=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- MIDI CRAFT -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- PAGE 1                               Orchestrating with MIDI                   A guide to the process and techniques               by Kevan R.Craft (c) 1993                    As MIDI music is being used more and more in professional productions for Film, Television and sound recording there has been a move towards making the MIDI music, especially classical orchestrations to sound more "real". With this in mind we'll examine the instruments of an orchestra then we'll discuss the difference between velocity and volume, and what that means to the orchestral composer, and then talk about other types of expressive control. Finally we'll discuss mixing and processing techniques. PERCUSSION AND TIMPANI The percussion section of a symphony orchestra is quite different from a rock 'n' roll drum kit. An orchestral bass drum is a big boomy thing, played with a relatively soft beater, and with less defined attack and far more decay than a rock kick. An orchestral snare is deeper sounding, with less punch, more snare rattle, and a longer decay than a rock snare. A 'military drum' is an even deeper snare, like something you'd find in a fife-and-drum corps. Tom-toms are almost never used in an orchestra, although you may sometimes see a part for ' snare drum with snare off'. Snare drum rolls are very common in orchestral music, and a true snare-roll sample is the best way to simulate it. A well-designed snare sample will allow real-time dynamic control using aftertouch or some other controller. If you, however, are stuck with a single snare sounds and want to create a roll, don't record it in step time into your sequencer or quantise it, because it will sound too mechanical. The best idea is to play the notes in by hand, at as slow speed as necessary. Use a MIDI drum pad, or if you haven't got one, play the roll on two adjacent black keys, like C# and D#, and then transpose the roll so that all the notes are on the key for the snare drum. If you must use step-time record, randomise the attacks after you've laid in the notes, smearing them between 15% and 25%. If the step-time function on your sequencer produces a constant velocity, randomise the velocities too. Whichever method you use, you can create a crescendo or diminuendo by drawing in velocity curves. Slightly randomising the velocities after the curves are drawn will make the dynamic changes more realistic. Cymbals on a stand, known as suspended cymbals, are sometimes used in orchestras, but far more common are Clash cymbals - which two cymbals banged together, also known as Piatti. For soft cymbal 'dings', you can get away with using a ride cymbal, but for those big crashes, these's no substitute for the real thing - a rock crash cymbal will sound pretty pathetic. (And don't even think of using a hi hat.) PAGE 2 Orchestral cymbal crashes have duration. A percussionist chokes a cymbal crash by pressing the cymbals against his or her body. (The notation 'Lv.' or 'Laissez viber' means 'let it ring'.) If your cymbal sample doesn't stop when you send a note off, here's how you can get it to cut off. * Set up the sound so that it is monophonic - only one cymbal note can play at a time. * Follow the initial crash with another crash with another note at an extremely low (but not zero) velocity, right at the point where you want the sound to stop. Suspended cymbals are used for playing cymbal rolls, so you can use your drum machine's large crash cymbal for them, keeping the velocity on the low side, and laying in the notes by hand like snare rolls. Orchestral tambourines have a head or skin, no just the ring of 'jangles' your everyday poseur waves around. They are played either by hitting them on the head (sometimes with soft sticks), or by shaking them. This means that the tambourine sound in you drum machine, which is no doubt a sample of a jangles- only tambourine being struck, isn't really appropriate for Bizet or Rimsky- Korsakov. If you don't have a real tambourine sample, try combining sounds: use a bit of high, damped tom-tom underneath your drum machine's tambourine, and the result will be much closer to the classical model. A shaken tambourine is trickier. One solution is to play your tambourine sample TWICE for each beat, with the two notes a few milliseconds (or tick) apart, and the first note at a lower velocity than the second. For continuous shaking, try rolling the tambourine using the same techniques as the snare rolls mentioned earlier. An orchestral triangle is a unique but sonically simple beast. Don't try to simulate it with a ride cymbal bell: that sound is much too complex. At soft volumes, a triangle is almost a pure sine wave, while at higher dynamics, it starts with a very short, metallic clank, which followed by a sine wave with several non-integral harmonics. Like cymbals, triangle notes often have defined durations, and the same technique for cutting off notes can be used. Temple blocks aren't wood blocks - they are more hollow-sounding and have a more definite pitch. So you should never use a 'sidestick' for these, but instead try to get the real thing. If you can't get a sample, they're actually quite easy to synthesize: use a sine wave mixed with a decent amount of noise, and a very short envelope. Marimba and xylophone, while they may look very similar, sound quite different. The marimba is a soft instrument, played with soft mallets, and with lots of fundamental and low harmonics in the sound, thanks to the tube resonators underneath the tone bars. Especially at low pitches, it can have fairly long decay, and soft rolls on low marimba notes, if executed properly, sound like one continuous note. The xylophone is made of much harder wood, has smaller resonators, and is played with very hard sticks - there's a lot more 'crack' in a xylophone sound, and it decays very quickly; a xylophone roll sounds like a series of discrete events. If you only have a xylophone sample and need to create a marimba sound, try these techniques. * Start the sample a little late to cut off some of the attack. * Slow the attack envelope slightly. * Add some decay. * Filter out some of the high harmonics, while boosting the fundamental. PAGE 3 Timpani samples sound in drum machines are generally of limited use. Usually you only get one or two, and they don't transpose particularly well, so you get only a short usable pitch range. You usually can't roll them, because each note cuts off the previous one, unlike real timpani, which ring. Far more flexible are multi-sample ROM or RAM timps, or even synthesized ones. Like cymbals, timpani notes often have durations, and they can be cut off with the same technique. However, be careful that you know what you're doing when using note-ons or note-offs to cut off a timp sound: sometimes you should let the sound ring, and that's not always obvious in the notation. Like snare rolls, timpani rolls should be randomised or laid in by hand, but you have to take the variable duration of a timpani note into account. In a convincing roll, the individual strokes overlap. If you are step-entering, make this so by setting the durations after the fact to 125-150% of their original values. If you are playing the roll, use the sustain pedal to keep the notes ringing. Be careful when using sustain, however, that you don't use up all your instrument's polyphony in the first 16 or 24 notes. You might have to use sustain sparingly, or not until the very end of the roll. Another useful technique is to put the timpani sound through a limiter (if you're using a multi-timbral synth, put it on its own output), so you can take advantage of the sound's wide dynamics without swamping the rest of the mix. REAL-TIME CONTROL Volume is one of the most powerful tools the MIDI orchestra has, but a lot of people don't know how to use it, or even what it is. Part of the reason is that it is often confused with velocity. The two however, are totally different animals. Velocity is part of the note-on command ((its also in note-off commands, but rarely has any meaning there). It represents how fast a key moves after it is struck, which usually corresponds with how hard it was struck. Most commonly, velocity controls the initial volume of a note, and often it controls its timbral qualities, such as brightness or attack time, as well. How velocity affects a note is entirely up to you: you can design a patch so that velocity changes the volume of the note after it speaks (the sustain portion of the envelope, not the attack), or you can design one in which it does nothing at all. The important thing to remember is that velocity is a one-shot event, specified at the beginning of the note, and after it is sent there's no way to change it. Volume however, is a continuous controller command, and as the term implies, it is used to continuously vary the volume of a sound any time - either when it first sounds, or as it plays. Volume commands override velocity bytes: if a volume command of zero precedes a note-on, the note will not sound, no matter how high the velocity is. Subsequent non-zero volume commands can then be used to fade the note in. In this scenario, the velocity can be thought of as controlling the maximum volume that sound will reach and the timbral quality, while the volume commands determine the moment-to-moment levels. Volume messages are almost universally responded to by MIDI instruments and are always responded to in the same way, although on some instruments you can program patches to ignore them. How does this tie into our discussion of orchestration? First of all, since volume is changeable over a note's duration, it can be used to create swells or fades in a way velocity cannot. Although a patch can be programmed to change in volume (through its envelope), that change will take a specific amount of time, and it cannot be altered after the note starts. volume lets you create dynamic changes of any type. level or length - in effect, an envelope with an unlimited number of segments, which can be altered in real time. Since orchestral wind and string players are constantly adjusting their volume, either for expressive purposes or to blend better with the ensemble, this lets you get very close to the real thing. PAGE 4 There's another, more subtle distinction between velocity and volume. Velocity can be thought of as corresponding with the 'dynamic1' level of an instrument, whether its piano, mezzo-forte, or fortissimo. These terms have as much do with intensity of a sound as with its actual volume, which is important, because in a real orchestra, notes with similar dynamics on different instruments will come out at very different volumes: a fortissimo bassoon is many dB softer than a fortissimo bass drum. So MIDI volume lets you adjust the level independently of the dynamic, as you were moving closer or further away, or moving a fader. So while velocity represents the intensity of the more, MIDI volume represents how much it moves the VU meter. When creating crescendo and diminuendos on top of phrases (as opposed to held notes), think about the effect you're aiming for. Do you want the instrument to sound as if its being blown or bowed harder and softer, or just to sound as if its moving closer or further away? If the former, use velocities; if the latter, use MIDI volume. If you're in mono mode (which is not a bad idea for simulating wind instruments), you can't use velocity at all with legato phrases, so you have to rely on volume. One very expressive effect is to generate a bright tone which is very soft, by playing a note with a high velocity and putting a low volume value right where it starts, then fading up the note with more volume commands. You can also use volume (or other controllers) to create a sforzando effect, in which a note gets a hard attack, then gets immediately softer, and then grows back up. There are plenty of pre-programmed sforzando brass patches available for most every synth, but maybe you don't want your sforzando to be timed exactly the way the patch envelope dictates, or you want to use the effect on a different instrument, like a clarinet or viola. In these cases, use a constant-level patch, and add the appropriate controllers after a high velocity keystroke. MIDI volume is most commonly added with a foot pedal or a slider on the keyboard. Aftertouch also makes a good controller for volume changes, because it is an intuitive way to effect a crescendo. Try designing a patch so that aftertouch is mapped to level (but be careful not to set it up so that the minimum level is zero, or your notes will never start on time). Or else record aftertouch information in a track and then, using your sequencer's controller- altering functions, change it to MIDI volume. One thing to be very careful of with MIDI volume however, is quantisation or 'zipper' noise: jumps in volume that are too large and too fast will sound very unnatural, so you'll need to pack your volume data relatively densely. On the other hand, data that is too dense can cause your MIDI line to choke up and induce timing delays, but that's a different can of worms. Follow this rule: Use as much as you need but no more. MIDI CONTROLLER FOR REAL EXPRESSION Besides volume, there are many other tools MIDI gives us to make sounds more expressive. The best ways to improve the realism factor of a synthesized sound is to give it a vibrato that changes over time, not just in depth, but also in rate. For example, aftertouch can be used to change vibrato rate while mod wheel controls depth, or vice versa. Since the apparent intensity of a note is often related to its vibrato speed or depth, aftertouch is a very natural way to control it - and if aftertouch is also controlling volume, you can get very expressive with it. Also try using foot pedals, data sliders or, if you're lucky enough to have one, a breath controller. PAGE 5 Not only vibrato but also timbre can be altered in real time. Wind instruments tend to get brighter when they get load, and string instruments noisier, so having a controller mapped to a particular layer, operator, or filter that deals with those aspects of the sound can create realistic effects. They can be especially useful when you need to make a sound emerge through a thick texture and you don't want to pump up the volume too much. Again, use real- time controllers that works for you. Ends of notes are important. Most patches can give you a fairly realistic release envelope, but sometimes you need to make it longer or shorter, and sometimes you need more to be going on than just a simple volume fade. very often before a note's level goes to zero, the timbre will get darker and the vibrato will stop. This is particularly noticeable with soft wind sounds. (To get a really good pianissimo flute fade-out, drop the fundamental tone first, and fade out the breath noise note that crescendos and then stops abruptly can be made much more dramatic by increasing the vibrato and brightness right up to the cut-off. The more control you have over all the parameters of the sound, the more realistic you music will be, MIXING Creating good orchestral sounds is only half the battle: you also have to mix them. How do you place the different instruments in the stereo (left-right) and reverb (near-far) perspective can go a long way towards helping the sense of realism. Since the strings are the largest instrumental group of the orchestra, they should be spread wide. Ideally, you would use separate channels or instruments for the different sections, in which case you should arrange them according to the conventional orchestra layout. SEE ENCLOSED 'Orchestra.Pic' (Figure 2.) for reference! (If the individual sections are in stereo, use the same panning, but make the stereo spread on each section very narrow.( It should be noted that some conductors and composers put the first and second violins on opposite sides, with the violas, cellos, and basses in the middle. If your piece is highly conceptual, and the first and second violins have equally important parts, this arrangement can be very effective. SEE ENCLOSED 'Orchestra.pic2' (Figure 3.) If all your strings are one one channel use a pitch-based stereo spread. In this case the high notes should be on the LEFT and the low notes on the RIGHT. Add reverb to taste: some synths don't need it, while others, which start out more 'in your face', can benefit greatly from it. Woodwinds like to be more centred - maybe 10 o'clock and 2 o'clock on the mixer. Individual instruments should be placed carefully, and stereo samples of flutes and clarinets are totaly unneccessary in an orchestral context. Woodwinds also can take a little more reverb than strings, to put them further away from the listener and help them blend. Brass can be spead out more, although French horns sound best right up in the middle. Again, if you're using a single channels for the whole section, pan the high notes to the LEFT and the low notes to the RIGHT. Lots of reverb on these instruments helps them blend and gives them more of a concert-hall ambience. Keep in mind that at lower frequencies our ears' sense of direction is not a good, and so won't help the image much, and may actually lead to trouble further down the road if your music is going to be broadcast (where mono compatibilty is always an issue) or cut into (remember?) vinyl. Timpani and bass drum, like other low instruments, work best panned towards the centre, while the rest of the percussion family can be spread around, athough not too far, lest they sound like they're being piped in from another planet. A cymbal climax is more effective if it doesn't come far off to one side or the other. PAGE 6 Stereo pan positions can be set on the synths themselves as part of each patch, or they can be determined at the begining of the sequence by sending the appropirate PAN controlller (#10) message on each track, if your synths respond to that. Zero is left, 64 is centre and 127 is right. Some instruments use 0 or 127 for disabling or reversing external pan control, or for setting up a random pan, so watch out for those. DON'T change the pan position of things in the middle of the piece. Instruments of the orchestra do not normally get up and walk around (at least not while the tape is rolling), and neither should yours. If you are using nothing but single-timbre synths, the process of panning and mixing instruments is relatively straight forward. With multitimbral units however, you have to make decisions about output assignments before you start mixing. If the unit has only two outputs, then you must try to create a stereo spectrum as best you can, and more or less forget about different delays, reverbs, or other processing on different voices. If the synth has multiple outputs, try to use them in stereo pairs: for example, the strings on one pair, the woodwinds on another, and the brass on a third. Percussion should be divided up into as many outputs as possible because timpani and bass drum, as mentioned earlier, might want some limiting, the snare drum might benefit from a diffuse reverb, and the wood blocks and tambourine might be happier with a bright, sharp reverb. If you don't have enough outputs, you can combine sections: the strings and woodwinds can go on the same output pair, with the winds panned closer to the centre. In a pinch, the brass can survive going out of a single output, but they will be helped greatly if you have a quality reverb that can create a good stereo image from mono. A solo instrument, like a concert-style violin or piano, should be panned right up the middle. If there is more than one soloist, they can each be set off a little away from centre, but should be ballanced around the centre. Reverb on solo instruments should be used sparingly: the illusion should be that these instruments are up at front to the listener. If you use too much reverb, they will sound like gigantic prehistoric monsters lurking behind the trombone section. SMALL ORCHESTRAS AND BIG BANDS The same principles apply to mixing acoustic-style instrumental ensembles other than symphony orchestras: use level, panning, and reverb to recreate the way the group looks on stage. generally speaking, melody voices can be spread fairly wide left and right, as long as they ballance each other, while bass and rhythm instruments belong in the middle. Soft instruments should have less reverb so they sound up front, while louder instruments can have more ambience around them. A string quartet is arranged in the same way as a string orchestra: from left to right, first violin, second violin, viola and cello. A wind quintet has the flute and obow on one side, the horn in the middle, and the clarinet and bassoon on the other. A mixed chamber enemble will often have strings on one side, with the highest-pitched in the front, while the horn, if there is one, sits at the back in the middle. A larger ensemble, such as for a baroque concerto, will put the strings in the fronts, then winds behind them, and the harpsichord and bass in the back or off to the sides. There's a lot of call these days for the sound of big jazz band. Big bands are set up so that the saxes (five of them) are in front, the trombones (four) behinbd them and the trumpets (four) in the back, on risers. Each section takes up the same amount of physical width, so in your MIDI big band, they can be spread about equally in the mix. SEE ENCLOSED 'Big Band.pic' (Figure 4.). Less reverb on the saxes gives a better definition, while more on the trumpets makes them more dramatic. bass should be in the middle, and piano and guitar can be off to the sides. PAGE 7 Big band drums are the exception to the 'what-you-see-is-what-you-hear' rule. on stage, the drums are usually on one side, but that sounds strange on a recording, so you should put them in the middle. How much you spread them out will depend largely on how much you want the drums featured: if you want to sound like Basie's band, they can be fairly narrow, but if you have delusions of being Buddy Rich, make 'em wide (and much too loud). BEYOND THE ORCHESTRA Now we've discussed how to construxt a 'real' orchestra, we'll now attempt a little deconstructing. Today's MIDI eqiupment can go a long way towards simulating an orchestra, but it can also go a long way beyond that role. In the rush to please clients, prefessed desire for 'realsim', let's not forget the flexibilty that electronic instruments give us. Those of us who have grown up with orchestral sounds in our ears can be counted on to have certain predictable emotional reactions when we hear them. Putting those sounds into contexts we're not used to can have a striking effect. Composers like Stravinsky, Messain and Varese had to ask orchestral players to do unconventional things with their conventional instruments to creat new sounds, but electronic musicians have available to them a palette of literally millions of unique and ecocative sounds, each as accessible and easy to play as any realistic sound. The juxtaposition of familiar sounds with unfamiliar ones can induce a strong reaction in the listener. Try mixing instruments that wouldn't normally be heard together, like a harpsichord and a gong, or a low, loud flute and a high, soft trumpet. USe vocal samples mixed in with instrumental ones to give a 'human' effect (Gustav Holst was one of several early 20th Century composers who used vowel singing vocalists as orchestral instruments). Double a natural sound, like a sampled violin, with an unfamiliar one, like a bell with a slow attack. Create a stereo image in which an oboe is imense (lots of delays and stereo spreading) while a tuba is tiny (no processing and low volume). Set up a harmonic pad using an exotic, shifting, evocative background, and play melodies in front of it on ordinary instruments, like piccolo or xylophone. Remember that Orchestral music was once only producable by symphony orchestras, and hundreds of years of listening to them have influenced the Western ear tremendously. With MIDI gear you can do a very convincing re-creation of an orchestra, if you pay attention to the details that make up the sound. And once you've learned how to do that, you can add in other sounds, giving your music a sense of dimension and context that you couldn't accomplish any other way, and that Stravinsky can only dream about.. Kevan R.Craft 12 Mount Road Halton Runcorn Cheshire. WA7 2BH. England. U.K. Tel: +44 0928 563762 4407 Words =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- MIDI CRAFT -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-