unity crossover design problem in akabak
Hi, I wonder if anyone has a way round this problem. As I understand it the normal way to design a two-way speaker in akabak is to model each driver or horn as a seperate system with filters to simulate the crossover; flatten the impedance of each section; and then synthesise a crossover network for each section into the load impedance.
the problem I have been having with my unity design is that it looks like it sums flat when simulating the mids and high as two seperate system each with its own horn but when I replace the filters with LRC components and put it all together into one system there is a 5dB dip at the crossover frequency.
I thought maybe I had made a mistake but it now occurs to me that the design process I'm using is flawed because the mids and highs are in reality acoustically coupled onto the same horn. The reason for the 5dB dip when I put the two systems together is that there is only half the horn mouth area radiating as in the two seperate systems I have used to design the crossover.
Filters can only be applied to a whole system or to every system. you can't put the mids and highs in the same system and have one filter apply to the mids and another filter to the high. also I have been flattening the impedance of mids and high systems seperately.
I'm thinking I could just design my two seperate filter systems to give a 6dB bump at crossover knowing it will return to flat when the system is put together. my worry is that this is still missing important aspects of the unity system because the mids and highs are acoustically coupled at the throat of the horn in the crossover region - possibly having a large effect on their phase relationship.
can anyone think of a way round this or should I just try and bodge it?
cheers
Phil
Hi, I wonder if anyone has a way round this problem. As I understand it the normal way to design a two-way speaker in akabak is to model each driver or horn as a seperate system with filters to simulate the crossover; flatten the impedance of each section; and then synthesise a crossover network for each section into the load impedance.
the problem I have been having with my unity design is that it looks like it sums flat when simulating the mids and high as two seperate system each with its own horn but when I replace the filters with LRC components and put it all together into one system there is a 5dB dip at the crossover frequency.
I thought maybe I had made a mistake but it now occurs to me that the design process I'm using is flawed because the mids and highs are in reality acoustically coupled onto the same horn. The reason for the 5dB dip when I put the two systems together is that there is only half the horn mouth area radiating as in the two seperate systems I have used to design the crossover.
Filters can only be applied to a whole system or to every system. you can't put the mids and highs in the same system and have one filter apply to the mids and another filter to the high. also I have been flattening the impedance of mids and high systems seperately.
I'm thinking I could just design my two seperate filter systems to give a 6dB bump at crossover knowing it will return to flat when the system is put together. my worry is that this is still missing important aspects of the unity system because the mids and highs are acoustically coupled at the throat of the horn in the crossover region - possibly having a large effect on their phase relationship.
can anyone think of a way round this or should I just try and bodge it?
cheers
Phil
unity in akabak
I think in akabak the radiators at the horn mouths are by default placed at position 0,0,0. the drivers are tapped in between different waveguide sections so this should give the correct physical alignment.
I think in akabak the radiators at the horn mouths are by default placed at position 0,0,0. the drivers are tapped in between different waveguide sections so this should give the correct physical alignment.
Hi All, Bill
I use Akabak a lot, it is a great program for modeling arbitrary shaped things like real horns. It was through Akabak that I found the “rules” that made the Tapped horn work also, a very handy program.
Like ANY computer prediction program, it is only as useful as it’s ability to predict what you will measure before you build the real thing and so like many programs, one has to model, build, measure and alter the model to fit reality. For example, even Akabak is usually WAY off predicting the upper frequency range of a Tapped horn and always predicts more “stuff” than you measure.
A proper measurement trumps a computer model, what you measure IS what you have to deal with and computer models often suffer from a lack of input details or oversimplifications.
Bill, I have not forgotten your idea either, I bring it up about every two weeks, while I thought there was progress, it’s still not where I can say it’s a go. To be clear, it is still something I would like to see happen.
I have been down at the shop recently for the company Christmas party and to catch up on things so I have not been on line as much as usual.
I am pleased to say that in the last three weeks I have heard from the patent office that the Synergy Horn, Paraline, Tapped Horn and Shaded Amplitude horn patents have been allowed so even if Santa brings me nothing, I have had a good Month.
So, in the Christmas spirit, I would offer a couple comments to the DIY’rs.
Bill is spot on, to figure out the crossover, you need to have measurements which include the phase shift caused by the different locations in space.
When two sources radiate in a space a quarter wavelength across or less, they add something like signals through resistors into one new signal depending on the magnitude and phase of each source.
When the spacing has reached about a half wavelength or greater, the two sources radiate independently and begins to produce an interference pattern (lobes and nulls in the polar plots) because now what you get is the vector sum based on the path lengths to each source plus each sources magnitude and phase. Add a crossover and you add another set of magnitude and phase relationships to the existing system.
We often think of a crossover as a free standing entity and for the simple cases we have names like Butterworth, L&R, Bessel and so on.
These are all based on mathematical relationships or other aspects of the filter and when one employs a text book case to a loudspeaker one may or may not get something resembling what the book said because loudspeakers are often far from a simple resistor as an electrical load.
Crossovers for these do not have names, they are all based on what is needed case by case to make the upper and lower bands sum without the normal phase shift.
These are not the normal way they work or how one designs a crossover and you NEED one of the advanced crossover designer programs like LSPcad but it is obviously possible and with some additional effort, this can be done with a passive crossover, the Synergy horn geometry provides the physical offsets roughly needed to offset the phase shift the crossover slopes impose individually and ideally sum without a step phase or in Group delay (as if it had just one driver in time/phase, radiation pattern etc).
My point here would be the crossover (especially if electronic) is where one has the greatest flexibility AND potentially you have a computer program than can curve fit with an infinite number of shapes and arbitrary time delays hint hint. If you have such a program now, start by making an imaginary crossover with just filters, what do you have to do to eliminate the step in phase shift and Group delay using a higher order crossover?
Merry Christmas all!
Tom
I use Akabak a lot, it is a great program for modeling arbitrary shaped things like real horns. It was through Akabak that I found the “rules” that made the Tapped horn work also, a very handy program.
Like ANY computer prediction program, it is only as useful as it’s ability to predict what you will measure before you build the real thing and so like many programs, one has to model, build, measure and alter the model to fit reality. For example, even Akabak is usually WAY off predicting the upper frequency range of a Tapped horn and always predicts more “stuff” than you measure.
A proper measurement trumps a computer model, what you measure IS what you have to deal with and computer models often suffer from a lack of input details or oversimplifications.
Bill, I have not forgotten your idea either, I bring it up about every two weeks, while I thought there was progress, it’s still not where I can say it’s a go. To be clear, it is still something I would like to see happen.
I have been down at the shop recently for the company Christmas party and to catch up on things so I have not been on line as much as usual.
I am pleased to say that in the last three weeks I have heard from the patent office that the Synergy Horn, Paraline, Tapped Horn and Shaded Amplitude horn patents have been allowed so even if Santa brings me nothing, I have had a good Month.
So, in the Christmas spirit, I would offer a couple comments to the DIY’rs.
Bill is spot on, to figure out the crossover, you need to have measurements which include the phase shift caused by the different locations in space.
When two sources radiate in a space a quarter wavelength across or less, they add something like signals through resistors into one new signal depending on the magnitude and phase of each source.
When the spacing has reached about a half wavelength or greater, the two sources radiate independently and begins to produce an interference pattern (lobes and nulls in the polar plots) because now what you get is the vector sum based on the path lengths to each source plus each sources magnitude and phase. Add a crossover and you add another set of magnitude and phase relationships to the existing system.
We often think of a crossover as a free standing entity and for the simple cases we have names like Butterworth, L&R, Bessel and so on.
These are all based on mathematical relationships or other aspects of the filter and when one employs a text book case to a loudspeaker one may or may not get something resembling what the book said because loudspeakers are often far from a simple resistor as an electrical load.
Crossovers for these do not have names, they are all based on what is needed case by case to make the upper and lower bands sum without the normal phase shift.
These are not the normal way they work or how one designs a crossover and you NEED one of the advanced crossover designer programs like LSPcad but it is obviously possible and with some additional effort, this can be done with a passive crossover, the Synergy horn geometry provides the physical offsets roughly needed to offset the phase shift the crossover slopes impose individually and ideally sum without a step phase or in Group delay (as if it had just one driver in time/phase, radiation pattern etc).
My point here would be the crossover (especially if electronic) is where one has the greatest flexibility AND potentially you have a computer program than can curve fit with an infinite number of shapes and arbitrary time delays hint hint. If you have such a program now, start by making an imaginary crossover with just filters, what do you have to do to eliminate the step in phase shift and Group delay using a higher order crossover?
Merry Christmas all!
Tom
Hi Tom,
Many congratulations on your patents! That's quite a bundle there. Nothing like official recognition that you've made history!
Thanks also for keeping the possibility alive for my idea. Anything I can do to help with the decision let me know, maybe a meeting or something?
Merry Christmas to you and to everyone at diyAudio and DSL.
Bill
Many congratulations on your patents! That's quite a bundle there. Nothing like official recognition that you've made history!
Thanks also for keeping the possibility alive for my idea. Anything I can do to help with the decision let me know, maybe a meeting or something?
Merry Christmas to you and to everyone at diyAudio and DSL.
Bill
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.