Very interesting! The program uses only omni-directional point sources, correct? So in order to simulate a directional 'speaker', one could simply add multiple point sources close together to form a dipole or cardioid?
Yes, it uses perfectly omnidirectinal point sources in the calculations. I haven't really thought about simulating dipoles in that way, but I guess that should work!
Do you get phase control with the delay function?Yes, it uses perfectly omnidirectinal point sources in the calculations. I haven't really thought about simulating dipoles in that way, but I guess that should work!
Do you get phase control with the delay function?
In a way you do, when you set the delay of a loudspeaker you also affect the phase response, but not in the same way as a phase control on a subwoofer usually do. I guess you want to know what the delay function is for. I added that mostly to be able to simulate the CABS system described earlier in the thread -- by using different delays on the subwoofers, it is possible to get a smoother response in the room. It can be an interesting parameter to play around with.
Nice program, I will be interested in testing/using it later in time. There's a lot of new members looking for this function in this simulator. Best Regards.
Is this a modal calculation? (Seems like that is how its done from the plots.) How high in frequency or how many modes do you use? Then what is done to get to higher frequencies?
If this is a time domain calculation, how many reflections do you consider?
If this is a time domain calculation, how many reflections do you consider?
Is this a modal calculation? (Seems like that is how its done from the plots.) How high in frequency or how many modes do you use? Then what is done to get to higher frequencies?
If this is a time domain calculation, how many reflections do you consider?
Hello Dr. Geddes, the calculation model is based on the image source method (J. Allen and D. Berkley, Image method for efficiently simulating small-room acoustics, 'Journal of the Acoustical Society of America, vol. 65(4), pp. 943-950, April 1979) i.e. time domain calculations.
The maximum image source order is configurable in the program, it defaults to 45 when you start it. The plots goes up to 20 kHz, so depending on what kind of plot the user is doing the calculated response at the highest frequencies will be more or less useful.
Neither 1440*900 resolution or higher works for me. Can't see the whole window. Using W7, 64 bit.
thanks for this incredible free program!
for those using win7, don't forget to calibrate the display so that text size is set to small (normal) otherwise you won't be able to see the complete program window.
for those using win7, don't forget to calibrate the display so that text size is set to small (normal) otherwise you won't be able to see the complete program window.
Hello Dr. Geddes, the calculation model is based on the image source method (J. Allen and D. Berkley, Image method for efficiently simulating small-room acoustics, 'Journal of the Acoustical Society of America, vol. 65(4), pp. 943-950, April 1979) i.e. time domain calculations.
The maximum image source order is configurable in the program, it defaults to 45 when you start it. The plots goes up to 20 kHz, so depending on what kind of plot the user is doing the calculated response at the highest frequencies will be more or less useful.
Thanks, I am very familiar with that work (Jont Allen is an old friend.)
I started doing my PhD on room modeling using this technique, but we found that its convergence and accuracy diminished as the room became more modal - lower frequencies. This is clearly because not all values of the wavenumbers are allowed and so the image method is inaccurate since not all images will lie in allowed wavenumber directions - i.e. in the modal region not all wave propagation directions are allowed. So I finished my PhD using the modal approach. Of course this approach becomes intractable as the modal density increased. Basically a fully accurate model has to have both with a blending of the two at about the Schroeder frequency.
Thanks, I am very familiar with that work (Jont Allen is an old friend.)
I started doing my PhD on room modeling using this technique, but we found that its convergence and accuracy diminished as the room became more modal - lower frequencies. This is clearly because not all values of the wavenumbers are allowed and so the image method is inaccurate since not all images will lie in allowed wavenumber directions - i.e. in the modal region not all wave propagation directions are allowed. So I finished my PhD using the modal approach. Of course this approach becomes intractable as the modal density increased. Basically a fully accurate model has to have both with a blending of the two at about the Schroeder frequency.
It's a small world, especially the acoustics world. 🙂 The image source model is clearly not perfect, but I believe it is good enough for experimentation and learning in this application (the Room Simulator app). In what application do you think the modal approach would be required?
I don't have the detailed mathematics of the modal approach fresh in my memory right now, but perhaps the problem you describe with the image source method is related to the spatial quantization of the image sources? For example, my Room Simulator uses a sampling frequency of 48 kHz, and reflections can only occur on sample times spaced by 1/48000 s. I tried using much higher sample rates but I did not see a relevant difference in simulated responses so I chose 48k for less computational demand.
The problem with image source, or ray tracing, at low frequencies is that it does not consider the quantized nature of the wave vectors direction. It's not the temporal quantization that matters.
Consider a room in a particular mode, say a 1-1-0 mode where the wave moves around the room hitting each of the four walls (but not the floor and ceiling). This implies a very specific direction of travel for the wave in that mode. Now if the direction of the image source to the receiver location is not in this same direction then its contribution to the result has to be weighted by the dot product of this direction and the modally allowed direction.
For example a source that is exactly aligned with the modal direction will contribute to the result fully, while a source that is orthogonal to the modal direction will not contribute at all.
Clearly as more and modes contribute, this problem just vanishes. But in the more discrete modal domain it can be significant.
The modal approach is just a sum over each mode weighted by the distance from each modes frequency to the frequency of excitation and the location of the source and receiver in the "mode shape" - cosines in the case of a rectangular room. It is a frequency domain calculation rather than the time domain. It is the usual method for calculation of the room response in the modal region. It fails as the modal density gets large and can no longer be tracked effectively.
There is a region where both techniques are effective and in that region they should agree and a blending of the results is possible.
One of the more bizarre effects of this quantization (as shown by Morse in "Vibration and Sound") is that at low frequencies, when a source is turned off, the decay does not occur at the frequency of excitation, but occurs only at the frequencies of the modes that contributed to the steady state response. This effect would never be seen with the image source method but does occur with the modal summation method.
Consider a room in a particular mode, say a 1-1-0 mode where the wave moves around the room hitting each of the four walls (but not the floor and ceiling). This implies a very specific direction of travel for the wave in that mode. Now if the direction of the image source to the receiver location is not in this same direction then its contribution to the result has to be weighted by the dot product of this direction and the modally allowed direction.
For example a source that is exactly aligned with the modal direction will contribute to the result fully, while a source that is orthogonal to the modal direction will not contribute at all.
Clearly as more and modes contribute, this problem just vanishes. But in the more discrete modal domain it can be significant.
The modal approach is just a sum over each mode weighted by the distance from each modes frequency to the frequency of excitation and the location of the source and receiver in the "mode shape" - cosines in the case of a rectangular room. It is a frequency domain calculation rather than the time domain. It is the usual method for calculation of the room response in the modal region. It fails as the modal density gets large and can no longer be tracked effectively.
There is a region where both techniques are effective and in that region they should agree and a blending of the results is possible.
One of the more bizarre effects of this quantization (as shown by Morse in "Vibration and Sound") is that at low frequencies, when a source is turned off, the decay does not occur at the frequency of excitation, but occurs only at the frequencies of the modes that contributed to the steady state response. This effect would never be seen with the image source method but does occur with the modal summation method.
Very nice program!
It really helps for tuning in subwoofers in the practical world, because you get a nice starting point to work from.
I once started something similar, but I never finished it 😛
Are you planning to update the program?
Different kind of sources would be nice.
Off course you can just use multiple omni source to create a different source.
But I think it would be easier and more manageable to directly use the right source.
It really helps for tuning in subwoofers in the practical world, because you get a nice starting point to work from.
I once started something similar, but I never finished it 😛
Are you planning to update the program?
Different kind of sources would be nice.
Off course you can just use multiple omni source to create a different source.
But I think it would be easier and more manageable to directly use the right source.
Vg,
Very nice program - thank you for sharing! I have another tool in my audio toolbox. 🙂
I just wish the displays could be made to work on a laptop with 1280 x 800 res. Perhaps having the input and results in separate windows can help alleviate req't for one large window? It installed and ran the first time on Win7 64 bit Enterprise.
Regards,
X
Very nice program - thank you for sharing! I have another tool in my audio toolbox. 🙂
I just wish the displays could be made to work on a laptop with 1280 x 800 res. Perhaps having the input and results in separate windows can help alleviate req't for one large window? It installed and ran the first time on Win7 64 bit Enterprise.
Regards,
X
b_force, xrk971,
Thanks, I hope you will have good use of the program. I like your suggestions. I will probably update the program some time, at the moment my attention is focused on some other projects though.
Thanks, I hope you will have good use of the program. I like your suggestions. I will probably update the program some time, at the moment my attention is focused on some other projects though.
The problem with image source, or ray tracing, at low frequencies is that it does not consider the quantized nature of the wave vectors direction. It's not the temporal quantization that matters.
Consider a room in a particular mode, say a 1-1-0 mode where the wave moves around the room hitting each of the four walls (but not the floor and ceiling). This implies a very specific direction of travel for the wave in that mode. Now if the direction of the image source to the receiver location is not in this same direction then its contribution to the result has to be weighted by the dot product of this direction and the modally allowed direction.
For example a source that is exactly aligned with the modal direction will contribute to the result fully, while a source that is orthogonal to the modal direction will not contribute at all.
Clearly as more and modes contribute, this problem just vanishes. But in the more discrete modal domain it can be significant.
The modal approach is just a sum over each mode weighted by the distance from each modes frequency to the frequency of excitation and the location of the source and receiver in the "mode shape" - cosines in the case of a rectangular room. It is a frequency domain calculation rather than the time domain. It is the usual method for calculation of the room response in the modal region. It fails as the modal density gets large and can no longer be tracked effectively.
There is a region where both techniques are effective and in that region they should agree and a blending of the results is possible.
One of the more bizarre effects of this quantization (as shown by Morse in "Vibration and Sound") is that at low frequencies, when a source is turned off, the decay does not occur at the frequency of excitation, but occurs only at the frequencies of the modes that contributed to the steady state response. This effect would never be seen with the image source method but does occur with the modal summation method.
I have been thinking a bit about your last paragraph. It would seem from it that a room is not a linear system? In practice I think it should be linear since the response from a source to a receiver can be simulated by a linear impulse response.
Update! RoomSim should now work with the latest OS X - download and install the new MCR installer and the new app linked to in the first post. =)
I have been thinking a bit about your last paragraph. It would seem from it that a room is not a linear system? In practice I think it should be linear since the response from a source to a receiver can be simulated by a linear impulse response.
I don't think it is related to nonlinearity but is simply the nature of resonance. While the source is on there is a forcing function causing oscillation at a frequency close to, but not on resonance. Turn the source off and the oscillation continues (energy is still present) but shifts to the frequency of resonance, where the impedance is higher.
Its a bit like noting that a trumpet player can bend notes away from the tuning of the instrument. It is easier to play on the resonance frequencies related to length, but you can force it off key with some effort.
David S.
The "boom boom room" syndrome defined.One of the more bizarre effects of this quantization (as shown by Morse in "Vibration and Sound") is that at low frequencies, when a source is turned off, the decay does not occur at the frequency of excitation, but occurs only at the frequencies of the modes that contributed to the steady state response.
Wasted many hours of my youth trying to fix those problems with EQ, but the modes always won 🙄.
I don't think it is related to nonlinearity but is simply the nature of resonance. While the source is on there is a forcing function causing oscillation at a frequency close to, but not on resonance. Turn the source off and the oscillation continues (energy is still present) but shifts to the frequency of resonance, where the impedance is higher.
I've knows about this phenomenon for quite some time, but I've never really understood how it is possible. This doesn't happen in the case of a resonance in an electronic system, or does it?
The "boom boom room" syndrome defined.
Wasted many hours of my youth trying to fix those problems with EQ, but the modes always won 🙄.
The way I see it, the phenomenon would lead to bass notes shifting out of key. I associate the "boom boom boom" with amplitude response issues and ringing in the time domain.
- Status
- Not open for further replies.
- Home
- General Interest
- Room Acoustics & Mods
- New program for Room Response Simulation