Fliterisk1
Non Posting Status
I've posted this on a few other places in the internet, with little response. Apologies for its length, and rather dry reading, but it is very important to the enjoyment of my stereo. It is a real solution to a problem. Stereos are important to a lot of us, and many here are do-it-yourselfers, so I posted it in this forum. There are two topics: the Zobel circuit, and speaker balancing, so this really should be split into several posts.
Since I am not active in electronics, my knowledge of the audiophile world is vague, to say the least. What is presented below is from information found on the internet, and from my own assumptions based on the lack of information. What is very odd is that, as popular as speakers are, there are NO news groups I know of that are specific to audio speakers. The question is have they been restrained?
At this point, no one has yet corrected this writing; which is essentially my perception of circumstances, that, apparently, are not favorable to the consumer.
A few references...
https://users.ece.gatech.edu/~mleach/papers/zobel/zobel.pdf
https://sound.westhost.com/lr-passive.htm
This is a set of instructions for a simple and inexpensive corrective circuit most speakers require, but are not commercially installed on any product I know of. If there are any passive crossover circuits (most are passive) in your system then you will probably need it, and badly.
If your speakers are balanced then they should sound almost as balanced as headphones. Are you so satisfied with your speakers? There is a chance you are not. I was not, and am angry as to why that is. This post is an act of war against those who cheated me, and are still cheating many of you. My speakers are now repaired. It is quite a relief!
If your speakers are not balanced then you will probably want to invest in more expensive systems, or different types of speakers, equalizers, etc. That's what audio companies want you to do, to spend money. But if you depend on their products then you will always be searching for that 'perfect' system. You may never find a balanced system off-the-shelf. That is done to you on purpose by those who know better. (I'm guessing that the very expensive systems do include this very inexpensive corrective circuit.)
The mental problem is in the greed of the audio companies. This repair would cost them, and the consumer, little. They are content to steal from you, by motivating you with a flawed product, and have been doing so for a long time. I am not an electrical engineer, so I do not depend on bloated audio companies to hire me, so I am not restrained to post this. I believe the circuit is rarely used.
The physical problem lies in the coil of the driver. It has impedance which varies considerably with frequency, and does not respond anywhere close to that of a resistor. Quite a while ago, I built speakers from a kit. The company which sold it said that speaker coils were poor inductors, and so will present a reasonably constant impedance to the crossover. You may have heard this also, but it is not true. If you are a little technically-minded; then with a tone generator, a VOM and a simple voltage divider circuit (one resistor), you can determine the actual impedance of your drivers at any frequency. It is anything but constant, and acts much as a coil would. They knew better.
Why is that important? Speaker sound output is nearly proportional to voltage, while crossovers are voltage divider circuits which include the driver as part of that circuit. What happens is that, with increase in frequency, the increasing impedance of the driver coil will increase the voltage applied to it, and so will force more high-frequency sound out of that driver. That sound was intended to be filtered by the crossover.
Active crossovers, however, use separate amps (so you spend more money). The driver, in such cases, is not part of any voltage divider circuit. The amp has full control of the voltage sent to the drivers. But if resistor attenuation, or a cap, is added then the driver may require it. (I probably wouldn't add it to the tweeter.)
The Zobel: •-----/\/\/\-----| |------•
Rz Cz
The corrective circuit required by all passive crossovers is called the Zobel circuit. It consists of a simple RC circuit in series. It is connected in parallel with its driver.
The woofer and midrange will certainly need this impedance compensation, though the tweeter may not be so seriously in need. Boosted ultrahigh frequencies are not as noticeable.. I have not installed it on my Klipschorn's, yet. They probably need it, (if there was a tweeter that did it would be the Klipschorn!). But the woofer and midrange do require this repair. You will notice the difference! I certainly did, and no longer fuss over my speakers.
___
Expected driver response: / \
volume vs frequency / \ (midrange)
___
Actual driver response: / ----____
without the Zobel /
The Zobel circuit will restore response to what we always expected and should have recieved. I've complained about my Kipsch horn tweeters, but found it was the increased output from my midranges which distorted the high end. With the Zobel, the midrange will be cleaner, and the woofer will be less distorted, since it won't be producing too much midrange sound. Balanced, with less intermodulation distortion.
If you decide to accept the project, first check to see that it is missing from your speakers before adding another. It probably isn't there. If they are present then it would be appreciated if you noted that fact.
For reference, search the internet for "impedance compensation", it may get better results than Zobel. There are other speaker anomalies mentioned with impedance, such as resonance. But impedance compensation is simple to correct and you will hear the difference immediately.
However, well-made, well-damped drivers should have little resonance problem. My drivers are well-damped. The manufacturer (Speakerlab) took measures to ensure that, and I have measured it. It is minimal, and does not resemble the large peaks shown in example drawings. The Klipschorns are also well-damped with magnetic fluid. (I refuse risk such a sensitive device by measuring them.) Resonance troubles would show up with midranges and tweeters. Resonance, as with coil impedance, causes an increase in impedance. But it occurs below the useful range of the driver, as opposed to above it.
The Zobel is attached electrically across the terminals of the driver, as if it were part of the driver itself. Except for a bit of wire, nothing electrically should be between the driver and its Zobel. It does not matter whether the resistor lead or capacitor lead connects to the hot wire of the speaker.
There are standard equations for determining the values of the resistor and capacitor. The standard values are Rz = 1.25 * driver dc resistance, and Cz = driver inductance/(Rz^2). Another source I've found suggests that Rz should equal driver resistance. BUT ANYTHING HELPS!
---->
Lacking information about driver characteristics (the inductance) should not prevent you from installing something. I would suggest...
Rz = 8 ohms (or 4 for 4 ohm speakers.)
Cz = 30-40 uf, for the woofer.
Cz = 10 uf for the midrange.
---->
If you set up a voltage divider circuit and test the driver with the Zobel connected, as was suggested earlier, you can refine the the values so that the impedance of the driver/Zobel combo is near the expected value. But, seriously, anything does help.
The ratings should be reasonably large for the amplifier being used, at least 10 watt resistors, and 50 volt caps. I would put two 16-ohm/10 watt resistors in parallel for woofers using large amps, and 100-volt caps. Bipolar capacitors will work. I use them, and they have acted predictably. The common polarized capacitors cannot be used with speakers.
One other thought...
It may be a good idea to check polarity of the wires to your drivers. I don't recall Speakerlab mentioning it, but reversing polarity may improve the sound. It did mine. Passive crossovers affect the phase of the signal. So much so that reversing polarity (180º shift) may help, though not completely correct it. That would be impossible with passive systems.
Phase error is relatively easy to check. Reverse the wires to the midrange of one of the speakers, then use a signal generator (such as WinISD, for free at https://www.linearteam.dk/default.aspx?pageid=winisd , which uses your computer) at the crossover frequency(s) to see which speaker is louder. The loudest speaker is the correct polarity.
You check by reversing the midrange, but can reverse the woofer or tweeter after determining which phase is off. Don't be surprised if reverse polarity doesn't make a lot of difference, as a 90º phase shift is not unusual. Do this after adding the Zobel, for it will affect phase somewhat.
---------------------------------
Speaker Balancing
Now that you've installed the Zobel circuit on your drivers, you may want to attempt to balance the speakers' output for a flat response. My system was a collection of unmatched drivers and crossover, so I had to develop my own method. I have one, and it is simple--darned near obvious--though a bit tedious.
What we want is the sound ouputs of the drivers to have the same graphical shape and relative amplitude as their filter's voltage output. The Zobel circuit should correct for most of the shape error. As for amplitude, resistor attenuation networks are used to level the differences in efficiencies between drivers.
But how much attenuation? Do we guess by just listening for balance? I do not have confidence in that method, and it's the only one I've heard suggested to those who do not have access to oscilloscopes and good quality microphones.
The Method: What we know about crossovers is that the voltage outputs of adjacent filters are equal at their crossover frequencies. It is logical that, if the speakers are balanced, then the amplitude of the drivers at those frequencies should also be equal. Equal voltages, and equal volumes (at the same frequency) are not so difficult to detect and correct.
You will need a signal generator for the test. I used the WinISD program in my computer, which was connected to the amp.
We want to know where the voltages are equal, with accuracy. The rated crossover frequencies provided by the manufacturer are not sufficient, as they are calculated into a resistive load, and not your particular driver/Zobel combination.
To make crossover outputs accessible so I could measure voltages easily, I installed taps to each filter output, at the point before attenuation. A tap is a single wire attached to one filter, and the other end attached to a screw which is visible from the back of the speaker so it could be touched by a meter's probe. There is one tap for each filter. Only one speaker needs the taps. Only one speaker is tested throughout, while the other is disconnected.
The AC voltage of each filter is measured by touching the meter's probes to its particular screw, and the other probe to the convenient cable ground--with the amplifier's volume set sufficiently high to get a reasonable voltage reading. Say 3 volts or more as measured from the speaker cable itself. Set amp volume by measuring voltage across the main speaker cable, for at some frequencies a particular tap will have zero volts. You may be surprised how loud 3 volts RMS can be. But, at this point, we don't care how loud, as long as you and the equipment can stand it. Generally, the higher the voltage then the more accurate the test.
Vary the frequency from the signal generator until you find that at which the bass and midrange voltages are equal. This is the first crossover frequency. If you have a 3-way system, then again vary the frequency until you find that at which the midrange and tweeter voltages are equal. This is the second crossover frequency. WinISD increments in 100 hz. But you can type in any particular frequency you want, such as 650 hz.
Now that you know what frequencies to test, the next thing is to determine the attenuation necessary to balance volume at those frequencies. You will need to install switches to one of your speakers. They are connected in series with the drivers. For the first test, one switch is connected to the woofer and the other to the midrange, and then midrange volume is adjusted with attenuation.. For the second test, the switch is removed from the woofer and connected to the tweeter, which has its volume then adjusted to match the midrange. Midranges are usually more efficient than woofers, and tweeters more than midranges.
The switches allow you to listen to each driver at the crossover frequency by turning the other off so that you can determine if they are at equal volume. The two switches are simple on-off, and are connected externally by using long wiring (~10 ft) so you can walk around while alternately switching each driver. Drilling a small hole in the back of the speaker for the wiring should not cause much damage, and easily filled after you are done.
Standing waves occur when listening to a single tone, and they will shift position when changing speakers, or just when moving your head. So you will have to determine an average volume by walking around with the switches.
WARNING: Higher order crossovers (than 1st-order) can resonate if they are disconnected; such as by switching, or connected to a burned-out driver. The amplifier itself may burn up under such conditions. So you might use an inexpensive audio source. You will not require a loud signal in this test. Power consumption will be low.
You can speed up the procedure by installing the test attenuation resistors to the external wires, instead of reopening the speaker box for each change. So you wll also need to bring a ground wire through the hole. When you have found the right combination of resistors then they can be installed permanently inside the speaker.
WinISD's signal generator includes a built-in attenuator, apparently rated in db, though it is not marked. So you can adjust it to determine how many decibels attenuation will be required to equalize. Since it is not marked, I did not think to use it in my tests. It sure would've been easier. But it does seem to be in decibels. What else? I could not find it in the Help section.
The Resistor Attenuation Network:
(from crossover) •------•------Rs----• (to driver)
|
Rp
|
• (to ground)
Attenuation...
db Rs (ohms) Rp (ohms) Impedance (~8)
---------------------------------------------------
-1.93 2 40 8
-2.76 3 29 7.97
-3.52 4 24 8
-4.21 5 21 8.02
-4.86 6 19 8.06
-5.46 7 17 7.96
-6.02 8 16 8
-7.04 10 14 7.87
-7.95 12 13 7.87
-9.17 15 12 7.88
-9.9 17 12 8.1
-11.8 23 11 8.11
-14 32 10 8
-14.8 36 10 8.14
-20 72 9 8.08
-24.1 120 8 7.53
-29.8 240 8 7.75
-35.5 470 8 7.87
The db attenuation rating for an 8-ohm network is...
db = 20 * log(8/(Rs+8)) (with log to the base 10)
Replace 8 with 4 for 4-ohm speakers if necessary.
Rp is calculated so that impedance always remains near the rated value, 8 ohms in this case.
Rp = 1/(1/8 - 1/(8+Rs))
The method did smooth out my speaker response, and it was an old _complaint_.
---------------------------------
The Resistor Attenuation Network:
(from crossover) •------•------Rs----• (to driver)
|
Rp
|
• (to ground)
Be sure and connect the attenuator with proper polarity; both when connecting to the external switched lines, and when permanently installing them. Mark the lines coming out of the access hole so you'll know which comes from the crossover and which goes to the speaker.
To change attenuation you can either adjust the values of one network, or add another network in series; treating the output from the attenuator as if it were the output from the crossover. In such cases the decibel reduction is simply added.
On the internet, this is called an 'L-pad.' But the L-pad I recall was a variable pot made to have a reasonably constant impedance. But those L-pads are not linear, and their impedance is not constant. I kept my L-pads in circuit, but used fixed resistors to do most of the balancing. The fixed circuit helps stabilize the L-pad's impedance.
L-pads tend to corrode, and break contact over time. Spinning them will help clean corrosion. But a good penetrating oil made for electronics has kept mine operating almost trouble-free. Since I've balanced my speakers, the value of a convenient variable L-pad has diminished considerably.
---------------------------------
Finally... Graphical Results
The taps are not absolutely necessary to determine crossover frequencies. But I use them so I could determine crossover response (particularly the woofer) any time I wanted to. The actual response of the midrange and tweeter, however, are more difficult to measure. That's because they are usually attenuated, and attenuation often reduces the voltage to levels not measurable by the typical volt-ohm meter. Attenuation will hide the signals that the driver is actually responding to.
Did you ever wonder why meters can measure small DC voltages accurately, while AC ranges are restricted to power-line levels? Maybe so we can't do our own audio work? Or at least discover what audio signals are all about. The restriction of AC ranges on VOMs has always been an irritatant. Apparently, it is purposed. At least I'm convinced they are hiding something.
Anyway, you can still measure audio signals, but only as long as you turn the volume up to uncomfortably high levels. I prefer it in the 4-5 volts range, having a voltmeter which measures to 0.1 of a volt.
But it is interesting, even useful, to measure the voltages coming out of the crossover at ANY frequency. By doing that one can determine the shape of the output curves from the filters. Audio curves are measured in decibels (db). db = 10*log(Pout/Pin), with P=power. But we can't always measure power directly. However, power is usually a function of the square of the voltage. That since current is often proportional to voltage, so both increase linearly. That changes the decibel equation to ...
db = 20 * log(Vout/Vin) (with log to the base 10)
You will need to record the input voltage at each reading, for it can and will change with frequency. If you change the volume in the middle of the test then the db value should remain about the same. Other factors, such as RMS as measured by a meter, will divide out of the ratio and do not affect db.
The following discussion refers to signals at the same frequency. A signal at 0 db is one that is not attenuated by a filter, or amplified by two speakers on the same frequency.
When adding the sound from two speakers then you are adding power, not voltage. Doubling power will add about 3 db; such as by adding the sound of two balanced speakers at the same voltage. But doubling the voltage to one speaker will add about 6 db to the sound heard.
Adding two signals with both at -3 db results in 0 db, unattenuated. While adding two signals at -6 db results in -3 db. Adding two unattenuated sounds at the same frequency will result in a volume of +3 db. Adding a -12 db signal to one at 0 db results in about +0.26 db sound. This may help to interpret the curves you get from testing your system.
For purposes of graphing a resultant output from all crossover filters...
Total db = 10 log( (Vw/Vi)^2 + (Vm/Vi)^2 + (Vt/Vi)^2 )
With Vw = woofer voltage, Vm = midrange V, and Vt = tweeter V;
all taken at the same frequency.
And assuming no phase variations exist, but that is not true.
Near the crossover freqs, the total volume will always be less.
The response curves are plotted db vs frequency. Decibels are on the y-axis, and it is linear; usually in a range of -30 to 0 db. While frequency is plotted on a log scale. You need to make your checks over the range of frequencies from 20 to 20k hz. I would suggest these frequencies: 20, 100, 200, 500, 1000, 1500, 2500, 5k, 10k & 20k. Then fill in with more tests in the region of the crossover frequencies.
If somehow you are able to accurately measure the voltages to the midrange and tweeter, then, to normalize the plot, you would add the db rating of the attenuation to the db you obtained from your tests. That assuming you have equalized the volumes of the drivers at their crossover frequencies, and assuming speaker output actually is proportional to voltage in its operating range.
I can't, but if you can accurately measure all voltages to all drivers, then the resulting plot might be the closest thing to a proper representation of sound output off your drivers that you can get without using an oscilloscope--which may still have difficulty with standing waves. Your plot would also show the resonance peaks, which occur below the driver's useful frequency range; as opposed to Zobel corrections which are at the high end. But it will not, however, show the actual phase losses between drivers. At this point, it just irritates me that phase variations exist. I think they are difficult to detect by hearing.
--
Bryan Rhodes
Since I am not active in electronics, my knowledge of the audiophile world is vague, to say the least. What is presented below is from information found on the internet, and from my own assumptions based on the lack of information. What is very odd is that, as popular as speakers are, there are NO news groups I know of that are specific to audio speakers. The question is have they been restrained?
At this point, no one has yet corrected this writing; which is essentially my perception of circumstances, that, apparently, are not favorable to the consumer.
A few references...
https://users.ece.gatech.edu/~mleach/papers/zobel/zobel.pdf
https://sound.westhost.com/lr-passive.htm
This is a set of instructions for a simple and inexpensive corrective circuit most speakers require, but are not commercially installed on any product I know of. If there are any passive crossover circuits (most are passive) in your system then you will probably need it, and badly.
If your speakers are balanced then they should sound almost as balanced as headphones. Are you so satisfied with your speakers? There is a chance you are not. I was not, and am angry as to why that is. This post is an act of war against those who cheated me, and are still cheating many of you. My speakers are now repaired. It is quite a relief!
If your speakers are not balanced then you will probably want to invest in more expensive systems, or different types of speakers, equalizers, etc. That's what audio companies want you to do, to spend money. But if you depend on their products then you will always be searching for that 'perfect' system. You may never find a balanced system off-the-shelf. That is done to you on purpose by those who know better. (I'm guessing that the very expensive systems do include this very inexpensive corrective circuit.)
The mental problem is in the greed of the audio companies. This repair would cost them, and the consumer, little. They are content to steal from you, by motivating you with a flawed product, and have been doing so for a long time. I am not an electrical engineer, so I do not depend on bloated audio companies to hire me, so I am not restrained to post this. I believe the circuit is rarely used.
The physical problem lies in the coil of the driver. It has impedance which varies considerably with frequency, and does not respond anywhere close to that of a resistor. Quite a while ago, I built speakers from a kit. The company which sold it said that speaker coils were poor inductors, and so will present a reasonably constant impedance to the crossover. You may have heard this also, but it is not true. If you are a little technically-minded; then with a tone generator, a VOM and a simple voltage divider circuit (one resistor), you can determine the actual impedance of your drivers at any frequency. It is anything but constant, and acts much as a coil would. They knew better.
Why is that important? Speaker sound output is nearly proportional to voltage, while crossovers are voltage divider circuits which include the driver as part of that circuit. What happens is that, with increase in frequency, the increasing impedance of the driver coil will increase the voltage applied to it, and so will force more high-frequency sound out of that driver. That sound was intended to be filtered by the crossover.
Active crossovers, however, use separate amps (so you spend more money). The driver, in such cases, is not part of any voltage divider circuit. The amp has full control of the voltage sent to the drivers. But if resistor attenuation, or a cap, is added then the driver may require it. (I probably wouldn't add it to the tweeter.)
The Zobel: •-----/\/\/\-----| |------•
Rz Cz
The corrective circuit required by all passive crossovers is called the Zobel circuit. It consists of a simple RC circuit in series. It is connected in parallel with its driver.
The woofer and midrange will certainly need this impedance compensation, though the tweeter may not be so seriously in need. Boosted ultrahigh frequencies are not as noticeable.. I have not installed it on my Klipschorn's, yet. They probably need it, (if there was a tweeter that did it would be the Klipschorn!). But the woofer and midrange do require this repair. You will notice the difference! I certainly did, and no longer fuss over my speakers.
___
Expected driver response: / \
volume vs frequency / \ (midrange)
___
Actual driver response: / ----____
without the Zobel /
The Zobel circuit will restore response to what we always expected and should have recieved. I've complained about my Kipsch horn tweeters, but found it was the increased output from my midranges which distorted the high end. With the Zobel, the midrange will be cleaner, and the woofer will be less distorted, since it won't be producing too much midrange sound. Balanced, with less intermodulation distortion.
If you decide to accept the project, first check to see that it is missing from your speakers before adding another. It probably isn't there. If they are present then it would be appreciated if you noted that fact.
For reference, search the internet for "impedance compensation", it may get better results than Zobel. There are other speaker anomalies mentioned with impedance, such as resonance. But impedance compensation is simple to correct and you will hear the difference immediately.
However, well-made, well-damped drivers should have little resonance problem. My drivers are well-damped. The manufacturer (Speakerlab) took measures to ensure that, and I have measured it. It is minimal, and does not resemble the large peaks shown in example drawings. The Klipschorns are also well-damped with magnetic fluid. (I refuse risk such a sensitive device by measuring them.) Resonance troubles would show up with midranges and tweeters. Resonance, as with coil impedance, causes an increase in impedance. But it occurs below the useful range of the driver, as opposed to above it.
The Zobel is attached electrically across the terminals of the driver, as if it were part of the driver itself. Except for a bit of wire, nothing electrically should be between the driver and its Zobel. It does not matter whether the resistor lead or capacitor lead connects to the hot wire of the speaker.
There are standard equations for determining the values of the resistor and capacitor. The standard values are Rz = 1.25 * driver dc resistance, and Cz = driver inductance/(Rz^2). Another source I've found suggests that Rz should equal driver resistance. BUT ANYTHING HELPS!
---->
Lacking information about driver characteristics (the inductance) should not prevent you from installing something. I would suggest...
Rz = 8 ohms (or 4 for 4 ohm speakers.)
Cz = 30-40 uf, for the woofer.
Cz = 10 uf for the midrange.
---->
If you set up a voltage divider circuit and test the driver with the Zobel connected, as was suggested earlier, you can refine the the values so that the impedance of the driver/Zobel combo is near the expected value. But, seriously, anything does help.
The ratings should be reasonably large for the amplifier being used, at least 10 watt resistors, and 50 volt caps. I would put two 16-ohm/10 watt resistors in parallel for woofers using large amps, and 100-volt caps. Bipolar capacitors will work. I use them, and they have acted predictably. The common polarized capacitors cannot be used with speakers.
One other thought...
It may be a good idea to check polarity of the wires to your drivers. I don't recall Speakerlab mentioning it, but reversing polarity may improve the sound. It did mine. Passive crossovers affect the phase of the signal. So much so that reversing polarity (180º shift) may help, though not completely correct it. That would be impossible with passive systems.
Phase error is relatively easy to check. Reverse the wires to the midrange of one of the speakers, then use a signal generator (such as WinISD, for free at https://www.linearteam.dk/default.aspx?pageid=winisd , which uses your computer) at the crossover frequency(s) to see which speaker is louder. The loudest speaker is the correct polarity.
You check by reversing the midrange, but can reverse the woofer or tweeter after determining which phase is off. Don't be surprised if reverse polarity doesn't make a lot of difference, as a 90º phase shift is not unusual. Do this after adding the Zobel, for it will affect phase somewhat.
---------------------------------
Speaker Balancing
Now that you've installed the Zobel circuit on your drivers, you may want to attempt to balance the speakers' output for a flat response. My system was a collection of unmatched drivers and crossover, so I had to develop my own method. I have one, and it is simple--darned near obvious--though a bit tedious.
What we want is the sound ouputs of the drivers to have the same graphical shape and relative amplitude as their filter's voltage output. The Zobel circuit should correct for most of the shape error. As for amplitude, resistor attenuation networks are used to level the differences in efficiencies between drivers.
But how much attenuation? Do we guess by just listening for balance? I do not have confidence in that method, and it's the only one I've heard suggested to those who do not have access to oscilloscopes and good quality microphones.
The Method: What we know about crossovers is that the voltage outputs of adjacent filters are equal at their crossover frequencies. It is logical that, if the speakers are balanced, then the amplitude of the drivers at those frequencies should also be equal. Equal voltages, and equal volumes (at the same frequency) are not so difficult to detect and correct.
You will need a signal generator for the test. I used the WinISD program in my computer, which was connected to the amp.
We want to know where the voltages are equal, with accuracy. The rated crossover frequencies provided by the manufacturer are not sufficient, as they are calculated into a resistive load, and not your particular driver/Zobel combination.
To make crossover outputs accessible so I could measure voltages easily, I installed taps to each filter output, at the point before attenuation. A tap is a single wire attached to one filter, and the other end attached to a screw which is visible from the back of the speaker so it could be touched by a meter's probe. There is one tap for each filter. Only one speaker needs the taps. Only one speaker is tested throughout, while the other is disconnected.
The AC voltage of each filter is measured by touching the meter's probes to its particular screw, and the other probe to the convenient cable ground--with the amplifier's volume set sufficiently high to get a reasonable voltage reading. Say 3 volts or more as measured from the speaker cable itself. Set amp volume by measuring voltage across the main speaker cable, for at some frequencies a particular tap will have zero volts. You may be surprised how loud 3 volts RMS can be. But, at this point, we don't care how loud, as long as you and the equipment can stand it. Generally, the higher the voltage then the more accurate the test.
Vary the frequency from the signal generator until you find that at which the bass and midrange voltages are equal. This is the first crossover frequency. If you have a 3-way system, then again vary the frequency until you find that at which the midrange and tweeter voltages are equal. This is the second crossover frequency. WinISD increments in 100 hz. But you can type in any particular frequency you want, such as 650 hz.
Now that you know what frequencies to test, the next thing is to determine the attenuation necessary to balance volume at those frequencies. You will need to install switches to one of your speakers. They are connected in series with the drivers. For the first test, one switch is connected to the woofer and the other to the midrange, and then midrange volume is adjusted with attenuation.. For the second test, the switch is removed from the woofer and connected to the tweeter, which has its volume then adjusted to match the midrange. Midranges are usually more efficient than woofers, and tweeters more than midranges.
The switches allow you to listen to each driver at the crossover frequency by turning the other off so that you can determine if they are at equal volume. The two switches are simple on-off, and are connected externally by using long wiring (~10 ft) so you can walk around while alternately switching each driver. Drilling a small hole in the back of the speaker for the wiring should not cause much damage, and easily filled after you are done.
Standing waves occur when listening to a single tone, and they will shift position when changing speakers, or just when moving your head. So you will have to determine an average volume by walking around with the switches.
WARNING: Higher order crossovers (than 1st-order) can resonate if they are disconnected; such as by switching, or connected to a burned-out driver. The amplifier itself may burn up under such conditions. So you might use an inexpensive audio source. You will not require a loud signal in this test. Power consumption will be low.
You can speed up the procedure by installing the test attenuation resistors to the external wires, instead of reopening the speaker box for each change. So you wll also need to bring a ground wire through the hole. When you have found the right combination of resistors then they can be installed permanently inside the speaker.
WinISD's signal generator includes a built-in attenuator, apparently rated in db, though it is not marked. So you can adjust it to determine how many decibels attenuation will be required to equalize. Since it is not marked, I did not think to use it in my tests. It sure would've been easier. But it does seem to be in decibels. What else? I could not find it in the Help section.
The Resistor Attenuation Network:
(from crossover) •------•------Rs----• (to driver)
|
Rp
|
• (to ground)
Attenuation...
db Rs (ohms) Rp (ohms) Impedance (~8)
---------------------------------------------------
-1.93 2 40 8
-2.76 3 29 7.97
-3.52 4 24 8
-4.21 5 21 8.02
-4.86 6 19 8.06
-5.46 7 17 7.96
-6.02 8 16 8
-7.04 10 14 7.87
-7.95 12 13 7.87
-9.17 15 12 7.88
-9.9 17 12 8.1
-11.8 23 11 8.11
-14 32 10 8
-14.8 36 10 8.14
-20 72 9 8.08
-24.1 120 8 7.53
-29.8 240 8 7.75
-35.5 470 8 7.87
The db attenuation rating for an 8-ohm network is...
db = 20 * log(8/(Rs+8)) (with log to the base 10)
Replace 8 with 4 for 4-ohm speakers if necessary.
Rp is calculated so that impedance always remains near the rated value, 8 ohms in this case.
Rp = 1/(1/8 - 1/(8+Rs))
The method did smooth out my speaker response, and it was an old _complaint_.
---------------------------------
The Resistor Attenuation Network:
(from crossover) •------•------Rs----• (to driver)
|
Rp
|
• (to ground)
Be sure and connect the attenuator with proper polarity; both when connecting to the external switched lines, and when permanently installing them. Mark the lines coming out of the access hole so you'll know which comes from the crossover and which goes to the speaker.
To change attenuation you can either adjust the values of one network, or add another network in series; treating the output from the attenuator as if it were the output from the crossover. In such cases the decibel reduction is simply added.
On the internet, this is called an 'L-pad.' But the L-pad I recall was a variable pot made to have a reasonably constant impedance. But those L-pads are not linear, and their impedance is not constant. I kept my L-pads in circuit, but used fixed resistors to do most of the balancing. The fixed circuit helps stabilize the L-pad's impedance.
L-pads tend to corrode, and break contact over time. Spinning them will help clean corrosion. But a good penetrating oil made for electronics has kept mine operating almost trouble-free. Since I've balanced my speakers, the value of a convenient variable L-pad has diminished considerably.
---------------------------------
Finally... Graphical Results
The taps are not absolutely necessary to determine crossover frequencies. But I use them so I could determine crossover response (particularly the woofer) any time I wanted to. The actual response of the midrange and tweeter, however, are more difficult to measure. That's because they are usually attenuated, and attenuation often reduces the voltage to levels not measurable by the typical volt-ohm meter. Attenuation will hide the signals that the driver is actually responding to.
Did you ever wonder why meters can measure small DC voltages accurately, while AC ranges are restricted to power-line levels? Maybe so we can't do our own audio work? Or at least discover what audio signals are all about. The restriction of AC ranges on VOMs has always been an irritatant. Apparently, it is purposed. At least I'm convinced they are hiding something.
Anyway, you can still measure audio signals, but only as long as you turn the volume up to uncomfortably high levels. I prefer it in the 4-5 volts range, having a voltmeter which measures to 0.1 of a volt.
But it is interesting, even useful, to measure the voltages coming out of the crossover at ANY frequency. By doing that one can determine the shape of the output curves from the filters. Audio curves are measured in decibels (db). db = 10*log(Pout/Pin), with P=power. But we can't always measure power directly. However, power is usually a function of the square of the voltage. That since current is often proportional to voltage, so both increase linearly. That changes the decibel equation to ...
db = 20 * log(Vout/Vin) (with log to the base 10)
You will need to record the input voltage at each reading, for it can and will change with frequency. If you change the volume in the middle of the test then the db value should remain about the same. Other factors, such as RMS as measured by a meter, will divide out of the ratio and do not affect db.
The following discussion refers to signals at the same frequency. A signal at 0 db is one that is not attenuated by a filter, or amplified by two speakers on the same frequency.
When adding the sound from two speakers then you are adding power, not voltage. Doubling power will add about 3 db; such as by adding the sound of two balanced speakers at the same voltage. But doubling the voltage to one speaker will add about 6 db to the sound heard.
Adding two signals with both at -3 db results in 0 db, unattenuated. While adding two signals at -6 db results in -3 db. Adding two unattenuated sounds at the same frequency will result in a volume of +3 db. Adding a -12 db signal to one at 0 db results in about +0.26 db sound. This may help to interpret the curves you get from testing your system.
For purposes of graphing a resultant output from all crossover filters...
Total db = 10 log( (Vw/Vi)^2 + (Vm/Vi)^2 + (Vt/Vi)^2 )
With Vw = woofer voltage, Vm = midrange V, and Vt = tweeter V;
all taken at the same frequency.
And assuming no phase variations exist, but that is not true.
Near the crossover freqs, the total volume will always be less.
The response curves are plotted db vs frequency. Decibels are on the y-axis, and it is linear; usually in a range of -30 to 0 db. While frequency is plotted on a log scale. You need to make your checks over the range of frequencies from 20 to 20k hz. I would suggest these frequencies: 20, 100, 200, 500, 1000, 1500, 2500, 5k, 10k & 20k. Then fill in with more tests in the region of the crossover frequencies.
If somehow you are able to accurately measure the voltages to the midrange and tweeter, then, to normalize the plot, you would add the db rating of the attenuation to the db you obtained from your tests. That assuming you have equalized the volumes of the drivers at their crossover frequencies, and assuming speaker output actually is proportional to voltage in its operating range.
I can't, but if you can accurately measure all voltages to all drivers, then the resulting plot might be the closest thing to a proper representation of sound output off your drivers that you can get without using an oscilloscope--which may still have difficulty with standing waves. Your plot would also show the resonance peaks, which occur below the driver's useful frequency range; as opposed to Zobel corrections which are at the high end. But it will not, however, show the actual phase losses between drivers. At this point, it just irritates me that phase variations exist. I think they are difficult to detect by hearing.
--
Bryan Rhodes
![[RotaryForum.com] - Audio Speaker Correction (The Zobel)](https://data3.rotaryforum.com/attachments/36/36711-58af7a3f8d3350fee7fb4694c7418837.jpg?hash=WK96P40zUP)
![[RotaryForum.com] - Audio Speaker Correction (The Zobel)](https://data3.rotaryforum.com/attachments/36/36712-74bf12acca745cf169b893c256f47f4e.jpg?hash=dL8SrMp0XP)
![[RotaryForum.com] - Audio Speaker Correction (The Zobel)](https://data3.rotaryforum.com/attachments/36/36735-b15c3186114eb1aab4ee29bbd0388890.jpg?hash=sVwxhhFOsa)
