Skip to the content

Websites we've worked on:
Sound tourism map link
sound around you link
sound101 link
Dr Andy Moorhouse on piano ...

BBC radio docs on sound:

- presented by Prof Trevor Cox

Concert Hall Acoustics: Art and Science

Testing Testing ....

Before a concert hall is built the design is tested using an acoustic model.  Two types of models are used for these tests: scale models and computer models.

Scale models are commonly between one tenth and one fiftieth of the actual size of the hall. In the exhibition, two models were on display:

  • An acoustic model of the Royal Festival Hall at 1:20
  • An architectural model of the Bridgewater Hall at 1:50

You can see more of the Royal Festival Hall model, by clicking on the thumbnails. To give an idea of scale, the model is 3-4 metres long. The 1:20 acoustic model for RFH was built by S&V Solutions of Sycamore, Illinois in early 1999 from survey drawings undertaken earlier in 1998, and from the archived information on the original construction project.

thumbnail of scale model thumbnail of scale model thumbnail of scale model thumbnail of scale model

Files are 50-100k. Pictures and details of testing of the model courtsey of Kirkegaard & Associates

The level of detail required for an acoustic model is greater than that for a simple architectural model.  In an acoustic model the interior surfaces must be made of material that represents the exact acoustic properties of the real, full scale hall. The Royal Festival Hall model is detailed in response to the frequency range of interest. Major architectural elements walls, columns, seating racks, box fronts - are accurately represented.  However, fine levels of detail handrails, lights, technical rigging, and surface textures do not interact with sound at the frequency range of interest.  Hence, it is not necessary to
reproduce these details in the model.  The interior of the model is quickly recognized as Royal Festival Hall, but stripped of the unrequired detail.

As the model is smaller than the real hall, the sound waves used in the testing must also be reduced in scale.  In a 1:10 scale model, the frequency of the sound must be 10 times higher than in the real hall.  A simulated double bass playing in the scale model sounds like a violin!  Any simulated violin sounds will be reproduced at ultrasonic frequencies which are inaudible to the human ear, though not to any dogs, cats or bats which may be nearby! For the Royal Festival Hall model, the useful frequency range is between 250Hz and 2000Hz - roughly the three octaves above Middle C.  Below this range most of the
absorption in the hall lies in the resonance of its thin finishes.  These low-frequency absorption characteristics are extremely difficult to model accurately since the thickness and stiffness of thin materials does not easily rescale with object size in the model.  Above 2000Hz (which is tested at 40,000Hz in the model) the air within the model becomes the predominant absorber of acoustic energy. Even with mathematical compensation for this effect, the results can be unreliable since the
amount of energy available to measure falls off dramatically with increasing frequency.

As the model was being constructed, acoustical testing of various samples taken from the hall was undertaken at Riverbank Acoustical Laboratories in suburban Chicago.  The samples tested include elm paneling, leather button paneling, copenhagen paneling, tapestries, and chairs - with and without people in them.  The data taken from these tests were used to calibrate
the corresponding surfaces of the model with the appropriate absorption characteristics.  The absorption of the samples was measured in a reverberation chamber of approximately 300 cubic meters.  The model materials were tested in a 1:20 model of that chamber.  The seats were particularly complicated to calibrate as their absorption must include that of their occupants.  The combination of felt and carpet bodies with ear-plug heads was found to give the best representation of a sold-out audience in the model.

When a model is tested, the impulse response is normally measured as this contains all the information about how sound travels from the musician to a listener, including reflections from the walls, ceiling, floor and other objects.

The impulse response shown is the sound received at a microphone in the audience when a short impulsive sound, such as a gunshot, is created on the stage for a British concert hall. The first sound to arrive is that which comes direct from the stage to the microphone. Then reflections arrive from all the walls, ceiling, floor etc.

impulse response

For the Royal Festival Hall, testing is undertaken with an instantaneous noise source (the source) - in this case a spark generator - and measurement microphones (the receivers).  The spark electrodes and microphones are inserted from below the model in various locations.  The source is located on the stage in positions corresponding to violin, soloist, cellist, woodwind, trumpet, and positions in the audience.  The receiver microphones can be located at any of these positions on the stage as well as at a number of audience positions distributed throughout the house.  Most of these positions correspond to source and receiver positions used during the 1995 tests in the hall itself.  In this way, it is possible to directly compare the impulse response of source/receiver relationships in the model with those same relationships in the hall.  A comparison of hall and model impulse responses is shown below. 

The impulse response contains all the information about how the sound gets from the stage to the audience and so is used to test the quality of halls. The impulse response graphs provide understanding of the reflection structure of the hall by showing when acoustic reflections arrive at a given receiver from a given source. The relative strength of these reflections are also displayed, providing insight regarding echoes (see diagram below) which might affect the clarity of sound in the hall.  Since sound travels at a known rate of speed, the arrival time of a problematic reflection corresponds to a fixed distance the sound had to travel to arrive at that point in time.  A simple string of the correct length with one end tied to the source and the other tied to the receiver can be stretched round a finger to identify the surfaces responsible for the reflection.  One or more of the offending surfaces can be reshaped or treated with absorption to mitigate the reflection. In a similar manner, holes in the impulse response, which indicate a relative lack of energy in a given time window, can be identified by the impulse response. By positioning new or relocated surfaces (such as reducing the canopy height) new reflections can often be added to fill in the holes and provide a more appropriate acoustic response. In addition the impulse response can be used to calculate important quality parameters, such as the reverberation time. In these ways, the impulse response can provide valuable information during the analysis and design processes.

impulse response with echo

Some actual impulse responses measured in the Royal Festival Hall and it's model are shown here. The impulse response of the calibrated model - essentially its signature closely resembles that of the hall itself as can be seen through comparisons with measurements taken in RFH before an audience in 1995 (green image, click to enlarge). The impulse response of the unaltered model becomes the basis for comparing the acoustics changes resulting from modifications of the model.  These modifications of course, reflect proposed changes in the configuration of the hall itself. The impulse response in blue indicates the direct and reflected sound which reaches a seat in the stalls from a source near the leader's position on the stage. Prominent individual reflections can be seen from the canopy, the side walls, and the ceiling.

impulse response impulse response

Computer models are also used to test how sound will behave in the hall.

Sound can be approximately modeled as particles of light bouncing around the hall – in the same way that a snooker ball bounces around a table. This enables the impulse response to be calculated and the hall acoustics tested. Testing is easier than with physical scale models.

The image shown is from Odeon, developed by the Technical University, Denmark.

Picture from computer model

Once a sound field has been predicted, it is possible to simulate the sound of the orchestra as it would appear within the real hall. This is auralisation, the acoustic equivalent of visualization. In theory this gives a chance for a concert hall to be heard and faults diagnosed by listening. In reality, however, correct concert hall auralisation is difficult to achieve and is still the subject of much research. Correct auralisation forms an important element of virtual reality systems.


In 1955, Joseph Wechsberg wrote:

[acoustic consultants] contend … that by applying certain laws of physics and using certain testing devices they can determine in advance how hospitable to sound a new auditorium will be. The fact is, however, that several auditoriums built in Europe recently under the guidance of consultants who presumably applied the laws of physics and used the testing device have turned out to have dreadful acoustics… The sad truth is that while scientists in many fields can foretell with unvarying accuracy what will result from a combination of known factors, those who specialise in acoustics seem to be on no surer footing in making their forecasts than meteorologists are in making theirs. From the evidence, it appears that no one can say for sure what the acoustical qualities of an auditorium will be until it is finished, furnished, heated, and filled with musicians, music, and listeners. And if the qualities turn out to be disappointing, it will very likely be expensive to correct them - if it can be done at all.”

Joseph Wechsberg. “Our Far-Flung Correspondents – A Question of Reverberation”, The New Yorker, 5th November 1955, pg90.

Is this still true today?

Back to concert hall exhibition homepage

(c) Bridget Shield, Trevor Cox 1999/2000. The material maybe re-used provided a link to the website is provided and a clear acknowledgement to the curators given.