Algorithms and computer-related questions.
A: There are two types of volume in Quick estimate estimated based on different methods.
When “user defined volume” is ticked, the number in the box to the right will be used as a basis for estimating the Quick estimate reverberation times.
If the user defined volume box is not clicked Odeon will use an estimated volume that is generated from the mean free path length of the defined source automatically. The automatically estimated volume can be seen if you click the Material overview in the quick estimate. If you change sources you can see that the estimated volume changes.
Another way to make Odeon help you guessing the volume of the room is to press the suggest box volume. This box volume is a box defined by the outer dimensions of your model (length-with-height) also given if you press the info button or (Shift+ Ctrl+ R). So you will see that this suggestion does not change if you use different sources as the estimated volume will.
If you know best you can enter a box volume from your own calculations and click the user defined volume to calculate the reverberation time.
So compare the different types of volumes with your own reasoning. And for calculating the reverberation time always use the point responses (Single point, Multipoint & Grid). In single point responses the volume is not included in the calculation of reverberation time as reverberation time is derrived from the simulated impulse response measurement.
Remember that the Quick Estimate gives only a very fast overview of your geometry.
A: Odeon and other Room acoustics software are energy based high frequency models. Calculations are in 1/1 octave bands because it gives better results for energy based calculations.
In cases with wall transmission it is possible to use 1/3 octaves as input, but then Odeon will recalculate these to 1/1-octave bands in order to fit with the best calculation principle for this type of model.
A: The calculation principles applied in Odeon is a combination of high frequecy models such as the image source method and ray-tracing, therefore validity of results depends on the frequency range of interest and if signals of interest are dominated by pure tones.
For Odeon simulations as with real measurements, the source and receiver should be at least 1/4th wave length from the walls. But at the very lowest resonance of the room the level can change a lot from position to position without Odeon being able to predict it. For investigation of low frequency behavior (resonances), indeed Odeon is not the tool.
Although small (Non diffuse) rooms is a challenge to Odeon, it might in some cases be the best tool for your calculations anyway. This report suggest that absorption should be chosen with care if predictions in rooms with very none diffuse sound fields are made, in particular be careful with extreme absorption coefficients (alfa > 0.9 and alfa < 0.05).
A: The reflectogram and the reverberation curve is shown as sound pressure level(dB) as the BRIR is given in pressure (P). P is a more sensitive measure than dB (10*Log10(p^2)), and therefore the BRIR shows small details more clearly. Also BRIR has all frequencies included in the response at the same time. And especially the low frequency contribution that are almost not audible can have a strong visual influence on the BRIR.
Due to the sensitivity of the BRIR also the differences between right and left ear can seems very strong compared to what one would expect from two measurements so close together.
Even when source, receiver and room are all ideal symmetrical there will always be a small difference between right and left ear due to the calculation method for sound distribution used in Odeon.
Finally the HRTF´s used in the BRIR calculations will be shaped to contain some asymmetries.
A: Select the unity HRTF in the auralisation setup – delete the text contents in the Headphone input box (turns read which is ok in this case) finally set the phase approximation to random. This will produce a BRIR where both channels are equal.
A: Try to bring the work files down to your computer for comparison. The speed of the calculation can have something to do with the speed of the network. Also if you make very heavy calculations it might take some of the network capacity from others.
A: In order to speed up point response calculations, try making a calculation at one of the most critical positions, e.g. where coupled room effects are present. (Auditorium and Combined editions only)
Before calculating the point response, set the calculation parameters in the Room Setup. Use the standard Engineering setup as a basis. Set the Impulse Response Length to a value between 2/3 and 1 times the expected Reverberation time. Set the Impulse Response Resolution to a value around 1/1000 – 1/500 of the Impulse Response Length.
Calculate the point response in the selected critical position.
Check the following in the Single Point Response (Auditorium and Combined editions only)
- Is the T30 calculated at all frequencies? If not, a result of 0 s appears at some frequencies. This indicates that the Impulse Response Length is not long enough; change this in the Room Setup.
- Compare EDT and T30. In simple, well-behaved rooms they should not differ very much; typically EDT is a little shorter than T30. However, in the case of coupled volumes and no direct sight from the receiver to the source, EDT can be longer than T30.
- Look at the decay curves; are they reasonably smooth? Look also at the squared impulse responses (Press A to switch from the display of integrated decays at all frequencies); the fluctuations should not have very strong spikes.
- Look at the Reflection Density; values between 30 and 100 / ms should normally be sufficient. If it is less than approximately 30 / ms the number of rays should be increased for reliable results.
If the quality of the results above is not satisfactory, the Room Setup should be changed as follows
- Increase the number of rays, (e.g. a factor of two should lead to the double reflection density).
- Increase the Desired late reflection density; in some cases the actually achieved density in the point response calculation may be significantly less than the desired value in the room setup.
If the quality of the results above is satisfactory, but you want to minimize the calculation time, try decreasing the number of rays and shortening the Impulse response length.
If you are in doubt whether the results are good enough or not, try to run a point response calculation using the Precision setup. In most cases this will create a very high reflection density, but in rare cases like an open air theatre the number of rays may need to be set even higher. Use the calculation results as a reference, i.e. the results obtained with the optimized setup should not deviate significantly from these results.
Grid Response calculations
For initial calculations consider using a grid with large distances between receivers or instead use Multi point responses with discrete receivers at strategic positions. Only for the final calculations make a detailed grid calculation.
A: Odeon 11 which will be released later in 2011 will support multi core CPU's for Multi point and Grid response calculations. The calculation engine has been rewritten to support this and even runs faster on a single core CPU.
A: Yes as well as scattering dependent of size of surface and distance between surface and source/receiver. read more in the following paper proceeding:
A: Yes much effort has been put into optimizing the calculation speed of ODEON over the years. It is not possible to mention all techniques implemented - here is just a few examples.
Odeon uses a dynamic sized 3D-cubenet in which it is stored, which surfaces intersect which cube in the cube-net, therefore Odeon only has to analyse a few of the surfaces in a room for each wall /ray collision instead of all surfaces.
From version 8.5 and up, ODEON make use of CPU specific instructions (MMX, SSE, SEE2…) in order carry out multiple calculations in one operation (e.g. multiplying x, y and z of a coordinate with a constant in just one operation instead of three). So indeed parallel processing is performed. Image sources are detected by use of ray-tracing. Rays will only detect image sources which are likely to be valid whereas the traditional image source method requires an enormous amount of image sources to be calculated of which only a few will be valid. Odeon keeps track that a given image source reflection is only added once to the impulse response by use of a tree data structure (image source tree). Therefore there are no cases where an image source is included twice in an impulse response as is the case for the Cone tracing method.
Apart from the above, the ‘Late-ray’ method is capable of producing many reflections at a receiver with just a few rays. For a description of the calculation method please see the ODEON manual.