Rbnr asks a question in an old post
There seems much less concern about lining up the voice coils/time alignment.
What are your thoughts/experiences on this?
Hi
I think it was Richard Heyser who first wrote about and then figured out how to measure a loud speakers acoustic phase. In his AES papers he referred to a frequency dependant phase shift as an effect similar to a forward / back change in the physical position of the radiation. I was fortunate to have met Dick and worked at a company that bought one of the “TDS” or TEF machines and had used them regularly developing and testing acoustics for the last…well approaching 30 years now.
Look at this situation from an alternative point of view;
For now, don’t think time, think frequency.
Phase is one cycle divided into 360 increments. A filter of a given type may impose a 180 degree phase shift at some point in it’s response. If you change the filter frequency, one finds the shape of the phase curve and magnitude is unchanged, only the frequency is shifted upward /downward. This is a “minimum phase” relationship where any and all changes in the amplitude curve also has a specific accompanying phase shift..
As a result, it is a mistake to look at a time figure without remembering where you are in frequency.
For example, people are often concerned about a subwoofer that has a very low corner but also has a large Group Delay. In reality, by having a low corner, one automatically has a longer time period wave and so everything, including Group delay is naturally larger.
Heysers View of phase is very useful, if the speakers phase lags say 90 degrees at frequency x, then, at that frequency, that signal is late or behind by the amount of 90 degrees at that frequency.
So, if one has in instrument like Heyser’s Time Delay Spectrometry that actually measures the speakers response in time, one finds that even a single driver generally spreads a broad band signal out in time and does so according to it’s acoustic phase shift vs frequency.
Adding two drivers together and you have new problems. Each driver has it’s own band pass response and phase. Each driver has a small internal delay relative to the time the signal arrives at the terminals, this delay is like the time delay that exists between the mic and driver except that this is within the driver, a result of it’s series inductance and mass.
So, what one see’s is the low frequency driver’s output begins a little after the hf driver even if their VC’s are coincident. A big woofer may have a time delay equal to a foot or more making it a requirement to actually measure the acoustic part and not with a ruler. With each driver having it’s own mag and phase response, how can they be added?
If you have electrical signals, you can add two signals with resistors, they sum each signal into one new signal, this is coherent addition the way most people picture things.
Take two subwoofer and place them close together and they sum coherently into one new source. That is coherent addition also and if one were to walk in a circle with a sound level meter, they sound find (outdoors) that the sound was radiating in all directions equally. In order to have coherent addition of acoustic sources, they MUST be ¼ wavelength apart or less.
AS soon as the spacing is larger than say ½ wavelelngth, then one has left Coherent addition and produced an interference pattern. For example now if one walked the circle, one finds a figure 8 radiation pattern, two lobes, two nulls. This (producing an interference pattern) is how the vast majority of multi-way speakers operate.
With coherent addition, if one were to have reversed one of the subwoofers, then one would have nearly killed all the sound by canceling it out with an equal but opposite source. When the spacing is larger than 1/3 wavelength then one produces an interference pattern. Now, with the two sources farther apart in incoherent addition, an interference pattern is produced and now if one were to reverse one of two sources, the total acoustic power is hardly changed, only the interference or radiation pattern changes. If one were to repeat the “walk in a circle” with a sound level meter, one would see the horizontal “Polar pattern” of the interference, a series of lobes and nulls. The farther apart the sources are (in wavelengths) the more lobes and nulls are produced. In the example at ½ wavelength, one finds a reversal of polarity only rotates the pattern 90 degrees.
So, in the real world, most drivers are not close enough together to sum coherently and are producing an interference pattern instead, each driver has it’s own magnitude, phase and time of origin so the idea of making a speaker “speak” over its entire band at the same instant is a daunting task.
The coax driver is not free of any of these issues, most have some pretty terrible interference where the cone and horn transition. A very few have properties that lend themselves to a seamless transition.
Here is a speaker I designed about 6 years ago that uses a nice coax driver. Here I added a short additional horn to the cone which acts as a horn up high, this produces a lower pattern loss point and adds some efficiency to the cone section.
http://www.danleysoundlabs.com/pdf/SH 100 Spec Sheet.PDF
Notice that (in addition to whoever measured it had it inverted) there is no big phase shifted where he crossover over would be (around 1k).
Normal crossovers sum flat but also produce a phase shift going from well above to well below crossover of 90 degrees per order (once above first order). This sum exhibits an all pass filters response, the lower part is progressively delayed in time.
To be “true” to the input signal, a theoretical goal might be to replicate the input signal looking at a measurement microphone.
To reproduce a tough broad band signal like a square wave, one needs to have flat magnitude and phase, an all pass response will wreck it.
In the case of that coax driver above (a B&C 8 inch fwiw), the hf driver is behind the woofer.
I was able to use a variation of the crossovers I use in our horn loaded speakers which uses the offset to eliminate the phase shift.
Those Horns might be interesting to, they are something like a coax driver and while made for a much larger room than a living room, they do preserve waveshape and are very hifi. The SH-50 can reproduce a square wave from near perfect to fair looking, continuously from 250Hz to about 2700Hz. Because all the drivers are less than a quarter wavelength apart, they all add coherently, there are no lobes and nulls because there is no interference. There is a cutaway and explanation here;
http://www.danleysoundlabs.com/pdf/danley_tapped.pdf
They are not quite as good in Time as the Manger I measured but they go at 30-40 dB louder and have about 1 / 400 the harmonic distortion in the mid band at 90dB SPL and have near constant directivity.
It’s funny, the reason I ended up with them acting this way (coherent addition, radiating as a single source) was because where most of our speakers go, there is a large area to cover and the “ideal” is to have the spectrum be the same everywhere in the pattern.
It is largely the interference pattern that multiple sources produce that causes the large variations in response across an audience. The larger the distance to the audience, the more acoustic power is needed, the more drivers are required, the interference worse. These horns allow large acoustic powers but no interference.
Anyway, if you’re a diy’r, look at the B&C coax, it’s a nice driver.
Best,
Tom Danley
There seems much less concern about lining up the voice coils/time alignment.
What are your thoughts/experiences on this?
Hi
I think it was Richard Heyser who first wrote about and then figured out how to measure a loud speakers acoustic phase. In his AES papers he referred to a frequency dependant phase shift as an effect similar to a forward / back change in the physical position of the radiation. I was fortunate to have met Dick and worked at a company that bought one of the “TDS” or TEF machines and had used them regularly developing and testing acoustics for the last…well approaching 30 years now.
Look at this situation from an alternative point of view;
For now, don’t think time, think frequency.
Phase is one cycle divided into 360 increments. A filter of a given type may impose a 180 degree phase shift at some point in it’s response. If you change the filter frequency, one finds the shape of the phase curve and magnitude is unchanged, only the frequency is shifted upward /downward. This is a “minimum phase” relationship where any and all changes in the amplitude curve also has a specific accompanying phase shift..
As a result, it is a mistake to look at a time figure without remembering where you are in frequency.
For example, people are often concerned about a subwoofer that has a very low corner but also has a large Group Delay. In reality, by having a low corner, one automatically has a longer time period wave and so everything, including Group delay is naturally larger.
Heysers View of phase is very useful, if the speakers phase lags say 90 degrees at frequency x, then, at that frequency, that signal is late or behind by the amount of 90 degrees at that frequency.
So, if one has in instrument like Heyser’s Time Delay Spectrometry that actually measures the speakers response in time, one finds that even a single driver generally spreads a broad band signal out in time and does so according to it’s acoustic phase shift vs frequency.
Adding two drivers together and you have new problems. Each driver has it’s own band pass response and phase. Each driver has a small internal delay relative to the time the signal arrives at the terminals, this delay is like the time delay that exists between the mic and driver except that this is within the driver, a result of it’s series inductance and mass.
So, what one see’s is the low frequency driver’s output begins a little after the hf driver even if their VC’s are coincident. A big woofer may have a time delay equal to a foot or more making it a requirement to actually measure the acoustic part and not with a ruler. With each driver having it’s own mag and phase response, how can they be added?
If you have electrical signals, you can add two signals with resistors, they sum each signal into one new signal, this is coherent addition the way most people picture things.
Take two subwoofer and place them close together and they sum coherently into one new source. That is coherent addition also and if one were to walk in a circle with a sound level meter, they sound find (outdoors) that the sound was radiating in all directions equally. In order to have coherent addition of acoustic sources, they MUST be ¼ wavelength apart or less.
AS soon as the spacing is larger than say ½ wavelelngth, then one has left Coherent addition and produced an interference pattern. For example now if one walked the circle, one finds a figure 8 radiation pattern, two lobes, two nulls. This (producing an interference pattern) is how the vast majority of multi-way speakers operate.
With coherent addition, if one were to have reversed one of the subwoofers, then one would have nearly killed all the sound by canceling it out with an equal but opposite source. When the spacing is larger than 1/3 wavelength then one produces an interference pattern. Now, with the two sources farther apart in incoherent addition, an interference pattern is produced and now if one were to reverse one of two sources, the total acoustic power is hardly changed, only the interference or radiation pattern changes. If one were to repeat the “walk in a circle” with a sound level meter, one would see the horizontal “Polar pattern” of the interference, a series of lobes and nulls. The farther apart the sources are (in wavelengths) the more lobes and nulls are produced. In the example at ½ wavelength, one finds a reversal of polarity only rotates the pattern 90 degrees.
So, in the real world, most drivers are not close enough together to sum coherently and are producing an interference pattern instead, each driver has it’s own magnitude, phase and time of origin so the idea of making a speaker “speak” over its entire band at the same instant is a daunting task.
The coax driver is not free of any of these issues, most have some pretty terrible interference where the cone and horn transition. A very few have properties that lend themselves to a seamless transition.
Here is a speaker I designed about 6 years ago that uses a nice coax driver. Here I added a short additional horn to the cone which acts as a horn up high, this produces a lower pattern loss point and adds some efficiency to the cone section.
http://www.danleysoundlabs.com/pdf/SH 100 Spec Sheet.PDF
Notice that (in addition to whoever measured it had it inverted) there is no big phase shifted where he crossover over would be (around 1k).
Normal crossovers sum flat but also produce a phase shift going from well above to well below crossover of 90 degrees per order (once above first order). This sum exhibits an all pass filters response, the lower part is progressively delayed in time.
To be “true” to the input signal, a theoretical goal might be to replicate the input signal looking at a measurement microphone.
To reproduce a tough broad band signal like a square wave, one needs to have flat magnitude and phase, an all pass response will wreck it.
In the case of that coax driver above (a B&C 8 inch fwiw), the hf driver is behind the woofer.
I was able to use a variation of the crossovers I use in our horn loaded speakers which uses the offset to eliminate the phase shift.
Those Horns might be interesting to, they are something like a coax driver and while made for a much larger room than a living room, they do preserve waveshape and are very hifi. The SH-50 can reproduce a square wave from near perfect to fair looking, continuously from 250Hz to about 2700Hz. Because all the drivers are less than a quarter wavelength apart, they all add coherently, there are no lobes and nulls because there is no interference. There is a cutaway and explanation here;
http://www.danleysoundlabs.com/pdf/danley_tapped.pdf
They are not quite as good in Time as the Manger I measured but they go at 30-40 dB louder and have about 1 / 400 the harmonic distortion in the mid band at 90dB SPL and have near constant directivity.
It’s funny, the reason I ended up with them acting this way (coherent addition, radiating as a single source) was because where most of our speakers go, there is a large area to cover and the “ideal” is to have the spectrum be the same everywhere in the pattern.
It is largely the interference pattern that multiple sources produce that causes the large variations in response across an audience. The larger the distance to the audience, the more acoustic power is needed, the more drivers are required, the interference worse. These horns allow large acoustic powers but no interference.
Anyway, if you’re a diy’r, look at the B&C coax, it’s a nice driver.
Best,
Tom Danley
