JEUX

These pages are devoted to my favorite music, and the JEUX SoundFont, which I designed to help me create MIDI realizations of pipe organ music written before 1750. The project has grown to encompass organ music from the Romantic era as well! —John W. McCoy

The Interpretation of Baroque Keyboard Music in the Age of MIDI

"Music is funny stuff" — Eva Heinitz, Grande Dame du Violincello

"La manque d'expression est peut-être la plus grand énormité. J'aimerais mieux que la musique dise quelque chose de différent de ce qu'elle devrait dire, plutôt que de ne rien dire du tout." — Jean-Jacques Rousseau

• The question of authenticity: Bringing the music to life, versus the "Urtext" performance
• The JEUX soundfont: MIDI sounds for baroque organ music, with a French accent
• Download the JEUX SoundFont, free of charge! Version 1.4 is now ready for distribution (July 4, 2000) - List of changes
• Click here to link to the stop list and technical details for the JEUX SoundFont.
• What does it sound like? That depends on the playback parameters relating to reverb and other options specified when the piece is recorded. In this case I used a SoundBlaster Live! card on a PC running Windows ME. The sound you hear also depends on your own speakers and soundcard. Notebook speakers probably won't be able to play the low notes, but you might be surprised at the quality delivered from a notebook PC when you use a good pair of stereo headphones! The background music for this page is Buxtehude's Toccata in F Mayor, BuxWV 156, in .MP3 format. It lasts almost 7 minutes and uses many of the capabilities of the JEUX soundfont. The .MP3 format is compressed, of course! For an even better example of what JEUX can do, here's an extended recording of a passacaglia by Buxtehude. The file is in CD-quality .WAV format, well over 50 MB, so don't try this unless you are using a high-speed internet connection!
• Who's using it?

  JEUX is already being used creatively for arrangements and new compositions. I am amazed by the effects and unexpected sonorities that others have been able to coax from it; indeed, I think the exploration of this virtual pipe organ has only just begun! Further, minds more industrious than mine have also begun to show us how to create high-quality realizations of MIDI files using the JEUX SoundFont and MP3 technology. As examples of what is possible, I must give special mention to Philip Goddard's beautifully crafted compositions and their MP3 recordings, and the organ music of Bach and other delightful confections of the "Lorin Swelk Orchestra".

• Ten years of JEUX ! After working with the JEUX soundfont for a decade, I am only beginning to exploit its resources. I begin to understand why some of my teachers stuck with one instrument for most of their careers: it does in fact take years of experience to get the most out of a particular instrument, especially one with registers. For my purposes, there is no need for more registers in JEUX. What I need most is time to develop my personal art.
• Download my harpsichord realizations for General MIDI (no special soundfonts required, but results will depend on your sound card) or for the "Hpschd" SoundFont
• Download my organ music realizations for the JEUX soundfont
• My muses
• JEUX has clones!

Note! In order to make room for a lot more music, the individual MIDI files are gradually being replaced with ZIP files containing extensive collections of MIDI files. While this modification is taking place, there might be a bit of confusion on this web site, but the results should be worth the inconvenience. Let me know if you find broken links.

The question of authenticity: Bringing the music to life, versus the "Urtext" performance

MIDI realizations can hardly make any claim to be "authentic" renditions of Baroque music. In this respect, they have a lot in common with live performances. There is only one way we will ever hear an authentic performance of Bach's organ music, and that involves having a time machine. Perhaps this fact explains the numerous claims for authenticity that have been advanced over the years: they cannot be tested! The most admirable musicians understand this. Making few claims, they let their music speak for itself.

Even though the "authentic" performance will remain elusive, it is possible, of course, to increase our understanding of music of the past by studying documents, scores, and instruments that survive from earlier centuries. The satisfaction this study brings more than makes up for the impossibility of discovering "the" authentic performance. The following brief essays reflect my experience as one who has played and listened to music of the baroque era for most of his life.

More thoughts on some dreadful "Urtext performances".

Registration

The worst criticism that can be leveled at a professional musician is that he or she "only plays the notes". For reasons incomprehensible to me, many harpsichordists now go out of their way to avoid coloristic effects in performance, in spite of the undeniable fact that the history of their instrument, like that of the organ, is littered with attempts from the earliest times to add registers and effects whose only possible purpose was to extend the palette of tone colors. Sometimes harpsichordists even include defensive or apologetic statements about this in their program notes. Yet we find that even some early fortepianos were equipped with janissaries, bassoon stops, drums, and other contraptions. Are modern harpsichordists afraid someone might think the music entertaining, instead of a deep, serious exercise in abstract counterpoint? Does that mean organists(sometimes playing the very same music) should play all early music drawing only the Montre or the plenum, on the grounds that the use of additional timbres would trivialize the music of Buxtehude, the Couperins, Pachelbel, or Bach?

The music itself seems to answer these questions. Pachelbel's best-known treatment of Vom Himmel hoch features joyous celestial trumpet fanfares over the cantus firmus. This delightful idea will be lost on the listener if the figures are not played like trumpet calls. To me, the effect sounds even better using the Clairon and Trompette. On the harpsichord, if I had enough hands, I would not hesitate to use the 4' register here.

Likewise, I find in one of his settings of An Wasserflüssen Babylon a very clear sound picture (in spite of the violent and bitter imagery of the other verses of this Psalm), in which the flowing waters wash away the sadness of the captive Israelites and carry them back in memory to their beautiful city. How to communicate this to the listener? Should not the registration be allowed to contribute to the effect?

The MIDI orchestrator, of course, has more hands than an octopus. Compositions intended for two manuals and pedal can be rescored for most improbable combinations of stops that could not possibly be produced on a real pipe organ. My guide in these matters is the intent of the music, to the extent that I can discern it. If, for example, the cantus firmus seems to play against three or more interlaced voices, or wanders to notes that can't be reached because my virtual organist's hands and feet are busy elsewhere, that will not deter me from giving the cantus firmus its own manual, if the result sounds better.

Did composers want their pieces played only in one "original" registration? I doubt it! Certainly there is little evidence of this except in France. The instruments differed significantly from one church to another, and from one town to the next. They differed greatly in size. Some were old, some had just been built or rebuilt. Some were built by great masters, some by unknowns. They differed in pitch by up to a fourth. Composers traveled long distances to study, to audition for a lucrative post as church organist, to curry favor from princes, or even to flee from various difficulties. We imagine John Bull leaving England and discovering the astonishing resources of the Dutch organs. Pachelbel, by no means unusual for his times, served as organist in his native Nürnberg only briefly before moving on to Altdorf, Regensburg, Vienna, Eisenach, Erfurt, Stuttgart, and Gotha, after which he returned permanently to Nürnberg. Along the way he encountered a plague and a French invasion which hastened him to his next venue. Thus, we have to assume that composers encountered a wide variety of instruments, some "modern", some decidedly antique with entirely different stops.

Except in France, there was little attempt to communicate how a piece should sound, even when it was published, or when it was copied for a distant patron. Manuscripts, too, traveled long distances, all over Europe and even into the New World. And in France, where the tradition of indicating registrations developed in the late 17th Century, the documents that survive indicate a range of options for such registrations as "Grand Jeu" and "Jeu de Tierce." There is enough consistency to allow the formulation of "typical" recipes, and enough variation to allow for different personalities and instruments.

Many recordings of historical instruments seem intended to show how the instrument sounds, even when not strictly contemporaneous with the featured composer. Why is so much Buxtehude recorded on Silbermann organs? Not for authenticity in the sense of recreating original registration, but for the pleasure of hearing some really nice sounds that combine great music with an equally great instrument, and in a way that shows both to advantage.

The Problem of Recordings

Recordings are the ultimate anachronism of the "authentic" instrument movement, since, of course, the required equipment did not exist in the Baroque era. We understand the need for recordings, but how some of them can be said to serve the cause of authenticity is simply beyond comprehension. We have two main complaints:

1. It has become very popular in "period instrument" chamber music for the recording engineer to fiddle with the sound levels in such a way that the harpsichord or organ seems to be on wheels. When the keyboard has a "featured" passage, we hear it suddenly loom forward toward the microphone, and then when the passage is completed, or sometimes earlier, it recedes into the background, lying in wait for its next solo. Whenever we hear such a recording, we know that, whatever the musicians were doing among themselves to achieve a good ensemble, it has not been captured on the recording. Would it be so terrible to put the microphones where they belong, establish the proper balance, and then leave well enough alone? My guess is that the engineers, and probably others involved in the recording process, are still laboring under the idea that in contrapuntal music, one voice must be louder than the others in order to prevent the listener from forgetting the subject of a fugue! This idea was wrong when it was adopted by pianists, and it is even more wrong when it is applied to real Baroque instruments. Above all, it is certainly inappropriate to give the illusion that large instruments such as harpsichords and organs are moving about the stage!

2. It has also become popular to perform the solo works of Bach with exaggerated rubato, apparently so that the performer will be recognized as having a more profound and meaningful understanding of the music than anyone else. This practice has been overdone to the extent that it is hard to find even four 16th notes played evenly. Typically, the recipe for sounding profound is to make the first 16th note quite long, and the next three progressively shorter. This rarely works very well; the result reminds us of a certain famous 20th Century composer of circus music whose dances seem intended for a little man with one leg shorter than the other. Instead of communicating whatever deep thoughts were behind, say, the Goldberg Variations, this method conveys instead mainly an appalling degree of narcissism, reducing the work of the master to an endless string of mannerisms.

   I can hardly believe that today's performers somehow missed the advice that they should have heard from their teachers, "When playing Bach, go for the long lines." Too much tinkering gives the same distracting effect that we might hear from a misguided actor reciting Shakespeare, if he tried to give a different inflection to each syllable, without regard for the meanings of the words and the additional layers of meaning of the passage in the context of the scene. It is more likely that some performers so deeply believe that everything they learned about Baroque performance practice in school was wrong, that they have thrown out the proverbial baby with the bathwater.

   There are ways of applying expressive rubato without detracting from the architecture and meaning of a piece. One has only to study the recordings of Jean Langlais playing the organ music of César Franck to understand that a flexible tempo can be highly expressive in an appropriate context. But to apply any sort of rubato to the music of Bach, one has to be clear about the context. In Bach's large-scale organ works, there are many passages that are immediately recognizable as written-out cadenzas or where the improvisational quality of the passage is obvious. These sections frequently come between sections featuring a strict counterpoint or dance rhythms. The dramatic rubati are best left to the improvisational sections; otherwise, the sense of the composition is lost. More subtle rubati may sometimes help to clarify the structure of the phrases and thus guide the listener through the structure of a piece. Applying exaggerated rubati at a more detailed level is simply distracting.

Tuning

Due to the difficulty of implementing meantone tuning with the MIDI software I use (Cakewalk Home Studio 8.0), I have made few attempts to duplicate historic tunings. Different tuning recipes are legion, as are different standards (historically justified or not) for concert pitch. I wonder if some of the tunings have not been used in recent times merely as pawns in a game of musicological one-upsmanship. On the other hand, my first piano was tuned far out of perfect temperament, rather like the tunings fashionable in the 18th Century, and I sometimes miss the interesting effects this had, making the sharp keys especially brilliant and the flat keys more subdued.

The tuning that matters most for the organ, and also for the MIDI that hopes to sound like it, is that mutations are tuned to their just intervals. It is possible to get away without doing this for the Nazard, perhaps, but the mutation pitches will not blend properly into a focussed Cornet timbre unless they are tuned. That means quints are raised by 2 cents, tierces are lowered by 14 cents, sevenths are lowered by 31 cents, and ninths are raised by 4 cents. (Yes, the virtual JEUX organ has a usable Septade III, Nonade IV, and a very nice Septième VI, a sort of ultra-cornet.) The JEUX soundfont accomplishes this fine tuning in the "Melodic Preset" definitions, so that the "stops" of the soundfont can be used without further adjustment.

The other important tuning, applicable to all soundfonts, is that all the sound samples have to be configured so that they are in tune with each other. This turns out to be exceedingly difficult if one is working with real sound samples. The SoundFont file structure supports fine tuning only to the nearest cent. For very high notes, it is easy to calculate on the back of an envelope that beats will be noticeable even if samples are within 1/2 cent of each other. Also, it is possible to hear strong heterodyne effects when matching simple tones at very low pitches. These problems occur on real pipe organs, too, accounting for an old tradition of drawing only one rank of each pitch level on a given manual.

In practice, I found it best to tune each sample to the same "Flute 7-3" sound sample that the AWEPOP soundfont was based on. I continue to make adjustments. I find JEUX is acceptably in tune with at least some GM soundfonts, so that it is possible to realize the Handel organ concertos.

The design of organs always involves compromises. With respect to tuning, if the mutations are true, they will be out of tune with the harmony if the organ is based on equal temperament. If we had a simple way to accomplish meantone temperament in the MIDI world, the tierces of the Grand Cornet and the Grand Jeu would be far more resonant in most of the intervals encountered in music of the 17th and 18th Centuries. However, the Cornet is most often used as a solo stop, and the Grand Jeu is supposed to be brash. We therefore opted to have pure tierces in preference to cleaner harmony.

Ornamentation

The indications of the composers provide the best guide for ornamentation. I generally begin trills on the auxillary note unless there is a strong reason to do otherwise. Melodic contour is not such a strong reason, because the notes of the trill are part of the melody they adorn. Authentic indications for the "tremblement appuyée" of some composers are a strong reason, provided we understand that the preceding note must then be slurred into the trill—this context provides evidence for the correct articulation of a delightful little Basse de Trompette by Dandrieu. It is widely appreciated that most of Beethoven's written-out trills start on the upper note, and the reasons for this are sometimes obvious—maintaining harmonic suspense, for example. Evidently, it is not true that all trills after Bach or even after Mozart should begin on the main note! Armed with some experience in French harpsichord music, as well as that of English and German composers who used their own ornamental vocabulary, I try to tackle ornamentation in a sequencing program such as Cakewalk without sacrificing these principles. Making the switch from keyboard to PC mouse makes one think hard about the nature of ornamentation and its musical function!

One ornament harpsichordists do not have to consider is tremolo. My readings have revealed that the use of the tremulant in early organ music was far more extensive than most organists are willing to use today. Who would suspect that a French fugue would call for multiple tierce registrations and tremulant? Why do we find so many specifications like "Jeux d'anches mais sans tremblemens", unless the tremulant was ordinarily used when the reed chorus was drawn? The tremulant of historic organs was a tricky thing to adjust, and would not have been maintained through countless restorations and alterations over the centuries if it were not important for the music!

A similar train of thought makes me suspect organists have always had a secret desire for what harpsichordists consider cheap tricks, in that Nightingale stops are found all over Europe. I would like to know when they were used! Other than a number of ornithologically suggestive passages in Handel's organ concertos, I am unable to guess. And what pieces were being played when churchgoers were so delighted to hear the Zimbelstern?

Rhythmic Alterations

"It don't mean a thing if it ain't got that certain je-ne-sais-quoi." — Peter Schickele

There are many situations in the Baroque repertoire that call for changes in note values. The extent to which these changes are necessary to make a piece sound well varies greatly. In the harpsichord music of D'Anglebert, I find it necessary to alter almost every note shorter than a quaver. On the other hand, many of Bach's chorales and fugues seem to demand (or even to accept) no alterations whatever. The strength of MIDI realizations is the possibility to achieve the precise effect desired, if one has the patience to do so. Perhaps the degree of patience this requires is one reason relatively few MIDI realizations exist for the French harpsichord repertoire. For French organ music, I posit additional causes, notably the lack of a suitable "instrument", a gap I have attempted to fill with the JEUX soundfont.

About the JEUX Soundfont

The JEUX soundfont is a collection of sounds designed for use with Creative Labs' AWE32 and similar sound cards that support the SoundFont 2 file structure. I have tried to duplicate the characteristic sounds of real pipe organ stops (including individual ranks, mixtures, choruses, and effects), with a particular goal of assembling a collection of stops suitable for MIDI realizations of organ music written before 1750. Through careful selection of stops, I have found it is possible to realize music of later composers as well, for the palette of tone colors I have included is very broad. The JEUX soundfont may be freely distributed but may not be sold or used commercially under any circumstance without my written permission.

What is a Soundfont?

A soundfont is a specific file structure used by AWE32, AWE64, and similar sound cards from Creative Labs and other vendors. Sound samples are organized into "Instruments", which in turn are organized into "Melodic Presets". The "Melodic Presets" are the "patches" you specify in your MIDI sequences with numbers 0 through 127, or by selecting from a "patch list" that assigns names to this same range of numbers. In JEUX, you can think of the soundfont "Instruments" as ranks or mixtures on a real pipe organ, and the "Melodic Presets" are the stop knobs, tabs, or buttons on the console.

There are actually two specifications for the soundfont format. The current one, for AWE soundcards, uses the file extension "sf2". Soundfonts are most easily edited with the Vienna Sound Studio software from Creative Labs. Vienna Sound Studio Version 2.3 is far more reliable than its predecessors. A large number of parameters can be specified, though most of them are not very useful for building organ stops! I have tried to use the parameters in a consistent manner in order to simplify maintenance and development tasks.

If you have the SoundBlaster Live! sound card, please be sure you have upgraded to version 2.0 or later of LiveWare from Creative Labs. This upgrade is required to fix errors in the original SoundBlaster Live! software that caused distortion in some soundfonts that worked correctly on AWE 32 or AWE 64 sound cards. JEUX was one of the soundfonts affected by this problem! The upgrade can be done by downloading files from the Creative Labs web site, but be prepared for a herculean effort, as the total size of the download approaches 30 MB! That should be enough to fix a lot of bugs and create some new ones besides.

Two products of Seer Systems, "Reality" and "SurReal", were designed to use soundfonts to generate MIDI realizations directly from software. "Reality" also recorded the output directly to .WAV format on the hard drive. Seer Systems was planning release a new product, "Wavemaker", to provide a more inexpensive method to write .WAV files from the output of "SurReal". This promising product does not seem to have caught on. Other software-based synthesizers using the soundfont format are available, but none actually duplicates the sound of JEUX.

The hardward from Creative Labs is undoubtedly the best vehicle for soundfonts, but in recent years the complications of drivers and operating systems have made it very difficult to obtain a workable environment. For example, my notebook PC, running Windows Vista, turns out to have a BIOS and an internal architecture that is incompatible with adding an external soundcard that supports MIDI synthesis.

Where does this leave the MIDI musician? If the original capabilities of the SoundBlaster cards are unavailable due to the absence of appropriate drivers, computer architecture, etc., then it will be necessary to work within more limited capabilities of software-based synthesizers. However, I have some suggestions on how this might be accomplished!

Measurement of the sounds produced by SoundBlaster cards shows that a number of the parameters defined in the published soundfont standard are implemented in a non-standard way. For example, the parameters that are nominally calibrated in decibels turned out to be implemented in units of approximately 0.4 decibels in the SoundBlaster cards. Some of the filters were found to differ from the standard as well. As a result, any part of the JEUX soundfont that depends on these parameters will necessarily sound different when a different synthesizer is used. Therefore, the only available method for ensuring reliable performance of the JEUX soundfont in a wide variety of hardward and software environments would seem to be to tweak the sound samples in such a way that they do not depend on any of these parameters.

For the future, then, the most reliable method of soundfont construction must be to prepare each sample so that it can be used without any adjustment of the internal soundfont parameters. All the filtering and all the volume adjustments will have to be incorporated into the individual samples. The sound samples will have to be adjusted to an appropriate volume level before they are added to the soundfont. In the case of JEUX, this approach will significantly increase the size of the soundfont, because many of the current samples are used at different volumes or with different filter threshholds in the various "stops". The parameters of the SoundBlaster synthesizer were a great idea, but until and unless the SoundBlaster cards can be made truly portable, we will have do without them. Unfortunately, converting the JEUX soundfont to this hardware and software independent format can only be done in an environment that fully supports the parameters that JEUX now uses. Will we ever have time to undertake this project?

Sound Samples

I gathered sound samples from all over, even my own kitchen. Most, however, were available free of charge or other stated restrictions on the internet. A few were included in the AWEPOP soundfont of Arend Zwaag, my starting point for the development of JEUX. Some of the samples have been synthesized by mathematical transformations of other sounds, and many have been subjected to filterings or digital transformations to bring them into the required form (44.1 KHz sampling rate, for instance). I would very much like to find better quality samples of some of the characteristic organ sounds, such as Bourdon 16', Doublette, Gambe 8', and some of the open flutes. Such samples need to be uncompressed, with a sampling rate of at least 22 KHz, free from extraneous noises.

History

As far as I know, the first attempt at a non-commercial, comprehensive library of organ stops for MIDI wasa set of specifications adapted by Benjamin Tubb from those of the Sound Canvas Pipe Organ Project (SCPOP) developed by Raphael Diodati and Filippo Tigli (1997). The Italian stop names of SCPOP have persisted in all of the spinoffs, and I have retained a few for convenience in the JEUX soundfont, too. These specifications were of the form, to make a "Fondi 16-8" registration, combine Flute at velocity 110, Shakuhachi at velocity 70, Whistle at velocity 60, and Recorder at velocity 100, the latter lowered by one octave. These specifications seem to have been used as a starting point by Arend Zwaag, who constructed the very nice AWEPOP soundfont (with similar Italian stop names). A somewhat different approach, using only ROM sounds, was taken by Ralph van Zant in the AWEORGAN soundfont. A few commercial soundfonts are also available, most recently from Andreas Sims, though I do not know of any that have targeted the baroque organ styles. The "Church Organ" patch usually encountered in General MIDI is, in my opinion, a miserable substitute for the sound of the pipe organ, to be used only as a last resort.

How I Designed JEUX

My intent has been to develop a soundfont for use with MIDI sequences created through Cakewalk or other sequencing programs. I have not added a MIDI keyboard to my home computer setup, but I see no reason why the JEUX soundfont should not be useful in that environment as well.

The JEUX soundfont will be easier to use if the design principles are understood:

• The stops are designed to have a reasonable and consistent level of reverb. I set my AWE64sound card for "Hall 2" and "Chorus 3" effects (chorus effect is used only in the Viola Celeste 4 stop).
• Samples are tuned to each other as well as possible. However, the soundfont parameters permit fine tuning only to the nearest cent, not enough to eliminate all heterodyne effects for very high ranks (Cymbale, mixtures), or for very low notes held longer than 5-10 seconds. These effects, of course, occur in real pipe organs, too. That's why many classical registrations contain only one rank at each pitch level.
• Mutation ranks are tuned to just intervals. Experimentation showed that failure to perform this step is a significant problem in the AWEORGAN and AWEPOP soundfonts. Without this adjustment, mixtures, especially those containing the tierce, do not sound well. Quints are raised by 2 cents, tierces are lowered by 14 cents, sevenths are lowered by 31 cents, and ninths are raised by 4 cents. JEUX has a very nice Septade III, a Septième VI cornet-like combination, and even a Nonade IV.
• Volume levels have been greatly reduced from those of the AWEPOP soundfont in order to restore some balance with General MIDI soundfonts and to reduce static noise generated within the AWE sound card or within the speakers' pre-amplifier. It is always possible to turn up the volume control on the speakers! Stops are designed to be used, whether singly or in combination, at a volume level of 100 (MIDI "velocity" parameter). If swell effects are desired, use the MIDI "expression" parameter (controller 11).
• Stops are based on classic French concepts, but with provisions for German, Spanish, Italian, and English tastes by providing additional flute and reed color stops and mutation ranks. At least a few of the sound samples (notably the Bombarde and Basse de Trompette) can be traced to French organs. It turns out that constructing a soundfont suitable for organ music of the German Baroque is far easier than duplicating French sounds. I have tried endless variations of the Plein Jeu and Grand Jeu. I feel these characteristic sounds are at last recognizable, at times with thrilling results.
• The second bank contains, among others, a series of foundation and reed combinations of the type usually called for in French organ music of the Romantic era, notably that of César Franck. Depending on the demands of the music and the polyphony limits of your sound card, some care is needed in using these combinations. When chords of more than 6 notes are encountered, and more than one track is required to play these same notes, the polyphony limits of the AWE64 sound card are exceeded, and some of the top notes may not be heard. When this happens, it is necessary to select a simplified registration or omit some of the duplicated notes in the chords.

The virtual Grand Orgue has more or less normal-scale principals (and a wide-scale Nazardos VIII for Spanish registrations). The principals on the virtual Positif use narrow pipes, based on the same flute samples from the AWEPOP soundfont. The resulting contrast between the virtual manuals provides interesting effects. The reeds on the Positif are also weaker than those assigned to the virtual Grand Orgue, though there is nothing in the design of JEUX to prevent the stops from being used as if they were on other manuals! (I would very much like to obtain better samples of authentic principals, particularly notes above the treble clef, sampled at 44.1 KHz and free of noise and reverb.)

The virtual Echo manual has a unique "Gobletflöte" created from a large wineglass I foundin my kitchen. It also has a selection of suitably weak reeds. Stops on the Echo can be used "as is" for the effects required in some of the Daquin noels.

The remaining stops can be considered assigned to the Récit, Bombarde, or Pedal divisions according to the needs of the music. My limited experience with the music of the French classical organ school, as well as the recent recording of the wonderful Dom Bédos organ at Bordeaux, have convinced me of the wisdom of assigning the Grand Cornet to the Récit and the largest reeds to the Bombarde manuals, for this repertoire. In the JEUX soundfont, these divisions have some special ranks of limited range, including the remarkable Bombarde 16, Basse de Trompette 16, and Gros Cromorne 8. For a more German rendition of chorales, the Posaune 16 is fairly well behaved, though like all low reeds its speech is a little slow. Organists know to add the Prestant to compensate for this problem.

The effects department, added purely for entertainment value, features huge bells, a very nice Carillon, music box ("Petit Carillon"), Zimbelstern (notes 60-71 give the effect ofthe Zimbelstern starting up and continuing its revolution with 7 tuned bells, while notes 48-59 give the effect of the apparatus coming to a stop), and real nightingales in stereo, very effective in certain passages in the Handel organ concertos.

The samples have varying lengths. The longest are the bells and nightingales, which could be removed if you find your sound card does not have enough memory!

The Vox Humana, following modern practice, has built-in tremolo. However, some early French vox ranks seem not to have had their own tremulant, and were used in combination with other ranks, so I included an early French registration. The other stops will generally respond to the MIDI "modulation" parameter (controller 1). Some may be surprised to learn that French classical organists favored the tremulant in fugues and with the Jeu de Tierce and Cornet. Many early recipes for the Grand Jeu require "tremblement fort", sure to raise eyebrows today.

The Grand Jeu turns out to be one of the more difficult sounds that had to be duplicated. To do it right would require more layers of sound than the AWE64 sound card could handle. Further, the combination of reeds and mutation ranks brings us face to face with a characteristic of the classic French organ that is not usually encountered in modern instruments: the reeds were much weaker in the treble than in the bass, while the cornet ranks tended in the opposite direction. Therefore, the Grand Jeu registration requires both the reeds and the Cornet, sounding like a Cornet in the high range and becoming more clarinet-like in the bass. When all of these pipes get going on a real classic French organ, like the one at Notre-Dame de Guibray in Falaise, France, there is a distinct wheezing sound, probably the accumulation of reverberations, overtones not quite in tune, and a substantial amount of wind, an exciting effect that the soundfont can only hint at.

Installation

JEUX can be loaded as a "user" soundfont into any available bank of the AWE sound card. My MIDI sequences assume it is in bank 42. It will be necessary to set the "bank selection method" to "controller 0" for each track (accomplished easily in Cakewalk software), and the list of stops will have to be defined to your sequencer, either by importing a jeux.insfile or by typing in the stop names in Cakewalk. My MIDI sequences use controller messages to set the banks for you—at least it works for me!

Stronger stops or combinations may still produce unwanted soundcard noise or other effects in some situations:

• If a high degree of stereo separation is used (the MIDI "pan" parameter). This was a major problem in the AWEPOP soundfont.
• If a large number of notes is played simultaneously.
• If strong bass stops are doubled in octaves (a situation which also tends to bring out the worst performance from small speakers—the worst bass "booming" comes from low notes of the wide-scale flutes and is exacerbated by high reverb levels).
• A variety of plenum registrations are provided. They will be most useful if attention is paid to how they are constructed (i.e., wide or narrow ranks, whether 5ths or reeds are included, etc.) and to which virtual manual they belong.
• It is not advised to use stops very far above their intended pitch level, especially when the track requires notes above the treble clef. This comes from internal processing limitations in the AWE sound cards, which can raise a sample sound no more than 2 octaves above its original pitch. Experimentation will reveal these limitations.
• Compound stops such as the Grand Cornet, Jeu de Tierce, Nasard III, etc. are built from ranks of the appropriate scale, at least over most of their range. Cymbale is of intermediate scale, Fourniture of relatively narrow scale. For very high mixtures such as Cymbale, it is physically impossible to achieve perfect intonation, because soundfont sample pitch can be adjusted only in increments of one cent.
• Good speakers make all the difference! I switched from Altec-Lansing ACS-40 (limited bass response, lots of booming and distortions) to Cambridge SoundWorks, which I find entirely satisfactory. The volume knob goes up to number 10, but I have not found any reason to turn it past 4.

A Note on 3-D Sound

For cathedral organ presence, a four-speaker system (two front, two rear) will probably not be any more realistic than a normal stereo system. Real pipeorgan sounds come from any part of the organ case, at different heights as well as left, right, front, or rear. In particular, the Echo division, if there was one, was frequently housed at the top of the organ case, well suited for celestial sounds.

If MIDI supported at least two dimensions of stereo separation (right-left and up-down), we could produce a reasonable approximation of realistic cathedral presence. Until such a MIDI structure comes along, one-dimensional stereo separation will have to suffice.

A Problem with Reverb

In the MIDI environment, reverb is an "effect" added to the output of the MIDI synthesizer. Reverb is usually created by "echoing" the "dry" MIDI output signal at a fixed time delay. Whatever the delay, there will exist a group of frequencies such that the reverb effect will be almost exactly out of phase with the dry signal. For JEUX and the SoundBlaster hardware and drivers, this happens at E in the bass clef. The result is that the low E's, especially those with fairly simply acoustical characteristics (such as principals and flutes) are significantly softer than the other notes. Further, when the note ends, there is an audible "bump" when the echo signal continues to sound after the dry signal has stopped.

This effect is sometimes heard on real organs, when the dimensions of some part of the case result in reflected sound cancelling out the sound produced by a particular pipe. Whenever you hear a "bump" at the end of an organ sound, you know that something like this has occurred.

In order to produce a smoother reverb, it seems to me that the software needs to "reflect" the sound from several different logical points (thus, several independent delays) chosen in such a way that there is less interference with the dry signal. Real reverb, of course, involves echoes from large, complex surfaces. Has anyone succeeded in modeling reverb in a realistic room? Does anyone know of a VST plug-in that might produce the sort of reverb we need, and that can be added to the dry signal in a digital audio editing program such as Audacity?