Note: The following is a paper I wrote for a studio recording class in 2003 back before I had switched to the all powerful GitaStudio. I was essentially limited to a small sized instrument samples, SoundFonts and a SoundBlaster audio card. I had to work around the problems of using low-fi instruments in order to create (as best I could) a convincing orchestra sound. I later adapted the paper in 2007 for an acoustics class which is what I have posted below.

Getting that Live Orchestra Sound Using MIDI and SoundFonts

My main goal is to accurately reproduce the natural acoustical sounds of a live orchestra by using MIDI and instrument samples. There are many problems with a standard MIDI setup that prevent it from achieving the convincing quality and realism of a live orchestra. I have tried to achieve a realistic orchestral sound using MIDI, the samples that are available to me, and my knowledge acoustics, instrument properties, music, and audio engineering. I have listened to a few choice recordings of orchestral music that seem to have that natural overall sound and tried to use them as reference as well. These recordings, as well as my experience from listening to live orchestras and my drive to find a realistic MIDI sound, have prompted me to customize the MIDI to approximate as acoustically accurate as possible the various aspects of a live orchestra.

Step 1: Listening to it Live

I attended a live orchestra rehearsal in Pope Joy Hall. The orchestra’s concert was the next day so the orchestra was in its prime – well rehearsed and properly staged. I milked this opportunity by taking readings with a SPL meter at different places in the hall during each section the orchestra rehearsed. Beethoven’s 5th Symphony was a perfect piece to listen to in order to get a good idea of typical orchestration, instrument balance, and dynamics.

I found the middle of the hall, in the center, was the best spot for all aspects of acoustics; balance, blend, dynamics, hall reverb, etc. The instruments became clearer as I moved closer, but the balance seemed to suffer a bit. Interestingly, there seemed to be a longer reverb time when I was closer to the stage than when I was in the middle of the hall. There, the instruments seemed to be more crisp than clear. Further back from the center there was also more and longer reverb, and the orchestra sounded darker with less higher frequencies but maintained the same clarity that they had when I was at the center of the hall. It was almost as if the center of the hall was the ‘hot-spot’. As I moved closer to the back corner the sound began to become very full in the lower frequencies, because larger waves tend to accumulate in corner spaces.

At A-weighting the decibels typically ranged a soft 60db (woodwind solo passages) to a bright and boomy 96 db (typically when the timpani and brasses were used). Woodwinds at mezzo-piano ranged from 60-70 db. My only complaint was that the rest of the orchestra covered the cellos most of the time (but I think it’s because the cellos were sissies). The violins at pianissimo in their upper register were at 52 db, but were still surprisingly clear. The French Horns I clocked from 70 db at mezzo-forte to 83 db at forte. The timpani at forte had a whopping 84 db and I imagine could go as high as 90 db, and I believe is responsible for the majority of the force in any particular fortissimo passage.

In addition to the live orchestra, I listened to many professional recordings of orchestras and found a few that had an excellent sound. I listened for good microphone placement techniques and good a type of good musicianship. I listened for, most importantly, a natural balance between the instruments (ex. a flute at forte should be covered by a trombone at forte). There seemed to be less reverb in the hall than in the recordings I had listened to. The reverb in the recordings felt more ‘attached’ to the instruments than in the concert hall, where the reverb was more of a separate entity that complemented the orchestra. I think this is because I have a better sense of special positioning when the reverberation is all around me, as opposed to the instruments and reverberation that are all coming from 2 speaker sources. Many times, the orchestral recordings were produced at lower levels than most popular music recordings. The recordings seemed to have compressed dynamics and I wanted to somehow avoid this and keep the balance and dynamics natural. I decided that the compression would come during the final mix to bring up the softer passages of the music, and should not affect the overall orchestral balance in the MIDI.

Step 2: Finding and Making Convincing Instruments

My next step was to collect all of the sampled MIDI instruments that were convincing to use. Aside from the adaptations I’ll make to the MIDI patches, I started by looking for instruments that had realistic timbres.

Wind instruments were the most difficult sounds to reproduce. The brasses from the standard MIDI set were bright, stale, and had splatty attacks. The woodwinds from the standard MIDI were also too bright, and some patches had vibrato that was not adjustable within the MIDI patch. In the end I found myself mostly limited to a handful of articulations, swells, and special effects. Some instruments were only convincing in particular registers. The MIDI clarinet had a disgusting sustain in each register except for its low chalumeau, which was appropriately dark and dry. Only a few of the instrument’s long tones were convincing enough to include.

The most difficult characteristic to match was the individual expressiveness and changes in an instrument’s timbre between dynamics. A natural characteristic of a trombone, for example, is the way its harmonic spectrum changes when moving between different dynamics. One way I attempted to recreate this was to take the instrument sample that was available to me and construct new piano and pianissimo patches with them. I tried to take the patches that were available and, using EQ, reduce the high frequencies and slightly reshape the harmonic spectrum from the recorded waveform to match a natural harmonic spectrum. In some ways this worked, but timbre is such a subtle effect that I found it very difficult to match convincingly.

I had a lot of percussion equipment that seemed to work well. A good group of pitched percussion instruments: Timpani, vibraphone, marimba, tubular bells, crotales, and temple blocks, as well as non-pitched percussion: Snare drums, tam-tams, bass drums, a modest assortment of cymbal rolls and crashes, ratchets, slapsticks, tambourines, cowbells, claves, and castanets. Each instrument included different patches with different mallets, note-lengths, rolls, bows, etc. All of these instruments had very convincing sounds with natural decays that would work well given the right context. Some already included a little bit of concert hall reverb. I found it weird that the triangle from the original general MIDI patches sounded exceptionally real once I took it away from the other nasty instruments, and took it off its default hard right panning (weird). All of the other general MIDI percussion sounds lacked both the authentic sounds and the natural decays that were necessary. A timpani patch had a comical thud instead of a nice resonant thum which would have been more convincing.

The most important part of the orchestra, the string section, was the weakest of all the instrument samples I had. The sustained cells were too boomy and lacked characteristic higher frequencies in the spectrum. The violins were useful only in the upper register sustained at piano, and the middle to lower registers with a short marcato articulation. The violas were scratchy, but had the nice dry characteristic that I thought to be useful. The contrabass, however, was excellent in its low and lower registers sustained from piano to mezzo-forte. It provided the right amount of bow-to-string sound and bass sound without being to boomy or too dull. Everything else was pretty much worthless.

Step 3: Mixing and Balancing the Orchestra Accurately

With this collection of instruments and my trusty chart of decibels from the live orchestra rehearsal, I proceeded to my MIDI setup. Each MIDI instrument needed to be calibrated so that they would balance correctly to each other. The instruments I have come from different sample packages, and each has a different dynamic range (usually not like the actual instrument). I carefully balanced each instrument to the others, trying different combinations to make sure I would get the proper results. For example, I would indicate a flute to play at forte, but got the sound of a flute at piano with the volume turned up. There were similar problems in other instruments so I made those adjustments as well. I thus had to make changes in volume so that each of the instruments would produce a close enough decibel level to the ones I had measured.

I also set up two patches of each instrument, one patched panned in the middle, and another panned at its correct position in the orchestra. Violins panned to the left, Cellos to the right, etc. Some instruments, like the piano, had default panning relative to pitch (lower notes were panned to the left and higher notes were panned to the right). I made sure to change this before I applied a global panning. I didn’t want to overdo the panning, as I heard on many recordings. I wanted the MIDI orchestra to sound like it would from the center of the concert hall, not from the conductor’s podium.

I also had to add different levels of reverb per instrument. For most of the instruments that were located in the back of the hall, I sent them to a separate channel with more reverb. The French horns were sent to their own channel, since the majority of French horn sound that we hear from the concert hall is from sound waves that are reflected from the back. Instruments like the trombone and trumpet were given only a little bit. Even though they are further back, their bells project most of the sound forward and so they did not need as much. The tuba was another instrument that needed more reverb like the French horn. It bounces off the ceiling before making its way around the room. I also added a little on the cellos, violins and trumpets, etc. to simulate the depth and balance of natural staging. I adjusted all these parameters to achieve a most accurate sound.

Conclusion

The end result was a major improvement over the traditional midi patches. The instruments that were most realistic tended to direct my compositions in such a way that hid the problems of the unrealistic instruments. I used a lot of MIDI to realize the compositions that I have written. The more I use MIDI, the more aware I become to know how poor and inflexible MIDI is in comparison to real live instruments. However sampling technology is quickly evolving (like all computer technology). It will soon be the method for presenting compositions once it meets that standard.


Read Comments (5)         Add a Comment