Restringing the Hubbard-Di Veroli harpsichord
Copyright © 2011 by Claudio Di Veroli, Bray Baroque, Bray, Ireland, April 2011
In a separate webpage we have told the saga of Hubbard-Di Veroli harpsichord.
1. THE RATIONALE. Why should one incur the risk, trouble and cost of restringing a harpsichord, even more so if well made, time-tested, resilient and nice-sounding? Because an even better sound quality will be the result. A detailed exposition on how to recreate the ancient harpsichord sound has recently been published (Irvin 2008). Needless to say, even the best modern replica of an ancient instrument cannot sound identical to the original: too many things are inevitably different, mainly the exact type of wood and centuries of ageing. However, a good replica should have a type and quality of sound approaching the original as much as possible, and the strings are paramount in tone production. Since historical instrument designs were developed to enhance and modify the output of historical wire to produce the sound desired historically, the historical sound cannot be achieved by using a different wire in an historical design (Irvin 2010).
Many ancient-type harpsichords are at present still strung with brass and steel produced following modern alloys. So was the Hubbard-Di Veroli instrument, built in 1972-1975 from a kit provided by the late Frank Hubbard of Boston, based on an original by Pascal Taskin, Paris 1769. In spite of the high quality of Hubbards kit, a few details differed from Taskin's original: harpsichord making has progressed enormously since the 1970s, and some of those details can now be significantly improved upon. For a masterclass on notes inégales in Buenos Aires in 2009, I used the beautifuland beautiful-soundingFrench double instrument recently built for Margaret Irwin-Brandon by the Argentinian harpsichord maker Leopoldo Pérez Robledo, shown in the picture below. This and other similar instruments were evidence for me that the sound of my Hubbard-Di Veroli harpsichord, however good and powerful, could benefit from further improvement. The harpsichord maker Paul Y. Irvin has verified in practice that modern harpsichords based on ancient models, but strung with modern-alloy wire, and very especially Hubbard kits, significantly benefit in sound quality by being restrung with the traditional-alloy wire now available.
Another important issue is the stringing schedule. Hubbard for his kits instructed to follow a generic "Northern European harpsichord" stringing list (Hubbard 1963 p. 6) which, unfortunately, differs significantly from any of the French 18th century extant schedules: there is thus matter for further improvement . (On section 8 below we will comment on the updated list issued in 2007 by Hubbard Harpsichords Inc.).
I started working in this restringing project in mid-2010: I had devised schedules and restrung harpsichords in the past, and Paul Irvin through a long email interchange kindly provided me with additional guidance. [Mr. Irvin did not revise the present text and is thus not responsible for any inaccuracies it may contain]. This project report is expected to help other owners who may wish to improve similarly their harpsichords.
2. THE REQUIREMENTS. This is a list of what you need to restring a harpsichord:
3. UNSUITABILITY OF ANCIENT OR GENERAL SCHEDULES. Ancient harpsichord makers, when stringing a new instrument, probably started with the stringing of a similar instrument and by trial and error improved wire alloys and sizes for some strings. Nowadays instead, when we are stringing a single instrument, we cannot experiment with alternative wires for each string: this would be expensive, impractical and certainly not good for the wrestplank. We need an "a priori" schedule of alloys and wire sizes, carefully calculated to guarantee a stringing so successful, that eventually very few stringsif anywill require further replacement. Luckily, modern musicology, technology and computers give us powerful tools that ancient makers would not even have dreamt of.
More than one reader may think: ""Wait a moment, my instrument is an exact copy of an ancient one, down to materials, dimensions and general pitch, and the ancient schedule is known! Following it is the best policy, right?". Unfortuntely it is not so because, inevitably, any modern replica departs significantly from this ideal situation. This is why:
Therefore, our instrument's final string lengths will always be some millimetres away from the original, yielding different Stresses and Pulls for every string (see THS section 8). Other factors intervene to further complicate matters:
At this point, blindly following the schedule of a similar instrumentor an average of ancient historical schedulesmay result in broken strings and an uneven or sub-optimal sound quality. We will see below the alternative way: how a schedule as authentically-sounding as possible, based on ancient sources, was devised for the Taskin-based Hubbard-Di Veroli harpsichord.
4. FINDING AN AUTHENTIC-SOUNDING SCHEDULE FOR OUR HARPSICHORD. Let me further clarify the approach followed here: based on the scaling of the instrument to be restrung, for each string, an alloy and size will be selected so that the overall Pull curve for the whole string choir matches as much as possible a historical pull curve. [There are other minor factors such as impedance that cannot be calculated beforehand: however, in the author's limited experience, only a handful of strings, if any, may need further "a posteriori" replacement to improve sound quality and evenness.] Finding the historical pull curve to be followedseparately for 8' and 4' choirsis a very important first stage, strongly dependent on the harpsichord model. We will not deal with this matter here: for the Hubbard-Di Veroli instrument I followed the curves determined in my recent Taskin harpsichord stringing (THS) essay. The historical pull curve is actually a range of pulls between two curves: the upper and lower envolvents (THS section 8).
The processbut not the final resultshown below in this text apply to any similar double-French harpsichord. For other models, the process described here and in THS can be used as a guideline for the reader in order to determine an historically-informed stringing schedule. For the Hubbard-Di Veroli harpsichord, eventually a schedule was found, optimised for the desired goals, contained within the historical envolvents and the constraints imposed by the instrument. This was not very difficult: it just required quite a few days of hard workmostly on the personal computer and attention to detail. Let us now see full details on how this was done and the results obtained.
5. THE BREAKING RISK ISSUE. Every restringing project has to include in the calculations the breaking risk R for each string. This will allow decidingstill at the planning stageabout the feasibility of the restringing project and, if affirmative, the optimal schedule to be followed. For the Hubbard French kit, it is worth noting that Taskin built his 1769 instrument for a pitch not far from A=400Hz. However, Hubbard supplied the kit as suitable for pitches up to modern A=440Hz. This did not cause issues with his "iron" strings, because he supplied modern steel instead. As for the brass strings, things were different: those at the top of the "brass range" would soon break, and Hubbard had to supply alternative modern alloyssuch as beryllium copperto mitigate the issue.
Tuning down the Hubbard double to "standard modern Baroque pitch" A=415.3 Hz certainly reduces the breaking risks, but they are still significant if we wish to restring the instrument with traditional alloys. This is because, from historical A=400 to A=415.3, we are increasing the stress of each string by 7.6%. In the Hubbard 1972 kit the situation is aggravated because the kit was designed with a treble scaling longer than Taskin's original, yielding a tension as much as 9.0% higher. The combined effects increase the stress by a factor of 1.076x1.090=1.173, i.e. a whopping 17.3% of additional stress in some treble regions. Thus, even though Taskin's original tensions for his iron wire were very safe, a Hubbard 1972 kit when strung in iron will frequently reach and even exceed a safe 80% of risk: therefore, as soon as we replace the kit's original steel strings with traditional iron wire, some strings will break frequently. To resolve these problemsin any instrument it is important to forecast breaking risks as accurately as possible. Therefore, as clarified in THS, we have to pay attention to the breaking limit B specified by wire suppliers, and even carry out our own tests: not infrequently we will be relieved to find that a coil of wire has higher resistance than expected.
Ever since Mersenne (17th century) it has been known that the Stress is independent of wire size (if we select a larger diameter for a string, in order to get the same pitch we have to increase the Pull to a point where the Stress reaches the same value as before). From this property it is easy to assume that breaking risk is also independent from wire size, but is not the case, becauseas you will find in the specs from the wire supplierthe breaking limit becomes higher as the wire becomes thinner (the so-called tensile pickup effect) . As for acceptable risk values, we have seen in THS that, for any harpsichord, a time-tested practical rule is as follows:
6. ALTERNATIVE SOLUTIONS FOR HUBBARD KITS. Paul Irvin has produced an elegant solution for the excessive breaking risks when restringing an old Hubbard French kit. He reduces the stress by retuning the instrument a full semitone lower than A=415.3, i.e. at "low French pitch" A=392 Hz. He compensates for this by simultaneously converting the keyboards, originally with range FF-g''' non-transposing, so that the instrument becomes transposing. Two alternatives are possible:
After the above conversion, the kit is restrung with a schedule devised for the A=392 pitch: the risk is now within very safe limits, and the instrument is significantly improved at a relatively low cost. Unfortunately, Irvin's solution was not practicable for the Hubbard-Di Veroli harpsichord, for many reasons:
I therefore preferred a solution that would allow for a successful restringing, yet keeping my keyboards and action unmodified, and also keeping, if possible, the present A=415.3 Hz pitch which allows playing with most present-day Baroque instruments. Following the envolvents from THS I devised a schedule and, as soon as I got to a string where the risk exceeded 80%, I decided between the two alternative mitigations:
Finally, risks were minimised by carrying out previous tests (see below section THE TESTS).
7. FINDING A FIRST SCHEDULE. For this purpose a special spreadsheet is neededbased on the formulae in section 8 of THSwhere one enters a schedule and obtains the tensions. As for alloys, one should follow historical use. Harpsichords in 18th C France had alloy transitions standardised as follows (varying up or down by no more than a semitone):
Having selected an initial set of transition points, one selects present-day traditional-alloy wire gauges. This "1st or initial" schedule may yield pull curves falling partially outside the envolvents. We resolve this by changing wire sizes, and even moving the above alloy transitions down one or two semitones in order to mitigate breaking risks in the brass. Thanks to spreadsheets it is easy to arrive, for each string choir, to a "2nd or compliant" schedule with a pull curve that
In the process we should also try to use only the wire sizes in use at the time in France:
Each spreadsheet chartone for the long 8' and the other for the 4'should include the following curves: lower envolvent, historical average (optional), upper envolvent, existing pull (if any), restringing pull. The breaking risk can also be included in the same chart, 1 Kg of pull in the axis also representing 10% of breaking risk.
8. CHANGES WITH RESPECT TO HUBBARD'S ORIGINAL STRINGING.
Lower 8' choir. We have seen above the departures of Hubbard kit's scaling from Taskin's original. Even more significant are the departures of Hubbard's recommended schedule (Hubbard 1963 p.6) from ancient ones (see the curve in magenta colour in section 10 below) and the implications in string pulls:
It is thus apparent that, for the purpose of our restringing project, we should ignore the non-historical Hubbard schedule of 1963. This was already apparent in 1974, when the Hubbard-Di Veroli harpsichord was strung with a thoroughly recalculated schedule, nearer to historical pulls.
4' choir. While the effect of restringingwith its changing pullson the Hubbard kit structure is not worrying for the 8', things are different for the 4'. In the kit at A=415.3 Hz the extreme bass has an average pull of 3.6 Kg, while restringing with historical values would average 5.0 Kg. The midrange is better but there is still one kilogram of difference for each string. Therefore, we cannot restring the 4' choir of an old Hubbard kit strictly following average historical pulls, because we would be increasing the total pull of the 4' choir by about 50 Kg. This would be borne by the bass and midrange sections of the 4' hitchpin rail ("boudin"), which in Hubbard's kit is a piece of soft poplar that is not particularly strong: the risk is to induce a significant distortion in the soundboard, affecting the functionality of the 4' bridge. The solution is a minor compromise: our "2nd or compliant" schedule still follows historical pulls, but for the bass and midrange of the 4', where the envolvents give us a relatively ample range, we select wire sizes yielding pulls nearer to their minimal rather than average values. This allows to increase the total pull over the boudin by only 13 Kg, from 208 Kg to 221 Kg (less than 6%): also, most of the increase lays in the bass range, where the boudin is more resistant because it is wider and nearer to its attachment to the spine liner.
9. THE TESTS. Once we have our new schedule ( "2nd compliant"), before buying a full set of string reels and restringing, it is good to carry out some simple tests.
A. TESTING PULLS OVER TRANSITIONS. This test shows whether the harpsichord will benefit from changes in the pull curve. It is a very simple test that does not require any change in a harpsichord, only knowledge of the existing schedule and good hearing. For every transition to a smaller wire size, you compare the sound of the two strings. The second one has a lower pull, the reduction being as small as 8% (from .024" to .022%) or as large as 27% (from .016" to .014"), 16% being the average. If the pull is what the instrument needs, the first note will be a few percentage points above the ideal, the second note a few points below, and no difference whatsoever in sound quality will be perceived. If however the first string sounds slightly better (or worse) than the other, then the string choir benefits from a higher (or lower ) pull in that region of the keyboard's range. It is now good to check that these higher (or lower) pulls are also found in the spreadsheet chart for the new schedule. The coincidence (or else) may not be easy to spot, because the new schedule may have changes in different places: what matters here is to check that there are no significant contradictions. In my Hubbard-Di Veroli, the transitions in steel strings showed no clear preference, but all the transitions in brass strings were consistent in having the first(lowest) note of each pair sounding slightly better than the other one. This proved conclusively that the instrument required higher pulls in the bass: this was also in full agreement with the new schedule, as shown in the charts below.
B. TESTING RESTRINGING RISK. These tests show the improvement to be expected by restringing with new alloys and sizes: it also allows gauging the "practical strength" of the different alloys, providing an accurate estimate of the breaking risk. For the Hubbard 1972 kit, the "2nd compliant" schedule showed "peaks" in the breaking risks curve for the following strings:
The test consisted of buying the required traditional harpsichord wire reels and restringing the six test strings listed above, keeping the old ones as a temporary "backup". The test strings were brought up to pitch in stages: first a tone below final pitch, a few days later a semitone below and days later up to final pitch. Two weeks laterto allow for the strings to settlethe following tests were performed:
Luckily, the 2nd and 3rd test above proved that my initial caveats about breaking risks for my instrument were too pessimistic, and I was able to produce a "3rd or final" schedule, less dependent on breaking risk and fully following historical pulls, especially in the 8'.
C. TESTING TONE IMPROVEMENT. As regards the tests A above, let us first note that the Hubbard-Di Veroli was strung in 1974 with a schedule significantly nearer to historical ones than Hubbard's 1963 list. Therefore, I expected for my instrument much less of an improvement in tone quality, compared with a typical Hubbard kit strung with his 1963 schedule. Therefore, I was happy with the results from the tests, showing a significant improvement in the brass, with a stronger fundamental. The improvement was found less significant in the upper half of the range. Even if iron and steel strings feel very different when handled, the change produced by restringing the treble with ironin this particular instrumentis merely a slight mitigation of the excessive brilliance of steel. For this reason I applied the "principle of diminishing returns", splitting the project in two stages, first restringing the range FF-d and leaving the completion of the restringing for a future second stage. An issue of my 1974 stringing was the transition between copper alloys and iron alloys: now the use of traditional iron helped to resolve this issue completely, and one of the welcome results of the restringing is that there are now no audible quality changes over any transition, whether in alloy or in size.
10. THE FINAL SCHEDULE. Find belowby decreasing frequencies as usual in the literaturethe "3rd or final" schedule for the Hubbard-Di Veroli harpsichord. The instrument is tuned sometimes in Standard French Temperament (Rousseau's ordinaire) others in Barnes or Vallotti: the latter was used for the frequency calculations. Wire sizes below are in inches.
The charts below show that the above schedules follow the traditional alloy transitions and yield pulls that closely follow Taskin's averages and envolvents (see THS section 8). For the high-risk red dots, 7 and 8 in the Y-axis mean 70% and 80% of breaking risk: no string reaches 80%.
It is apparent from the chart above that (a) the Hubbard 1963 schedule on the 8' choirs [ | ] yielded pulls that were far away from any historical value, (b) the schedule to which the Hubbard-Di Veroli was strung in 1974 [ | ] was significantly better, though pulls were still too low in the bass and two high in the top treble, and (c) the final 2011 schedule [ | | | ] closely follows the historical pulls.
The chart above shows that, in the 4' choir, the 1963 and 1974 schedules produced bass and tenor pulls significantly lower than historical values. This issue was resolved with the new 4' schedule: however, to avoid a significant increase in the load over the boudin, the restringing kept the pulls slightly below the historical average in the bass and tenor range. As expected, no change whatsoever was observed in the soundboard around the boudin.
Having completed the first part of the project, with the instrument fully restrung from FF to d following the above chart, a significantly improvement in sound quality has been achieved. Further, a very smooth timbre is observed, being virtually impossible to spot any of the alloy transitions in the instrument. After a thoroughly revised voicing, the restrung Hubbard-Di Veroli was successfully premiered in Bray in the Italian-French Harpsichord Workshop and Recitals in July 2011.
The final schedule shown above should fit reasonably well any harpsichord tuned to A=415.3Hz and based on 18th century French-models, provided the scaling of the instrument lies within 3% of Hubbard-Di Veroli's string lengths in mm, i.e. within the following tolerances in mm:
11. THE PROJECT PLAN. Once all the technical information is ready, the tests are completed, the final schedule is calculated, and you have all the necessary tools and string coils, you should have a project plan. This is a simple list of stages that will prevent forgetting something until it is too late, and will include details particular to the specific instrument, such as repair work or a peculiar restringing order. You may simply copy the plan below, introducing the changes or simplifications you think fit. Please note that in the text below there is no explanation about string installation and voicing: the person carrying out the restringing is assumed to be conversant with these operations. The following was the Project Plan for the Hubbard-Di Veroli harpsichord:
RESTRINGING (for each string)
Appendix. MAKING A STRING EYE. Harpsichord strings are individually attached to the hitchpins by means of loops normally called eyes. Some webpages by different harpsichord makers show how to make an eye using only your hands and the hook included in a traditional tuning hammer. This is instructive of traditional practices and looks deceptively simple, due to the skill acquired by full-time harpsichord makers. However, most harpsichord owners and tuners very seldom change a string: without the necessary practice, the traditional procedure easily yields very irregular eyes, prone to slip or break under tension.
Some string vendors carry eye-making machines, but they are not easy to operate in order to guarantee the even tension required to make a tightly- and evenly-coiled eye. I prefer a more involved but foolproof operation, that yields very even eyes and requires a minimum of tools and skills. It is an elaboration after the basic directions that Hubbard included on p.46 of his Harpsichord Kit Instructions manual (c.1970). The final eye should look like in the following picture. The real-life length of this eye is about 2cm or 3/4in.
The procedure above looks complicated, but requires a few minutes per string, is easy to follow for amateurs, and guarantees very good results.
Adlung, Jakob. Musica mechanica organoedi. Berlin, 1768, vol.II p.105. Facsimile by Bärenreiter, Kassel, 1931.
Di Veroli, Claudio. The Hubbard-Di Veroli harpsichord. Bray, Ireland 2010.
Di Veroli, Claudio. Taskin's Scalings and Stringings Revisited. Bray, Ireland 2010.
Hubbard, Frank. Harpsichord Regulating and Repairing. Tuner's Supply Inc. Boston, MA (USA) 1963.
Hubbard, Frank. Three Centuries of Harpsichord Making. Harvard University Press, Cambridge, MA (USA) 1965.
Hubbard, Frank. Harpsichord Kit Instructions. Boston, MA (USA) 1965.
Hubbard Harpsichords. Stringing Lists from both versions of Harpsichord Regulating and Repair. P.1: Typical Stringing of Northern European Harpsichord. Framingham, MA (USA) 2007.
Irvin, Paul Y. "An Approach to Recreating Historical Sound" in Harpsichord & fortepiano Vol. 12 No.2 and Vol. 13 No. 1. England 2008.
Irvin, Paul Y. "Understanding Musical Wire" in Clavichord International. Bennebroek (Netherlands) 2010.
Raymond Russell. The Harpsichord and Clavichord. Faber and Faber, London 1959, 2nd. Ed. rev. 1973.
Page last updated: 23-Oct-2016