Virtual acoustic, the recalculated sound.
Jul 1, 2003 - by Franck ERNOULDTranslated by Dominique ZBIEGIEL.
The computer-assisted acoustic have been existing for years. The progress carried out, in particular in the field of processors, made it possible nowadays to synthesize very complex acoustic environments, integrating in real time multiple factors.Virtual acoustic will change our sound practices.
Acoustics "applied" to the building trade, from the concert hall to the railway station, is due at the same time to art and science. The propagation of the sound waves is a very complex phenomenon, utilizing the three dimensions of space, the dynamic profile of the sound and its spectrum, the atmospheric conditions, the materials used, the shape of the room. Some theoretical formulas allow, on the basis of "objective" criterias (dimensions, characteristic of absorption of the materials), to predict the reverberation time, the average absorption, the distribution of the reflexions between the walls, the floor and the ceiling. But unpleasant surprises are not rare: let's mention the example of the new Bundestag, in Bonn, whose opening had to be delayed because of enormous acoustic problems. There always remains a certain difference between the "calculated" room and the reality. The equations do not take into account all subtleties. In addition, even if the acoustic expert starts from objective data to carry out his calculations, the perceived sound, in the end, is always judged according to subjective criterias. In other words, the terms used to describe acoustics of the same room will be probably different for each person. Each one has its own jargon, which give sometimes place to a dialogue of the deaf between the acoustic expert and his partners, architect, scenographer, musicians, financier.
|Waves and ultrasons tanks.|
The first tools of behavior forecasting for the rooms required a physical realization on reduced scale. Wallace Sabine, founder member of the modern acoustics, used a waves tank for a long time. He immersed in a few centimeters of water a plane profile of the room, cross-sectionned from the ground to the ceiling. The plane waves generated by a small vibrator on the surface of the water collided against the shapes of the model, and made it possible to immediately realize the problems of the sound focusing in certain places, for example.
According to the observed results, Sabine then modified the volumes without "fine" calculations , by listening to his experiment and his instinct, and nevertheless, the changes he asked, remained sufficiently reasonable for the designer to recognize his project . Let's admit that generally, the architects start from preestablished forms visually significant, without taking into account in their choice the acoustic properties of these forms. To take an extreme example, to build a great glazed surface is satisfying on the visual level: the light enters flowing in, creating a total transparency. On the acoustic level however, nothing brings more disorders than a reflective surface like glass, which causes echoes, focusing phenomena. Acoustician and architect does not have the same way of looking at things!
Another way of anticipating the response of a room was to produce a reduced model at 1/20 of it for example, then to bombard it with "on the scale", wavelength sounds, therefore of frequencies multiplied by 20. The experimenter works then to a great extent in the field of the ultrasounds, for which the absorption by the air is not any more negligible. Air can be replaced by nitrogen, very complicated corrections can be one. In short, the results, there again, are marred with errors. And what' s more, these two techniques require the long and expensive construction of detailed models, which can only start when the designers have already a very advanced idea of the realization of the room. In front of the obtained results, acoustics expert is then sometimes reduced to play the "fireman", and must use all the options of his art to make satisfactory an acoustics at the beginning really bad, due to bias which could have been avoided by discussion or an earlier modeling.
A room, whatever it is, always comes down to a limited number of elementary geometrical surfaces, slopes and known absorption coefficients. These elementary forms bear the name of "facets". The possible occupants of the room, from the public to the seats, have too quantifiable acoustic properties. There are libraries gathering the characteristics of the hundreds of principal
materials used. The more factors are taken into account, by applying the "elementary" formulas of acoustics, the more the reality can be approached.
The computers quickly were thus called upon service to carry out these million of calculations. It is the way chosen by Bose since 1985 for its professional software Leveler, dedicated to the acoustic study of rooms, within the well defined framework of the PA system of spaces by loudspeakers. The data captured by hand of all the characteristics of the few hundreds of facets (Leveler admits some up to 768) constituting a room, from the fittings of this one, to the position of the diffusion speakers (up to 100, whose characteristics of output and directivity appear in a data base separately) is certainly tiresome, but it is rewarded in the moving sight of a central processing unit "crushing" calculations (during a whole night, on Macintosh at the time) before delivering the results in the form of curves, evaluations ,"factors of intelligibility", calculations of arrival times , of colored graphic representation of the sound cover.
On today's PowerMacs, the tenth version of this software provides its detailed diagnosis in a few minutes. It still remains however in the field of figures and complex diagrams (for example, representation of the ways of the sound waves at the time of tens of first reflexions,), which can be deciphered only by specialists. Once more, we speak about the sound of the room, we represent it more and more accurately, but it would be much more immediate of being able listen to it.
This precise modeling, carried out at the beginning to apply simple acoustic formulas, can also be used to calculate what is called the "impulse response" of the room. Without entering in mathematical concepts, let's simplify by saying that any room can be considered as some kind of filter, equal to a distributor filter in a speaker, for example: a black box, whose behavior is linear.
In the digital domain, the more samples are processed, the better the definition is (it is called the "length" of the filter). Any filter is perfectly characterized by what is called its impulse response. In other words, any room is perfectly defined, in a given point, by its response in time to a short sound impulse (a cinema clap , for example).
More extremely: the French mathematician Fourier proved that an unspecified sound can always be broken up into a sum of sinusoids. In other words, any sound can be represented as well by its evolution in time as by its contents in frequencies. Let's apply this principle to our room : if we have recorded its behavior in time due to a sound impulse, we can deduce its behavior in the frequential domain from it, the colors that it brings, etc.
Still more extreme : other mathematical theories allow, starting from the impulse response of a room and a "neutral" signal, not acoustically coloured, to obtain by an operation called convolution the response of the room to this signal. In clear words : starting from the sound signature of the room, and a sound recorded in soundproof room, it is possible to calculate what this sound in this room would give, at the point of reference. And this restriction on a signal recorded in a non resonant environment should fall before long, due to the progress of the techniques known as deconvolution, aiming at isolating a signal from its acoustic context.
|From arithmetic to sound : acoustic simulations.|
Going from the models to an audible signal is the reserved field of the signal numerical processors and the sound cards. That's what Bose carried out with the Auditioner system (including DSP and listening units), which came as of 1991 in addition to the Leveler software. At the time, the processing part of the signal consisted of some reverberations, echoes and electronic filters connected to each other. It is from now on a DSP card (developed by the Australian Lake) which takes again the file of the calculated data of the room, integrates the position of the listener decided by the operator, and the filter deduces the matching filter by calculation.
The signal (CD of voice or music, computer) present at the audio input of the DSP is taken into account and integrated in the simulated acoustics of the room. The DSP provides two different signals, one by ear, and works in real time, with however a two seconds shift. These signals are then reproduced by a Bose unit (derived from the famous Acoustimass speakers) whose tolerances are draconian : 0,1 dB of difference in response between the two loudspeakers.
The position of the head of the listener is fixed by a chin-rest, the angle and the distance from the satellites are fixed too, which imposes a position of listening identical to all. The sound results are amazing, the simulation of the room is credible and, according to Bose, very often close to the final result. Auditioner is not available for sale, but implemented by Bose in collaboration with service providers.
The Leveler is used to estimate levels of sound coverage in a direct field: from this perspective of sonorisation the reverberation of a room is a defect from which it is necessary to be freed. There are many other systems, combining for example the Catt Acoustic software, for PC, and a Turtle Beach card or, to relieve the computer's microprocessor and to make it possible to work in real time, dedicated Lake FDP1 plus processors (2 channels of reproduction) or Huron (up to 256 channels!). These systems function all on principles of numerical convolution in real time and try hard, as for them, to synthesize as accuratly as possible the reverberation of a simulated room. It should be pointed that in fact the first hundred reflexions in a room subjectively characterize the quality of a reverberation. The most precise temporal knowledge possible of the behavior of a sound launched in a room makes it possible to detect defects of design and to remedy to it even before the work is started. The response of an existing room can directly be entered to try to improve it. Thanks to the sound cards the acoustics expert can with concrete sound proofs justify a particular installation to the architect.
The modification of such detail finds thus immediately its sound expression. This form of virtual acoustics goes thus in the direction of the feelings, of a greater sensitivity: the recourse to words inevitably vague is not necessary any more, it is enough to listen to the result, on two channels, with headphones or speakers.
From a room simulation to the simulation of more complex evolutionary environments, there is only one step to cross. By distributing the signals corresponding to elementary reflexions on multiple loudspeakers dispatched a little everywhere in an unspecified room, one can give to the latter an acoustic character completly different. The IRCAM is very committed in this domain, with the "Spatialisateur": the possibility of varying instantaneously acoustics of a concert hall, according to the moods of the work, opens to the contemporary composers new horizons. In the same kind of idea, but in a more "gadget" way, recently in a Central Park concert that took place in the open air, loudspeakers that were suspended above the head of the spectators made it possible to recreate a church acoustics, in harmony with the played works.
The DSP calculating the elementary signals of the reflexions can also be programmed to integrate dynamic modifications of the site of listening. In other words, the synthesis of the reverberation can follow in real time the virtual movements of the listener All you have to do is to provide the headphone with a position encoder, that transmits the movements of the head of the listener! The applications are innumerable and promising, in the booming domain of the 3D computer-generated images for example.
These techniques are unfortunately still confidential, with the result that recently, with a showcase in Montpellier, nobody was astonished to see a re-creation by extremely successful computer-generated images of an ancient church, in which one could move with the help of a joystick, accompanied by a stereo organ music , whose sound position did not follow the displacements on the screen! The tools however exist, and without any doubt will spread themselves quickly.
At the same time as the synthesis of unbelievable sound environments, the researches undertaken in virtual acoustics are also interested in the manner of collecting as simply as possible spatialized sound information. An example of application is provided by the CEA : at the time of the
dismantling of the nuclear thermal power stations, robots will work in irradiated zones where man cannot get to. Their piloting is carried out for the moment by a mainly visual interface, exploiting the information provided by remote-controlled cameras. This interface however shows its limits in the banal case of the loss of a tool, that takes sometimes a very long time to find by randomly directing the cameras. The engineers of the CEA wish to introduce acoustic perception in the command interface. Correctly collected, and transmitted to the operator in a headphone developed for the applications of virtual acoustics, the sound would make it possible to know instantaneously where to seek the lost tool, in distances as in position. This sound sensor could be the English Soundfield Mk V microphone unique in its kind, which comprises four pressure pick-ups assembled in tetrahedron. These four ways are absolutely necessary to transport all the information necessary to a three-dimensional location. By an external processing, and from this single microphone, it's possible to emulate the directivity characteristics of any stereo couple.
The professional studios have had at their disposal already for a long time numerical reverberations, capable of recreating easily a great number of "standard" acoustics .Their processors are however programmed in a radically different way from those from the virtual acoustics softwares. For power saving reasons, the calculations use loop settings of reflexions, with the result that these reverberations do not "fool" an experienced ear, that perceives their imprecisions on certain consonants for example. The reverberations recreated according to virtual rooms modeling are finer, and take into account the space and temporal distributions of the first reflexions in the room, the transients of the instruments.
In short, the acoustic simulation that they offer is more credible. In Japan, the "Reservation" service of a concert hall has proposed for a few years to the customer to immediately hear the sound that they will perceive accordingly to the chosen place . This rather rudimentary simulation system of listening in situation "was arranged" by means of some professional studio devices signed Sony and Yamaha, intelligently connected to each other. The service as it is today seems to receive a warm reception by the customers, and could be easily improved by the recourse to "real" dedicated processors. Useless to specify that in the near future, the information highways could propose the interactive equivalent of this service at home.
The NASA works a lot on the location in space, in conditions of weightlessness. It appears that the recreation, by a movements sensor headphone of a credible acoustic universe following accurately the evolutions of the astronauts, helps those to a better location and to fight against the spacesickness. The NASA thus dedicates important funds to the researches in virtual acoustics. The European telephone companies , involved on the juicy market of teleconference, also follow closely these developments. A teleconference gathers participants around a table of virtual meeting: there too, the sound must "follow the image". When several speakers speak at the same time, they saturate the sound communication channel, no discrimination is anymore possible, the message is lost.
Replacing these speakers in a sound environment spatially credible makes it possible to the listener to take again processes of intelligent binaural listening : its two ears work again efficiently and can recreate the "cone of presence" (in other words to be focused on a zone of precise emission in a noisy environment: it is called the " cocktail party" effect, and the profit of intelligibility is enormous, equivalent to a profit from 6 to 10 dB.
This principle of spatialization of the sources to avoid confusion is also applied in the air controllers. Those work on an already modelled visual interface: the light signals corresponding to the planes appear on their monitor. The vigilance of the visual sense requires a constant effort, whereas the auditive sense acts to some extent as a "background task": auditive vigilance is passive. In a crisis situation, where speed prevails, a signal of alarm will thus be more quickly felt and discriminated by an auditive signal than by a visual signal. It is useless to recreate a sound field sophisticated in the headphone of the air controllers, it is enough to spatialize the signals corresponding to the six or eight planes followed simultaneously. These three last applications require a headphone equipped with position encoders, which return to the processors in charge of synthesizing the acoustic environment of the signals modelling the position of the head of the bearer. Our brain functions also that way to define the source of certain sounds: unconsciously, our head is animated by small rotational movements which are used to compare the differences in information arriving to our ears. The sufficient angular resolution of the position encoder is approximately 10, and the listening with the headphone is thus cleared of its large defect: an impression of "sound in the head".
The binaural hearing, by the two ears, is then possible. Another application is the rehabilitation of people becomed blind due to an accident. Developping and educating their auditive direction gradually, by using acoustics of virtual rooms diffused in a headphone , enable them to find a mean of apprehending volumes and the distances, that the blind from birth integrated in their youth. Lastly, there is no doubt that the fanatics of video games would be delighted to be able to hear a sound in connection with the situation that they play. There are besides already machines with the look evoking Mad Max, combining display screens, ear-phones, and position encoders. The market is enormous, but the efforts required in the field of sound creation too. While integrating progress of virtual acoustics, the world of the video game will enter another age!
The computing power increase with the possibilities of the microprocessors, and the costs constantly drop . It is thus reasonable to think that configurations costing more $15,000 today will be accessible in a few years by music lovers. The techniques resulting from virtual acoustics will propose new ways of perceiving the music at home, in the sense of a greater sound immersion. The impulse responses of the recording rooms could be made
available to the public in the form of computer files placed on the first trak of the CD of a classical music recording , for example.
The CD itself would propose the "spatialized by the sound engineer" version of a musical work, but it would be permissible to the music lover to take again the process with its own taste. The computer would initially isolate the musical signal from the room, then virtually replace it in another room whose characteristics would be provided to him.
In theory, this operation is not very far away from the acoustics processors already available on the market, offering the choice between various acoustics : jazz clubs, of cinema or concert hall. The difference is that in this case, this coarsely synthesized environment comes to be superimposed on that existing already on the recording, whereas the recourse to virtual acoustics would lead to "natural" results more. CD-ROMS proposing the acoustic signatures of the most famous concert halls of the world would make it possible to the amateur to vary ad infinitum the pleasures. Matsushita presented a significant prototype of its global design of the future listening living room. It consists of a wall at facets, integrating at the same time acoustic and electronics processings : the television set, the various speakers, the devices are laid out in a calculated way, nonmodifiable, for the best possible sound diffusion. The techniques of virtual acoustics make it possible to personalize the listening to the extreme. Until recently, the equalization curves for the binaural listening synthesized by the Catt Acoustic software were gauged for a STAX Lambda headphone .
But in fact, any headphone can have its sound behavior estimated, its answer finely evaluated and modelled as a filter, which makes it possible for the processors to adjust the message that they calculate according to the defects of the headphone .
Then you can be certain of the linear quality of the reproduction. The headphone measurement laboratories thus have one's work cut out. Let's go even further. Each person has a head with its own shape and size, and a drawing of auricle specific to him. These characteristics, in virtual acoustics, are modelled in the form of different filters for each one, that are transmitted to the processor. Being able to establish at home the response of its own filters would thus make it possible to refine more the spatialization when listening through headphones. It appears that manufacturers like Sony and Yamaha show themselves very interested in front of the commercial prospects revealed by this idea. Let's leave for a moment our living rooms for the film sets ! Everyone knows that the majority of the realizers, some purists leaving aside, remake a part of their direct sound after the shooting, in postsynchronization.
The problem for the sound engineer, at the time of the mixing, consists in " connecting " the sentences remade in the auditorium with the original sentences, even certain scenes between them. Let us imagine that the sound engineer at the shooting has collected the impulse response of the place where the recording was made. By injecting it into a dedicated processor, as well as a voice recorded in close field, without resonance (in an outside cabin for example), you can have the realistic impression that the voice was recorded at the place of the shooting.
The joint will be thus be immediately almost perfect. Other applications could relate to the restoration of files, the recovery of too coloured or too reverberated recordings, etc...
One realizes, with the reading of this nonexhaustive list, that it is our whole listening conception of the sound restitution that will be changed by the advent of virtual acoustics. A major problem remains : to develop a user-friendly interface for a general audience, allowing him to manage the various elements simply allowing to simulate the most complex acoustic phenomenas. Once this problem solved, it will be very difficult to go back in many fields. But won't a musical home listening of a recording carefully carried out in "a true" room by a well equipped talented sound engineer remain for many audiophiles quite higher than a re-creation, as finely calculated as is it?
Intented for the few specialists needing this kind in recordings, exist some CD including tracks recorded in a soundproof room. Let us recall that such a room is entirely covered with sound absorbing materials, and that no sound reflexion can occur there. The felt sound impression is sometimes close to faintness, the instruments as the voices sound very oddly, but the soundproof room is the only means of freeing itself from the influence of a room. It is primarily used to measure noise levels, but musicians can also be recorded there. B&O ("Music for Archimedes"), Denon ("Orchestral Anechoic Music Recording") or Technics thus propose to the spatialization amateurs some raw material to make the processors work.
Thanks to Bruno SUNER, from the Euphonia company, specialised in virtual acoustics, and toEric CHIQUET and Dominique MARPHAY, from the Bose France professional department, for their precious assistance to elaborate this article.
© 1998 - Franck ERNOULD
|Be the first to comment on this article|