Acoustic Modelling 9.....
CLARINET and FLUTE MODELS.
This model is the saxophone model with one fundamental change. A major difference in the sound of a clarinet is that, due to its construction, generally the instrument behaves as a stopped pipe, and therefore produces an overtone series of predominantly odd harmonics whereas the saxophone generally produces an overtone series of all harmonics. So it follows that reversing the overall delay loop feedback from positive to negative should have the same effect. On my delay lines there is a switch for doing this. Tried on the saxophone model, it gave some promising results. The sound quality, once again, was very lifelike, being very much of the clarinet family. Accidental ‘overblowing’ was again something of a problem here also with certain bright sounding parameters. The interesting phenomena with the clarinet is the contrasting and conflicting effects of a ‘reed’ producing even harmonics as it squares repetitively round a delay loop which is trying to organise this into a series of odd harmonics.
Much work in the digital world using large mainframe computers had already been done on clarinets by J. O. SMITH and others when I started doing this. Quite elaborate software models have been written with much more accurate modelling of the reed and the delay lines (usually called wave guides in these research papers). I had only a handful of analogue modules but I was surprised how acoustic the results were from my really basic “no frills”model.
The clarinet is an interesting instrument as it sounds quit different in different registers. The complex key system changes the resonant response significantly as it enables the player to sound a wide range of notes. Composers make good use of the variations of tonal quality. Doing this is well beyond the capability of a simple analogue synthesiser based model. In fact all the ‘pipe’ based models here just simply lengthen and shorten the ‘pipe’ to reach the different pitches. What happens in real instruments is of course a lot more complicated than this. My main concern was to see if a simple analogue model could just get the acoustic sound so I could play around and experiment with it in my music. I never intended to create say a saxophone in a box type of thing.
By reducing the signal level feeding one of the multiplier’s inputs, the control input, and changing the filtering in the delay loop in the saxophone patch, the non-linear overtone series was gradually reduced and softened until a soft flute-like sound was produced. Once again the sound had that real acoustic quality about it. The flute is a pipe stopped at both ends so it will tend to resonate mainly at all harmonics. The valve, this time, is merely a mass of air the is being carefully blown across a mouthpiece that is being resonated by vibrating air in the pipe. A mass of air’s quality of resonance is obviously quite different than a reed. Pipe organs work basically in the same way but the design and sophisticated valve geometry can be radically different from blowing air over a hole in a flute. The wide range of tones available from the stops of an acoustic organ shows how much can be done with this machnism.
To show the saxophone and the flute are related, the next sound example demonstrates that I managed to perform a gradual transformation of the sound from one to the other by just altering the patch setting knobs. To do this, after playing a phrase, I held on to one note and twiddled various knobs. It was, however, extremely difficult to keep everything in balance and the transition smooth. I even used the computer (see “Synthesiser adventures” section) to help out a little but it didn’t do that much. In this example you can hear the saxophone sound gradually collapse as the non-linear valve effect is turned down and then right at the end the flute’s pipe is brought to life and made very live adding some tremolo to enhance and vary the layering of sound in the loop as it decays. Creating a good flute or clarinet sound is quite difficult with such a simple model and I’ve had less success at my first try digitally. The clean ultra-pure mathematical digital environment definitely makes some things a lot harder to achieve. Analogue dirt definitely helps.
[SOUND EXAMPLE 13: Transformation from saxophone to flute on one note of a musical phrase... MP3 (128kbit/sec)].
Performing with any of these ‘pipe’ models as substitutes for real instruments is really challenging. You are really up against it. Think about it. First you have a very simple model that just gets the basic sounds but is devoid of most things that can give expression. In fact to add expression with the array of twiddleable knobs and a keyboard, a whole new level of skill has to be mastered. You could add joysticks, foot controllers and several hands to the patch to help but you still need to get your brain around how exactly to operate them all. Even with the best possible model, to compete, you also have to think and play just like a real player of the instrument and do things that they would do on a real instrument. You would have to translate all the intuitive input they put in to operating a set of electronic controls. Then... even if you manage this, the frustrating downer is that everyone has heard really good players many times performing on top quality instruments. So anyone not knowing that you spent hours of work on it, would probably say something like, “nice, but it’s just a saxophone solo: so what. It sounds OK I suppose but I’ve heard better” and they no doubt have. If you tell them it was produced by a synthesiser they would probably still remain fairly unimpressed saying “yep it’s amazing what these little boxes can do these days”.I think this is why physical acoustic modelling as a synthesis tool has not been that popular for conventional musicians. Whatever the success and accuracy of a particular model you are still faced with learning a new instrument each time and that requires a big effort for many instruments.
This is why I did it just to experiment with the great acoustic type of sounds I got from modelling and use the sounds as as source material to put in my electronic music. In fact I still very rarely use non-percussive models to play melodies and solos. Life’s too short. It’s much easier and quicker to use a sample based synthesiser.
However those of you who want to experiment with processing and transforming acoustic type sounds, acoustic modelling does open up an amazing new world of sound exploration.
For the last example I took the reed model a little further and added a second filter element that was wired in anti-phase as a kind of crossover circuit. The idea was to change the delay loop feedback from positive to negative at some point in the frequency range. I think I was trying to improve the pipe model’s range of sounds. Here’s the result...
These days it’s so much easier to experiment with acoustic modelling than when I did it in the mid eighties. There are several software packages that allow the user to delve into the this world to some extent or other. These will happily run on just about any reasonably modern desktop computer or laptop. Compared to the cost of a mainframe computer in the eighties, it costs peanuts to do it. I experimented a bit with this as few years ago and found Native Instruments “Reaktor” to be particularly good at providing all basic building blocks to do it. What follows in the next section is my attempts with “Reaktor” to duplicate those early analogue patches and the all important sound clips of examples of those models.