As a composer and a recording artist, my favorite part of thearranging process has always involved the string section. Strings arethe backbone of a symphony orchestra; they have a wider dynamic range,are more versatile, and possess greater expressive power than any othersection of the orchestra. Unfortunately, recording with a fullcomplement of strings is prohibitively expensive and highly problematicon a number of levels.
During the past two decades, many composers and arrangers turned tosample libraries to realize their musical goals, and a realistic andprotean string library soon became the elusive Holy Grail of thesampling world. I, too, dreamed of having the perfect string-samplelibrary at my fingertips, and before long I had embarked on a quest todevelop such a library.
The endeavor to produce the ultimate string library was indeed likethe pursuit of the Holy Grail. The adventure was beset with trials,tribulations, and occasional disappointments. I planned the journey,made the commitment, set out on the mission, encountered unexpecteddetours, and met many interesting and amazing people along the way.
THE JOURNEY BEGINS
Garritan Orchestral Strings (GOS) is the product of years ofcreative work. Planning for the library began in 1998 when I engagedthe Moravian Philharmonic Orchestra in the Czech Republic to recordsome of my music. (See “The Real Thing” in the October 2000issue of EM for more on my experiences working with EasternEuropean orchestras.) Earlier that year, I had purchased several of theleading string libraries, but none of them met all of my criteria:impeccable sound quality, the finest instruments available, chromaticsampling, wide-ranging multiple dynamics, numerous articulations, andgreatly expressive capabilities. Although I had resigned myself tohiring a real orchestra for the project at hand, the idea sprang tomind that someday I should develop my own string library based on mycriteria.
In scouring various newsgroups and forums, I learned that many othermusicians were also dissatisfied with their string libraries and wantedmuch more. So I began to imagine what the ideal sample library wouldencompass. Recent advances in high-resolution audio technology and theincreased memory capacity in PCs were making it possible to overcomethe limitations of existing libraries. With the advent of Tascam'sGigaSampler and GigaStudio (which was originally developed by Nemesys),I could sample the string sections chromatically at multiple Velocitiesand I could sample every articulation and playing technique imaginable.I was beginning to see the light; the odyssey had begun.
MAPPING THE TERRITORY
To guide me through my quest, I needed a reliable map. Obtaining thehelp of others who knew the orchestral and sampling terrains wouldprove invaluable as I charted the journey. I spoke with many composers,arrangers, and players, and each of them told me what he or she wantedmost in a string library. String players were consulted to determinewhich articulations to record and to explore the capabilities of theirinstruments.
I analyzed existing libraries to ascertain what was alreadyavailable and what was lacking. I consulted many other specialists,including orchestrators, string experts, conductors, sample developers,audio engineers, university professors, and sound designers to glean asmuch information as possible before the recording sessions. I evenreturned to the method books that I used as a young violin student andread various textbooks on orchestration. Another part of the researchinvolved attending concerts and listening to great orchestral stringrecordings. I also sat in on orchestral recording sessions to gatherideas about string-recording techniques. The more I sought the adviceof the experts, the more prepared I felt.
The most important attribute of any sample library is the quality ofits samples. In my pursuit of the best string samples, I knew that Iwould have to search for the finest stringed instruments. My initialplan was to record an orchestra in Eastern Europe. I investigatedseveral orchestras in various countries and inquired about theirinstruments. Unfortunately, the countries were still struggling afterthe Cold War, and the players didn't have the high-quality instrumentsthat I needed for my library. In addition, European orchestrasfrequently tune to a different standard (A 443) than Americanorchestras (which typically tune to A 440). It was clear that I neededan alternate plan.
Frank Spitznagel, the musical director for the project and a veteranartist at New York's Lincoln Center for the Performing Arts, arrangedfor a rare opportunity to record some of the world's finest stringedinstruments in a large rehearsal hall at the Lincoln Center complex. Iwas thrilled. As a young boy, I was deeply moved by the sound of aStradivarius violin played by Itzhak Perlman at Lincoln Center; littledid I know how far that inspiration would carry me.
Lincoln Center is the largest and most respected performing-artscomplex in the world (see Fig. 1). It is the Camelot of culture.To record there meant my quest was off to a great start. As goodfortune would have it, several other factors guided me along the way,including the necessary contacts, the latest recording technology, andmany good and generous people who were willing to assist me.
As with the knights of the Round Table, each string player for theGOS library was handpicked. Finding top-notch performers with fineinstruments was a challenge. Dave Eggar, the conductor, and PaulineKim, the concertmaster, worked diligently to select and bookworld-class string players with incredible instruments.
In time the instrument roster grew to an impressive size: twoStradivarii, a Guarneri, three Gaglianos (see Fig. 2), twoTestores, a pair of Montagnanas (see Fig. 3), a Vuillaume, aKlotz, a Calcanius, and a Betts, among many others. These famousinstrument makers are revered for the quality of their workmanship andrepresent the pinnacle of instrument design. Collectively, the stringedinstruments in the GOS library are worth millions of dollars. Despitebeing hundreds of years old, the instruments possess a powerful andbeautiful sound. I'm still amazed that we were able to assemble suchfine players with their exquisite instruments in the same room at thesame time. Providence was indeed smiling down on this quest.
Assembling the appropriate tools for the recording sessions andattending to the technical aspects of those sessions were vitallyimportant tasks. Choosing the right recording equipment, microphones,recording format, sampling rate, and resolution demanded considerableplanning. When you're sampling exceptionally fine instruments, you mustcapture each instrument's full spectrum and tonal characteristics oryou'll seriously compromise the samples. That meant using superiorequipment and paying careful attention to recording technique. Thesubtle details, overtones, and resonances would all be lost if we optedfor anything less.
We started with a pair of high-quality professional Brüel andKjaer omnidirectional microphones (Model 4006) and a Millennia Mediamicrophone preamp. The Brüel and Kjaer microphones are favored bymany audio professionals for recording orchestras, and they'reespecially good at capturing the upper regions of the frequencyspectrum.
Finding the ideal placement for the mics — the “sweetspot” — required experimentation and careful listening.Locating the microphones too close to the players would cause one ortwo instruments to dominate the ensemble, destroying the ambience ofthe sound. Too far away would result in excessive room noise,muddiness, or loss of detail. It would also make it harder for thesample user to control the sense of space. We therefore settled on amore moderate mic placement to optimize the audio quality and characteras well as flexibility.
In order to avoid phasing and to achieve a proper stereo spread, weplaced a Jecklin Disk between the two microphones. In addition, we puttwo Neumann KM 84 cardioid microphones farther back in the hall. (Thefour-microphone configuration allowed for possible 4-channel orsurround-sound releases in the future.) We also used a Crown SASS-Pstereo microphone to capture room ambience and impulse responses.
Once the microphones were in place, we routed the signals to anApogee PSX-100 analog-to-digital converter and recorded directly tohard drive with 24-bit resolution and a sampling rate of 88.2 kHz.Stringed instruments have a rich harmonic content that 16-bit, 44.1 kHzequipment cannot adequately record. A higher sampling rate captures thefull range of the music more accurately with less chance of distortion(or aliasing) and fewer phase problems.
Moreover, a higher sampling rate better preserves the transientresponses such as the attacks. (The high-resolution recording alsoallows the string library to stay current when 24-bit sampling becomesthe industry standard.) We opted to use the 88.2 kHz sampling rateinstead of the 96 kHz rate to avoid errors in downsampling. We feltthat a sample conversion based on a 2:1 ratio instead of a 2.177:1ratio would be more accurate (less prone to error).
I was concerned about what to do if anything went wrong,particularly in the event of a computer failure. We therefore devised abackup plan, adding a separate external archival drive and a Tascam24-bit DAT recorder to the rig — just in case.
SLAYING THE DRAGON
The thousands of samples in the GOS library were obtained after manydays of intense and demanding recording sessions at Lincoln Center. Wearrived early to set up the equipment and troubleshoot any problemsthat might arise. The conductor, the musical director, and theconcertmaster attended every recording session. Each day we informedthe players of what to expect. In preparation for the sessions, themusicians played various classical selections and some of my originalmusic while the recording staff performed the requisite sound checks.Once everyone felt comfortable, was warmed up and in tune, and had agood “group” frame of mind, the sampling began.
Dave Eggar, a Harvard- and Juilliard-trained musician, was theconductor in charge. Although I had prepared extensively, he had hisown expert opinions on common usage and on the articulations and rangesthat we should sample. He led the string sections as though he wereconducting an actual performance. He interpreted and approached theproject as a major musical work, tying the various bits and pieces ofthe project into a cohesive whole. Eggar had a good rapport with theplayers, which brought out the best in everyone and made for a betterperformance. He was brilliant, and my recording sessions were clearlyin good hands.
The musicians played through many scales using differentarticulations. We recorded myriad long bowings and short bowings aswell as additional techniques and effects. At times the players startedat the lowest note of their instruments and worked their way up thescale chromatically. We recorded as many as four dynamics levels foralmost all articulations, bowings, and playing techniques. We sampledalternative bowings and various finger positions. Occasionally, themusicians played through the circle of fifths to break up the monotony.They also played intonation notes followed by the desired target note.That technique not only allowed for proper tuning but also provided asense of connectedness from note to note. We recorded second and thirdtakes to obtain alternative sets. Throughout the process, we paidconstant and close attention to monitoring in order to obtain the bestrecording levels and signals.
At the time of recording, we had to decide how best to use thelimited and expensive time available. Of course, we had businessmatters to take care of, such as signing releases and payments, and wekept a detailed logbook of the recordings. Periodically, we tookbreaks, which helped to avoid fatigue and provided a wonderfulopportunity to talk with the players. Their patience and stamina wereamazing; imagine being a virtuoso player and having to play scales formost of the day. Overall, the recording sessions were a success. Afterinvesting so much effort and money, the major challenge was over, but agreat deal more work lay ahead.
THE NEXT BATTLE
Producing a playable sample library from a series of sourcerecordings proved to be a daunting challenge involving a mountain ofediting and programming work. It was a laborious and tedious task,requiring a great deal of care and unfaltering attention to detail.Each of the thousands of individual notes had to be sliced up, trimmed,and named — one at a time. Because multiple takes were recorded,the editing team had to choose the best takes for each note. We madedecisions and compromises on a note-by-note basis. We had to deal withfalse starts, bow scrapes, coughs, moving chairs, levelinconsistencies, mistakes, tuning problems, and more.
For editing, we used several advanced programs, each capable of24-bit, 32-bit, and 64-bit processing. When necessary, we applied noisereduction using the HDA AudioCube to process the samples with 64-bitfloating-point precision, thereby preserving the sonic fidelity. SonicFoundry Sound Forge 5 was our main editing program, though it was inbeta at the time; needless to say, this library put the program througha rigorous test. Other editing programs included Steinberg WaveLab,Syntrillium Cool Edit Pro, and Sek'd Sequoia.
Neither equalization nor compression was applied to the sound files,but we did use some normalization to provide smoother transitions inGigaStudio. As the final step, we used Apogee's UV-22 process to ditherthe samples from 24-bit to 16-bit, preserving as much detail aspossible. During the editing phase, I was fortunate to have access tostate-of-the-art equipment and to have several good people by my side.It took more than eight months just to edit the samples — aHerculean feat!
We now had meticulously edited samples of the finest stringedinstruments and the most comprehensive selection of articulations andtechniques available. Still, the journey was far from over. We knew wehad to make the samples musically playable and expressive. Mere notescannot convey the act and art of musical performance. When emulatingthe sound of a real string section, the more control options you have,the more successful you will be at creating a realistic performance. Inthe past, sampled instruments were considered the antithesis ofexpressive performance. The time had come to change thatperception.
Real string performers play their instruments in a variety of ways.For example, they bow and pluck the strings, they change articulationsand bow strokes at a moment's notice, they alter dynamics instantly orgradually over time, they slide to a note or across many notes, andthey impart a lush, expressive vibrato to a tone. What we needed was away to allow a keyboard player to emulate those and other performanceactions with relative ease. That would require an innovative solution.I therefore thought it best to assemble a round-table of experts tohelp me succeed in my quest.
KNIGHTS OF THE ROUND-TABLE
As I continued my journey to produce the ultimate string library, Ienlisted the help of a creative beta team. In the pursuit of musicalexcellence, as in other areas, collaboration often beats working inisolation. Fortunately, the modern age permits us to linkelectronically to a great many people. So I extended an invitation tomembers of the Northern Sounds GigaSampler Users Forum (www.northernsounds.com) to join a beta-testproject. Previously, I had asked the forum members what they wanted ina sampled-string library, and many shared their wishes and providedvaluable input (see Fig. 4).
Seventy-five forum members applied for the test project. From theinitial group, I carefully chose beta testers based on their knowledgeand experience. The resulting beta team consisted of 20 professionalsfrom six countries. Their backgrounds were diverse — the groupincluded Hollywood composers, educators, computer programmers, stringplayers, professional musicians, engineers, scientists, and even oneperson working on the Human Genome Project. They were a gifted andspecial team and maintained a good group dynamic throughout thebeta-test period.
With all of the creative input from the beta testers, I needed someexceptional programming and perhaps a few miracles. Enter Tom Hopkins,a fine musician, composer, and programmer extraordinaire. Hopkins wasoriginally contracted to do what we thought would be several weeks ofprogramming. As new possibilities and new horizons began to emerge,however, it soon became evident that this project was not what we hadthought it would be.
Hopkins developed a workable version of the string library, which wesent to the beta testers. They experimented, played, and composed withthe library, submitting comments and suggestions almost immediately. Wesoon established a unique private forum on the Northern Sounds site,where all of the beta testers exchanged ideas and interacted on a dailybasis. Hopkins's musicianship and programming magically transformed thelibrary into a coherent and musical whole. (For more on theprogramming, see the sidebar, “The Programmer's Tale.”)
To my delight, the beta testers were amazed by the library'scomprehensiveness and the quality of its sounds. The energy theydevoted to their appointed tasks was astounding and unexpected. Theyput the library through its paces, scrutinized every note, and evenused it in their films, television shows, and compositions. They werebrutally honest in telling us what they liked and what they didn'tlike.
We received general and specific ideas, as well as criticisms, allof which we wanted so that we could further improve the library. Bugfixes, new instruments, additional experiments, and more suggestionssoon followed. We tried to examine all of the ideas the team membershad to offer, adopting the best ones while continuing to implement newideas of our own. Some excellent ideas were simply not practical (atleast not initially), but most could be incorporated into theprogramming and were therefore included.
A period of many weeks ensued that could best be described asintense brainstorming or “idea jousting.” Three weeksbecame five months. With the help of the beta team, we began to thinkabout sample libraries in new ways. As we delved into the library, wesystematically considered what makes strings sound the way they do. Abeta member who uses the GOS library in a television series mentionedthat the strings, being as detailed as they are, tended to stand out abit during underscoring. We responded by implementing Warmth Control tolet the user adjust the warmth of the strings as needed. Some betamembers wanted tighter pizzicatos, whereas others wanted the looser,more natural-sounding pizzicatos as recorded in Lincoln Center.Undaunted, Hopkins programmed both options, allowing users to selectfrom between the two.
The beta team pointed out that one telltale sign that a performanceis sampled (or synthesized) is that each note sounds exactly the sameregardless of its context. They stressed that variation is essential toany musical performance. In response to that idea, we developed up-bowand down-bow Instruments. At first the presets were separated on thekeyboard by three octaves to permit manual alternation of bow strokes.However, that was before the development of a standalone MIDI programthat was called MaestroTools.
Jeff Hurchalla, who was one of the beta testers, brilliantlyprogrammed MaestroTools with an Auto-Alternator feature to supplementGigaStudio's built-in capabilities (see Fig. 5). TheAuto-Alternator feature lets you alternate up-bow and down-bow samplesautomatically and effortlessly, like real string players do. It alsoenables you to switch between tight and loose attacks for certaininstruments or to switch between other types of samples.
Some beta testers also observed that an important problem withsamples was the difficulty of playing legato. One beta tester posted,“If we can get [GOS] to yield an easily playable and realisticlegato phrase, we will have achieved what no one has donebefore.” Legato is a natural technique for string players, butit's difficult to emulate with a keyboard because of the lack of anaturally smooth connection between successive tones. Playing stringsamples on a keyboard usually creates an awkward, forced effect,especially if each sampled tone has a conspicuous swell.
The beta-team members put their heads together over the legatoproblem for weeks. What we needed was a method whereby only the firstnote of a legato phrase would have a full, natural attack, allowing thesubsequent notes of the phrase to sound more closely connected. Weexplored and tested several options until we came up with an effectivesolution: the MaestroTools Legato Mode feature, which overlays“masking-attack samples” and smooths out the transitionsbetween connected notes.
With over 888 postings spanning 160 threads, we continued tocollaborate, strategize, and discuss and to follow interestingtangents. There was a constant flow of ideas. We witnessed outcomespreviously thought unattainable and began to think that anything mightbe possible. It almost became a game for the beta testers to try tostump Hopkins and me to see what we couldn't deliver. We werecontinually asking, “What can we do to make this evenbetter?” The beta team became a driving force behind some ofGOS's best features, and we often relied upon the expertise ofindividual beta testers. It's hard to imagine how we could have pulledthis off without the beta team. I thoroughly enjoyed the process.
THE HOLY GRAIL
As a result of my odyssey into the world of sample-librarydevelopment, a unique and special orchestral sample collection hasemerged (see Fig. 6). While on the journey, I learned what ittakes to make a fine string library, though the newly formedfriendships were perhaps the best consequence of all. The quest,however, won't end here, because I have chosen to make this an evolvinglibrary rather than a final one. In my journey, I came to realize thatthe Holy Grail was not simply an end result. The journey itself, theprocess, was the true Holy Grail.
Gary Garritan is the creator of the GigaHarp samplelibrary and inventor of the MIDI Harp. You can reach him at firstname.lastname@example.org. Formore information on the Garritan Orchestral Strings library, visitwww.garritan.com.
The Programmer's Tale
by Tom Hopkins
When developer Gary Garritan first described to me the extent of theoriginal recording sessions, I could see that this project presented anunusual opportunity to develop the kind of string library that I as acomposer had always wanted. The challenge was to take a massive numberof raw samples and organize them into a usable and expressivecollection of specialized Giga-format Instruments.
From a technical standpoint, construction of the library'sInstruments took place at four levels. At the sample level, programmingincluded things such as tuning, editing, and looping. The note-regionlevel involved mapping the samples across the keyboard, makingnote-to-note volume adjustments, creating the amplitude envelopes, andso forth. At the Instrument level, I worked with layer crossfades,response curves, filters, key switches, and controller assignments,among other things. Finally, there was the overall organization of theInstruments into groups that were intuitive to use and that werepractical for CD replication.
With its array of programming aids, the GigaStudio Instrument Editormade the job easier. To assist in the programming process and to ensureconsistency, I devised a number of MIDI files to test the evenness ofattacks, note-to-note levels, and Velocity-split transitions. To keepGarritan abreast of my progress, I improvised short music files thatdemonstrated each new feature and sent them in as MP3 files.Eventually, these short “progress” files were posted asdemos on the Garritan Web site.
The beta-testing process led to the further refinement of theInstruments and the addition of new ones; it turned the evolution ofthe library into a real team effort. Perhaps the most interesting phaseof the project involved Jeff Hurchalla's programming of theMaestroTools utility, which allowed us to extend the capabilities ofGigaStudio.
From the start, Garritan and I were of one mind about trying toretain as many human qualities as possible in the final product.Because the library was so large (with numerous articulation and layerchoices), we felt we had the luxury of allowing many smallrealism-enhancing idiosyncrasies to remain in the library. The betatesters helped us walk the line between human variability anddistracting flaws.