Crossover Network

[ladyphoebe15]

I plan to use this Crossover Network design for my power amplifier with an output of 100W. Is this design good enough.

 

[audioguru]

The circuit is very simple so it will sound bad and it might damage your speakers.

The inductor in series with the woofer speaker is useless because the woofer is already an inductor. You need to add a capacitor parallel with the woofer speaker to make a lowpass filter.
The series resonant LC in series with the midrange speaker provides a peak at only one frequency instead of being a bandpass filter.
The capacitor in series with the tweeter is so simple that it cuts low frequencies gradually then the tweeter will be damaged by midrange and low frequencies.

 

[eTech]

Hi

The component values should be adjusted for crossover frequencies that match the response of the each speaker.

The circuit is very simple so it will sound bad and it might damage your speakers.
The inductor in series with the woofer speaker is useless because the woofer is already an inductor. You need to add a capacitor parallel with the woofer speaker to make a lowpass filter.

But it does help block higher frequencies before reaching the woofer voice coil where the power would be wasted.

The capacitor in series with the tweeter is so simple that it cuts low frequencies gradually then the tweeter will be damaged by midrange and low frequencies.

But chosen at the proper value, it does help block low frequencies.

eT

 

[Nigel Goodwin]

AG is completely wrong about the circuit - it's perfectly fine as long as it's designed correctly - it's just a poorer quality and cheaper crossover than used in better designs.

 

[audioguru]

Most woofers produce a "shriek" sound at about 4kHz due to a resonance and cone "breakup". A single inductor in series barely reduces this bad sound.
You do not want a series resonant LC feeding a mid-range speaker.
Tweeters are fragile and can be damaged if only a single series capacitor cuts low frequencies. OOPs, Now I notice that it has a proper LC highpass filter but its phase is reversed at its crossover frequency.

I wanted to post the frequency response of a woofer and a good crossover network but my PDF program is stuck showing the datasheet for a Japanese dual Mosfet.

 

[Nigel Goodwin]

Most woofers produce a "shriek" sound at about 4kHz due to a resonance and cone "breakup". A single inductor in series barely reduces this bad sound.
You do not want a series resonant LC feeding a mid-range speaker.
Tweeters are fragile and can be damaged if only a single series capacitor cuts low frequencies.

There are many millions of speakers working perfectly well, and sounding 'reasonably' decent, on crossovers as cheap as this one, and many even considerably worse.

As an absolute minimum, you feed the bass unit directly, and have a single capacitor feeding the tweeter - ensuring that the 'crossover' occurs high enough to prevent excessive excursion of the tweeter cone. Most cheap speakers are as crude as this.

 

[audioguru]

An 15" woofer sound is worse than I thought. An 8" woofer shrieks at 4kHz and it is not difficult for a simple lowpass coil to reduce but a 15" woofer "honks" at about 1.2kHz.
It would take a pretty good lowpass network to reduce it and still have some bass bandwidth.

Here is from the datasheet of a typical cheap American 15" woofer:

 

[Nigel Goodwin]

Here is from the datasheet of a typical cheap American 15" woofer:

I suggest you try listening to commercial speaker designs (particularly cheap ones), rather than reading datasheets - which are pretty meaningless.

But regardless, stick a 15" or 8" speaker in a box, play music through it (no crossover of any kind) and it will neither 'shriek' nor 'honk'

You might be interested in more details of practical crossovers, with the simplest here:

https://colomar.com/Shavano/crossover6db.html

 

[audioguru]

I hear cheap commercial speaker systems in grocery stores and bowling alleys. They honk and shriek. Most people think that is how a normal speaker is supposed to sound.
Experts say that this mid-high peak in the frequency response of a woofer is caused by cone breakup.

They say,"At low frequencies a cone moves as a whole. This is the 'pistonic' area of operation.
At higher frequencies the cone starts to flex, leading to resonances. This is what is referred to as 'breakup'.
These resonances are at fixed frequencies and are thus not harmonically related to the input signal.
Non-harmonic frequencies like this sound bad, and so should be avoided like the plague."

The experts also say,"In my opinion, if you can hear woofer break up, it is the worst sounding noise an otherwise undamaged speaker can make. It means the crossover is poorly designed. Start over.
Most people blame woofer break up noise on the tweeter only because the noise is high frequency."

One expert shows that the frequency response peak caused by cone breakup also causes horrible distortion:

 

[eTech]

I plan to use this Crossover Network design for my power amplifier with an output of 100W. Is this design good enough.

Hi

Crossover networks range in complexity from simple to complex filters.
In your diagram, remove L3 and you have a simple 3-way crossover network.
Whether it is "good enough" depends on what you expect to hear.
If it sounds good to you...then its good enough..

eT

 

[audioguru]

Using the very simple crossover network with a 100W amplifier then the "twitter" will be the first speaker to burn out.
You have 4.7uF for C2 to block low frequencies from damaging the twitter. If the twitter is 8 ohms and L3 is missing then the maximum power will be 50W at 4300Hz. At 2150Hz the power will be 25W. At 1075Hz the twitter will probably be resonating and destroying itself with 12.5W. I looked at the datasheet of a pretty good tweeter and it recommends a crossover higher than 1800Hz using an LC 12dB/octave network.

 

[audioguru]

I suggest you try listening to commercial speaker designs (particularly cheap ones), rather than reading datasheets - which are pretty meaningless.
But regardless, stick a 15" or 8" speaker in a box, play music through it (no crossover of any kind) and it will neither 'shriek' nor 'honk'

Hi again, Nigel.
I listen to my car radio a lot. It sounds pretty good until I listen to its details then I hear the speakers in the front doors Shrieking!
The cause is probably cone breakup from each wideband speaker (there is a tweeter above each one).
The speakers in the front doors are supposed to be woofers but they probably do not have a lowpass filter or notch filter to reduce the shrieking sounds. I would add filters if the speakers or radio were easier to access.

On the car radio I like some songs and hate other songs. One song that I like a lot became very popular recently so it is played a lot on FM radio stations. The shrieking sound from the door speakers is in tune with that song!
Then do I like the song or do I like that it is in tune with the shrieking?? Most of the songs that I hate are country (hee, haw) , RAP (somebody talking to the beat of background noises) or are out-of-tune with the shrieking.

My son bought a round little speaker for his tablet. It is about 2.5 inches in diameter, has a built in power amplifier and rechargeable battery. It is heavy and sounds pretty good.
I can make it HONK by cupping it in my hands. Then music or speech honks at one frequency peak something like a car horn.

 

[eTech]

Hi

Here is a sample 3 way crossover network with values generated using a crossover calculator.
I didn't spend any time choosing specific speakers. Just trying out the calculator.

eT

 

[audioguru]

Hi E-Tech,
Your midrange has a sharp shrieking peak in the response at 8kHz. It will be very noticeable to anybody with normal hearing.
I could not find a Dayton DC-60FA midrange speaker but their DC-50FA does not have that peak. Maybe the peak is caused by the crossover network feeding the inductive high impedance of the midrange speaker. Then a Zobel Network parallel with the midrange speaker will reduce its impedance at 8kHz to 8 ohms so that the crossover network can work properly. Does the crossover calculator assume that the speakers have Zobel Networks installed?
The crossover frequency between the midrange and tweeter is very high at about 9kHz. Maybe it should be reduced to about 4kHz?

Likewise, the midrange response shows its honking resonance at 400Hz where it might destroy itself. Then maybe the crossover frequency between the woofer and midrange should be increased to 800Hz?
This midrange speaker is causing problems at both ends of its response, maybe it should be removed and use a two-way crossover instead?

I like the response of the little 8" woofer giving excellent output at 20Hz. I do not like the tweeter having too much output.

Lastly, the crossovers are 2nd-order. Then when the speakers are in-phase like you show then they form a notch at each crossover frequency instead of adding.

 

[ronsimpson]

eTech,
Where did you get "tweeter.lib"?
Please attach you spice and lib files. I would like to give it a go.
You can plot V(tweeter)+V(mid)+V(wolf) to get the total frequency response.
Ron,

----edited----
I don't know the frequency response of the raw speakers so we really don't know the total frequency response of the system.

 

[ronsimpson]

Here is the frequency response of a Dayton 385-8. There is a nasty bump at 1.2khz. Hopefully the cross over network will role that off.
The frequency response below 200hz will depend greatly on the enclosure.

AND
Here is the impedance of this speaker. I don't know if the SPICE model takes into account the inductive effect above 200hz.

 

[eTech]

Hi E-Tech,
Your midrange has a sharp shrieking peak in the response at 8kHz. It will be very noticeable to anybody with normal hearing.
I could not find a Dayton DC-60FA midrange speaker but their DC-50FA does not have that peak. Maybe the peak is caused by the crossover network feeding the inductive high impedance of the midrange speaker. Then a Zobel Network parallel with the midrange speaker will reduce its impedance at 8kHz to 8 ohms so that the crossover network can work properly. Does the crossover calculator assume that the speakers have Zobel Networks installed?

I used the DC-50FA-8 in the simulation. Might be the model.
I've added the Zobel Network componets to each speaker model and attached a new simulation.
Let me know what you think.

The crossover frequency between the midrange and tweeter is very high at about 9kHz. Maybe it should be reduced to about 4kHz?

Likewise, the midrange response shows its honking resonance at 400Hz where it might destroy itself. Then maybe the crossover frequency between the woofer and midrange should be increased to 800Hz?
This midrange speaker is causing problems at both ends of its response, maybe it should be removed and use a two-way crossover instead?

I like the response of the little 8" woofer giving excellent output at 20Hz. I do not like the tweeter having too much output.

Lastly, the crossovers are 2nd-order. Then when the speakers are in-phase like you show then they form a notch at each crossover frequency instead of adding.

I didn't spend much time choosing the speakers, I pretty much chose them at random.
I just wanted to play with the calculators and compare them with network simulations.
Thanks for your input.

eT

 

[eTech]

eTech,
Where did you get "tweeter.lib"?
Please attach you spice and lib files. I would like to give it a go.
You can plot V(tweeter)+V(mid)+V(wolf) to get the total frequency response.
Ron,

----edited----
I don't know the frequency response of the raw speakers so we really don't know the total frequency response of the system.

Hi Ron

Sorry for delay but I'm not retired yet....

I made the symbols and libraries myself and I'm consolidating the libraries into a single speaker.lib file. The speakers are modeled using specifications from speaker datasheets such as Re, Le, Fs, Qms, Qes, to compute values for Cone suspension, Cone Mass, and Cone and Suspension losses. They also contain Zobel Network impedance correction. I use the Parts-Express web site to obtain speaker data, because they specialize in speaker building components and most of the speakers they carry also have data sheets. All the speakers I used in the sample simulation are available at Parts-Express.

Thanks for your input.
I let you know when I've completed the lib.

eT

 

[eTech]

Here is the frequency response of a Dayton 385-8. There is a nasty bump at 1.2khz. Hopefully the cross over network will role that off.
The frequency response below 200hz will depend greatly on the enclosure.
View attachment 88875
AND
Here is the impedance of this speaker. I don't know if the SPICE model takes into account the inductive effect above 200hz.
View attachment 88876

Hi Ron

Can you provide the test parameters that were used to get those plots?

thx

eT

 

[ronsimpson]

No,

This is off the data sheet.

I've added the Zobel Network componets to each speaker model and attached a new simulation.

You know more than I do.
I would have used a 8 ohm resistor and a inductor to get the impedance effect above 25hz.
The 20hz bump, maybe a cap and 80 ohms resistor, (maybe). I would have to play with it for a while. I think your method is better.

I see that the Ls and Cs in the cross over are working into the impedance of the speaker so I would like the impedance in the model to be some what close.