A degree of "realism" of an organ model

It is often heard - and we also say this - that the sample sets in conjunction with a software sampler such as Hauptwerk offer a high degree of realism. This means that the organ "model [is] specifically designed to reproduce the features and sound of a pipe organ, and is thus able to achieve much more realistic results" (a quote from M. Dyde website). The degree of realism of the highest quality Sonus Paradisi sample sets is such that it is very hard to distinguish the recordings made virtually from those made on the place itself.

But some clarification is needed to explain the term realism employed here. What is the "reality" to which the model tends to? If we think about this, we come to the conclusion that the concept of realism is a vague one. If you visit a real organ, you will always make quite obvious observations. The organ is: 1) never in perfect order 2) changing in time 3) the sound perception changes in space.

1) There is a number of tuning, voicing errors, sometimes there are mechanical errors, the blower noise tends to be higher than one would like to and the noise and hiss emanating from all the parts of the organ is also a problem. However, these and other features are, of course, very real. Does this mean that the organ model should reproduce faithfully all these to be realistic? Or, to what degree should it reproduce these to be realistic? Should not we rather attempt to "correct" all the errors so that we get some ideal state of the instrument? Do we really know what this ideal state would be? These questions have no universal answer and therefore the reality laying behing the model is vague.

2) The organ is subject to deterioration or improvements over the time. Its sound is therefore not a constant. It is even argued that each time you press a key of the same tone, its sound is a bit different. This is especially true from a longer perspective. So, you would need infinite number of samples to represent this. Sonus Paradisi tries to fight this difficulty by applying more than one sample per pipe and Hauptwerk chooses at random which one is played back when you press the key. The variability of sound is therefore maintained. But still, the amount of samples available is finite.

Also, the acoustic conditions vary greatly over the time (winter, summer, changing of the furniture, wall painting and so on). So, it is not even possible to say that the room acoustics has been captured precisely when making the model since you capture just one "moment" of the acoustics, neglecting the others. To what extend is one competent to say that the organ model offers a high degree of realism? An answer to these questions could be, that the organ model reflects the state of the instrument it was found in at the moment when the recording was made. So, if we record an instrument on 27th October 2006, our model of the instrument will be very close to the reality of that instrument on that particular date. But as many say, what is the purpose of reproducing the instrument to its state on that particular date which is absolutely insignificant from the historical perspective? Should not the model, again, reflect some ideal state of the instrument in its history?

3) It makes difference if you listen to the instrument close to it or if you are seated at farther place. It is the same as looking at the object. From various perspectives, it can look a bit different. Therefore, if you capture an instrument from a specific position, no matter how many channels you use during the recording process, you capture just one "prospective" of its sound, neglecting intentionally the others. How can one say, then, that his model of an organ faithfully reproduces the acoustics of the original ambient? Or, should we look for an ideal place where the sound is heard best? Do we really know what this place would be and what is the best sound perception of the sound of a particular organ?

In addition to these trivial observations, there are other features of an organ sound which cannot be reproduced faithfully by a model. For example, it is often argued that the pipes interfere with each other producing a new sound quality. There is, therefore a difference in the sound of a Principal 8' when speaking alone and when speaking together with other pipes. This is called "coupling" but the terminology may be not precise. As one German specialist puts it: "Nehmen Sie z.B. auf einem Manual einen Principal 8' Ton auf, wird dieser in Zukunft immer so wiedergegeben [in a digital model]. Spielt man aber bei einem guten historischen Instrument diesen benannten Ton und ergänzt dann einen Ton z.B. im Pedal, dann verändert dieser hinzutretende Ton bereits den vorhandenen Ton im Principal 8'." (a quote from the letter to me) This is true and this behavior is not possible to reproduce in digital world where the samples are "stand alone". A complex wind model (such as in Hauptwerk) can give good results in certain aspects, but not all is the matter of a wind model. The frequencies and harmonics of pipes influence each other. A nice example of how pipes interfere with each other is the Example no. 5 of my authenticity quiz.

Listen to the beginning of the composition. There is a chord F-dur heard and then after few moments a high C breaks in. Listen carefully to this high C in all the examples you have in the quiz. What the pipes do with each other in the real recording is quite incredible. They seem to fight with each other as their frequencies try (unsuccessfully) to come to a harmony. Their tones seem to "walk around" each other producing very intense beating. This beating does not have "sinusoid"of "linear" characteristics which would be the result of the two constant frequencies sounding close to each other. The frequencies of pipes vary in time being "pulled" by the other sounding pipe. I suspect that this behavior has some sort of a chaotic pillow shape. This cannot be reproduced having two independent samples, as is shown in the corresponding recordings made in Hauptwerk. An organolog's note: I have serious doubt if this and similar behavior can be really ascribed to the type of the windchest used. Some organists argue that only the key-chambered type of a windchest  allow this behavior. This is questionable as some others suggest. It may be simply one of the organological "myths" which one often hears in these circles. I think it has more to do with the oscilations producing the actual tone which excite or "pull" the other pipe in some direction than with the fact that the pipes sit on the same wind supply channel. But this is only a hypothesis which I cannot prove now.

Also the reverberation of short "staccato" notes is known issue. This can be reproduced by an impulse response reverberation model or by the means of multiple release samples but it is always an approximation and other problems are rised. Of course, the questions one has to ask is, to what extend these features are substantial for a "realistic" organ model. The answer again will be oriented towards a "degree" of approximation. And again, these features which are described theoretically represent certain "ideal" view of the behavior of an organ which in practice may produce variable results.

Towards a conclusion

Many times in this exposition, we employed the term "ideal". Indeed, many features of our sample sets reflect some sort of an "ideal" state of an instrument. This is necessary since no user wants to listen to errors and casualties reflecting the state of the sampled instrument in a specific moment of history each time he presses certain key or so. Also it is practical since to reproduce many features (especially those with a "chaotic" and unpredictable behavior) would take enormous amount of energy while their contribution to the model could be controversial or negligible. The sample set producer has to impose limits on the "realism" of the model.

Is the weakening of the relationship with reality not perceivable enough? If our models tend to an ideal organ rather than to a real one, is it not more just and honest to speak of a degree of "idealism" rather than "realism"? Of course, it sounds better and attractive to say that the sample set sound "realistic" rather than to confess that the sample set sounds "ideal" and that its features are only approximations of the real behavior of an organ.

A corollary: Is not, after all, the term "realism" in conjunction with the sample sets used (and abused) for purely commercial reasons, to help selling of the sample set?