Creating Orchestral Arrangements On A Budget

Use of a full symphony orchestra remains a luxury afforded only to the world's top‑selling artists or film composers. But with a little help from technology, the sound of convincing strings and blaring brass is achievable on a budget. John Rowcroft reports, and offers advice to anyone seeking to follow his example...

How on earth do aspiring composers convince hard‑nosed film bods that they are the right choice to write the score for a forthcoming blockbuster? This is the question one such aspiring composer found himself pondering recently at the tail end of a Masters course in Composing Film and Television Music. Finishing off my showreel with some tasty orchestral showpieces, I thought, might just set me apart from the crowd, but there was one small problem... symphony orchestras don't come cheap, with upwards of 60 musicians to pay, and other high costs to meet such as studio and engineering fees... Many of today's lower‑budget productions (cartoons, computer games and so on) opt for the orchestral 'mock‑up' which can offer very passable results with just a couple of samplers and a small studio setup (see Philip Meehan's feature on the subject in SOS March '99). This might fool most of the punters most of the time, but there are still limitations. Because some sampled instruments are more convincing than others, you even tend to write the music to avoid those sounds that will sound unrealistic. Pizzicato strings? Fine. Tinkly glockenspiel? No problem. Exposed oboe solo? Umm... perhaps not. Any instrument with such enormous expressive potential as the oboe still defies emulation in an exposed setting.

One solution to this problem lies somewhere between the MIDI mock‑up and a full live recording. It is now relatively common practice to beef up acoustic recordings with synthesized or sampled instruments which will probably go unnoticed by even discerning ears — adding sampled basses to a small string section is one example. My problem was, however, that I couldn't even afford a small string section at full Musician's Union rates — and trying to organise such a gathering on a charitable basis is no mean feat.

Lateral Thinking

Thinking positively, I looked at what I did have at my disposal:

• Limited studio time using my college's facilities — a hall, isolation booth, well‑equipped control room and a good selection of mics.

• My own small setup at home comprising a Kurzweil K2000 synth/sampler, Korg X5DR module, PC running Emagic's Logic Audio with Audiowerk8 soundcard, a DAT machine, a desk and a couple of effects units.
  
  One of the author's finished 'orchestral' arrangements — audio and MIDI tracks side by side on the Arrange page of Emagic Logic v4.0 for PC.

My cunning plan was to recruit the best players I possibly could from London's various music colleges, and offer their group/ensemble a nicely recorded demo in return for playing for me. I threw in a Travelcard and a couple of beers as well!

Putting up notices at The Royal College of Music, The Guildhall School of Music and The Royal Academy of Music, I soon had the phone ringing. My initial plan was to get a string quartet (two violins, viola and cello), a brass quintet (two trumpets, horn, trombone and tuba) and a wind quintet (flute, oboe, clarinet, horn and bassoon). These three groups between them could cover all the acoustic instruments I needed; basses and percussion would be provided by my trusty K2000. Each group would have a three‑hour session in the hall in which to record their own demos. I would then record individual players in the isolation booth for my material (ie. dry and totally isolated), and edit it together later in Logic Audio.

Of course, things didn't quite pan out as I'd hoped. The wind quintet turned into a solo oboe due to exam pressures on the other players, and the string quartet cancelled at the last minute in favour of earning £800 playing at some garden party in a stately home... some people.

Very Lateral Thinking

Fortunately, a couple of weeks previously, I'd spotted a pretty hot string quartet playing at Brighton train station and (as you do) bowled over to get their card. In fact, I'd been collecting these kind of contact numbers all year — some of the string quartets busking at Covent Garden and other tourist traps are seriously good players (usually final year or post‑grad students at one of the London music colleges) and the possible classical stars of tomorrow! Anyway, I booked the Brighton station lot for £200 for 6 hours (a serious bargain), decided to play clarinet myself and badgered a friend into playing flute (she also doubled piccolo). Bassoon would suffice from the Kurzweil, and the brass quintet were still up for it.

You may have noticed that including myself, I had only booked a total of 12 players — and you might fairly conclude that this hardly constitutes the 'big Hollywood sound' I had been looking for. This, however, is where technology comes to the rescue. There's nothing new about the idea of overdubbing, but today's MIDI + Audio sequencers really do allow a different level of control in this type of application. Building up a full orchestral piece to multitrack tape is certainly possible, but having tried it myself I can assure you that the timing and tuning problems alone are enough to knock the idea firmly on the head, let alone the vast number of tape tracks you need. Computers on the other hand open up a multitude of possibilities, and even a modest 16‑track setup like mine can be coaxed into handling an orchestra's worth of sound. The rest of this article explains how I went about it, with a few pointers as to how you might recreate what I did. For more general advice on how to approach this kind of project yourself, see the three 'Sound Advice' boxes dotted around this article.

The Recording Sessions

This was a nice easy gig for the college engineer, because everything was recorded straight from one mic, through a preamp and onto DAT in mono. We opted to use a Neumann U87 mic through a Focusrite Green preamp on most of the instruments, but for info on choice and positioning of microphones I refer you to Hugh Robjohns' previous SOS articles on miking up brass, winds and strings (see SOS January and April 1999). I sent a continuous click from a sequencer into the player's headphones, having deliberately opted for pieces with no tempo changes, and then sat in the booth with them as they played. Because everything was to be edited together later, the players could tackle their parts one phrase at a time until they got a good take. This really allowed me to get note‑perfect performances, even though the players had never seen the music before.

Seeing as the players were having to play along to nothing but a click, you might well wonder whether the feel of the music suffered. The answer, for a number of reasons, is no. Everybody in a real orchestra plays a relatively small expressive role, but due to the sheer number of players, a huge expressive range results. The 'overdub and re‑edit' approach to constructing orchestral arrangements wouldn't work nearly as well on, say, string quartet music, because all the subtle, intimate communication which takes place between just four players interacting would be lost. On the other hand, intimate is one thing orchestral music ain't.

My decision to sit in with each player also meant that whatever the limitations of the notated music might have been, I could always sing them a line as I wanted it, with just the right phrasing and rhythmic feel. I did also make sure that they heard a MIDI demo version of each piece so that they had an overview of the style I wanted. Most of the players responded beautifully, but of course it helps if your musical lines really suit the instrument in question — getting string players to swing, for example, is something that just cannot be taken for granted!

When recording the woodwinds, I made sure I got a couple of good takes of each phrase just for safety. Good players tend to breathe naturally in much the same place each time they play, but I had to keep an eye out for inconsistencies between any parts which were supposed to move together.

I tried to obtain at least four good takes of each phrase for the editing process later, though this wasn't always possible due to time constraints. I also paid great attention to the string players' bowing, ensuring sometimes that they were consistent between overdubs, sometimes making sure they weren't. Why? Well, different string players vary their phrasing by means of different bowings, but if you watch a real orchestral string section in action, you'll soon notice that each group (such as the first violins) all play with the same up and down bow strokes together. If you want to build up a convincing overdubbed string section sound, therefore, your single player's changes of bow direction must remain consistent over all four takes. On the other hand, for long notes which require a change of bow direction in the middle, players in a real orchestra will naturally stagger their individual bow direction changes, thus helping to hide the slight break in sound which results. Varying the point at which bow changes took place in each of my takes helped me greatly at the editing stage.

Post‑Production

My first task was to get the recorded DAT material into my PC. I simply recorded each instrument into Logic Audio through the S/PDIF digital in on my soundcard. Once I was happy with an edited line, I would use the glue tool to digitally mix down into a new WAV file, delete all the debris of other sequences to preserve hard disk space, and move on to the next instrument. Orchestral arrangements can very quickly become huge, so limiting the number of WAV files that your sequencer has to look at will reduce the workload for your PC's hard disk and processor, and help to keep playback glitch‑free.

Strings

Strings were the hardest orchestral family to deal with because of the challenge of creating a string section sound from just one instrument. Having been through this process, I would really recommend trying to get at least four good takes of each phrase, even though I often found I could only obtain three or fewer. (When really pushed for time, I found I could make do by giving priority to the first violin parts, as these are the most prominent parts for most of the time.) Having recorded the track into Logic Audio, it was quickest to align the whole lot to a click at the same tempo at which the players were recorded, so that only small tweaks were needed to get individual phrases exactly in time with the click later. The worst problems here were with really tight 'chugging'‑type accompaniment figures; the various takes often required very detailed editing to bring them in time with one another. In some cases, cutting and shifting of individual notes proved unavoidable.

In the few cases where I actually had about six good takes of a string phrase, simply layering them up one on each track created a lovely rich ensemble sound. However, an element of chance came into getting the best sound, and this was where I really had to use my ears. For example, muting one or more of the six tracks sometimes actually yielded a better sound due to the chance distribution of tuning variations in each take.

Another screen grab from Logic Audio. This one shows a copied viola line being shifted up by four cents using Logic's Time Machine module, to give a more realistic spread of pitches.

Where I only had a couple of good takes, I might create a new copy of each, take them into Logic's Time Machine processor and retune one up and one down three or four cents (see screenshot, above). This gave me four sequences to layer, and with the copied sequences fractionally delayed or advanced, a reasonable ensemble sound usually emerged. Occasionally, though, the result would be too 'synthy' with chorus or phasing effects creeping in. Experimentation with retuning and time‑shifting timing is the only answer here, but always I found it worthwhile to try muting individual tracks first, since often only two of the takes were 'incompatible'. Two notes with tuning and timing too close together, for example, often introduced an undesirable flanging effect.

I generally aimed for a tuning 'spread' of around 6‑8 cents, although if I could hear that a take was particularly sharp, I might just try to guess how many cents were needed to flatten it, aiming for concert pitch. This is amazingly good practice for your ears, but on refelction I could have cut corners by using the likes of Antares Auto‑Tune.

The whole layering process is a slightly hit‑and‑miss affair, but here are a few more points you might wish to consider:

• Violin — higher notes (especially on the E string) started to make a beautiful rich sound quite easily after only three layers were combined. The lower G and D strings were more problematic though, with a tendency towards 'boxiness'. A little mid‑EQ cut around 900Hz helped.
• Cello — fewer layers than with viola or violin proved necessary to create a balanced section sound. Three or four usually did the trick.
• Synth strings/samples — I also tried layering together some synthetic strings and samples for good measure! I actually found that subtle use of my Korg X5DR's 'Marcato Strings' patch worked better than my ultra‑realistic Peter Siedlaczek samples, perhaps because a little tonal contrast helped create a real ensemble effect. Careful use of programmed volume curves in Logic's Hyper Edit helped mask the effect (see the screenshot, bottom left). I had to keep an eye on levels, panning, and reverb levels for the extra samples, to keep them consistent with the audio tracks. In the end, the effect was 'felt' rather than heard, though it definitely added something.
• Care with detune — based on my experiences, I would advise anyone attempting a similar project not to go mad with detuning effects. Chorus is likely to make strings sound artificial, and a bigger tuning spread than about eight cents will make your strings sound like a school band!
• Belt and braces — if you're not sure whether to double, say, your first violins with flute then go ahead. You can easily remove the flute later if you want to.

Wind

Editing the woodwind and brass parts was a much simpler process than the strings because most wind orchestration is on a one player‑per‑part basis. Where for example I had scored specific effects such as four horns in unison, I just layered as appropriate. Where I only had one good take, I found that adding some subtle, slowly modulating chorus worked well for brass, but not so well on woodwind.

As I started combining brass and strings, I found that it was actually the slight tuning discrepancies between the two which started to make things sound really big! Where brass and wind were working as a pad to the strings I liked this effect, and left the tuning well alone. Where solos or counter‑melodies were appearing out of tune, though, I did tweak — sometimes cutting out individual notes to be processed by the Time Machine' in Logic.

Mixing

Once again, I would refer you to Philip Meehan's article in SOS March '99, since mixing should really take place according to the layout of a real orchestra (see diagram on page 136). That means panning instruments as they appear in a real orchestral layout, and adding more reverb to instruments at the back such as percussion and brass, to simulate the hall in which a real orchestra might be playing.

Subtle shifts in volume have been introduced into four duplicated string parts for greater realism, by drawing in vector information in Logic's Hyper Edit.

Adding in the MIDI instruments — basses, bassoons and percussion — was a relatively straightforward affair. For the basses, I used the Korg X5DR's 'Marcato Strings' patch, layered with a Kurzweil K2000 solo bass sample for edge and a little warm vibrato. I also added ensemble pizzicato samples just to fatten things a little.

Orchestras are usually heard from some distance, and so I found that panning anything too hard left or right sounded unnatural — especially with strings. I used a stereo 'Concert Hall' preset on a Zoom 9010 reverb unit, and no other processing except a little corrective EQ was necessary. In terms of balance, referring to CDs of other orchestral recordings helped greatly. It was tempting to crank woodwind solos up to artificially high levels, but the best results were obtained in the end by ensuring that the accompanying instruments were sympathetically quiet instead — just as a good conductor would. I used a combination of Logic's automated volumes for individual audio tracks and 'burned in' gain changes to balance tracks.

Does It Actually Work?

It takes a lot of time but this approach really can work. Securing realistic‑sounding results depends largely on the content of your instrumental arrangements (I found the most successful passages were those with plenty of doublings), but I hope this article has at least hinted as to what and what not to expect. I decided to attribute my finished recordings to the 'LCM/RCM Studio Orchestra' — a cheap trick, but who's going to know it doesn't really exist?