AllenB,
Thanks very much for this. I have to admit that speakers were my first effort at DIY audio and almost the end as I nearly gave up. I could never get them to sound quite right. I moved to amps, but alway wished I could someday make some good speakers. Your explanation on tweaking will have me pulling out those dusty old half built speakers and modifiying those crossovers with more insight.
Thanks again
Jim
Thanks very much for this. I have to admit that speakers were my first effort at DIY audio and almost the end as I nearly gave up. I could never get them to sound quite right. I moved to amps, but alway wished I could someday make some good speakers. Your explanation on tweaking will have me pulling out those dusty old half built speakers and modifiying those crossovers with more insight.
Thanks again
Jim
Regarding the baffle step, What we are actually talking about here is directivity rather than loss. There wouldn't be much sense in giving you a number between 1 and 6dB as the 'lost' sound is still interacting via the room.
So when tweaking by ear it is helpful to know that a baffle step can be dealt with in a crossover, but that the room can change things again to the point where it is hard to tell what's going on or it may even be the more significant issue. Sometimes when the room changes the sound it cannot be fixed again through a crossover, and must be dealt with in other ways.
I'll describe the baffle step, some crossover suggestions and the room issue.
A baffle is often made a little wider than the woofer, and woofers are often crossed just below their breakup region which is related to the size of the cone... so, the baffle step usually occurs within the woofer's passband and reaches up to somewhere near the crossover. I'll assume this is the case here.
Crossover Tweaks.
The inductor, resistor and capacitor as used in this thread can often deal with this issue at the same time as crossing over the woofer. All three components get used as free points of adjustment. From the basic crossover you'll want to reduce the upper woofer response gradually starting an octave or two below the crossover point. Below this the response shouldn't need to change. This region just below the crossover is referred to as the knee of the filtered response, as that's just what it looks like.
Reducing the knee is fairly simple. You can increase the value of the resistor. In doing this I might start by trying a value up to twice the original, and see if I can hear what changes. If you feel you want a value much higher than double you should try changing the other components because as the resistor gets higher, the woofer's highest frequencies above the crossover point will come back into play. It may not be significant, and if you can't hear it then there's probably something else worth chasing instead. By the way, a very large resistor effectively disables the capacitor.
You can increase the value of the inductor for a similar effect. Taken too far, this may reduce lower frequencies too, but won't have the issue of the higher ones coming back in. The filter rolloff will probably be more correct for the lowered top end of the woofer response as well. The inductor is also best not overdone unless your individual speaker responds well to it.
To reduce the knee and maintain the balance with larger adjustments, you should change all three components at once. For example, you might double the inductor and resistor values and halve the capacitance. The response should be similar to before above and below the knee, but the knee will be reduced. You can now change the resistor up or down a bit from that point.
Don't forget to lower the tweeter level as required. You'll notice this if it gets too far out of line because the tweeter will begin to stand out as if unsupported. If you listen to the tweeter with the woofer disconnected you'll hear this sound, it will seem out of place at it's lower end where it cuts off. When a woofer is properly integrated below it this bad sound disappears.
The baffle itself.
The speed of sound will be the same for all frequencies. The low frequencies cycle so slowly that by the time the cone has moved in and back out again, the pressure waves have moved significantly beyond the baffle. Effectively they can't 'see' it as it is too small. This pressure can then radiate freely in all directions.
The higher frequencies will already be radiating in a dome pattern from the front of the flat baffle (nothing to do with a tweeter dome), before they've even reached the edge.
The distance sound travels during one cycle is called its wavelength. The sound pressure measured from the front (say, outdoors in an open place) would be 6dB higher for the high frequencies, so if not for the room I'd probably say just use 6dB.
The room
Rooms give reflections. As the frequencies get lower and reach around the baffle, they can bounce off the rear wall, but all walls have their effect. Since there is a delay involved in the extra travel time, they may meet back up with the direct sound and add constructively, cancel, or anywhere in between. Usually you get a bit of everything.
Under some circumstance some of this energy may fill a part of the baffle step back to where it should be, and some will create dips in the response. Another concern is the midrange frequencies, that when exposed to the room, may give the impression of a lack of bass. The location of your speaker, the way it's facing, and room treatments can be worthwhile.
So when tweaking by ear it is helpful to know that a baffle step can be dealt with in a crossover, but that the room can change things again to the point where it is hard to tell what's going on or it may even be the more significant issue. Sometimes when the room changes the sound it cannot be fixed again through a crossover, and must be dealt with in other ways.
I'll describe the baffle step, some crossover suggestions and the room issue.
A baffle is often made a little wider than the woofer, and woofers are often crossed just below their breakup region which is related to the size of the cone... so, the baffle step usually occurs within the woofer's passband and reaches up to somewhere near the crossover. I'll assume this is the case here.
Crossover Tweaks.
The inductor, resistor and capacitor as used in this thread can often deal with this issue at the same time as crossing over the woofer. All three components get used as free points of adjustment. From the basic crossover you'll want to reduce the upper woofer response gradually starting an octave or two below the crossover point. Below this the response shouldn't need to change. This region just below the crossover is referred to as the knee of the filtered response, as that's just what it looks like.
Reducing the knee is fairly simple. You can increase the value of the resistor. In doing this I might start by trying a value up to twice the original, and see if I can hear what changes. If you feel you want a value much higher than double you should try changing the other components because as the resistor gets higher, the woofer's highest frequencies above the crossover point will come back into play. It may not be significant, and if you can't hear it then there's probably something else worth chasing instead. By the way, a very large resistor effectively disables the capacitor.
You can increase the value of the inductor for a similar effect. Taken too far, this may reduce lower frequencies too, but won't have the issue of the higher ones coming back in. The filter rolloff will probably be more correct for the lowered top end of the woofer response as well. The inductor is also best not overdone unless your individual speaker responds well to it.
To reduce the knee and maintain the balance with larger adjustments, you should change all three components at once. For example, you might double the inductor and resistor values and halve the capacitance. The response should be similar to before above and below the knee, but the knee will be reduced. You can now change the resistor up or down a bit from that point.
Don't forget to lower the tweeter level as required. You'll notice this if it gets too far out of line because the tweeter will begin to stand out as if unsupported. If you listen to the tweeter with the woofer disconnected you'll hear this sound, it will seem out of place at it's lower end where it cuts off. When a woofer is properly integrated below it this bad sound disappears.
The baffle itself.
The speed of sound will be the same for all frequencies. The low frequencies cycle so slowly that by the time the cone has moved in and back out again, the pressure waves have moved significantly beyond the baffle. Effectively they can't 'see' it as it is too small. This pressure can then radiate freely in all directions.
The higher frequencies will already be radiating in a dome pattern from the front of the flat baffle (nothing to do with a tweeter dome), before they've even reached the edge.
The distance sound travels during one cycle is called its wavelength. The sound pressure measured from the front (say, outdoors in an open place) would be 6dB higher for the high frequencies, so if not for the room I'd probably say just use 6dB.
The room
Rooms give reflections. As the frequencies get lower and reach around the baffle, they can bounce off the rear wall, but all walls have their effect. Since there is a delay involved in the extra travel time, they may meet back up with the direct sound and add constructively, cancel, or anywhere in between. Usually you get a bit of everything.
Under some circumstance some of this energy may fill a part of the baffle step back to where it should be, and some will create dips in the response. Another concern is the midrange frequencies, that when exposed to the room, may give the impression of a lack of bass. The location of your speaker, the way it's facing, and room treatments can be worthwhile.
Last edited:
That's why I said in theory the higher the frequencies the woofer has to reproduce, when wavelength get small enough to see the baffle it changes from 4P radiation to 2P radiation and would pick up another 6 dB in level on axis.
In our usual room then the speakers will be close enough to the floor and other boundaries, so that lower frequencies reflected from there will still be in phase with the initial wavefront to fill up the "loss".
So the 6 dB in reality never really happen.
If the woofer is crossed over too high, even if it has a very good cone and it does not break up quick, its surface will get too large to radiate higher frequencies because the frequencies from middle od the cone to the side of the cone will seen from an angle be severly out of phase. They will cancel out gradually and will cause a narrow coverage. A 6.5" woofer will radiate 2 kHz with 90° and above that narrow rapidly. A 8" woofer will be there at 1700 Hz.
That is lower than intuition would lead us. Constructions with a 8" woofer crossed over to a tweeter at 3K will have to deal with severe problems in coverage over frequencie and will sound different in every room.
The higher the woofer runs, the more periods away will the tweeter be to the center of the woofer and phase coherent crossing will be impossible to find.
Yes, it's not very exacting when doing things this way...you just need a good starting point. For many the baffle step will happen a bit lower than the crossover, requiring extra components. Not sure where I might have suggested this but in some cases adding a second inductor in series with the first, with a resistor across it the same value as the nominal driver impedance can give a starting point for speakers with a lower frequency baffle step.
@ Jack.
The thing with directivity through cone size is that it can come with side lobes. By the way, my first floor cancellation happens at just under 200Hz.
@ Jack.
The thing with directivity through cone size is that it can come with side lobes. By the way, my first floor cancellation happens at just under 200Hz.
Of course there are sidelobes. Another reason why the woofer should be crossed over not to high. There will be also a lobing between woofer and tweeter. Not useful for a good reproduction.
Floor cancelation @ 200 Hz maybe meassured on axis but will change under angle and will be averaged out by the room. Still not to relevant for stereo. Do you know about your cancelation or do you really hear it?
Floor cancelation @ 200 Hz maybe meassured on axis but will change under angle and will be averaged out by the room. Still not to relevant for stereo. Do you know about your cancelation or do you really hear it?
A crossover is by definition, a compromise. If we didn't need them we wouldn't have them, but it's not all that bad. Some things we just can't hear and these are being proven even in recent times.
I can't agree with your first two points. I run some woofers right on to the edge of their first breakup mode so I can exploit their narrowing pattern (and the response behaviour to increase the rolloff). The side lobes ask only to be plotted and accounted for, plus a nice round edge on the baffle if it's possible. My tweeters are also more than a wavelength above the woofer centre (44cm).
This may be getting a smidge off topic, but 200Hz is below the Schroeder frequency. Pick your room mode weapon of choice. Mine is the multi-sub approach.
I can't agree with your first two points. I run some woofers right on to the edge of their first breakup mode so I can exploit their narrowing pattern (and the response behaviour to increase the rolloff). The side lobes ask only to be plotted and accounted for, plus a nice round edge on the baffle if it's possible. My tweeters are also more than a wavelength above the woofer centre (44cm).
This may be getting a smidge off topic, but 200Hz is below the Schroeder frequency. Pick your room mode weapon of choice. Mine is the multi-sub approach.
You can run a midbass higher than a wavelength of its diameter. You can also have the tweeter further away than a wavelength at crossover. For listening music in a average room this can work. But if you look (listen) at your onaxis response compared to the overall generated magnitude vs frequencie, the diffuse field will not be flat when onaxis is.
If my speakersystem must work in a given room which is live, I depend on the reflected and direct sound to be same.
Maybe I should mention that I am constructing sound systems for professional use where people use microphones and speach intelligibility is essencial.
We can not have side lobes as they can catch in the microphones and will colour the allover tone. My crossovers often are in the 2k range and it is important that the drivers are time alligned and close as possible.
If my speakersystem must work in a given room which is live, I depend on the reflected and direct sound to be same.
Maybe I should mention that I am constructing sound systems for professional use where people use microphones and speach intelligibility is essencial.
We can not have side lobes as they can catch in the microphones and will colour the allover tone. My crossovers often are in the 2k range and it is important that the drivers are time alligned and close as possible.
Great post - extremely valuable. Bookmarked.
I wish there were a way to measure the signal-out of my dbx Driverack 260 and translate that into accurate x-over components.
I understand I should use the actual measured impedance of the driver opposed to the nominal impedance given:
Do you know this to be an accurate method to acquire impedance values from each driver?: placing a 10 ohm resistor in series with the negative of the loudspeaker, running a tone at the chosen x-over freq (i.e. 1.6k Hz for a 2-way x-over) at 1V (max) output and with a digital multimeter
completing the following formula?:
Vr = Vin - Vs
Where Vr is voltage across resistor, Vin is unloaded voltage, Vs is voltage across speaker
I = Vr / R
Where I is current, and R is the value of the resistor (10 ohms is suggested)
Z = Vs / I
Assume an input of 1V, R = 10 ohms and Vs = 400 mV
Vr = Vin - Vs = 1 - 0.4 = 0.6 V
I = Vr / R = 0.6 / 10 = 0.06 A
Z = Vs / I = 0.4 / 0.06 = 6.67 Ohms
I found this information HERE
I don't have WT2, WT3 or DATS and I don't feel I have enough time to learn and troubleshoot the software I have installed on my computer. As long as I have time to get this crossover in the ballpark for a show on the 20th, I can tune it for higher accuracy afterward, while re-measuring the components after they're broken-in.
I'm considering using Xover Pro (same company as BassBox Pro) to get me in the ballpark for now. The program offers impedance correction as well as EQ correction. But, I would like to acquire accurate impedance values to start with.
I wish there were a way to measure the signal-out of my dbx Driverack 260 and translate that into accurate x-over components.
I understand I should use the actual measured impedance of the driver opposed to the nominal impedance given:
Do you know this to be an accurate method to acquire impedance values from each driver?: placing a 10 ohm resistor in series with the negative of the loudspeaker, running a tone at the chosen x-over freq (i.e. 1.6k Hz for a 2-way x-over) at 1V (max) output and with a digital multimeter
An externally hosted image should be here but it was not working when we last tested it.
, 
completing the following formula?:
Vr = Vin - Vs
Where Vr is voltage across resistor, Vin is unloaded voltage, Vs is voltage across speaker
I = Vr / R
Where I is current, and R is the value of the resistor (10 ohms is suggested)
Z = Vs / I
Assume an input of 1V, R = 10 ohms and Vs = 400 mV
Vr = Vin - Vs = 1 - 0.4 = 0.6 V
I = Vr / R = 0.6 / 10 = 0.06 A
Z = Vs / I = 0.4 / 0.06 = 6.67 Ohms
I found this information HERE
I don't have WT2, WT3 or DATS and I don't feel I have enough time to learn and troubleshoot the software I have installed on my computer. As long as I have time to get this crossover in the ballpark for a show on the 20th, I can tune it for higher accuracy afterward, while re-measuring the components after they're broken-in.
I'm considering using Xover Pro (same company as BassBox Pro) to get me in the ballpark for now. The program offers impedance correction as well as EQ correction. But, I would like to acquire accurate impedance values to start with.
The way I'd normally do this is to loop it in with a sound card for the response, then overlay this with a crossover simulator then play with some values. Of course this would imply having a proper impedance plot already done.
You could sweep it with a generator and measure with a voltmeter and plot the response by hand if you preferred.
Sounds reasonable. One way I used to do this was to use a larger resistor (more than 100 ohms and preferrably 1k). Connect a 10 ohm resistor in place of the speaker and adjust the voltage across the 10 ohm resistor until it is 10mV (or some convenient multiple thereof). Replace the speaker and the voltage you measure in millivolts across the speaker will equal the impedance. You could sweep the frequency and gauge the trend in the curve.
AllenB,
Thanks very much for your tutorial! Your post #67 is amazing.
An idea that I have to refine designing crossovers without measurement is this. A better approximation of the actual impedance of a driver at the cross-over frequency than the stated, rated or nominal impedance of the driver is equal to 1,2 times Re.
Take for example the Morel MDT-12 tweeter. This tweeter is described as 8 Ohm. However, at frequencies above resonance, the actual impedance of the tweeter is less than 8 Ohm from resonance at about 1 kHz, to 10 kHz. Multiplying DCR or Re of the tweeter equal to 5,1 Ohm times 1,2 equals 6,1 Ohm. From 2 kHz to 5 kHz impedance of the tweeter equals about 6,0 Ohm.
Would you agree with this, or do you think that calculating in this manner would result in only a marginal improvement?
Regards,
Pete
Thanks very much for your tutorial! Your post #67 is amazing.
An idea that I have to refine designing crossovers without measurement is this. A better approximation of the actual impedance of a driver at the cross-over frequency than the stated, rated or nominal impedance of the driver is equal to 1,2 times Re.
Take for example the Morel MDT-12 tweeter. This tweeter is described as 8 Ohm. However, at frequencies above resonance, the actual impedance of the tweeter is less than 8 Ohm from resonance at about 1 kHz, to 10 kHz. Multiplying DCR or Re of the tweeter equal to 5,1 Ohm times 1,2 equals 6,1 Ohm. From 2 kHz to 5 kHz impedance of the tweeter equals about 6,0 Ohm.
Would you agree with this, or do you think that calculating in this manner would result in only a marginal improvement?
Regards,
Pete
Hi Excellent tutorial, Ive only just come across it after Months of experimentation.
What is the main reason for very poor off axis treble response? and roughly how much attenuation would I get with a 4 ohm resistor in series before the tweeter? (no L pad)
I am am crossing over at quite a high fq as the woofer goes to 4K (and pretty smooth) and I prefer a smoother less fatiguing sound at the top end. I think the woofer is an old monitor audio modified version of the Seas P17RCY and the Tweeter is an Audax TWO25 series. I'm using a 3rd order X-over due to the high tweeter fs of 1090 but I seem to be getting really bad off axis response and poor punch in the mids.
What is the main reason for very poor off axis treble response? and roughly how much attenuation would I get with a 4 ohm resistor in series before the tweeter? (no L pad)
I am am crossing over at quite a high fq as the woofer goes to 4K (and pretty smooth) and I prefer a smoother less fatiguing sound at the top end. I think the woofer is an old monitor audio modified version of the Seas P17RCY and the Tweeter is an Audax TWO25 series. I'm using a 3rd order X-over due to the high tweeter fs of 1090 but I seem to be getting really bad off axis response and poor punch in the mids.
This seems reasonable but it comes down to what accuracy means and how important it is.
Variations in the impedance can result in variations in the response. When you measure the impedance as a curve, including phase, and do the same with the response taking into account the acoustic nature of the driver and its mounting by measuring it at various angles, then you should get few surprises in the result. Flattening the impedance is simply a way of reducing one of the more significant variables in this case.
If you have less information that this then obsessing over a component value can sometimes be hit and miss. It's not like one value will necessarily be perfect. Often you'll find that one sounds good in some ways and another in other ways, but neither may fix everything.
Component accuracy often recieves more attention than it deserves. In many cases with electronics in general, there is less point in matching components to greater than 10% than there is in ensuring a competent design. With regards to speakers this means that it is more important to control the propagation of sound, ie. that it gets out into the room smoothly and in a way that suits the room, the driver and the application. Only then will finer control of the response through the crossover give more of the expected result. (Not saying you can't make a difference with small component changes, just that it will be more experimental and less certain of the result.)
You also need to consider the facts regarding perception and common sense. Would you rather have a few peaks in the response in the bass range or in the sensitive upper midrange region? If there was a peak in one of your speakers would you rather the other side be perfect on its own, or match the other one?
Component tolerances were once in the order of plus and minus 50%. This was not only due to manufacturing errors but also due to changes as they age and changes with the weather, as well as drifting as they warmed up. This wasn't enough to prevent good products from being made. Having built numerous of these valve amplifiers using antique parts, I normally don't worry about better than 10% in resistances/voltages except in specific parts of a circuit, where I wouldn't bother with any better than 1%.
You also need to consider the facts regarding perception and common sense. Would you rather have a few peaks in the response in the bass range or in the sensitive upper midrange region? If there was a peak in one of your speakers would you rather the other side be perfect on its own, or match the other one?
Component tolerances were once in the order of plus and minus 50%. This was not only due to manufacturing errors but also due to changes as they age and changes with the weather, as well as drifting as they warmed up. This wasn't enough to prevent good products from being made. Having built numerous of these valve amplifiers using antique parts, I normally don't worry about better than 10% in resistances/voltages except in specific parts of a circuit, where I wouldn't bother with any better than 1%.
Last edited:
There are a couple of common reasons that the top end might sound rough or fatiguing. Since you seem to prefer the woofer to play higher in frequency, I'd assume it's due to the off axis sound up high. This can come down to personal preference but I prefer less.
As a woofer goes higher in frequency, it produces less off axis sound. Just to illustrate, imagine taking a small stone and dropping it into a pond. Note the width of the ripples. Now take some sand and drop it over that wider area. The ripples will have trouble forming as the parts do not add properly and they partially cancel one another. Woofers will lose most of this energy off to the sides, more with increasing frequency.
Having been in this situation I understand the temptation to try and run the better driver as wide in frequency as possible to get the best out of it. What you end up with in this case will be reflections that are more loud above the crossover frequency than below due to the tweeter and its wider radiation.
When you listen to speakers in a room, you can expect the lower frequencies to be coming from all around, and you can make a choice to taper this off so that the higher frequencies are mostly coming to you directly. What you want to avoid is reducing the off-axis sound at some frequency and allowing it to be present again at higher frequencies, which is what you have when you cross a woofer quite high to a dome tweeter.
It could be worth exploring some different types of speaker, and there are two common alternatives that can achieve this end. One of them is to use a larger driver for the treble. An example would be using a small full range driver. The advantages of this are that the treble will be more direct (meaning less reflected sound) and therefore more relaxed even when designed for the same level on your listening axis as a dome. You'll also be able to cross lower in the midrange (good), and you'll be able to use a larger woofer. This format has come to be known here as a F.A.S.T.
The other is to use a horn to control the dispersion in the treble. A waveguide is a kind of horn that has been designed primarily for controlling the directivity. It is usually conical or similar. If you know the woofer and the waveguide angle you want to use, you find the frequency where the woofer radiation narrows to that amount and make sure the waveguide and tweeter can be crossed that low.
Last edited:
But may be it receives less attention than it deserves in a complex crossover. For high order filter (3rd order electrical with correction and above), measurement and simulation cannot tell everything what the drivers do. So attention must be paid more here (voicing), resulting in non-standard component values.