the silences. Jacks are placed into all register slots, black delrin plectra are installed into all tongues, and the side-to-side position of the register is adjusted until the tongues are about 5 mm from the strings. The plectra are then trimmed from the tip until they just brush the strings when the jack is raised, without actually plucking in the normal manner, though in the bass they are allowed to weakly pluck the strings to adjust for the wire thickness. This process yields a consistent plectrum length of around 5 mm. As each quill is trimmed, the tip (as viewed from the side) is given a chisel-shaped profile to assist the backward tilt of the tongue when the jack descends.

Next, the register is advanced forward so that the plectra project past the strings by about one string diameter. This is the correct "on" position. The correct "off" position is set slightly behind the position at which the plectra ghost the strings.

With the register turned on, the tongue springs in each jack are adjusted so that the tongue tilts back as the jack descends. The severest test is to let the jack down extremely slowly: if the plectrum doesn't get hung up on the string, then the jack will work properly in every playing scenario. If the spring is too firm, stopping the tongue from tilting adequately, the tongue spring is bent back slightly and tested until the jack descends reliably.

Voicing can now begin. The intention is to produce plectra that pluck the strings consistently but not so strongly that the tone is harsh. To the fingers, the plectrum should be neither too firm nor too weak. A gradation in touch from bass to treble is required. The bass needs stronger quills because the string feels like it stretches somewhat before it gets plucked, due to the deeper plucking point in this region. In the treble, the short strings and close plucking point makes things feel stiff, so a weaker quill is needed.

This process can be expedited by using quills of differing thicknesses. Before ghosting, I experimented with 8 different quill thicknesses, discovering where it helped to change from one thickness to the next. I erred on the heavy side, since of course a heavy quill can be thinned to maker it weaker, but nothing can be done to strengthen a weak quill. Ultimately I used 6 different thicknesses, changing approximately every tritone to a thinner quill.

Voicing requires controlled cuts on the sides, arrises and underside of the quill, all of which thin the quill and reduce its stiffness. When done sensitively, the resistance of the quill as it raises the string seems to yield slightly just before the pluck, which I find feels quite controllable to my fingers.

The customary tools for voicing are shown below:


A set of wire cutters helps to trim excessively long plectra from the back. The hemostat acts like a set of pliers for installing and removing quills. Either a #11 scalpel blade (shown above) or a #11 Xacto knife blade are used for cutting the plectra. A voicing block—in this case the hard end-grain surface of an ebony block—acts as a little cutting board.

Once both registers received their preliminary voicing—a lengthy process that took two days of on and off work—the jack end screws were adjusted to locate the plectra a consistent distance below the strings. I first unscrewed each screw considerably, lengthening each jack to the point that the plectrum was actually above the string even with the jack fully lowered. Then I screwed back in until the plectrum just barely slipped under the string. At this point, I gave the rear row of jacks (the "back 8") 3 more half-turns, and the front row (the "front 8") 6 more half-turns. This difference separates the plucking instant between the two rows so that, if they are both turned on and played together, they won't pluck at exactly the same time. Otherwise the touch would be extremely heavy.

With these preliminaries out of the way, the instrument was played for several days and any discrepancies were corrected by replacing plectra, shaving down plectra, or adjusting the jack end screws further. These screws should not be cranked around much once they are set, but sometimes one or two more half-turns make a positive difference.

Keyboard hold-downs

As the keyboard was designed to transpose, the key frame must stop slightly short of the lower bellyrail to allow side-to-side motion. This means there's a bit of front-to-back slop when the keyboard is in place. In addition, the light weight of the keyboard means it might not stay still if the instrument is ever moved, which often involves flipping it onto its spine.

I made two hold-downs that attach to the baseboard inside the keywell, just behind the front rail of the key frame where the head stop cloth is stapled down. Here's one of them:


It's basically a scrap block and a mirror support clip used to attach mirror glass to walls. When the keyboard is in place, the mirror clip ensures the key frame is kept fully forward. It overlaps the rear edge of the front rail, holding the keyboard down:


The block can pivot out of the way if the rear screw is loosened. This releases the keyboard and allows it to be removed.

Jackrail supports installed

Both jackrail supports were glued down after the finish was applied to the case. Where the supports were to go, the gluing surface was protected with masking tape, as I felt it would be easier to varnish without the supports getting in the way. So here's what the spine side support looks like glued down and with the jackrail in place:

Case finish applied

I've been busy working on the harpsichord this past month, but remiss in updating this blog. Quite a bit of work has been done: in fact, the harpsichord is practically finished.

The exterior of the case, and all remaining interior portions, were finished with two coats of paste varnish. After drying, each coat was buffed lightly with a finishing pad (a kind of steel wool substitute) to smooth and level it.

Cap molding

The cap molding covers the top edge of the case, overhanging about 1 mm on each side.

Here's the profile used for the cap molding:


It's based on the same profile used for the bridge and soundboard moldings, except it has two decorative edges instead of one.

Originally I had planned to apply a set of moldings to the inside upper edge of the case, but I discovered, before getting to that point, that Italian harpsichords don't necessarily have an inner molding unless they're made in the false inner-outer style (i.e. a set of veneers are glued to the interior of a painted wooden box to make it look like a separate wood-coloured instrument resides within the box). It's a good thing I didn't go ahead with this, not only because of the workload, but because the resulting cap molding would have been extremely wide. As it is, the molding has to cover and overhang two thicknesses of case material, which requires a width of 3/8" plus 2 mm.

Before gluing the cap molding down, I cleaned up the case edges with a special edge-trimming plane. Since the case edge is so narrow, I thought a normal hand plane would be hard to keep flat. The built-in fence on the edge-trimming plane helps the plane cut a square edge.

I steam-bent the bentside cap molding to get it approximately to the right curvature, then glued it down:


Tiny 1/2" brads were nailed through the molding and partly into the case edge to help resist the shear forces of the molding trying to straighten out. I think I will clip off the heads of these and drive the remainder of the shaft below the surface of the molding, just to provide some extra holding power.

The brads also help to act as positioning pins so that the overhang of the molding is equal on both sides. There's no easy way to keep the molding in the correct position without using positioning aids of some sort.

All the other cap moldings were done the same way, except that the brads were pulled out once the glue dried. Glue alone should be enough to secure moldings that aren't bent to shape.

Next, I'll swell the nail holes with water to close them up as much as possible, sand the molding, and plug any remaining holes with yellow wax.

Cosmetic details

A little fragment of molding decorates the back edge of each keywell block:


This piece of molding covers the slightly inaccurate joint between the bentside and cheek:


An identical piece covers the bentside-to-tail joint. I found these case joints tricky to make absolutely tight, without gaps: something to improve upon in future. In the meantime, these little sins are now safely hidden.

Keywell blocks

Since the keyboard is narrower than the space it fits into, the remaining space is taken up with a pair of keywell blocks plus a transposing block. The keywell blocks are shown below, glued up against the decorative front edges of the case:


In order to make sure the first and last keys don't rub against these blocks, 2 mm shims were glued to the left and right edges of the key frame:


The removable transposing block, shown below, allows the pitch of the instrument to be changed by shifting the keyboard sideways to fill the space left over when the block is removed. The block is held in place with two rare-earth magnets that are attracted to nail heads inside the keywell.


The keyboard shifts by one jack position, which allows a pitch of A=415 Hz when to the left and A=440 Hz when to the right.

Determining the proper thickness of the keywell blocks is an important part of making sure that the keyboard is properly aligned with the jacks: if not, the jacks could slip sideways off the key ends. I did this by progressively planing the blocks down and putting them in place, then sliding the keyboard left and right and observing how the jacks sat on the keys in both the left and right positions. When everything looked good, I glued the blocks in, then planed down the transposing block to fit the leftover space.

Register controls

The on/off positions of the registers are controlled by capstan screws at the left and right sides of the harpsichord. Here is the set located at the cheek:


The thickness of the brown block is the same as that of the jackrail supports, which ensures the screw heads won't be blocked by the supports.

By inserting a thin rod into the holes, the screws can be turned in and out. The register ends butt up against them, controlling their sideways motion.

To locate these screws I had to cut down the registers, which were still a bit too long, and this meant I had to have a fair idea of where the on and off positions would be. I determined this by progressively trimming the register ends and trying my prototype jacks in various slots after each cut.

Shifting the registers on and off is controlled by this pair of register levers on the spine side of the instrument:


The handles are brass lamp finials, used to screw down lampshades in a decorative manner. They fit into two brass bars, which I drilled for the necessary screw sizes and polished. At the far end, the registers were drilled for #8-32 machine screws which self-tapped their holes as they were screwed in.

Since the jacks will face in opposite directions, the on and off motions for the registers are opposed. Moving both levers in the same direction silences one stop and turns on the other, while pinching the levers together turns both on.

Jackrail

The jackrail sits over the register gap and keeps the jacks from flying out of the instrument when the keys are pressed. Its ends slide into grooves milled in the jackrail support blocks.

I made the jackrail from a scrap piece of the quartered western red cedar used for the baseboard, and glued pieces of Alaska yellow cedar to the three visible sides. Then I routed a decorative molding along the top edges, using the same bit that provided the profiles for the soundboard moldings and bridge:


The jackrail is 3/8" thick. The ends were reduced to 1/4" to fit the jackrail support grooves.

On the underside I'll staple one or two layers of 1.5 mm felt cloth to stop the upward travel of the jacks. The cloth doesn't need to absorb the full force of playing the keyboard because the head stop under the keys takes some of that impact too. What's needed is for the jackrail to arrest any further travel of the jacks once the keys have stopped moving. This avoids the unpleasant feeling of the jacks hopping upwards, bouncing off the jackrail and landing back on the key ends, which results in an odd jiggling sensation under the fingers.

Jackrail supports

The jackrail keeps the jacks from flying out of the instrument when it's played. Two supports glued to the case walls above the register gap hold it at the correct position.

Now that the jacks are done, I've checked to see how high they rise when the keys are fully depressed. I will set the jackrail position based on this measurement.

Here's one of the two jackrail supports:


Both supports are made from Alaska yellow cedar about 11 mm thick. A 1/4" groove in each will receive the ends of the jackrail. I plan to cut off the non-scrolled end at a 45 degree angle, which will look nicer.

Damper kerfs

The jacks need a slot for the felt damper that mutes the string. This is a simple matter to take care of at the bandsaw:


A stop block controls the length of the kerf, which reaches just below the position of the plectrum mortise. Instead of centring the kerf I positioned it closer to the edge of the jack. This keeps the jack body from bending inwards into the tongue slot when the damper is inserted: instead, it will bend outwards slightly.

The kerf must be tight enough to hold the damper securely; otherwise the damper will work loose over time as it thumps down onto the string.

I tested the inexpensive 1/4" bandsaw blade I had and found its kerf was too wide to provide a good grip. So I used a little trick I read about to reduce the set of the blade teeth: I ran the saw while pressing lightly on the teeth from both sides with a pair of grindstones. After making lots of sparks and several test cuts, I succeeded in getting a kerf of about 0.85 mm, down from the original 1.37 mm. This is tight enough to hold the damper without shifting.

Here is the first prototype jack to be complete in all respects (tongue, axle, spring, end screw, damper and plectrum):


Note that the top of the slot was eased with a triangular file to help insert the damper more easily.

Now it's just a question of completing the jacks by tapping the bottom holes and installing the end screws. I won't cut any more dampers at this time because the damper lengths depend on having all the plectra installed and voiced first, so that the register on and off positions can be properly determined. Voicing must wait until the project is basically done: it's the last thing to do before playing and enjoying the finished instrument.

I am indebted to the renowned organologist and harpsichord maker Grant O'Brien for publishing, on his web site, a detailed description of the construction of several Neapolitan harpsichords, including drawings and photos of his jacks, which I have shamelessly borrowed from. I adjusted his jack dimensions to suit the cutters and router bits I could easily get, but my final product is not significantly different from his except in the choice of wood and my decision to use only one damper, instead of a pair.