Talk delivered by Peter Bavington at a British Clavichord Society meeting at Edinburgh, 23 – 25 August 2002, and subsequently re-printed in Clavichord International, vol. 7 no. 1

A clavichord is a kind of box, and its action has only one moving part; so it seems like an easy instrument to make – easier than, say, a harpsichord or spinet. The truth is, though, that it is harder to make a half-way decent clavichord than a harpsichord. The factors in design and set-up which in combination produce either an excellent or a mediocre instrument are subtle and hard to pin down. In this talk, I plan to examine some of them, and the way in which they interact.

By ‘design’ I mean anything that is permanently built into the instrument and can’t be changed after it is made; by ‘set-up’ I mean those things which can be changed again and again without damage, such as the strings and listing.

The process of producing a note starts with the player’s finger, so let’s begin there. The finger presses on the front end of the keylever; the lever rotates about its pivot, and the tangent, somewhere near the rear end, rises to strike the strings.

Now unless the tangent is moving fairly fast when it strikes, little or no sound will be produced; yet the keylever starts from rest, and the finger has only a short distance over which to speed it up. What factors, under the control of the designer, affect the speed of the tangent?

Firstly there is the fall-back weight, arising from the way the lever is balanced, which the finger has to overcome before it can move the key at all. This can be measured by putting weights on the front of the key until the lever just moves. The heavier the fall-back weight, the harder the finger must work to accelerate the lever; so a heavier fall-back weight leads, other things being equal, to a quieter sound. Heavy fall-back weights are typically found on twentieth-century revival clavichords, which is one reason that their sound is generally so much quieter than that of old instruments.

Closely connected with the fall-back weight is the inertia of the keylever, which depends on its mass and length. Longer levers will tend to have a greater inertia, which suggests a reason why smaller fretted clavichords, with their short levers, often seem more lively and responsive than big unfretted ones. If, in a particular design, the keylevers need to be long in order to reach their strings, then they had better be made as light as possible; and indeed, this is what we find on old unfretted clavichords, where sometimes the levers seem lightweight to the point of flimsiness. The traditional carving of the back part of the keylevers may also play a part: as well as being decorative, it has the effect of reducing their mass, and hence their inertia.

The inertia also depends on how the mass is distributed along the keylever. If mass is concentrated near one end, for example in the form of a substantial tangent or a lead weight added to the lever, it will increase the inertia more than the same mass distributed evenly along the length of the lever. It seems, therefore, that to reduce inertia you should avoid lead weights and use the lightest possible tangents; however, there is another factor which works the other way.

Some concentration of mass at or near the tangent is needed to sustain the vibration of the strings, once a note has been produced. The cure for a note which ‘barks’ – that is, which produces a coarse, rather loud, sound without sustain – can often be to add mass to the lever by inserting a lead weight as close as practicable to the tangent. I wonder if we see evidence of this in the Hubert clavichord in this collection, which has lead weights in only some of the keylevers, and not always where they are needed simply to ensure a reliable fall-back of the lever.

Another vital factor in the design of the keylever is the position of the pivot, that is, where along the length of the lever to drill for the balance pin. This determines the leverage ratio, which is the ratio between the front and back parts of the lever or, more precisely, between the distance from the point of application of the player’s fingertip to the pivot (distance A) and the distance, measured perpendicularly to the axis of rotation, between the pivot and the tangent (distance B).

leverage diagram

I define the leverage ratio as B divided by A: thus, the closer to the player the balance pin is put, the greater will be the leverage ratio, and the faster the tangent will move for any given finger speed. A high leverage ratio therefore seems desirable; but again, there are factors acting in the opposite way.

A high leverage ratio, while it leads to a faster-moving tangent, at the same time reduces the force with which the tangent, once it has made contact, presses up against the strings. Now a certain minimum force is required to keep the tangent from being bounced off by the strings’ vibration: when this happens, it leads to a stifled note and the phenomenon we call ‘chucking’ or ‘blocking’. Where an excessive tendency to chuck is a fault of the instrument (rather than of faulty playing technique), I am inclined to think incorrect key leverage is the principal cause.

Another consequence of an excessively high leverage is a hard, board-like touch, over which the player feels he has no control. Hardness of touch actually varies with the square of the leverage: thus, a very small re-positioning of the pivot point can have a surprisingly large effect on the perceived hardness.

leverage of a Hass clavichord

Looking at the leverage ratio of old clavichords shows up some surprising facts. This graph shows the leverage ratio of every note on a clavichord by Johann Adolph Hass dated 1763: not the instrument in this collection, but a very similar one. You will notice at once two things:

Firstly, the leverage is higher for the sharps than the naturals. I don’t know the reason for this, but I suspect that the player unconsciously adapts to it.

The other striking thing is how the leverage ratio increases towards the treble, something again that you find on most old clavichords. This may be because the old makers had discovered that treble notes need a faster tangent speed than bass notes if they are to sound in balance. A consequence is that the treble notes will have a harder touch. However, this seems to be mitigated to some extent, in big unfretted clavichords like the Hass, by the flexibility of the long treble keylevers, which has the effect of reducing the shock of the impact as felt by the player’s finger. You will remember that I said these treble levers were sometimes lightweight to the point of flimsiness; it seems it would probably be a mistake to make them heavier and stiffer.

We said the tangent had to be moving quickly in order to produce a note. It starts from rest, and as it moves the short distance before making contact with the strings, it is accelerating all the time under the pressure of the player’s finger. It follows that one way of making the tangent go faster is to increase the distance it travels before it strikes: a longer distance will mean a louder sound.

Now a larger distance between the tangent and the strings means that the front end of the key will go further down before the player’s finger feels that contact has been made. This descent from rest to contact point can be measured at the front end of the key: I call it ‘contact distance’ rather than ‘depth of touch’, since the finger continues to descend after contact is made, and the true depth of touch is therefore greater than the contact distance.

A larger contact distance means a louder sound; but, once again, there is a limiting factor. If the contact distance is too great, players report that the instrument is hard to control, and specifically that it is hard to play softly without chucking.

Contact distance does not have to be uniform throughout the compass – it is often somewhat deeper in the bass than in the treble – but it ought not to vary in a random way from note to note. If a note sounds weaker than its neighbours, it is possible to increase its loudness by reducing the height of the tangent, creating a larger contact distance. This certainly works, but it is something of a last resort, since the player’s fingers will undoubtedly notice the irregularity. A better way may be to reduce the fall-back weight on that note.

We have come a long way with the keylevers, but before I leave them I would like to explore one other aspect.

Many clavichord keylevers must be cranked to the left or right in order to put the tangent in the correct position along the strings. If supported with a simple cloth washer, such levers have a slight tendency to rotate laterally, and perhaps in order to resist this, old makers nearly always place the pivot directly behind the point of application of the player’s finger; they often also support the key at its pivot with a lateral edge, such as a piece of leather or gut.

However, this results in the pivot being to one side of a line drawn between the fingertip and the tangent. Once the tangent has made contact with the strings, they press down on it whilst the finger, of course, continues to press on the key. The effect of this is to create a force tending to push the keylever down on the side of the cranking. The only resistance to this force is the balance pin. Not only does this increase the friction at the pivot, but it leads to excessive wear on one side of the balance pin mortice, something which can often be found on old instruments which have had a great deal of playing.

cranked clavichord keylevers

Cranked clavichord keylevers
(1) conventional; (2) pivot in line with tangent and fingertip

Perhaps it will be possible, in future, to devise a keyboard in which the pivot is directly in line with the fingertip and tangent. I suspect the extra complexity of making such a keyboard would be justified by the reduced wear and friction, and an improvement in the player’s feeling of control.

Let’s move on from the keylevers to the strings themselves. They have three sections, which in certain respects function independently:

  1. the part between the tangent and the bridge, normally called the sounding length;
  2. the part between the hitch-pin and the tangent, which I shall call the after-length; and
  3. the part between the bridge and the tuning pins, which I shall call the over-length.

The lengths of all three are under the control of the designer, and in practice they vary quite a bit. Most writing on the clavichord tends to focus on the sounding length, because it is the vibrations set up in this section that produce the sound we hear when the note is played.

Along with the material, gauge and tension of the wires, the sounding length controls the pitch of the note in accordance with well-known physical laws, which apply to the clavichord along with all other stringed instruments. One way of expressing the string laws is to say that the longer the sounding length, the tighter the wires must be pulled in order to achieve any particular pitch. If the sounding length is made too long for the desired pitch, the wires will simply break before they get there.

Now, metal wires need to be pulled up fairly tight if they are to produce a good, clear note. Anyone who has fitted a new string on to a clavichord, sounding the note repeatedly as they tune up the wire to pitch, will agree, I think, that the sound becomes markedly clearer and louder as the wire approaches its designated pitch.

Just how close to breaking point that pitch should be is a controversial question which I do not intend to pursue here, except to say that I suspect – it is no more than a suspicion at present – that for a clavichord it may be slightly below the maximum practical pitch.

It is impossible, in any case, for all the wires to be pulled up close to their ideal pitch, because this would mean impractically long bass strings. Accordingly, as we descend the compass, there is generally some foreshortening, which means that the lengths get further and further from the ideal. On some instruments the bass strings are so foreshortened that solid wires will not give a good note and overspun strings are necessary.

The choice of wire diameter or gauge is vital. It is largely independent of the sounding length: for any given sounding length, you have almost a free choice of gauges. A wire of larger diameter will require greater tension, of course, to reach the required pitch; but this is almost precisely balanced by the fact that the thicker wire is stronger. Thus the two wires, provided that they are made of the same material, will be roughly the same interval away from their breaking point.

I have to say ‘almost’ and ‘roughly’ because, in fact, additional strength is imparted to wire by the actual process of drawing. A thin wire has been drawn through the die more times than a thicker one, and thus is somewhat stronger, in proportion to its cross-sectional area. So, if a wire continually breaks when tuned to pitch, the solution (perhaps counter-intuitively) is to fit a thinner wire, not a thicker one.

Most clavichords, however, are strung in brass, and with brass this ‘tensile pick-up’, as it is called, is less significant than it is with iron.

If not sounding length, what other factors must be taken into account when deciding which gauge of wire to fit?

  • A thicker wire, because it will be at a higher tension, will produce a harder touch. It will also lead to a louder note. If a clavichord has been under-strung, fitting slightly thicker gauges can quite dramatically increase the loudness of the sound. The quality of the notes will also be affected: essentially, the thicker wire will have a smaller proportion of high partials in the sound it produces. It may be described as ‘solider’ or ‘warmer’.
  • Once again, though, there are limiting factors. As you fit thicker and thicker wires, not only will the touch become unpleasantly hard and board-like, but beyond a certain point the notes will begin to sound shorter and shorter, until eventually you produce a thump rather than a proper note.

This loss of sustain is mainly because of what happens at the place where the wires meet the bridge. A sustained note requires a steady vibration (called a ‘standing wave’) to be set up and maintained in the sounding lengths of the wires. For this to happen, most of the energy reaching the bridge has to be reflected back along the string. Only a small proportion is passed to the soundboard, to be transmitted through the air to our ears.

Now, reflection requires a difference in what is called the impedance of the string and of the bridge: if the impedances are the same, or too close, sound energy will flow freely across the junction, and it will not be possible to sustain the vibrations in the wires. Without going into the definition and calculation of impedance, you can take it that the impedance of the soundboard and bridge is generally greater than that of the wires; but the thicker the wire, the greater its impedance, until you reach the point where it begins to approach the impedance of the soundboard. Less and less sound energy is reflected with each increase in gauge, until you reach the point where it is not possible to produce a sustained note. In practice, you will want to use a wire well short of this point.

To sum up, thicker wires produce a louder sound, but beyond a certain point, unacceptably reduce the sustain.

Let’s move on to consider the other two parts of a clavichord string.

The after-length, to the left of the tangent, carries the damping or listing. The length of the after-length varies a good deal in historical clavichords: the shorter it is, the harder will be the touch. The presence of listing also hardens the touch somewhat, particularly if it is tightly woven.

The main function of the listing is to stop the vibration of the strings when the key is released, but it also seems to play a role in producing the note. Everyone who has set up a clavichord will agree, I think, that listing is vital, but exactly how it works and what kind of listing is ideal are still quite mysterious. In the absence of any systematic understanding, we are thrown back on random trial and error as the only way to achieve a good result.

It has several times been reported that too much listing stifles the sound. For example, in a letter printed in our Newsletter no. 15, Edmund Handy described the surprising result of replacing ‘ruffle’ listing with over-tight woven listing:

I was amazed to find that the tone of the instrument had been dramatically stifled… [it] had become weak and lifeless, and the strings had lost much of their sustaining power.

Much the same effect is reported in an interesting article in the May 2001 issue of Clavichord International by Paul Irvin and Richard Troeger.

Another observation is that the length of the after-length seems to affect the quality of the note. Short after-lengths lead to a brighter sound; the effect is somewhat similar to that of plucking a string close to the bridge on an instrument such as the guitar. An example is the small anonymous German clavichord in this collection, in which the bass strings, with their short after-lengths, have a noticeably brighter sound than the treble.

In both these cases it seems that the arrangements to the left of the tangent can influence what happens to the right of it, in the sounding part of the string. It is hard to imagine how this can be possible, and as far as I know no-one has yet attempted an analysis.

However, I think I can identify one factor, which depends on the fact that the damping of the wires not actually being played is not complete. You can demonstrate this by running a finger lightly across the string band without depressing any key: you will hear a short-lived sound of indefinite pitch. Now the bass strings are particularly poorly damped. They have a greater mass than the treble strings, and to damp them effectively would need more turns of the listing cloth; but actually in most designs there is room for very little listing. Players have to exercise care in playing these notes, to avoid them continuing to sound after they are wanted, spoiling the harmony.

I suggest that the unplayed strings, and the bass strings in particular, contribute some sympathetic vibration when notes higher in the compass are played. The bass strings will vibrate sympathetically not only at their own fundamental pitch but also at the octave, twelfth and so on, so most pitches in the upper part of the compass should be able to evoke some sympathetic vibration. It may only last for an instant, but it probably assists the initiation of the note.

An experiment will reveal whether this is happening on your own clavichord. Try the effect of completely damping the lowest four or five courses by resting on them a cloth-covered block: now play any music that does not use these damped courses. The chances are that you will notice a loss of quality in the sound of the instrument.

If this explanation is right, it follows that the poor damping of bass notes which is a feature of nearly all historical clavichords may not necessarily be a design fault.

I should be interested in other ideas about the mechanism by means of which the after-lengths and the listing affect the quality of a clavichord’s sound. Maybe we shall eventually be able to find someone prepared to attempt to elucidate the question by experiment.

Finally, let’s consider the overlengths between the bridge and the tuning pins. It seems to be generally agreed that these should be left undamped to vibrate sympathetically. Now a problem arises when, on some clavichords, many of the after-lengths resonate at the same pitch, or several pitches very close together. The notes which evoke this sympathetic vibration will then sound different from other nearby notes which do not.

This often happens in those clavichords which have the tuning pins in straight rows at the right-hand end of the case. The bridge is generally further from the tuning pins at its treble end, but approaches them more closely towards the bass: thus the overlengths of the lower notes get progressively shorter. Now you will remember that I said that the sounding lengths of the lower notes were generally foreshortened, that is, shorter than their theoretically ideal length. A consequence of this foreshortening is that for any given pitch a shorter length of wire is required. Now it can happen that the sounding lengths are foreshortened at exactly the same rate that the overlengths are reduced, with the distressing result that the overlengths all produce the same note.

This is the main reason, I think, why more sophisticated designs introduce an angle into the rows of tuning pins, and also do not pack the pins too closely together; this way, a variety of pitches can be obtained and the sympathetic resonance of the overlengths can be adjusted so as to reinforce each note of the chromatic scale.

Let’s review the factors we have considered in this by no means comprehensive talk:

  • Keylever length, mass and stiffness;
  • fall-back weight;
  • leverage ratio;
  • the distribution of mass along the lever;
  • the contact distance;
  • the sounding length, after-length and over-length of each string;
  • the material, gauge and pitch of the wires; and finally
  • the listing.

There are other vital aspects that I haven’t even begun to consider, such as the material, shape and thickness of the soundboard and of the bridge and soundbars attached to it. The clavichord maker has to use what skill and judgment he or she has in order to decide on all these matters. As Derek Adlam has put it in an article in the November 2000 issue of Clavichord International:

The difference between success and failure in the manipulation of these variables can be remarkably small.

I do not think a maker can avoid the use of judgment by attempting to copy old instruments because there are too many uncertainties involved. In practice judgment comes into play as soon one piece of wood is chosen rather than another.

How can a maker’s judgment be guided? In practice, a body of workshop lore has been built up which has been passed around by word of mouth. Much of this is doubtful and possibly inaccurate; certainly very little of it has been written down. Published studies have tended to concentrate on detailed descriptions of surviving instruments: these are obviously valuable additions to knowledge as far as they go, but the problem with them is that they can tell us what the old makers did but not why. We need an equal emphasis on studies which seek to develop a better understanding of how the clavichord actually works.

To return to Contents, click here.