A rarely discussed crossover is the Series Crossover (in contrast to the universally popular Parallel Crossover). It offers considerable advantages especially with a subwoofer and a bass midrange. I used a series passive crossover in the Bitches Brew Live Edge Dipoles (which are featured in the new October 2023 issue of AudioXpress). Here is the schematic:
ABOVE: In a series crossover, you wire the woofer and midrange in series. You short the woofer with a capacitor. You short the midrange with an inductor.
The circuit above is what I used in the Bitches Brew design. The Bitches Brew has (2) SB Acoustics 15OB350 subwoofers and (1) B&C 15CXN88 coax used as a bass midrange. Note the reverse polarity of the midrange; usually but not always required.
This is in contrast to the usual, BELOW, where the woofer is in series with an inductor and the midrange in series with a capacitor (and both sections in parallel).
ABOVE: Look closely because this is super interesting.
I'm showing the woofer half of the coax. I'm using actual measured impedance of the drivers in my model, and showing the filter voltage at the driver terminals. The parallel circuit above is not "optimized" for anything except to illustrate the problems with standard parallel crossovers. It just happens to have the same values I used in the series circuit.
In the subwoofer section, the inductor interacts with the impedance of the 15OB350s to create a peak around 50Hz. Then it rolls off above 100Hz.
In the midrange section, the capacitor interacts with the impedance of the 15CXN88s to create a dip at 100Hz, followed by a 6dB peak at 35Hz. Very bad!!! This +5.5dB peak at 35Hz is HIGHLY undesirable and happens ALL THE TIME in subwoofer-mid passive xovers. The reactance of the capacitor cancels the inductance of the bass-mid just below the bass-mid resonance and creates a peak exactly where you least want it.
(The average speaker designer is likely not aware that a passive crossover can BOOST, rather than cut the voltage below the crossover frequency! I discuss this phenomenon in detail in the thread Open Baffle Bass Boost: +4 to +7dB w/ Passive Xover, No DSP where I show some cool ways to use this to your advantage. Especially in dipoles.)
Usually, nobody aware it is even going on. You build a 100Hz crossover for your subwoofer and satellite and you don't even know the satellites are getting a few dB EXTRA bass at, say, 60Hz. Right where it's most harmful.
There is no easy fix for this in a conventional parallel crossover.
However in a series crossover all of the impedance peaks work in your favor instead of against you.
#1: The impedance peak of the subwoofer cuts the signal received by the bass-mid by over 20dB at 28Hz. Which is exactly where we most want to protect it. Don't forget, this is 'just a 6dB per octave crossover' yet the electrical slope between 60Hz and 30Hz is closer to 24dB per octave (!) which is fantastic. It's effectively a steep 18dB shelf filter.
#2: The impedance peak of the bass-midrange (which is about 60Hz) interacts with the crossover to boost response 3dB. While this is not desirable per se, the roll off above that point is immediate. In my opinion, the mild 6dB/octave slope above 100Hz helps the subs blend better with the mids.
I corrected the peak at 60Hz with DSP. In most bass-mid crossovers, the midrange would have a higher resonant frequency than 60Hz and would be even easier to work with.
Above: Impedance of the circuit. In this design the 60Hz 2ohm minimum is OK for me. It's possible to choose different crossover values to achieve a higher Zminimum. Actual measurement is below (I've added a Zobel to flatten the impedance.)
In an earlier discussion with @Lynn Olson we discussed the difficulty of getting subwoofers and bass-mids to seamlessly integrate. This is a very good way to do it especially because the cone sizes are all the same. In this design, both subs and bass-mid are 15" and both in the same dipole array.
The Bitches Brews are bi-amped with DSP, with the active crossover between the coax midrange and tweeter. I could have triamped them, but I felt that a series crossover as shown here was much more simple and elegant. Fewer amps, fewer DSP channels, fewer signal cables and speaker cables.
I feel air core inductors are worth the extra money for an "ultimate" no compromise design like this; I didn't want the hysteresis of iron core inductors. So I used Madisound air core copper foil inductors with DCR below 0.6 ohm (low DCR is very important, the higher the DCR the less attenuation of low frequencies that you get in the midrange signal).
That's especially important in this design, because the DSP boosts low bass 12-15dB around 30Hz and the deep notch in the midrange filter is very desirable there. I can shove hundreds of watts at 30Hz into these, and the midrange cone only moves +/- 1mm or 2mm.
The Bitches Brew has one of the most seamless crossovers I've ever heard. The acoustic crossover between subs and mids is at about 90Hz, all the cones are 13" in diameter, and the coax tweeter integrates perfectly with the midrange. The radiation pattern is extremely well behaved and there are no lobing problems at any frequency.
Above: Measured polar response, +/-90 degrees, 1/3 octave. I don't have a good way to measure these at low frequencies, but VituixCad calculations are below.
Above: VituixCad vertical directivity calculations for 2x 15OB350 and 1x 15CXN88. Variation is only about +/-2dB.
Expanded details on the Bitches Brew crossover are here. @Yourmando
ABOVE: In a series crossover, you wire the woofer and midrange in series. You short the woofer with a capacitor. You short the midrange with an inductor.
The circuit above is what I used in the Bitches Brew design. The Bitches Brew has (2) SB Acoustics 15OB350 subwoofers and (1) B&C 15CXN88 coax used as a bass midrange. Note the reverse polarity of the midrange; usually but not always required.
This is in contrast to the usual, BELOW, where the woofer is in series with an inductor and the midrange in series with a capacitor (and both sections in parallel).
ABOVE: Look closely because this is super interesting.
I'm showing the woofer half of the coax. I'm using actual measured impedance of the drivers in my model, and showing the filter voltage at the driver terminals. The parallel circuit above is not "optimized" for anything except to illustrate the problems with standard parallel crossovers. It just happens to have the same values I used in the series circuit.
In the subwoofer section, the inductor interacts with the impedance of the 15OB350s to create a peak around 50Hz. Then it rolls off above 100Hz.
In the midrange section, the capacitor interacts with the impedance of the 15CXN88s to create a dip at 100Hz, followed by a 6dB peak at 35Hz. Very bad!!! This +5.5dB peak at 35Hz is HIGHLY undesirable and happens ALL THE TIME in subwoofer-mid passive xovers. The reactance of the capacitor cancels the inductance of the bass-mid just below the bass-mid resonance and creates a peak exactly where you least want it.
(The average speaker designer is likely not aware that a passive crossover can BOOST, rather than cut the voltage below the crossover frequency! I discuss this phenomenon in detail in the thread Open Baffle Bass Boost: +4 to +7dB w/ Passive Xover, No DSP where I show some cool ways to use this to your advantage. Especially in dipoles.)
Usually, nobody aware it is even going on. You build a 100Hz crossover for your subwoofer and satellite and you don't even know the satellites are getting a few dB EXTRA bass at, say, 60Hz. Right where it's most harmful.
There is no easy fix for this in a conventional parallel crossover.
However in a series crossover all of the impedance peaks work in your favor instead of against you.
#1: The impedance peak of the subwoofer cuts the signal received by the bass-mid by over 20dB at 28Hz. Which is exactly where we most want to protect it. Don't forget, this is 'just a 6dB per octave crossover' yet the electrical slope between 60Hz and 30Hz is closer to 24dB per octave (!) which is fantastic. It's effectively a steep 18dB shelf filter.
#2: The impedance peak of the bass-midrange (which is about 60Hz) interacts with the crossover to boost response 3dB. While this is not desirable per se, the roll off above that point is immediate. In my opinion, the mild 6dB/octave slope above 100Hz helps the subs blend better with the mids.
I corrected the peak at 60Hz with DSP. In most bass-mid crossovers, the midrange would have a higher resonant frequency than 60Hz and would be even easier to work with.
Above: Impedance of the circuit. In this design the 60Hz 2ohm minimum is OK for me. It's possible to choose different crossover values to achieve a higher Zminimum. Actual measurement is below (I've added a Zobel to flatten the impedance.)
In an earlier discussion with @Lynn Olson we discussed the difficulty of getting subwoofers and bass-mids to seamlessly integrate. This is a very good way to do it especially because the cone sizes are all the same. In this design, both subs and bass-mid are 15" and both in the same dipole array.
The Bitches Brews are bi-amped with DSP, with the active crossover between the coax midrange and tweeter. I could have triamped them, but I felt that a series crossover as shown here was much more simple and elegant. Fewer amps, fewer DSP channels, fewer signal cables and speaker cables.
I feel air core inductors are worth the extra money for an "ultimate" no compromise design like this; I didn't want the hysteresis of iron core inductors. So I used Madisound air core copper foil inductors with DCR below 0.6 ohm (low DCR is very important, the higher the DCR the less attenuation of low frequencies that you get in the midrange signal).
That's especially important in this design, because the DSP boosts low bass 12-15dB around 30Hz and the deep notch in the midrange filter is very desirable there. I can shove hundreds of watts at 30Hz into these, and the midrange cone only moves +/- 1mm or 2mm.
The Bitches Brew has one of the most seamless crossovers I've ever heard. The acoustic crossover between subs and mids is at about 90Hz, all the cones are 13" in diameter, and the coax tweeter integrates perfectly with the midrange. The radiation pattern is extremely well behaved and there are no lobing problems at any frequency.
Above: Measured polar response, +/-90 degrees, 1/3 octave. I don't have a good way to measure these at low frequencies, but VituixCad calculations are below.
Above: VituixCad vertical directivity calculations for 2x 15OB350 and 1x 15CXN88. Variation is only about +/-2dB.
Expanded details on the Bitches Brew crossover are here. @Yourmando
Last edited:
Not scaling the L and C values in the parallel example is confusing for many folks.
This is sort of an apple to oranges comparison.
In either case, these are large value components required at this frequency.
When you add up all the pro's and con's, parallel-configured crossovers are generally better. That's why the vast majority of folks use them.
Dave.
This is sort of an apple to oranges comparison.
In either case, these are large value components required at this frequency.
When you add up all the pro's and con's, parallel-configured crossovers are generally better. That's why the vast majority of folks use them.
Dave.
Last edited:
@Davey per your suggestion, I scaled the capacitor from 400uF down to 100uF; I changed the inductor from 6mH down to 3mH.
I still get this nasty peak in my midrange driver at 48Hz. This is the last thing we want from a high pass filter. I also get a 7dB dip at 100Hz.
I can't find any values that result in a good crossover. I don't see that the parallel configuration offers any advantages here. It doesn't seem to reduce the necessary size of passive components, which at low frequencies are costly.
I have attached my VituixCad project files if you'd like to run a simulation on these components.
I still get this nasty peak in my midrange driver at 48Hz. This is the last thing we want from a high pass filter. I also get a 7dB dip at 100Hz.
I can't find any values that result in a good crossover. I don't see that the parallel configuration offers any advantages here. It doesn't seem to reduce the necessary size of passive components, which at low frequencies are costly.
I have attached my VituixCad project files if you'd like to run a simulation on these components.
Last edited:
That peak is associated with the driver itself. Looking at the B&C 15CXN88 data sheet it appears to be at 47Hz and well over 100 ohms. You can't do anything about that unless you add a large-value conjugate filter.
So, when you look at the electrical drive on the downstream side of your high-pass capacitor.....that's what you see.
This is an inherent characteristic when doing passive crossovers down low in frequency like this. In that sense, I understand the premise of your thread.
Preferably....in my opinion....I would always do something like this with an active-crossover bi-amp setup.
Dave.
So, when you look at the electrical drive on the downstream side of your high-pass capacitor.....that's what you see.
This is an inherent characteristic when doing passive crossovers down low in frequency like this. In that sense, I understand the premise of your thread.
Preferably....in my opinion....I would always do something like this with an active-crossover bi-amp setup.
Dave.
Yes off course you can always opt for DSP instead. It's like using nuclear weapons instead of a bow and arrow. But DSP is not always practical or economical. The point of my post is that in a series network, all the impedance disadvangages flip to advantages.
In this application the series network is far superior. You solve solve the entire problem with one inductor and one capacitor.
I wager the same is true for many other subwoofer satellite passive crossovers.
In this application the series network is far superior. You solve solve the entire problem with one inductor and one capacitor.
I wager the same is true for many other subwoofer satellite passive crossovers.
That's right, the series crossover does have a special tendency to turn some unfortunate circumstances around. This is why it has been so popular in the past.
It should be understood that if you have enough time and components, you can eventually produce the same result with either style of crossover. When you boil it down though, where series shines is that it can give a more complex interaction for fewer components. In a regular parallel crossover the fewer components, the simpler the shaping that can be achieved, but a series crossover has greater interaction with and between the driver impedances. If you want to control that interaction you need to go back to adding more components but it's fascinating how the natural impedances sometimes already seem to be in the right place at the right time.
It should be understood that if you have enough time and components, you can eventually produce the same result with either style of crossover. When you boil it down though, where series shines is that it can give a more complex interaction for fewer components. In a regular parallel crossover the fewer components, the simpler the shaping that can be achieved, but a series crossover has greater interaction with and between the driver impedances. If you want to control that interaction you need to go back to adding more components but it's fascinating how the natural impedances sometimes already seem to be in the right place at the right time.
You can't do anything about that unless you add a large-value conjugate filter.
Wrong!
You can do exactly as Perry suggested and reap all the benefits.
I use only 6dB Series filters in all my builds and this is exactly the behavior I see.
I know people are hesitant using Series XO, especially 1st order, because they think they lose the imaginary control of the parallel filters. Actually it is exactly the opposite, the simplicity and elegance of the 6db Series XO makes life so much easier as Perry and Allen explained above.
Try it in practice and see for yourself.
@Stanislav I was talking about the parallel-configuration. (You might have missed that point.)
In his speakers, with series crossover, the subwoofers function as sort of a quasi-conjugate network.
Series crossovers have their advantages and disadvantages........just like parallel crossovers do.
Series crossovers have been around for a looooooong time. There's nothing new here.
I'm sorry, I don't need to try it in practice and see for myself. I've built many series-crossover speakers through many decades.
The speakers that I'm listening to while I type this have a series crossover.
Let's not get carried away with the rhetoric.
Dave.
In his speakers, with series crossover, the subwoofers function as sort of a quasi-conjugate network.
Series crossovers have their advantages and disadvantages........just like parallel crossovers do.
Series crossovers have been around for a looooooong time. There's nothing new here.
I'm sorry, I don't need to try it in practice and see for myself. I've built many series-crossover speakers through many decades.
The speakers that I'm listening to while I type this have a series crossover.
Let's not get carried away with the rhetoric.
Dave.
Last edited:
Yes, exactly, you were talking about the parallel configuration hence my reply.
If there's nothing new as you say you should have refrained from disputing the obvious.
You better follow your own advice re. rhetoric.
I was glad that finally someone was willing to discuss 1st order series xo and their inherent advantages but alas there's always someone else who'll try to spoil the fun with uninvited arguments.
Stan
If there's nothing new as you say you should have refrained from disputing the obvious.
You better follow your own advice re. rhetoric.
I was glad that finally someone was willing to discuss 1st order series xo and their inherent advantages but alas there's always someone else who'll try to spoil the fun with uninvited arguments.
Stan
I didn't dispute the obvious. I pointed out the obvious.
DIYaudio.com is filled with numerous threads on series crossovers. (All the way back to this forum's beginnings in 2001.)
Also, there are numerous textbooks from many decades that outline series crossovers.
If you have a problem with rhetoric, I think you might talk to the fella who mentioned nuclear weapons and bows/arrows.
Dave.
DIYaudio.com is filled with numerous threads on series crossovers. (All the way back to this forum's beginnings in 2001.)
Also, there are numerous textbooks from many decades that outline series crossovers.
If you have a problem with rhetoric, I think you might talk to the fella who mentioned nuclear weapons and bows/arrows.
Dave.
Last edited:
Let's try a super common application that begs for a series Xover: A subwoofer satellite system. Here I'd like to show examples of both series and parallel Xover networks, using a ported subwoofer and both sealed and ported satellites.
The subwoofer: This is a Tang Band 8" subwoofer in a ported box, tuned to 32Hz resonance. Impedance below:
Satellite #1: Boston Acoustics HD5, which is a perfect example of a typical small satellite. Acoustic suspension enclosure. Fb = 100Hz. Impedance below:
Satellite #2: Fountek FE85 3.5" with passive radiator tuned to 90Hz. Impedance below:
Here are the driver terminal voltages using a parallel xover network with 8" TB Subwoofer + Boston HD5, below:
Comment: My #1 priority is protecting the satellite from low bass. Above, the satellite signal is -12dB at 50Hz.
Here are the driver terminal voltages from a series network, TB8 sub + Boston HD5 satellite, below:
Comment: With the series network at 50Hz the satellite drops to -15dB, which gives 5dB more protection than the parallel network. The dropoff below 70Hz is steeper. The overlap between the drivers is also better behaved.
Here is a parallel network, TB8 sub + Fountek 3.5" in ported box:
Comment: The 90Hz resonance of the vented satellite enclosure is the frequency where it can most effectively produce sound. But then vented boxes go downhill RAPIDLY below resonance. Excursion goes through the roof. We should try to remove everything below 75Hz [=Fb*0.85] because otherwise you just overload the system, producing minimal SPL and generating tons of distortion.
The parallel network is only -1dB at 65Hz and only -6dB at 50Hz. Not very effective. You can play with the component values all you want, but you still never get around this without adding extra components.
Below is a series network, TB 8" sub + Fountek ported satellite:
Comment: The behavior of this network is close to perfect. Signal to satellite drops like a rock below 70Hz. The reason for this is that the TB8 sub's impedance starts rising rapidly below 70Hz, peaking at its upper resonance of 50Hz. It interacts with the crossover to make the slope steeper.
So... any ported subwoofer tuned between 30-40Hz will work very nicely with this particular satellite.
A sealed sub might not have that convenient impedance peak at 50Hz, so will it still make a good crossover?
I simulated that too. Impedance of a sealed sub with 40Hz resonance:
Below is the series crossover using the 40Hz acoustic suspension sub and the 3.5" Fountek with passive radiator:
Comment: In this network above, the satellite also drops like a rock below 70Hz because of the rising impedance of the subwoofer's 40Hz resonance. This is exactly what we want it to do. It's nice that we can pull this off with only two passive xover components.
Yes, series crossovers must surely be 100+ years old. Nothing new about them...
...but I don't hear people talking about them much.
And with the modeling tools we have today like VituixCad, for most folks it's like discovering a whole new kind of crossover network. Cuz even though there are probably books from 1955 that talk about this, a 1955 reader couldn't run computer models like we can now. They would just have to build and measure.
A parallel network would require a lot of expensive components to fix these errors. But as @AllenB said, "it's fascinating how the natural impedances sometimes already seem to be in the right place at the right time."
The subwoofer: This is a Tang Band 8" subwoofer in a ported box, tuned to 32Hz resonance. Impedance below:
Satellite #1: Boston Acoustics HD5, which is a perfect example of a typical small satellite. Acoustic suspension enclosure. Fb = 100Hz. Impedance below:
Satellite #2: Fountek FE85 3.5" with passive radiator tuned to 90Hz. Impedance below:
Here are the driver terminal voltages using a parallel xover network with 8" TB Subwoofer + Boston HD5, below:
Comment: My #1 priority is protecting the satellite from low bass. Above, the satellite signal is -12dB at 50Hz.
Here are the driver terminal voltages from a series network, TB8 sub + Boston HD5 satellite, below:
Comment: With the series network at 50Hz the satellite drops to -15dB, which gives 5dB more protection than the parallel network. The dropoff below 70Hz is steeper. The overlap between the drivers is also better behaved.
Here is a parallel network, TB8 sub + Fountek 3.5" in ported box:
Comment: The 90Hz resonance of the vented satellite enclosure is the frequency where it can most effectively produce sound. But then vented boxes go downhill RAPIDLY below resonance. Excursion goes through the roof. We should try to remove everything below 75Hz [=Fb*0.85] because otherwise you just overload the system, producing minimal SPL and generating tons of distortion.
The parallel network is only -1dB at 65Hz and only -6dB at 50Hz. Not very effective. You can play with the component values all you want, but you still never get around this without adding extra components.
Below is a series network, TB 8" sub + Fountek ported satellite:
Comment: The behavior of this network is close to perfect. Signal to satellite drops like a rock below 70Hz. The reason for this is that the TB8 sub's impedance starts rising rapidly below 70Hz, peaking at its upper resonance of 50Hz. It interacts with the crossover to make the slope steeper.
So... any ported subwoofer tuned between 30-40Hz will work very nicely with this particular satellite.
A sealed sub might not have that convenient impedance peak at 50Hz, so will it still make a good crossover?
I simulated that too. Impedance of a sealed sub with 40Hz resonance:
Below is the series crossover using the 40Hz acoustic suspension sub and the 3.5" Fountek with passive radiator:
Comment: In this network above, the satellite also drops like a rock below 70Hz because of the rising impedance of the subwoofer's 40Hz resonance. This is exactly what we want it to do. It's nice that we can pull this off with only two passive xover components.
Yes, series crossovers must surely be 100+ years old. Nothing new about them...
...but I don't hear people talking about them much.
And with the modeling tools we have today like VituixCad, for most folks it's like discovering a whole new kind of crossover network. Cuz even though there are probably books from 1955 that talk about this, a 1955 reader couldn't run computer models like we can now. They would just have to build and measure.
A parallel network would require a lot of expensive components to fix these errors. But as @AllenB said, "it's fascinating how the natural impedances sometimes already seem to be in the right place at the right time."
Last edited:
I need a series crossover for a Polk-style SDA arrangement. 4 Ohm speakers; parallel the amp sees 2 Ohms at some frequencies. One speaker is bandpass (say, 100 to 2-5kHz), the other is full range. The bandpass one needs to appear as a resistor to the full range it's wired in series with.
I have no clue how to implement this. Seems if I put a cap in parallel with one series speaker, that effects the FR of the other. So it's parallel network I go - easy-peasy!
DSP? I'd need 6 channels of it; only have 4.
I have no clue how to implement this. Seems if I put a cap in parallel with one series speaker, that effects the FR of the other. So it's parallel network I go - easy-peasy!
DSP? I'd need 6 channels of it; only have 4.
What a coincidence Other than shifting the crossover point slightly higher, this would be a really good start.
I did only one quick shot measurement and did not get the expected extension down to 3-4 kHz, which I noticed too late.
Any suggested DCX crossover I have found and simulated was producing a huge wide dip at the crossover point. The drivers are bought new, so I do suspect my measurement, not the product.
Other than shifting the crossover point slightly higher, this would be a really good start. I did only one quick shot measurement and did not get the expected extension down to 3-4 kHz, which I noticed too late.
Any suggested DCX crossover I have found and simulated was producing a huge wide dip at the crossover point. The drivers are bought new, so I do suspect my measurement, not the product.