Reverse engineering a crossover

[Zek]

It seems to me that something is not right here.

 

[Zapper]

The schematic:

Does this make any sense?

Can somebody point out in the schematic which components, or group of components are responsible for the actual High-pass and low-pass filtering? I assume that there is also an attenuator (or 2) in the schematic. If so, where is it?

View attachment 335257

The tweeter has a 1st-order high pass, consisting of the 3.3uF capacitor which blocks low frequencies from the tweeter, followed by an attenuator comprised of the 4.7 ohm and 15 ohm resistors.

The woofer also has a 1st-order low pass filter consisting of a single large inductor, which blocks high frequencies from the woofer.

The midrange has the most complicated circuit. There is an initial 2nd order bandpass filter consisting of 9.6uF capacitor (two 4.8uF in parallel) in series with a medium inductor and a 4.7 ohm resistor, plus a 10 ohm resistor to the return. In simple terms, the capacitor blocks low freqs, the inductor blocks high freqs, and the middle freqs are allowed through. The 4.7 ohm resistor provides some damping of the resonance that the series L and C produce, and together with the 10 ohm provides some attenuation. Next is a filter with the two 25uF caps and two red inductors. It appears to introduce a phase shift. At high frequencies there is a path through the first 25uF to the + speaker lead, out the - speaker lead, through the second 25uF, to the return. At low frequencies there is a path through the 1st red inductor, to the - speaker lead, out the + speaker lead, through the second red inductor, to the return. Notice that the direction of current in the midrange speaker is in the reverse direction at low frequencies and in the forward direction at high frequencies. This is called an all-pass filter - the purpose is to change the phase of the signal, not its amplitude. The phase of the midrange shifts 180 degrees from inverted to non-inverted over its frequency range. This is done to better merge the phase response of the midrange with both the woofer and tweeter, at opposite ends of its frequency range.

Edit: The bandpass filter introduces phase shifts across frequency. The current leads the voltage in a capacitor by 90 degrees. At low frequencies, the parallel 4.8uF caps have the highest impedance, so their characteristic dominates the current in the midrange. So at low frequencies, the current in the midrange leads the voltage by up to 90 degrees. At high frequencies, the medium inductor has the highest impedance, so it controls the current in the midrange. The current in an inductor lags the voltage by 90 degrees. So at high frequencies, the current in the midrange lags the voltage by up to 90 degrees. The total phase shift between low and high frequencies is 180 degrees, from +90 to -90 degrees. The all pass filter also introduces a 180 degree phase shift, but in the opposite direction. So the all-pass filter acts to maintain the phase coherence of the midrange speaker.

 

[vdH_83]

It seems to me that something is not right here.

You're right... See below for v2.

 

[MAB]

You're right... See below for v2.

View attachment 335559

The tweeter is 5.3 Ohms according to the datasheet I posted above:

Combine with the 4.7 Ohm a nd 15 Ohm resistors to make 7.9 Ohm.

With the 3.3uF cap and 7.9 Ohm effective resistance, 6.1kHz crossover. Your measurements with the driver seem to agree!
Since this is well away from the tweeter's natural rolloff, you should see about 6dB/octave in the crossover region, steeper below about 2.6kHz.
This fits Dynaudio's 6dB/octave ethos. You can do better for sure by actually taking the driver's response into account!

Do you have an LCR meter or DATS or other way of measuring inductance? You can measure the woofer inductance directly on the crossover board. You could also do a near-field measurement of the woofer's response with UMIK and REW.

 

[Zapper]

You're right... See below for v2.

View attachment 335559

You moved some parts around, but it looks to me that all the connections are the same. Am I missing something?

 

[Zapper]

The tweeter is 5.3 Ohms according to the datasheet I posted above:
View attachment 335567
Combine with the 4.7 Ohm a nd 15 Ohm resistors to make 7.9 Ohm.
View attachment 335565
With the 3.3uF cap and 7.9 Ohm effective resistance, 6.1kHz crossover.

You need to calculate the resistance in series with the capacitor to find the crossover. The 4.7 is in series with the 15 ohm when looking from the capacitor. And the tweeter impedance is in parallel with the 15 ohm resistor. If the tweeter is 5.3 ohm that gives 3.9 ohm. Then add the 4.7 ohm series resistor gives you the effective resistance in series with the cap. 4.7+(15||Rs) = 8.6 ohm if Rs=5.3 ohm.

Your measurements with the driver seem to agree!
Since this is well away from the tweeter's natural rolloff, you should see about 6dB/octave in the crossover region, steeper below about 2.6kHz.
This fits Dynaudio's 6dB/octave ethos. You can do better for sure by actually taking the driver's response into account!

Do you have an LCR meter or DATS or other way of measuring inductance? You can measure the woofer inductance directly on the crossover board. You could also do a near-field measurement of the woofer's response with UMIK and REW.

 

[terryforsythe]

That's in series.

In parallel, it's the opposite with a parallel coil blocking/reducing low frequencies and a parallel capacitor blocking/reducing high frequencies. Usually you'll see both series & parallel components. (You can't use them in-parallel alone in this application because low inductive reactance (Ohms) or low capacitive reactance can "short out" the signal from the amplifier. You also need something in series.)

You research "LC" filter to see how high-pass and low-pass filters are made.

I was speaking about the individual components.

Capacitor reactance equation: Xc = 1/(2*pi*f*C)
Inductor reactance equation: Xl = 2*pi*f*L

Vary the frequency in these equations and see what you get for the reactance of the individual components.

E.g., a capacitor will always operate as I stated, but in series it will pass the high frequencies to the driver. In parallel it will pass the high frequencies to ground, the net effect being to shunt those frequencies around the driver. Put passive components together in various filter topologies and you end up with a voltage divider circuit that is frequency dependent.

No need to research filters - I designed passive filters, as a full time occupation (BSEE), for many years.

 

[MAB]

You need to calculate the resistance in series with the capacitor to find the crossover. The 4.7 is in series with the 15 ohm when looking from the capacitor. And the tweeter impedance is in parallel with the 15 ohm resistor. If the tweeter is 5.3 ohm that gives 3.9 ohm. Then add the 4.7 ohm series resistor gives you the effective resistance in series with the cap. 4.7+(15||Rs) = 8.6 ohm if Rs=5.3 ohm.

Thanks, you are absolutely correct.
I dropped .7 from 4.7 Ohms.

 

[Keith_W]

Stupid question here (I know nothing about circuit design) ... does the impedance of the speaker driver need to be taken into account when calculating the effect of the filter? What about the changing impedance when the voice coil heats up? Also, does "back EMF" affect the performance of the XO?

 

[Zapper]

Stupid question here (I know nothing about circuit design) ... does the impedance of the speaker driver need to be taken into account when calculating the effect of the filter? What about the changing impedance when the voice coil heats up? Also, does "back EMF" affect the performance of the XO?

Yes, the impedance of the drivers makes a big difference in interacting with the crossover components. For example see MAB's note above and my response.

The self heating of the voicecoil will have some effect, but I think it's relatively minor. However the temperature of the speaker surround can make a significant difference, as the cone suspension gets stiff at cold temps which can change the frequency response a noticable amount.

Back-EMF is a component of the speaker's (complex) impedance - "complex" signifying that its phase is important. In other words it has inductive or capacitive components. Back-EMF raises the impedance.

 

[Keith_W]

Yes, the impedance of the drivers makes a big difference in interacting with the crossover components. For example see MAB's note above and my response.

The self heating of the voicecoil will have some effect, but I think it's relatively minor. However the temperature of the speaker surround can make a significant difference, as the cone suspension gets stiff at cold temps which can change the frequency response a noticable amount.

Back-EMF is a component of the speaker's (complex) impedance - "complex" signifying that it's phase is important. In other words it has inductive or capacitive components. Back-EMF raises the impedance.

If I am interpreting you correctly, you are saying that the mechanical effects of ambient temperature on speaker drivers is more significant than the relatively minor changes that take place with crossovers?

Follow up question: how many % variation from design parameters on the electronic crossover network would you expect from the voice coil heating up from ambient to (say) 60C? Actually, how much does driver impedance rise when it heats up to operational temperature?

 

[NTK]

Follow up question: how many % variation from design parameters on the electronic crossover network would you expect from the voice coil heating up from ambient to (say) 60C? Actually, how much does driver impedance rise when it heats up to operational temperature?

There is actually a Wikipedia page on this subject.

Power compression - Wikipedia

en.wikipedia.org

 

[Zapper]

If I am interpreting you correctly, you are saying that the mechanical effects of ambient temperature on speaker drivers is more significant than the relatively minor changes that take place with crossovers?

I don't want to overstate the case, or my knowledge of it. Amir has posted some speaker measurements he made at cold temps, where the frequency response was significantly different than when measured at normal room temps. He measures the speakers in his garage, and some tests were done when it was cold. Some other data was presented showing that the low frequency speaker resonance changed significantly with temperature.

As for voice coil heating, I don't really know how hot they can get when driven hard. Erin at https://www.erinsaudiocorner.com/ measures Dynamic Range (Instantaneous Compression Test) for the speakers he tests. He describes the test as follows: "The purpose of this test is to illustrate how much (if at all) the output changes as a speaker’s components temperature increases (i.e., voice coils, crossover components) instantaneously". But heating is not the only source of compression - mechanical nonlinearities also can cause reductions in output (suspension gets stiffer as displacement increases), as can the electromagnetic nonlinearity of the coil in the magnetic gap (the force produced decreases as the coil is displaced). So I don't know how to separate the contribution of those various effects (thermal, mechanical, electromagnetic) in Erin's compression data.

Follow up question: how many % variation from design parameters on the electronic crossover network would you expect from the voice coil heating up from ambient to (say) 60C? Actually, how much does driver impedance rise when it heats up to operational temperature?

I would estimate the effect by Googling the temperature coefficient of resistance of copper, such as here: https://www.allaboutcircuits.com/textbook/direct-current/chpt-12/temperature-coefficient-resistance/ . It says the coefficient is 0.004041. So an increase from 20C to 60C increases the resistance by 1.16. That could produce a change of amplitude of 1.3dB, or less, depending on the circuit.

 

[radix]

This is pretty old and likely not an issue any more, but it is fun to read about the old McIntosh speakers. Just to be clear, the fire below was started intentionally with a match to show what happens with flammable speaker filling.

McIntosh Loudspeaker Division Part 2

One of the reasons for concern about fire is that several amplifiers on the market can supply 30 or 40 volts of direct current to a speaker system if the output transistors fail. This high current can heat the voice coil sufficiently to ignite, particularly if the voice coil form is made of paper or other combustible materials. The main concern is for woofers as there is normally no series crossover capacitor to block the flow of DC. The resistance of the series crossover coil is typically less than one ohm and most of the heating occurs at the woofer voice coil. Gordon Gow learned that a few dealers had an actual fire in their showroom from this cause. One salesman had thrown his cup of coffee on the speaker to put it out. Fortunately it is not due to McIntosh amplifiers or speakers. Underwriters Laboratory approval is not required for speakers.

I think it best to show what can happen if a combustible acoustic material is used in a speaker system. Did you ever wonder how this would burn? In this test Doron is used to fill an XR-5 system and it is placed in a dirt area outside of our building in Hillcrest. A match is attached near the center of the woofer cone and ignited, much like what happens if the voice coil form and spider are to ignite. As a result, the fire from the cone ignites the Doron and gradually spreads. The entire system eventually catches fire and is completely destroyed. Impressive! The acoustic material burns with a vengeance. The 8" speaker in the lower left corner had just fallen on the ground. You can see the 12" woofer basket outline at the bottom of the enclosure. The basket and magnet assembly later fall back into the cabinet. The fire ias so hot it melts the thick aluminum extrusion covering the crossover.

 

[MAB]

I don't want to overstate the case, or my knowledge of it. Amir has posted some speaker measurements he made at cold temps, where the frequency response was significantly different than when measured at normal room temps. He measures the speakers in his garage, and some tests were done when it was cold. Some other data was presented showing that the low frequency speaker resonance changed significantly with temperature.

As for voice coil heating, I don't really know how hot they can get when driven hard. Erin at https://www.erinsaudiocorner.com/ measures Dynamic Range (Instantaneous Compression Test) for the speakers he tests. He describes the test as follows: "The purpose of this test is to illustrate how much (if at all) the output changes as a speaker’s components temperature increases (i.e., voice coils, crossover components) instantaneously". But heating is not the only source of compression - mechanical nonlinearities also can cause reductions in output (suspension gets stiffer as displacement increases), as can the electromagnetic nonlinearity of the coil in the magnetic gap (the force produced decreases as the coil is displaced). So I don't know how to separate the contribution of those various effects (thermal, mechanical, electromagnetic) in Erin's compression data.


I would estimate the effect by Googling the temperature coefficient of resistance of copper, such as here: https://www.allaboutcircuits.com/textbook/direct-current/chpt-12/temperature-coefficient-resistance/ . It says the coefficient is 0.004041. So an increase from 20C to 60C increases the resistance by 1.16. That could produce a change of amplitude of 1.3dB, or less, depending on the circuit.

There are many effects.
I wonder about the audible impact.
Klippel for example has lots of literature on it.

https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_18_Measurement_of_Linear_Thermal_Parameters.pdf

https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_19_Nonlinear_Thermal_Parameters_%28Convection%20Cooling%29.pdf

The implication is that these things get really hot:

I'm a bit unclear on how often I operate my speakers in this regime. Given recent discussions about drivers, break-in, and temperature, I thought I would try to see how much a reasonable application of heat affects a driver.

I got some Audax HP210GO woofers in storage for some time. After confirming it operates correctly, and breaking it in (another thread) I measured the parameters at ambient room conditions. I then dangled it in front of a space heater, careful to make sure it was on a gentle setting that didn't create hot spots or overly heat the driver:

I heated the driver for 10 minutes, turned off the heater and measured. I repeated the 10 minute heating and measurement cycle 8 times. After the 5th cycle I moved the heater a bit closer, then after the 7th cycle I moved it really close. I then let the speaker rest unheated at ambient room Temperature for 1.5 hours, then made a 9th and final measurement. Here are the f(s) and Q(ms) trends:

The result is f(s) drops 2.8% in the first ten minutes. Note this is twice the change I observed on this same driver under break-in! After the first ten minutes, further cycles with the heater have little impact to f(s) even as Q(ms) is changing. Even moving the heater slightly closer after run-5 has little additional effect. Moving it really close after run-7 does drop f(s) a bit more to 3.6% lower than initial. I can change f(s) and Q(ms) by over 3% with a space heater.

This rough heater experiment is likely much different than heating with amplifier power to a voice coil. It's more like being exposed a hot environment, I'm pretty sure direct sunlight is more aggressive than what I did. It's likely I never got the woofer to a steady state in any of these measurements, need an oven to do that properly. But I think we can at least see that moderate changes in temperature like what might be experienced in a room will have a few percent impact on the driver's parameters.

Here are R(e) and L(e) trends for this experiment. It does look like the voice-coil was steadily increasing in Temperature, reinforcing that the driver was never in a steady state and different parts were likely warming at different rates. I was directing the hot air at the woofer-cone, so no surprise if the VC is going to respond over a longer period. I am not surprised by this, most divers have dramatically different thermals across different components. I was actually surprised to see these drivers return to baseline after only 1.5 hours after all of this heating. These drivers have small motors, more robust gear like the Dynaudio will heat up and cool down much more slowly. And take lots more heat than a small motor like in these modest Audax units.

Is all of this audible? I have long claimed break-in is real but tiny, and for all practical purposes inaudible. In this case break-in is maybe half the effect of the modest heating I applied here.

 

[MaxwellsEq]

There are many effects.
I wonder about the audible impact.
Klippel for example has lots of literature on it.

https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_18_Measurement_of_Linear_Thermal_Parameters.pdf

https://www.klippel.de/fileadmin/klippel/Files/Know_How/Application_Notes/AN_19_Nonlinear_Thermal_Parameters_%28Convection%20Cooling%29.pdf

The implication is that these things get really hot:
View attachment 335836
I'm a bit unclear on how often I operate my speakers in this regime. Given recent discussions about drivers, break-in, and temperature, I thought I would try to see how much a reasonable application of heat affects a driver.

I got some Audax HP210GO woofers in storage for some time. After confirming it operates correctly, and breaking it in (another thread) I measured the parameters at ambient room conditions. I then dangled it in front of a space heater, careful to make sure it was on a gentle setting that didn't create hot spots or overly heat the driver:
View attachment 335839
I heated the driver for 10 minutes, turned off the heater and measured. I repeated the 10 minute heating and measurement cycle 8 times. After the 5th cycle I moved the heater a bit closer, then after the 7th cycle I moved it really close. I then let the speaker rest unheated at ambient room Temperature for 1.5 hours, then made a 9th and final measurement. Here are the f(s) and Q(ms) trends:
View attachment 335847
The result is f(s) drops 2.8% in the first ten minutes. Note this is twice the change I observed on this same driver under break-in! After the first ten minutes, further cycles with the heater have little impact to f(s) even as Q(ms) is changing. Even moving the heater slightly closer after run-5 has little additional effect. Moving it really close after run-7 does drop f(s) a bit more to 3.6% lower than initial. I can change f(s) and Q(ms) by over 3% with a space heater.

This rough heater experiment is likely much different than heating with amplifier power to a voice coil. It's more like being exposed a hot environment, I'm pretty sure direct sunlight is more aggressive than what I did. It's likely I never got the woofer to a steady state in any of these measurements, need an oven to do that properly. But I think we can at least see that moderate changes in temperature like what might be experienced in a room will have a few percent impact on the driver's parameters.

Here are R(e) and L(e) trends for this experiment. It does look like the voice-coil was steadily increasing in Temperature, reinforcing that the driver was never in a steady state and different parts were likely warming at different rates. I was directing the hot air at the woofer-cone, so no surprise if the VC is going to respond over a longer period. I am not surprised by this, most divers have dramatically different thermals across different components. I was actually surprised to see these drivers return to baseline after only 1.5 hours after all of this heating. These drivers have small motors, more robust gear like the Dynaudio will heat up and cool down much more slowly. And take lots more heat than a small motor like in these modest Audax units.


View attachment 335846
Is all of this audible? I have long claimed break-in is real but tiny, and for all practical purposes inaudible. In this case break-in is maybe half the effect of the modest heating I applied here.

Fascinating, thanks for doing the experiment!

 

[vdH_83]

Do you have an LCR meter or DATS or other way of measuring inductance

Thanks for your input! Its like a sea of information that is starting to make sense. Very exciting! The DATS V3 is on its way

 

[vdH_83]

You moved some parts around, but it looks to me that all the connections are the same

There was a mistake in the midrange path. I triple checked the schematic, Im pretty sure that its right now. Is there anything out of the ordinary?

 

[Zapper]

There was a mistake in the midrange path. I triple checked the schematic, Im pretty sure that its right now. Is there anything out of the ordinary?

What was the mistake? I've compared the two versions of the schematic several times and haven't found an electrical difference. The components are drawn differently but the network is the same.

 

[vdH_83]

What was the mistake? I've compared the two versions of the schematic several times and haven't found an electrical difference. The components are drawn differently but the network is the same.

Dear Zapper, please see picture below. Just checked the PCB for the seventh time. Im pretty sure this is it Used DATS tonight and the inductors are measured now. I'll upload the DEF schematic right away.