Is it real or is it synthesized? PAUL FARRER passes on some tips to help you inject some real feel into your MIDI programming.

 

Morphing a kazoo into a male voice choir, or mapping the pysically-modelled shape of an orchestral hit onto the parameters of a sample of your granny knitting, may well be the kind of tricks that equipment manufacturers like to dazzle us with. But one thing that we continually demand from sequencers, sound modules and samplers is their ability to sound more 'real'. For the kind of work I do, the majority of my time is taken up with trying to encourage my 19-inch, rack-mounted box of circuitry, SIMMs and SCSI boards (otherwise known as a sampler) to talk to my over-stuffed source of many compatibility headaches (otherwise known as a PC) in such a way that the result will sound just like a 40-piece symphony orchestra, or a guitar trio, or maybe even just a rock 'n' roll band. A fruitless waste of time, or an exciting programmer's challenge? Read on...

DIFFERENT STROKES

One of the first things to bear in mind when thinking about MIDI instruments is that they are absolutely nothing like the real musical instruments they're pretending to be. The player/instrument relationship between a violinist and their violin, and the keyboardist playing a violin sample via MIDI from a keyboard, is so wildly different that it would seem silly to pretend otherwise. Both require discipline and musicality to coax the best sound possible from each instrument, but there the similarity ends. Once we acknowledge and respect this fundamental difference, I believe the process of trying to emulate real instruments via MIDI will become much easier.

PIANO FORTE

Preset number 1 in nearly every synth and sound module since the year dot has been the piano. This must be the most duplicated, sampled and synthesized sound in MIDI history, and yet many of us still find it hard to make what we programme into our sequencers sound like a real pianist. The main trick is to try to utilise all the MIDI controllers that are relevant to the particular sound you are trying to re-create:

• PEDAL POWER: The number of people who still try to play a MIDI keyboard without a sustain pedal is astonishing. This is the main expression controller that a piano has, so to try and program piano parts without one is a sure-fire way to make your performances seem flat and one-dimensional. As MIDI controllers go, sustain pedals are both versatile and extremely cheap (I picked mine up for a fiver from a bargain-bin at a music store!), but it's worth remembering that pianos are not the only thing to benefit from the sustain pedal. Long, sustained synth pads work very well if you play a chord, push the sustain pedal, then move your fingers to the next chord (whilst the notes are still playing); then, just before you play the second chord, release the pedal. Repeat this process as many times as you need to and this produces an almost seamlessly legato chord sequence. Don't worry if you've never used a pedal before -- you'll be surprised at how easy it is to pick up, and once you've used one, you'll wonder how you ever managed without it.

• PUMP UP THE VOLUME: The volume pedal is another valuable piece of MIDI gear that is often overlooked. Swelling brass or string parts is great fun with a pedal, and this saves hours of fiddling around in a sequencer's real-time MIDI Generator page getting the crescendo you want. Bear in mind that, particularly within an orchestra, the ability to increase and decrease volume (often very quickly) is one of the keys to a good arrangement. Although pop music generally has less dynamic variations, the volume pedal can be an invaluable tool, particularly for live work, where a high degree of control is needed but you don't have time to fiddle with countless edit screens.

IMPERFECT PERFECTION

No instrumentalist (no matter how good they are) plays exactly in time or in tune. There will always be tiny imperfections and inaccuracies in every performance, and that is what gives a piece its emotional content. One clever way of adding these imperfections artificially is to try and picture in your head not only the kind of instrument you are emulating, but also the player and all the relevant playing styles. General MIDI has given us a couple of pointers in this respect, adding 'Fret Noise' and 'Breath Noise' to the list of GM sounds. But why stop there? If you're recording an acoustic guitar through a conventional microphone, there are loads of unwanted scrapes, string buzzes and fingerboard thumps that, whilst being the bane of the recording engineer, can be the salvation of the MIDI programmer. I recently sampled a selection of these thumps and scrapes and dropped them at random places throughout a sequenced nylon guitar part. Providing they aren't too loud, the effect can be quite dramatic. I even managed to fool a guitarist friend of mine into thinking I'd got a real player in to record the part (the ultimate benchmark of a programmer's success!).

The same approach could be adopted for all sorts of instrument sounds. Breath noises are an obvious way of giving a touch of authenticity to wind instrument parts, but it's worth looking at the other factors involved in a realistic saxophone, clarinet or flute solo line. Physical modelling has given us a big boost in this respect, and anyone who has used Yamaha's VL series of synths or a Korg Prophecy will tell you just how realistic many of their woodwind sounds are. But for those of us without access to the latest generation of hardware, pretty good results can be obtained from a standard synth module, as long as you keep in mind a few simple rules of thumb.

• A CASE OF THE BENDS: One of the most common forms of expression in any instrument is its ability to 'bend' the pitch of the notes it plays. With wind instruments, the most common form of pitch bending is a form of portamento, such as when a saxophone or trumpet bends up to reach its note. Players usually start between a tone and a semitone below the note they're going for and, depending on how jazzy they're feeling, often start a fraction ahead of the beat as well. This is very often coupled with an increase in volume as the 'true' note arrives. Working with MIDI, this can be duplicated with a combination of pitch-bending (plus or minus two semitones), and use of the expression pedal and/or MIDI volume level adjustments (see Figure 1). A good way to program these is to record the part using only the pitch wheel, then listen back to it and decide what other expressive elements could be added. Move to a separate record track (on the same MIDI channel) then add other features, such as modulation or volume changes. Once you're happy with these, simply merge the two tracks together. Of course, if you can get away without having to quantise a solo of this kind, all the better, but if not, there's always the option of quantising more difficult sections and leaving the rest untouched.

• ALL MOD CONS: Moving onto modulation, I've noticed that a good number of manufacturers' factory preset sounds seem to come with the modulation wheel set firmly on the task of modulating the pitch of a sound, rather than modulating its amplitude. If you think about it, and certainly with wind instruments, vibrato is achieved through a subtle mixture of both slight changes in pitch and variations in volume. If you're programming a flute sound, for example, it's worth listening to a real flautist to see just how they use vibrato. Very often the vibrato doesn't appear until quite a way into the played note, and therefore modulation that you programme from the start of the note will always sound unnatural. Also it's important to decide just how much vibrato you need, and of what kind (amplitude or pitch). This is where a volume pedal could come in handy for programming the diminuendos and crescendos so crucial to the expressive element of any wind instrument performance.

BRASSED OFF

Of all the sounds that your average synth module can produce, surely the most shockingly artificial of them all has to be the 'Brass Section' preset. They all stink, and nearly every module has within it the innate capability to give any sequenced track that 'Children's BBC' feel, simply by playing a few choice lines on what the manual describes as "...a rich multisample featuring all the instruments of a contemporary brass section...". One of the main reasons for this inherent cheesiness could stem from the fact that a brass 'section' is something of a contradiction in terms. There are as many different configurations for a brass section as there are records made using one, but the traditional line-up usually consists of at least one of each of the following: trumpet, trombone, and saxophone. Depending on the kind of music you're recording, the line-up will change accordingly, and it's not unusual for a trumpet player to also double on flugelhorn, and the saxophonist to also play all four types of sax (soprano, alto, tenor and baritone) -- not forgetting the all-important French horn, which, although mainly a 'classical' instrument, is often seen lurking within the ranks of the brass section even for pop songs these days. When you consider this collection of instruments (and all their amazingly versatile playing styles) you can begin to see just how surprising it is that any synth patch can claim to even try to re-create their sonic range in a single preset.

The trick when programming any brass section is to view each instrument separately. Most sound modules give you a good number of the individual sounds as solo patches (Sax section, Trumpet, Single Trombone, and so on) and by assigning one to each of the parts in your multitimbral set, you not only have a greater degree of control over the individual sounds (particularly with reference to pan positions, volume levels and tunings), you actually have a much more accurate representation of what a brass section actually consists of (see 'Brass Tack' box for a suggested brass section setup). Programming the parts, of course, takes a little more time, in that it requires you to think a little more about what each individual part is doing, but you'll probably find that working in this way encourages you to be a bit more creative with your brass parts. And should you need to get live players in to replace any of the lines for real -- in a live situation, for example -- the process of transforming what has been programmed into notes that the players can work with is speeded up greatly.

STRUNG OUT

When programming any orchestral string parts, as when programming brass parts, you should always try to think of the orchestra as a collection of individual instrumentalists, as opposed to simply a collection of different instrument sections.

We've all got our favourite string preset or multisample, and when we play a big G minor chord they sound great, but this is quite different from the workings of a real string orchestra. Violins, cellos and violas normally only play monophonically, and the arranger's job is to allocate each note of a chord to a particular section -- for example, cellos could play the root (A), violas could play the third (C#), and the violins might play the fifth (E). The resulting sound is a complete chord of A major, and the richness and depth of any arrangement is in the way in which these elements interact and combine throughout a piece. Again, as with the brass parts, it can take a bit more time to program string parts in this way, and it also requires a greater knowledge of the chord structures you're using, but the end result is nearly always richer, more interesting and undoubtedly more realistic than if you had simply 'strummed' the chords polyphonically using your keyboard.

If, however, you feel your playing really isn't up to the job, or you don't have the time, you can always cheat a little by taking a chord track and using the command within your sequencer software that extracts the top and bottom lines of any given chord track. You can then assign these individual monophonic lines to different tracks, and therefore different sounds. For example, drop the lowest note of the chord by an octave (12 semitones) and assign it to a cello sound, then perhaps take the top line, make a copy of it, raised by one octave, and assign both tracks to the same violin sound. This certainly 'widens' the sound of a chord and is much more in keeping with a real string ensemble.

When you're programming these individual single string lines, you may find that there are slight gaps between the notes you've played (caused by releasing the keys a bit early, in order to move to the next note). This can be cured by using your sequencer's 'Force Legato' command. This 'stretches' the programmed notes within a given section and triggers MIDI note off information only fractionally before the start of the next note. The effect is to 'glue' the notes together without any gaps, and can often be very effective.

If the string part is still feeling a touch hesitant, you can always try going into your edit pages and creating a slight overlap between the individual notes, so that each one plays slightly into the next (see Figure 2).

Another thing you might like to try, particularly if you're limited by polyphony or 'timbrality', is using a string preset with a fairly fast attack time. This helps to ensure timing accuracy. With most samplers and sound modules, it's a relatively simple procedure to programme the attack time to be proportional to key velocity. This gives you a good amount of flexibility for playing both soft 'legato' phrases (at a lower velocity) and sharper or 'spiccato' lines when playing the keyboard a bit harder.

STAR GUITAR

The success of the guitar (arguably the world's most-played instrument) lies in its sheer versatility, range of sounds, and expressive potential. But it's these very qualities that make it so difficult to recreate realistically via MIDI.

If you'd like to try anyway, the first thing to think about is what kind of guitar sound you'll be working with. Whether it's a nylon, steel-strung acoustic, or electric will have as much bearing on the playing style as it does on the final sound. Most sound modules have a brave stab at nylon and steel-strung acoustic sounds, with varying degrees of success, but despite the fact that a distorted guitar patch is on the General MIDI list, I've yet to hear one that sounds like anything other than a pair of wasps in an aluminium jam-jar. The key to guitar realism mainly rests in the choice of sounds, so sampled guitars seem to be the only way to approach the problem seriously. Sample CDs such as the Steve Stevens' Guitar Collection offer us some interesting options, namely splitting the guitar into a number of playing styles (lead sounds, power chords, strums, etc) and letting us work with those. Certainly, a distorted set of guitar power chords in fifths work very well to lend body to the average rock track, and (just like a real guitar) expression, through pitch-bending up to the note, modulation and volume changes, all help to add to the realism.

Picked acoustic guitars can be a little trickier, as they generally sit slightly louder in a track and are usually more exposed. The key here is to try to get away from the idea of programming the sound in the same way that you would play a keyboard. The tunings of the guitar strings are such that a standard G major chord on a piano is very different to a standard G major chord played on a guitar. For example, a simple G major guitar chord in MIDI-speak looks something like this:

Not many people would play a keyboard chord in this way, but if you're to emulate a guitar successfully, these are the kind of things to think about. Timings are also crucial; generally the plectrum hits the lower notes first, so really any guitar chord is nothing more than a very quick arpeggio, starting from the lowest note up. The faster the track, the closer together these notes will be, obviously.

Another favourite form of expression amongst guitarists is to bend certain notes within a passage. Using traditional MIDI pitch-bending techniques, this would mean bending every single note being played on that particular sound at the time, which gives the impression of a 'whammy-bar' as opposed to a fingered bend. A good way around this problem is to assign more than one of the MIDI channels in your multitimbral setup to the same sound. In this case, if MIDI channels 1 and 2 are both hosting the same acoustic guitar preset, the majority of the work can be done on channel 1, and any detailed picking involving bending single notes can be recorded using channel 2. As the two channels are independent of each other, bends for the solo parts on channel 2 do not affect the pitch of notes on channel 1. The added benefit of working in this way is that you can also easily control volume and any other performance data (such as modulation, detuning, expression, and so on) of a solo part using the same sound. As I've already mentioned, any amount of 'sampled guitar ambience', such as string squeaks and buzzes, that you can throw in will all help with the overall sound.

External processing can also often be a godsend for giving that all-important bite to even the tackiest and thinnest of synth sounds. Many of today's samplers and modules come with tasty built-in effects processors, but these can generally only deal with one or possibly two effects at a time, so if your sound source has extra outputs, it may well be time to brush off that old guitar pre-amp or fuzz box and let rip with some unconventional MIDI guitar sounds of your own. Try starting with a clean electric guitar sound, assign it to a separate output, and route it through an external effects processor, in much the same way as you would if you were working with a real guitar. The results may surprise you.

The actual notes you play are surely the most important element to any performance, so what if your playing isn't up to it? This is where quantising can be a great help. We're always being told that quantisation kills expression, but there are many situations in which it can be both time-saving and creative. Quantisation is a tool, and like any other musical tool, if used correctly it can produce great results. The basic quantise levels of early sequence packages have now been replaced by much more 'intelligent' quantising systems. Emagic's Logic is a good example of this, but later versions of Creator and Notator were also well-suited to the less robotic 'groove' style of timing correction.

It's worth noting that within any given piece, certain sounds will always quantise better than others. For instance, bass lines and bass drums seem the obvious things to quantise first. They are the foundation upon which the rest of the track rests, and timing errors here will almost certainly give the whole song an unstable or sloppy feel.

Sounds with longer attack times (such as synth pads) seem to quantise less well. You'll probably find that when you play a string line with a slow attack time, you automatically play a little ahead of the beat to compensate for the lateness of the note. This is fine, but if you then quantise this line, the notes start exactly on the beat and therefore sound a little behind the beat. You can cure this to some extent by quantising the whole track, then either using a negative delay or, on the edit pages of your sequencer, selecting all the relevant notes and dragging them forward slightly until they sound right. Be warned, though: if the first note of a line edited in this way starts at the very beginning of the song or pattern (bar 1, beat 1) it often 'drops off' the front of a song and can disappear altogether, so in this case you would have to start editing from the second note in the line.

Some software packages offer the happy compromise of first quantising a MIDI performance, then 'humanising' it by scanning through the inputted data and adding small timing inaccuracies. Although this sounds good in practice, I've never really managed to get a huge degree of success using it, and I can't help feeling that this feature is an ironic and somewhat 'soulless' approach to the whole business. If something isn't sounding right, don't immediately leap for the quantise icon: why not slow the track down a little or try to record it in easy-to-manage sections? Like many things in life, the quickest and easiest answer is not always the best.

Despite going out of vogue at about the same time as they were invented, breath controllers have always been the unsung heroes of the MIDI world -- partly because it's always been a bit unnerving not knowing who was the last person to demo the unit in the shop before you bravely strap one onto your head and put it into your mouth, and partly because wind players have always played second fiddle to the keyboardist when it comes to inputting musical data into a sequencer. Keyboards are a relatively simple series of On/Off switches known as keys, whereas dealing with variables such as wind pressure and tonguing effects makes the maths a good deal more complex, and manufacturers in the past (with the notable exception of Yamaha and Akai) have generally moved away from breath controllers as an accepted method of inputting expressive MIDI data.

Thankfully, with the rise of the physical modelling synth, breath controllers are now seen as the ideal way to add realism to any MIDI composition. Another great thing about breath controllers is the feeling of interactivity that they give you whilst playing a sound. Depending on how your sound is programmed to deal with the BC data (volume, pitch, modulation, etc) breath controllers can be the difference between simply inputting data and really making a track sing!

 

BRASS TACK: BUILDING A BRASS SECTION IN A MULTITIMBRAL SETUP
PART	INSTRUMENT	VOLUME	PAN (-64 to +64)
1	Solo Trumpet	100	-10
2	Solo Trombone	90	-40
3	Tenor Sax	115	+30
4	Alto Sax	90	+15
5	Baritone Sax	75	+50
6	French Horn	100	+00
7	Flugel Horn	95	-20
8	Trumpet 2	90	-15

 

 

Home | Search | News | Current Issue | Digital Editions | Articles | Forum | Subscribe | Shop | Readers Ads

Advertise | Information | Links | Privacy Policy | Support