Home |
Search |
Today's Posts |
#82
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. As I said the diffraction effects that DO exist may be compensated for in the filter. If I try do trace the dimensions of the baffle for the tweeter (gosh, I HAVE to do something about the user interface in my program) I end up with a transfer function with a 2-3 dB dip at 3 kHz, which I think I can see in your response graphs (not as deep though). I think this dip can be made smaller by moving the tweeter slightly to either side. The dip at 3.2 kHz could either be a small diffraction effect caused by the edge, but the form of the 5 inch driver on the front baffle could also account for the dip. The bevel edge is 12 cm at tweeter axis and 17 cm at woofer axis and should have a good effect to "cancel" diffraction effects from the tweeter. Nevertheless, the dip is found only in the 0 deg axis, and is actually not doing very much on the energy at 3 meters. On-axis measurements say little of the total sound from the speaker. In addition there is a wish for me to have a slight dip in the 3-4 kHz regions compared to 1-2 kHz, since the ear is more sensitive in the direct sound vs diffuse fields in the 3-4 khz region. Listen to soprano voices at high volumes, and I think you would prefer a slight dip in the 3-4 kHz region. But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. I have not seen your program (mac user) so I don't know if it calculates on the two corners. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com I've lost the direct link to the Spreadsheet, unfortunately. http://hem.bredband.net/b113928/spea...asurements.htm The actual dimensions of the speaker is: http://hem.bredband.net/b113928/ritning.htm A sphere is bad in one sense, internal standing waves. But it could be fixed by making asymmetric internal cabinet. Yes, anyone who has hit a basketball knows the sound of standing waves inside a sphere. |
#83
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. As I said the diffraction effects that DO exist may be compensated for in the filter. If I try do trace the dimensions of the baffle for the tweeter (gosh, I HAVE to do something about the user interface in my program) I end up with a transfer function with a 2-3 dB dip at 3 kHz, which I think I can see in your response graphs (not as deep though). I think this dip can be made smaller by moving the tweeter slightly to either side. The dip at 3.2 kHz could either be a small diffraction effect caused by the edge, but the form of the 5 inch driver on the front baffle could also account for the dip. The bevel edge is 12 cm at tweeter axis and 17 cm at woofer axis and should have a good effect to "cancel" diffraction effects from the tweeter. Nevertheless, the dip is found only in the 0 deg axis, and is actually not doing very much on the energy at 3 meters. On-axis measurements say little of the total sound from the speaker. In addition there is a wish for me to have a slight dip in the 3-4 kHz regions compared to 1-2 kHz, since the ear is more sensitive in the direct sound vs diffuse fields in the 3-4 khz region. Listen to soprano voices at high volumes, and I think you would prefer a slight dip in the 3-4 kHz region. But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. I have not seen your program (mac user) so I don't know if it calculates on the two corners. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com I've lost the direct link to the Spreadsheet, unfortunately. http://hem.bredband.net/b113928/spea...asurements.htm The actual dimensions of the speaker is: http://hem.bredband.net/b113928/ritning.htm A sphere is bad in one sense, internal standing waves. But it could be fixed by making asymmetric internal cabinet. Yes, anyone who has hit a basketball knows the sound of standing waves inside a sphere. |
#84
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. As I said the diffraction effects that DO exist may be compensated for in the filter. If I try do trace the dimensions of the baffle for the tweeter (gosh, I HAVE to do something about the user interface in my program) I end up with a transfer function with a 2-3 dB dip at 3 kHz, which I think I can see in your response graphs (not as deep though). I think this dip can be made smaller by moving the tweeter slightly to either side. The dip at 3.2 kHz could either be a small diffraction effect caused by the edge, but the form of the 5 inch driver on the front baffle could also account for the dip. The bevel edge is 12 cm at tweeter axis and 17 cm at woofer axis and should have a good effect to "cancel" diffraction effects from the tweeter. Nevertheless, the dip is found only in the 0 deg axis, and is actually not doing very much on the energy at 3 meters. On-axis measurements say little of the total sound from the speaker. In addition there is a wish for me to have a slight dip in the 3-4 kHz regions compared to 1-2 kHz, since the ear is more sensitive in the direct sound vs diffuse fields in the 3-4 khz region. Listen to soprano voices at high volumes, and I think you would prefer a slight dip in the 3-4 kHz region. But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. I have not seen your program (mac user) so I don't know if it calculates on the two corners. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com I've lost the direct link to the Spreadsheet, unfortunately. http://hem.bredband.net/b113928/spea...asurements.htm The actual dimensions of the speaker is: http://hem.bredband.net/b113928/ritning.htm A sphere is bad in one sense, internal standing waves. But it could be fixed by making asymmetric internal cabinet. Yes, anyone who has hit a basketball knows the sound of standing waves inside a sphere. |
#85
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. As I said the diffraction effects that DO exist may be compensated for in the filter. If I try do trace the dimensions of the baffle for the tweeter (gosh, I HAVE to do something about the user interface in my program) I end up with a transfer function with a 2-3 dB dip at 3 kHz, which I think I can see in your response graphs (not as deep though). I think this dip can be made smaller by moving the tweeter slightly to either side. The dip at 3.2 kHz could either be a small diffraction effect caused by the edge, but the form of the 5 inch driver on the front baffle could also account for the dip. The bevel edge is 12 cm at tweeter axis and 17 cm at woofer axis and should have a good effect to "cancel" diffraction effects from the tweeter. Nevertheless, the dip is found only in the 0 deg axis, and is actually not doing very much on the energy at 3 meters. On-axis measurements say little of the total sound from the speaker. In addition there is a wish for me to have a slight dip in the 3-4 kHz regions compared to 1-2 kHz, since the ear is more sensitive in the direct sound vs diffuse fields in the 3-4 khz region. Listen to soprano voices at high volumes, and I think you would prefer a slight dip in the 3-4 kHz region. But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. I have not seen your program (mac user) so I don't know if it calculates on the two corners. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com I've lost the direct link to the Spreadsheet, unfortunately. http://hem.bredband.net/b113928/spea...asurements.htm The actual dimensions of the speaker is: http://hem.bredband.net/b113928/ritning.htm A sphere is bad in one sense, internal standing waves. But it could be fixed by making asymmetric internal cabinet. Yes, anyone who has hit a basketball knows the sound of standing waves inside a sphere. |
#86
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. I correct myself is that the measurements are made on my new speakers, using approximately the same cabinet but with different drivers. The original NG8 did not measure this flat. T |
#87
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. I correct myself is that the measurements are made on my new speakers, using approximately the same cabinet but with different drivers. The original NG8 did not measure this flat. T |
#88
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. I correct myself is that the measurements are made on my new speakers, using approximately the same cabinet but with different drivers. The original NG8 did not measure this flat. T |
#89
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message . com... (Svante) wrote in message . com... (Thomas A) wrote in message Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. PS. Have you seen or heard the Mirsch Rondo? They are made as perfect spheres and measured quite good, if I remember correctly. Articles in EV, around 1993-1994. The Allegro NG8, around 1990, are beweled cubes, also by Mirsch, to counteract diffraction. See No I have not heard them. Mmm... As I am beginning to understand it, diffraction cannot be avoided in ANY kind of (finite size) design. But for sure the design of the baffle is very important, and not always very intuitive (at least not to me). "Avoid symmetry" seems to be the only simple advice that holds, so far. http://hem.bredband.net/b113928/inde...r/image002.jpg At the first look at the images, they may look asymmetrical, but when I try to model this baffle shape in my little program it appears as if diffraction would affect the response of this speaker a lot (well, peaks and dips 3-4 dB apart). I think it is the symmetrical left-right placement that is responsible for this. Of course I cannot say anything about how they sound based on this simulation, and the effects may be compensated for in some way, also, I have only rough estimates of the sizes from the photo. But at a first look I would say that the placement of the speakers on the baffle is not optimal. I made some simulations long time ago using the baffle diffraction simulator, and it looks ok. The beweled edges act as "filters". I cite from the baffle diffraction simulator manual: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter." The actual measurements of the speakers are rather good. This is an old measurement which I made (I have increased the response now in the 1-2 kHz region to take take care of lost energy in the extreme angles). The drop 10 khz is due to microphone error. The peak at 1.2 kHz is not as bad as it looks at 3 meters listening. Yes, you are probably right in that the speaker is good. I correct myself is that the measurements are made on my new speakers, using approximately the same cabinet but with different drivers. The original NG8 did not measure this flat. T |
#90
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote:
: have not seen your program (mac user) so I don't know if it calculates : on the two corners. Have you compared your result with the Excel : Spreadsheet at : : http://www.pvconsultants.com : : I've lost the direct link to the Spreadsheet, unfortunately. http://www.pvconsultants.com/audio/d...ownloadbds.htm linked from http://www.pvconsultants.com/audio/frdgroup.htm (Not very easy to find from the main page ;-)) |
#91
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote:
: have not seen your program (mac user) so I don't know if it calculates : on the two corners. Have you compared your result with the Excel : Spreadsheet at : : http://www.pvconsultants.com : : I've lost the direct link to the Spreadsheet, unfortunately. http://www.pvconsultants.com/audio/d...ownloadbds.htm linked from http://www.pvconsultants.com/audio/frdgroup.htm (Not very easy to find from the main page ;-)) |
#92
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote:
: have not seen your program (mac user) so I don't know if it calculates : on the two corners. Have you compared your result with the Excel : Spreadsheet at : : http://www.pvconsultants.com : : I've lost the direct link to the Spreadsheet, unfortunately. http://www.pvconsultants.com/audio/d...ownloadbds.htm linked from http://www.pvconsultants.com/audio/frdgroup.htm (Not very easy to find from the main page ;-)) |
#93
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote:
: have not seen your program (mac user) so I don't know if it calculates : on the two corners. Have you compared your result with the Excel : Spreadsheet at : : http://www.pvconsultants.com : : I've lost the direct link to the Spreadsheet, unfortunately. http://www.pvconsultants.com/audio/d...ownloadbds.htm linked from http://www.pvconsultants.com/audio/frdgroup.htm (Not very easy to find from the main page ;-)) |
#94
|
|||
|
|||
Baffle edge diffraction model, comments?
Look at the old Olsen measurements, should be a reference on the web
somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio |
#95
|
|||
|
|||
Baffle edge diffraction model, comments?
Look at the old Olsen measurements, should be a reference on the web
somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio |
#96
|
|||
|
|||
Baffle edge diffraction model, comments?
Look at the old Olsen measurements, should be a reference on the web
somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio |
#97
|
|||
|
|||
Baffle edge diffraction model, comments?
Look at the old Olsen measurements, should be a reference on the web
somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio |
#98
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message
(Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. I have not seen your program (mac user) so I don't know if it calculates on the two corners. No. If you desperately want to run it I have heard of windows emulators for mac, but I know nothing about them. But the program is very simple and mostly a test to try if the baffle "step" can be calculated by means of edge sources. It should not be seen as a "end user product". I posted it to get real measurements to compare with, and input from others (which I have, thank you all). In several cases I have been positively surprised to see features of the real measurements coincide with the model. Since I wrote it, I have also learned that others have written similar things too, both more and less complicated. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com No, I can't find anything about loudspeakers there. Is the link correct? |
#99
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message
(Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. I have not seen your program (mac user) so I don't know if it calculates on the two corners. No. If you desperately want to run it I have heard of windows emulators for mac, but I know nothing about them. But the program is very simple and mostly a test to try if the baffle "step" can be calculated by means of edge sources. It should not be seen as a "end user product". I posted it to get real measurements to compare with, and input from others (which I have, thank you all). In several cases I have been positively surprised to see features of the real measurements coincide with the model. Since I wrote it, I have also learned that others have written similar things too, both more and less complicated. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com No, I can't find anything about loudspeakers there. Is the link correct? |
#100
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message
(Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. I have not seen your program (mac user) so I don't know if it calculates on the two corners. No. If you desperately want to run it I have heard of windows emulators for mac, but I know nothing about them. But the program is very simple and mostly a test to try if the baffle "step" can be calculated by means of edge sources. It should not be seen as a "end user product". I posted it to get real measurements to compare with, and input from others (which I have, thank you all). In several cases I have been positively surprised to see features of the real measurements coincide with the model. Since I wrote it, I have also learned that others have written similar things too, both more and less complicated. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com No, I can't find anything about loudspeakers there. Is the link correct? |
#101
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message
(Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. I have not seen your program (mac user) so I don't know if it calculates on the two corners. No. If you desperately want to run it I have heard of windows emulators for mac, but I know nothing about them. But the program is very simple and mostly a test to try if the baffle "step" can be calculated by means of edge sources. It should not be seen as a "end user product". I posted it to get real measurements to compare with, and input from others (which I have, thank you all). In several cases I have been positively surprised to see features of the real measurements coincide with the model. Since I wrote it, I have also learned that others have written similar things too, both more and less complicated. Have you compared your result with the Excel Spreadsheet at http://www.pvconsultants.com No, I can't find anything about loudspeakers there. Is the link correct? |
#102
|
|||
|
|||
Baffle edge diffraction model, comments?
(Edu Silva) wrote in message . com...
Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio Yes, thanks. That's a good example. T |
#103
|
|||
|
|||
Baffle edge diffraction model, comments?
(Edu Silva) wrote in message . com...
Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio Yes, thanks. That's a good example. T |
#104
|
|||
|
|||
Baffle edge diffraction model, comments?
(Edu Silva) wrote in message . com...
Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio Yes, thanks. That's a good example. T |
#105
|
|||
|
|||
Baffle edge diffraction model, comments?
(Edu Silva) wrote in message . com...
Look at the old Olsen measurements, should be a reference on the web somewhere. Olle Mirsch should have an article in MoLT; No 1, 1994. Yes, I will. A good article at True Audio by John Murphy, based on Harry Olson's work: http://www.trueaudio.com/st_diff1.htm HTH, Edu Silva ES2 Audio Yes, thanks. That's a good example. T |
#106
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " |
#107
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " |
#108
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " |
#109
|
|||
|
|||
Baffle edge diffraction model, comments?
(Svante) wrote in message . com...
(Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " |
#110
|
|||
|
|||
Baffle edge diffraction model, comments?
Svante wrote:
Interesting! Could you share these measurements, if possible? I'll try to remember this as something to come back to - possibly on my website - in case I resume development of system8, but they will be unconclusive because the baffle measured is "non-simple", consequently the cabinet effects are eqally non-simple. Thx. -- ******************************************* * My site is at: http://www.muyiovatki.dk * ******************************************* |
#111
|
|||
|
|||
Baffle edge diffraction model, comments?
Svante wrote:
Interesting! Could you share these measurements, if possible? I'll try to remember this as something to come back to - possibly on my website - in case I resume development of system8, but they will be unconclusive because the baffle measured is "non-simple", consequently the cabinet effects are eqally non-simple. Thx. -- ******************************************* * My site is at: http://www.muyiovatki.dk * ******************************************* |
#112
|
|||
|
|||
Baffle edge diffraction model, comments?
Svante wrote:
Interesting! Could you share these measurements, if possible? I'll try to remember this as something to come back to - possibly on my website - in case I resume development of system8, but they will be unconclusive because the baffle measured is "non-simple", consequently the cabinet effects are eqally non-simple. Thx. -- ******************************************* * My site is at: http://www.muyiovatki.dk * ******************************************* |
#113
|
|||
|
|||
Baffle edge diffraction model, comments?
Svante wrote:
Interesting! Could you share these measurements, if possible? I'll try to remember this as something to come back to - possibly on my website - in case I resume development of system8, but they will be unconclusive because the baffle measured is "non-simple", consequently the cabinet effects are eqally non-simple. Thx. -- ******************************************* * My site is at: http://www.muyiovatki.dk * ******************************************* |
#114
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message . com...
(Svante) wrote in message . com... (Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: I can understand that, don't worry. I found the program, it seems to do good stuff. I have not found out how to place the driver on the baffle, but I guess that would be found in the documentation. "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " I realise that this is not your explanation, but BDS's. I have some problems with this viewpoint, I think it is confusing. For sure, if we think of a circular baffle with the driver at the centre, the situation outlined above would occur. There would be one reflection from the first edge, and one from the second edge. And possibly this could be viewed as two separate comb filters working in parallel. But I think this explanation is unnessecarily complicated, combining two filters that run in parallel is not a very intuitive task. Some zeroes may be cancelled, that is easy to realise, but new would occur when the two signals are of equal magnitude but opposite phase. The model seems especially non-intuitive in the case of a non-circular baffle, where the durations of the two impulse responses would overlap. Since the only difference between the first and the second reflection is the time at which they occur, and that this also is the difference between different loactions along the (possibly non-circular) edge, it is IMO easiest thought of as ONE impulse response, and the fourier transform of this single impulse response would correspond to the frequency response. It makes little sense to separate the two reflections and then combine the filters. It is not wrong, but IMO confusing. But of course, for those who like that viewpoint, it is OK, and possibly it can also help to understand things from different aspects. Thanks for your input! |
#115
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message . com...
(Svante) wrote in message . com... (Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: I can understand that, don't worry. I found the program, it seems to do good stuff. I have not found out how to place the driver on the baffle, but I guess that would be found in the documentation. "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " I realise that this is not your explanation, but BDS's. I have some problems with this viewpoint, I think it is confusing. For sure, if we think of a circular baffle with the driver at the centre, the situation outlined above would occur. There would be one reflection from the first edge, and one from the second edge. And possibly this could be viewed as two separate comb filters working in parallel. But I think this explanation is unnessecarily complicated, combining two filters that run in parallel is not a very intuitive task. Some zeroes may be cancelled, that is easy to realise, but new would occur when the two signals are of equal magnitude but opposite phase. The model seems especially non-intuitive in the case of a non-circular baffle, where the durations of the two impulse responses would overlap. Since the only difference between the first and the second reflection is the time at which they occur, and that this also is the difference between different loactions along the (possibly non-circular) edge, it is IMO easiest thought of as ONE impulse response, and the fourier transform of this single impulse response would correspond to the frequency response. It makes little sense to separate the two reflections and then combine the filters. It is not wrong, but IMO confusing. But of course, for those who like that viewpoint, it is OK, and possibly it can also help to understand things from different aspects. Thanks for your input! |
#116
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message . com...
(Svante) wrote in message . com... (Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: I can understand that, don't worry. I found the program, it seems to do good stuff. I have not found out how to place the driver on the baffle, but I guess that would be found in the documentation. "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " I realise that this is not your explanation, but BDS's. I have some problems with this viewpoint, I think it is confusing. For sure, if we think of a circular baffle with the driver at the centre, the situation outlined above would occur. There would be one reflection from the first edge, and one from the second edge. And possibly this could be viewed as two separate comb filters working in parallel. But I think this explanation is unnessecarily complicated, combining two filters that run in parallel is not a very intuitive task. Some zeroes may be cancelled, that is easy to realise, but new would occur when the two signals are of equal magnitude but opposite phase. The model seems especially non-intuitive in the case of a non-circular baffle, where the durations of the two impulse responses would overlap. Since the only difference between the first and the second reflection is the time at which they occur, and that this also is the difference between different loactions along the (possibly non-circular) edge, it is IMO easiest thought of as ONE impulse response, and the fourier transform of this single impulse response would correspond to the frequency response. It makes little sense to separate the two reflections and then combine the filters. It is not wrong, but IMO confusing. But of course, for those who like that viewpoint, it is OK, and possibly it can also help to understand things from different aspects. Thanks for your input! |
#117
|
|||
|
|||
Baffle edge diffraction model, comments?
(Thomas A) wrote in message . com...
(Svante) wrote in message . com... (Thomas A) wrote in message (Svante) wrote in message But, anyway, these are small effects, and the program does not take the response or directivity of the driver into account etc, so probably there is nothing to worry about. Your edges have a greater angle than 90 degrees, which should decrease the amplitude of the edge reflections, which is good. On the other hand you will have a second edge reflection from the back of the box. Well there will be a beat pattern from this arrangement where the second edge which will cancel the effect of the first diffraction edge. With two corners there are two comb filters. There will be a destructive phase alignment of the two reflecting filters with a center f and bandwith which is determined by the bevel edge size. Errhh... I don't think I understand this. In my world edge reflections will add up regardless of if they come from the first edge or the second edge. It may very well be that a reflection from the first edge and some direction, con occur at the same time as another reflection in some other direction from the second edge. It does not matter if the reflections come from the first or the second edge, only the time (and amplitude) in important. Add up the source, it's mirror and all reflections (regardless of origin) and you will have an impulse response. The fourier transform of this impulse response would be the transfer function of the system (and include the baffle step and ripple). This is what my program does. I could include a second border for the second edge, we'll see about that. The links for the Excel spreadsheet and the manual is given by another person in the thread. At the moment I am a little overworked, but I cite a little more from the Baffle Diffraction Simulator manual. Maybe it helps: I can understand that, don't worry. I found the program, it seems to do good stuff. I have not found out how to place the driver on the baffle, but I guess that would be found in the documentation. "If you sharply turn a baffle corner, there is a pressure change that sends a reflection back at the listener, which arrives late and causes a comb filter frequency effect. If you turn that same corner, and instead of finding a single 90 degree sharp turn you pass over two 45 degree sharp edges forming a bevel, you create two such filters. The distance to the median point between those two edges is the center of a notch-filtering band. The distance between those edges is a width of notch effect. The item filtered by that notch filter is the cancellation of the comb filter like warbling. In all actuality, it is a beat frequency pattern, not a notch at all. The first nodal destruction is a half wavelength sum and the periodicity is every harmonic thereafter. But it still resembles a notch-filtered band, and as that is the way it appears when you view its effects in the frequency domain, so I improperly refer to it as a notch filter. So. If you sharply turn a baffle corner there is a reflection that comb filters. If you turn that same corner with a two sharp edges forming a bevel you create two filters. The distance to those two edges is the center of a notch-filtering band. The item filtered by that notch filter is the cancellation warbling. For that narrow region there is significant cancellation of the comb filtering effect because there is a destructive phase alignment of the two reflecting filters. Above that frequency and below that frequency you see increased evidence of the combing effect. Below that region the phase difference is not significant (they remain correlated) so the two reflections sum more or less. Above that region the phase difference is again changing with frequency and so appear to be uncorrelated, so they sum and difference at the harmonics of the notch but for only tiny frequency windows and with only a modest amplitude effect in total. " I realise that this is not your explanation, but BDS's. I have some problems with this viewpoint, I think it is confusing. For sure, if we think of a circular baffle with the driver at the centre, the situation outlined above would occur. There would be one reflection from the first edge, and one from the second edge. And possibly this could be viewed as two separate comb filters working in parallel. But I think this explanation is unnessecarily complicated, combining two filters that run in parallel is not a very intuitive task. Some zeroes may be cancelled, that is easy to realise, but new would occur when the two signals are of equal magnitude but opposite phase. The model seems especially non-intuitive in the case of a non-circular baffle, where the durations of the two impulse responses would overlap. Since the only difference between the first and the second reflection is the time at which they occur, and that this also is the difference between different loactions along the (possibly non-circular) edge, it is IMO easiest thought of as ONE impulse response, and the fourier transform of this single impulse response would correspond to the frequency response. It makes little sense to separate the two reflections and then combine the filters. It is not wrong, but IMO confusing. But of course, for those who like that viewpoint, it is OK, and possibly it can also help to understand things from different aspects. Thanks for your input! |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
More cable questions! | Tech | |||
When did the dual-woofer trend catch on? | Tech | |||
Comments on Polk/Momo 10" boxed sub model MM10 | Car Audio |