Very useful and easy to use software ! My only problem that the following formula (Oslon's ?) would give you a different value at the same baffle width : L=Wb*R/1.021
For example taking a 0.25m (Wb) wide baffle and 8ohm (R) load would equal to 1.8mH using the formula above.
While your program calculates 5.1mH || 10ohm with 8 ohm load for flat response.
The difference is something 2.3x. Of course baffle height, speaker height and size, etc. cannot be applied in the formula above, but it's still a large difference. What do you think ?
Zozo
For example taking a 0.25m (Wb) wide baffle and 8ohm (R) load would equal to 1.8mH using the formula above.
While your program calculates 5.1mH || 10ohm with 8 ohm load for flat response.
The difference is something 2.3x. Of course baffle height, speaker height and size, etc. cannot be applied in the formula above, but it's still a large difference. What do you think ?
Zozo
Zozo:
I am not sure of what f1 and f2 you used to end up with the 10 ohm 5.1 mH, but for sure f2 was not 2*f1, right? In order to compensate for the baffle step of 6dB this should be the case.
Anyway I do agree that the deviation is large, and I assume that setting f2=2*f1 would only bring this factor to 2.0 .
I vaguely recall that I have seen this mismatch before between my program and the baffle step compensation formulas, and that I checked my program without finding any errors. This does not necessarily mean that there are no errors in my program, but anyway I *think* it is OK. I mean, take a circular (eg by using ~40 corners) baffle of 1 m diameter. given that the edge source has a reversed phase, there should be minima when the path via the edge source is a multiple of the wavelength. 0.5 m corresponds to 690 Hz and indeed that is where the first minimum occurs in The Edge.
Has anyone else encountered mismatches of a factor 2 between formulas, or do you think that The Edge i wrong?
I am not sure of what f1 and f2 you used to end up with the 10 ohm 5.1 mH, but for sure f2 was not 2*f1, right? In order to compensate for the baffle step of 6dB this should be the case.
Anyway I do agree that the deviation is large, and I assume that setting f2=2*f1 would only bring this factor to 2.0 .
I vaguely recall that I have seen this mismatch before between my program and the baffle step compensation formulas, and that I checked my program without finding any errors. This does not necessarily mean that there are no errors in my program, but anyway I *think* it is OK. I mean, take a circular (eg by using ~40 corners) baffle of 1 m diameter. given that the edge source has a reversed phase, there should be minima when the path via the edge source is a multiple of the wavelength. 0.5 m corresponds to 690 Hz and indeed that is where the first minimum occurs in The Edge.
Has anyone else encountered mismatches of a factor 2 between formulas, or do you think that The Edge i wrong?
I suddenly remembered what convinced me that alast time I checked. I simulated the measurements of Linkwitz labs
http://www.linkwitzlab.com/diffraction.htm
and I think the agreement between his measurements and The Edge is pretty good.
http://www.linkwitzlab.com/diffraction.htm
and I think the agreement between his measurements and The Edge is pretty good.
Zozo said:
OK, so setting the values to 150 and 300 Hz at least makes the resistor 8 ohms and the high frequency response 0 dB. In this case I get am inductance of 4.2mH which is slightly more than twice the value your formula suggests. I'll double check the formulas for my inductance this evening. I am pretty convinced that the response of the program is OK, and also that the f1 and f2 produces the response that The Edge shows. We'll see about the inductance.
I'll be back...
/A
OK, I am not a baffle step expert. The formula has taken from John Murphy's site, but it's seems to be correct. BTW, Loudspeaker Lab 2 also gives the about the same value as the formula.
http://www.trueaudio.com/st_diff1.htm
Zozo
http://www.trueaudio.com/st_diff1.htm
Zozo
Svante said:
Hello -
First of all, let me say that this is probably one of the best programs I have ever seen, of any sort, for any application. I think you are a true genius!
To write this program required several important skills to be brought together:
1) Programming -- lot's of people can do this.
2) User interface -- this is rarely done well, but is unbelievably good in this program!
3) Understanding of what the task is -- this program could only be written by someone with a lot of experience in speaker design. This allows for a clear understanding of what capabilities are required.
So congratulations for the outstanding work, and sincere thanks for sharing it! (Just so you know who is heaping praise on you, I was the founder and original designer of Avalon Acoustics.)
Regarding the choice of a 180 degree corner, I think that is good for a starting point. The next thing to do would be to add successive arrays of diffractive sources.
So to do a rectilinear box you would choose 90 degree sources at the edge of the front panel (baffle) and then another array of 90 degree sources at the edge of the rear panel. The user would then specify a cabinet depth that would determine the time delay to the second array.
The next step would be to add two arrays, for a total of three. This would apply to cabinets where there was a 45 degree chamfer between the front panel (baffle) the side panel. The first array would be at the edge of the front panel and would have 45 degree sources. The second array would also be 45 degrees, and placed at the junction between the bevel and the side wall. The third array would remain at 90 degrees and still placed at the edge of the rear panel.
The next step would be to add a group of arrays that simulate a radiused transistion between the front panel (baffle) and side panel.
I think all of this would be fairly trivial from a programming standpoint. The user interface would be a bit trickier, but based on what you've already achieved I think you could do a bang up job!
Thanks again!
Charles Hansen
Zozo said:
Hmm, I've seen this page referred to many times, so I'd say it is a highly respected site. This makes me hesitate to critisise it, but I think I have found a flaw on it. The author derives his formula from the sphere. Then he assumes that a baffle of the same width as the diameter of the sphere would have about the same effect on the diffraction. However, I think it is clear from the Olson graphs that the 3dB frequency is lower for the non-spherical baffles.
This is also what I find using The Edge. Fooling around with the edge I find these formulas as alternatives to the ones on the TrueAudio page:
For a narrow baffle (H>>W):
f(3) = 33/W(B)
For H=2W
f(3) = 53/W(B)
For H=W
f(3) = 69/W(B)
These equations yield considerably lower f(3) values than the TrueAudio values of 115/W(B). Taking the square shape as an example the 24" (610 mm) size would give an f(3) of 113 Hz using the last of my equations, and 190Hz using the TrueAudio equations. I don't know if 113Hz or 190Hz is nearer the 3dB point, and I dont know the precision of the tracing of Olson's original data either.
Could someone please tell me I am wrong, I'd hate to see that a lot of speakers have been designed using a too high frequency for the baffle step compensation!
PS I double-checked my calculation of the inductance and it came out OK. I also tried to simulate the circuit in a circuit simulator using the inductance, the transfer function came out identical to the compensation curve of The Edge.
Hello -
I think the answer is "it depends". Olson's work is very accurate, and has the advantage of not relying on modeling. In other words, since his measurements are real, we can check the accuracy of our models by comparing them to his real results.
Olson's work included a 24" cube with a small driver mounted in the center of one face. If I simulate these conditions with the Edge, there is generally good agreement. I would guess that the observable small residual discrepancies are due to the fact that the Edge uses 180 degree sources at the edge of the baffle, while Olson's cabinet essentially had 90 degree sources, followed 24" later by more 90 degree sources. (Remember that these 90 degree sources are not equivalent. I am referring to them that way simply for the sake of convenience.)
On the other hand your description of the True Audio formula's result of 190 Hz is more consistent with Olson's baffle called "truncated pyramid and rectangular parallelopiped combination". This has the same basic dimensions except that there is a large 45 degree chamfer where the baffle meets the sides.
In this case the large fluctuations are averaged out, leaving a "smoothed" response version of the 24" cube. Now f3 is in very close agreement with the True Audio result of 190 Hz.
So perhaps the True Audio formula is making a generalization that would apply if the edges were chamfered, while the Edge calculates a more exact result.
Hope this helps (and makes sense!),
Charles Hansen
I think the answer is "it depends". Olson's work is very accurate, and has the advantage of not relying on modeling. In other words, since his measurements are real, we can check the accuracy of our models by comparing them to his real results.
Olson's work included a 24" cube with a small driver mounted in the center of one face. If I simulate these conditions with the Edge, there is generally good agreement. I would guess that the observable small residual discrepancies are due to the fact that the Edge uses 180 degree sources at the edge of the baffle, while Olson's cabinet essentially had 90 degree sources, followed 24" later by more 90 degree sources. (Remember that these 90 degree sources are not equivalent. I am referring to them that way simply for the sake of convenience.)
On the other hand your description of the True Audio formula's result of 190 Hz is more consistent with Olson's baffle called "truncated pyramid and rectangular parallelopiped combination". This has the same basic dimensions except that there is a large 45 degree chamfer where the baffle meets the sides.
In this case the large fluctuations are averaged out, leaving a "smoothed" response version of the 24" cube. Now f3 is in very close agreement with the True Audio result of 190 Hz.
So perhaps the True Audio formula is making a generalization that would apply if the edges were chamfered, while the Edge calculates a more exact result.
Hope this helps (and makes sense!),
Charles Hansen
That makes sense. Chamfered corners like the ones of the speakers on your website (there were a few of them there!) would surely behave more like the spherical baffle, and the trueaudio formula would be more OK. But for all of the flat-baffled speakers, especially the floorstanding "towers", wouldn't the TrueAudio compensation be far too small?
I'm starting to doubt my sanity.
I'm starting to doubt my sanity.
Good question, but I don't know the answer. I think it depends on how the ear/brain hears.
On the surface level we can look at the curves of the square baffle as the sum of two different effects. One is the 6 dB rise due to the change in radiating impedance. The other is the large fluctuations imposed by the symmetry of a point source in the middle of a square baffle.
Now it is possible that the ear/brain tends to average out the fluctuations in response caused by the second effect. In that case one would want to equalize only for the first effect, and thereby use the True Audio formula. The answer really isn't clear.
If we dig deeper, the answer is even less clear. Specifically, we know that if the delayed sound sources have enough delay that the ear/brain can "separate" them from the original sound source. In that case, we wouldn't need to apply any equalization. I would guess that the answer is dependent on how large the baffle is.
In the meantime, I think that making The Edge more accurate by including the 90 degree sidewalls would be a good start. Then adding the capability of beveled and rounded edges would be really nice...
I'd even be willing to pay for such a program!
On the surface level we can look at the curves of the square baffle as the sum of two different effects. One is the 6 dB rise due to the change in radiating impedance. The other is the large fluctuations imposed by the symmetry of a point source in the middle of a square baffle.
Now it is possible that the ear/brain tends to average out the fluctuations in response caused by the second effect. In that case one would want to equalize only for the first effect, and thereby use the True Audio formula. The answer really isn't clear.
If we dig deeper, the answer is even less clear. Specifically, we know that if the delayed sound sources have enough delay that the ear/brain can "separate" them from the original sound source. In that case, we wouldn't need to apply any equalization. I would guess that the answer is dependent on how large the baffle is.
In the meantime, I think that making The Edge more accurate by including the 90 degree sidewalls would be a good start. Then adding the capability of beveled and rounded edges would be really nice...
I'd even be willing to pay for such a program!
Hello Svante, I downloaded your program and it runs with no problem. I'm on my laptop now, it's a 200 Mhz Pentium MMX with an Integrated Technology Express ITE8330G chipset. This also is running Windows ME. I've been running the program for about an hour without not even a hiccup.
I also downloaded the FPU test program, that ran with no problems.
I'll have to try them both on my Pentium 4 with 256MB RAM, Intel 845 chipset and Windows ME and see what happens.
The laptop is;
200Mhz Pentium
64MB RAM
ITC8330G chipset
2 GB hard disk
wireless network card (I have DSL on a router)
It seems like a cool little program and I like it.
Take care and keep up, I'll let you know how it works on the P4.
I also downloaded the FPU test program, that ran with no problems.
I'll have to try them both on my Pentium 4 with 256MB RAM, Intel 845 chipset and Windows ME and see what happens.
The laptop is;
200Mhz Pentium
64MB RAM
ITC8330G chipset
2 GB hard disk
wireless network card (I have DSL on a router)
It seems like a cool little program and I like it.
Take care and keep up, I'll let you know how it works on the P4.
Hello again Svante, I'm on my P4, and again, no problem with your program or the FPU test. This machine consists of;
Pentium 4 1.7Ghz
256 MB RAM
Intel 845 Chipset
Windows ME (both the laptop and this machine have the updates).
60 and 20 GB hard drives.
Your program looks like a winner, keep up the good work!
Pentium 4 1.7Ghz
256 MB RAM
Intel 845 Chipset
Windows ME (both the laptop and this machine have the updates).
60 and 20 GB hard drives.
Your program looks like a winner, keep up the good work!
Hmmm, another way of looking at it:
A box is made for a dual subwoofer, and the box is wide and very flat so the edge of its baffle approximates a 180 degree step. There is one transducer at the front, and an identical one on the opposite side producing an identical sound, in phase. As an approximation, the sound does not curve around the baffle at all because there is no change in radiation impedance, so each speaker only produces one hemisphere of sound but at twice the amplitude. At low frequencies this is equivalent to having one speaker that has an outrageously large baffle, or a pair of speakers in close proximity to each other in a full space configuration - two separate boxes even.
I still think that subsequent interactions between point sources in the simulation could be vital. If the total energy generated by the speakers simulated in The Edge could be measured, then the frequency response would be flat, because all the peaks and troughs, and baffle step would cancel out. It's only when a true resonant system is added (T/S parameters for the speakers and recursive calculations) when some frequencies start to resonate.
CM
A box is made for a dual subwoofer, and the box is wide and very flat so the edge of its baffle approximates a 180 degree step. There is one transducer at the front, and an identical one on the opposite side producing an identical sound, in phase. As an approximation, the sound does not curve around the baffle at all because there is no change in radiation impedance, so each speaker only produces one hemisphere of sound but at twice the amplitude. At low frequencies this is equivalent to having one speaker that has an outrageously large baffle, or a pair of speakers in close proximity to each other in a full space configuration - two separate boxes even.
I still think that subsequent interactions between point sources in the simulation could be vital. If the total energy generated by the speakers simulated in The Edge could be measured, then the frequency response would be flat, because all the peaks and troughs, and baffle step would cancel out. It's only when a true resonant system is added (T/S parameters for the speakers and recursive calculations) when some frequencies start to resonate.
CM
CeramicMan said:
I think somebody suggested this as a way of doing baffle step compensation a week ago or so in another thread.
No, T/S would do more harm than good (if I understand what you're aiming at). If you are looking for adding the effect of the altered radiation resistance due to the mutual coupling between sources, the principle of superposition takes care of that.
If we take the example of two point sources to explain this, we know that two sources give a raise of 6 dB if the sources are placed close to each other. There are at least two ways to explain this.
1. The power is doubled, since we have two sources (+3dB) and it is also doubled once more due to the doubled radiation resistance (from source interaction). This adds another 3 dB, net effect is +6dB.
2. The sound pressures from the two sources are superimposed, which leads to a doubled sound pressure and a level raise of +6dB.
Note that the second explanation implicitly takes care of the increased radiation resistance without having to do the calculations.
Or am I way off here with respect to what you were saying?
Svante said:
I would say that below a certain frequency the amplitude can double at most, but that doesn't mean that the simulation takes into account physical interactions between speakers.
If you start off with one transducer, add a second one, then a third one, and so on, eventually the output power will exceed the input power if we don't take into account the amount of space that each speaker takes up. With a large array of speakers, they will only be mutually in phase below a certain frequency due to distance offsets, and that frequency will drop every time another speaker is added to the array. I don't think there's any disagreement here.
But... Adding T/S parameters would allow the speakers to lose/gain efficiency in a similar way to putting them in a different box. The current superposition of diffraction effects only seems to take care of lobing and interference effects, whereas the cone of the speaker doesn't "feel" any difference regardless of the size of the baffle or how many other speakers are near it. Obviously, box resonances are likely to be a lot worse than baffle resonances, but that doesn't mean that the sound bouncing back and forth between the baffle edges and the cone doesn't influence its excursion at all.
CM
Ok, so you mean to introduce T/S parameters for the *loudspeaker* source? That makes me more comfortable with the question. For a moment I thought you wanted to apply it to the edge sources. 
Anyway, the volume flow out of the source (=loudspeaker) is approximately independent of the radiation resistance for any sane number of speakers. This is because the radiation resistance is small compared to the other mechanical impedances in the speaker, primarily the cone and voice coil masses. The fluctuations in the radiation resistance will the cause the radiated power to fluctuate, and by this we can realise that the radiation resistance fluctuates exactly like the simulation of The Edge. In fact, the very definition of the radiation resistance as being the resistance that dissipates the acoustic power tells us this.
The fact that there are *a lot* of sources in the model does not imply that the condition above (radiation resistance/mass reactance ratio) falls, since each edge source weakens with an increasing number of sources. The sum of all edge sources is *never* stronger than half the driver source (excluding the open baffle). Only when the number of speakers is really large, we will see a limiting effect on the cone velocity.
As you say, the effect is smaller than the effect of standing waves inside the box (and I say it is very much smaller), the reason that standing waves can have an effect at allis that the sound pressure inside a box can reach very high levels at a standing wave (~130dB
for 90 dB @1 m outside the box is not uncommon if the box is not stuffed) and of course such high pressures may affect the velocity of the cone, but still only a little. The edge diffraction "back pressure" is very much smaller.
I wonder if I made anything clearer...
Anyway, the volume flow out of the source (=loudspeaker) is approximately independent of the radiation resistance for any sane number of speakers. This is because the radiation resistance is small compared to the other mechanical impedances in the speaker, primarily the cone and voice coil masses. The fluctuations in the radiation resistance will the cause the radiated power to fluctuate, and by this we can realise that the radiation resistance fluctuates exactly like the simulation of The Edge. In fact, the very definition of the radiation resistance as being the resistance that dissipates the acoustic power tells us this.
The fact that there are *a lot* of sources in the model does not imply that the condition above (radiation resistance/mass reactance ratio) falls, since each edge source weakens with an increasing number of sources. The sum of all edge sources is *never* stronger than half the driver source (excluding the open baffle). Only when the number of speakers is really large, we will see a limiting effect on the cone velocity.
As you say, the effect is smaller than the effect of standing waves inside the box (and I say it is very much smaller), the reason that standing waves can have an effect at allis that the sound pressure inside a box can reach very high levels at a standing wave (~130dB
for 90 dB @1 m outside the box is not uncommon if the box is not stuffed) and of course such high pressures may affect the velocity of the cone, but still only a little. The edge diffraction "back pressure" is very much smaller.I wonder if I made anything clearer...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.