SERIES CROSSOVERS
Series crossovers, how do they work? The fundamentals of series crossovers
Ranting on
Instead of another finished speaker design I thought it would be interesting to try and explain how the series crossover actually works. Which components are the active ones and how much do they influence the other components and/or the drivers.
With a standard parallel filter the filter functions are easy to comprehend and implement. After all each driver has a separate dedicated network to tailor each driver individually. These separate networks are then connected next to each other (thus in parallel) and bingo: music! (Maybe not music, but at least sound). This is what I believe (in my humble opinion) is the main problem with parallel filters. You feed a signal, the music* from the source, through an amplifier to the drivers via separate networks! People actually split up the signal into nice bite size pieces for each driver to cope with. Just think of it, it would be like recording a song where each instrument and vocalist was recorded separately on different tracks at different times. Like first you record the drum and bass parts, then a few days later the keyboard player comes along and does his bit, hey wait a minute, this sounds like modern popular music! How can they ever make music* if they weren’t there together at the same time? I am ranting on a bit here, sorry, but you get my drift. This is why I believe or feel that series crossovers in basis must be better in reproducing the “one-ness” of music – the crossover and drivers work and interact together instead of beside each other. I know this is not scientifically based and just a “feeling” but for what its worth I believe in it. With the article below I will try and explain the electrical function of series crossovers. I will not go into a deep technical explanation supported by complex mathematical formulae but try and explain the fundamentals of series crossovers in an easy to comprehend manner supported with clear graphics showing the direct consequences of each alteration to the filter. This first article if focused on a standard 2-way speaker starting with a basic 1st order crossover, via a hybrid design and 2nd order crossovers to 3rd order crossovers with full compensation networks.
* Music is more than notes varying in length and loudness coming from an instrument or a voice placed in time. Music is an elementary form of expressing emotions. It should also include every surrounding detail that was present during the recording. Things like room acoustics, breathing musicians and most important of all: musical interaction. Unfortunately most of the time we have to listen to conserved recorded music but on the other hand the whole thing about this hobby is to retrieve the musical event in the most satisfying way possible – without it costing too much financially! Luckily I am in the position of living only 25km away from the Royal Concertgebouw in Amsterdam, so every now and then I go along to check that my hobby is still going in the right direction – and enjoy the performance of course!
Driver choice for a standard 2-way configuration
Often it is stated that you have to use drivers with a smooth roll off to well above and below the crossover frequency, so no cone break-ups etc. This is correct for 1st order networks because if you place a notch filter parallel to the woofer it will automatically be in the signal path of the tweeter resulting in suck-outs in the tweeter response (all a notch filter does is to drop the impedance at the resonance frequency of the notch). If you move up to 2nd order for at least woofer then it is possible to implement a notch filter as can be seen with the DD8 speaker. When the notch filter is connected parallel to a driver the electric energy will choose the path with the lowest resistance and therefore draw away the energy at that point from the woofer. The relatively large inductor in series with the woofer acts as a sort of protection for the tweeter. Lets say the notch filter makes the impedance drop by 6 ohms at the cone break-up frequency of the woofer. At this point the inductor has a resistance of say about 70 ohms so the dip of 6 ohms created by the notch filter is relatively small and therefore won’t effect the tweeter.
To keep things simple and also to be able to explain 1st order filters I have chosen easy to filter drivers for this article:
● The woofer is the Seas CB17RCY/P as used in the Proteus
● The tweeter is the Scanspeak D2905-9700 as used in the Geers-eVe-II (hey that would make a nice little bookshelf speaker)
● The drivers are time aligned so that the graphs show full integration of the units without phase shifts added due to different distances from the speakers to the listener (this means sloping the baffle for example)
● Compared to the 1st order networks, with 2nd and 3rd order networks (versions 4a+b+c and version 5) the woofer had to set back 2cm to maintain the best phase integration at fx
● To maintain the best phase integration at fx the tweeter is connected out of phase in all cases except for the 3rd order network
● I have taken the liberty of varying the tweeter L-pad sometimes to match the sensitivity of the tweeter to that of the woofer
● The low frequency response below about 350Hz is a free-field closed box simulation. The graph doesn’t take into account the lift of the bass and lower mid-range region when placed in a room
1a – simple basic 1st order without any form of compensation
1st order network with values often found in designs on the net
● Very high crossover point at about 5700Hz
● Woofer is hardly filtered at all
● Tweeter has a gradual high-pass function; the SPL curve flattens out around 1kHz before it drops again below 600Hz
● Relatively smooth impedance curve
● Total output level shows a bump around fx
1b – same as 1a but with larger values for L1 and C1
Trying to lower fx
● Crossover point still rather high at 4200Hz even though large values for L and C are used
● Woofer is better attenuated
● Tweeter has to work a bit harder; the SPL curve still flattens out around 1kHz before it drops again below 600Hz
● Relatively smooth impedance curve
● Total output level shows a bump around fx
1c – a combination of 1a and 1b with large value for L1 and small value for C1
Trying to protect the tweeter better
● Very high crossover point at about 5700Hz
● Woofer is hardly filtered at all but its output level is attenuated a little compared to version 1a
● Tweeter has a reasonable high-pass function but the SPL curve still flattens out around 1kHz before it drops again below 600Hz
● Relatively smooth impedance curve although a little higher than the former two versions
● Total output level shows a bump around fx
2a – crossover 1a with a CR-network added for the woofer
Trying to get the overall output level smoother
● High crossover point at about 5100Hz
● Woofer is better attenuated; it actually looks like it has a “roll-off”
● Tweeter has a smooth and gradual high-pass function and the output around 1kHz doesn’t flatten anymore
● Nice smooth impedance curve
● Total output level shows no bump; the pass band of both drivers shows +/- 1dB
● A Zobel network for the woofer looks like a must with 1st order networks
2b – same as 2a but with larger values for L1 and C1
Trying to lower fx
● Crossover point lowered to about 3000Hz
● Woofer output lowered and flattened
● Tweeter has to work harder; the output around 1kHz still doesn’t flatten anymore
● Smooth impedance curve
● The pass band of both drivers shows +/- 1dB but is about 1,5dB’s lower than version 2a
● A Zobel network for the woofer looks like a must with 1st order networks
3 – adding a second inductor, baffle step compensation?
Trying to get a lower crossover point and by coincidence compensate baffle step diffraction at the same time
● Crossover point is now lowered to a classic 2400Hz (note that L1 is now small and C1 large)
● Woofer output shows a perfect low pass function
● Tweeter has a sharper attenuation below fx but now the SPL curve flattens out again around 1kHz before it drops again below 600Hz
● The impedance curve shows a steeper and higher peak
● The pass band of both drivers shows +/- 1dB but is about 1,5dB’s lower than version 2b
● Acoustic and electric phase show greater fluctuations
● Overall output level looks ideal considering that the graph doesn’t take into account the lift of the bass and lower mid-range region when placed in a room.
4a – basic 2nd order crossover with a Zobel network on the woofer
Trying to give the tweeter maximum protection.
● Crossover point is a classic 2400Hz (L2 and C1 had to be changed a little to get a flat overall response)
● Woofer output shows a perfect low pass function
● Tweeter has strong attenuation below fx and the SPL curve doesn’t flatten out as much around 1kHz before it drops again below 600Hz
● The impedance curve shows a steeper and higher peak
● The pass band of both drivers shows +/- 1dB but is about 1dB lower than version 3
● Acoustic and electric phase show greater fluctuations
● Overall output level looks ideal considering that the graph doesn’t take into account the lift of the bass and lower mid-range region when placed in a room.
4b – 2nd order crossover with a Zobel network on the woofer and a LCR on the tweeter
Trying to make the tweeter high-pass function more “correct”.
The Scanspeak D2905-9700 tweeter has no Ferro-fluid in the air gap and therefore has a high impedance peak at fs (nearly 30 ohms).
Adding an LCR network parallel to the tweeter should flatten its impedance so that the high-pass can function properly.
The 10-ohm resistor in the L-pad obviously wasn’t enough.
● Crossover point is a fraction lower at 2200Hz (C1 had to be enlarged a little to get a flat overall response, therefore fx is a little lower)
● Woofer output shows a perfect low pass function (note that it is not effected by the tweeters LCR network)
● Tweeter shows a near perfect high pass function
● The impedance curve shows a steep and high peak
● The pass band of both drivers shows +/- 1dB and is the same as version 4a
● Acoustic and electric phase show great fluctuations
● Overall output level looks ideal considering that the graph doesn’t take into account the lift of the bass and lower mid-range region when placed in a room.
4c – 2nd order crossover with full options
Fully corrected impedance curve except for the woofer peak at resonance. This peak will vary depending on the cabinet load of the woofer and therefore I have not compensated it.
● Crossover point is the same as version 4b
● Woofer output is the same as version 4b
● Tweeter output is the same as version 4b
● The pass band of both drivers is the same as version 4b
● Acoustic and electric phase are the same as version 4b
● Overall output level is the same as version 4b
● The impedance curve is now flat with a smooth phase curve (5,5 ohms +/- 0,5 ohms from 200Hz – 20kHz)
5 – 3rd order crossover with full options
Trying out the not so often used 3rd order series crossover.
● Crossover point at about 2400Hz
● Woofer output has a steep roll-off above fx
● Tweeter output has a steep roll-off below fx (note that the tweeter is now connected in phase)
● Tweeter has maximum protection, which raises power handling
● The pass band of both drivers shows +/- 1dB although a fraction less smooth than the other crossovers
● Acoustic phase shows large fluctuations
● Impedance dips a little from 4-5kHz
● Rather complex to get right due to the large amount of variables
Conclusions?
These examples purely indicate the electric functions of various series networks. How they actually sound, I have left out on purpose to stay objective. As with all crossovers (parallel or series) the higher the order of the network the better the attenuation outside the high-pass and low-pass functions, the easier it is to get a flat overall response, but the steeper the phase shifts and the lower the sensitivity becomes. In a series filter the energy is divided across all the components and drivers, so if you add a component the energy it takes is taken from the other components. That is why the efficiency of the system dropped each time the crossover became more complex. With parallel filters the extra components just drop the impedance so the speaker/filter combination just draws more current from the amp to maintain efficiency. There you go again, parallel filters try and add their own things to music*
Tony Gee
The Netherlands
November 2002