KP-302/3002 High Frequency Speaker - Intermittent & Scratchy Sound - Resolved, Bad Cap' Connection

[Chief bonehead]

So is the goal a good looking voltage transfer curve or a good looking freq response?

[captainbeefheart]

14 hours ago, Chief bonehead said:

So is the goal a good looking voltage transfer curve or a good looking freq response?

 

Ideally both.

 

The transfer function comes down to each different type of filters cutoff characteristics, and is mathematically defined.

 

First order Butterworth (-6db/octave slope) is phase coherent, the outputs combine perfectly. But it's very demanding on the drivers, they need to be able to handle and faithfully reproduce frequencies slightly outside their sweet spot. A second order Butterworth steepens the slope up to -12db/octave but the outputs are 180° out of phase which is why you see the tweeter leads flipped around. Most Klipsch crossovers are third order Butterworth which has fast rollof of -18db/octave with good phase characteristics. The outputs phase sum together well in either polarity, and the phase response is a gradually changing shift over the audio range. The pitfall of this design is it can not be corrected using time delay for the case where the loudspeakers do not radiate from the same vertical plane. I kinda like the fourth order linkwitz-Riley Crossover. The outputs sum with flat frequency response, the outputs are in phase at the crossover frequency, and the phase relationship of the output allows time correction for drivers that are not in the same acoustic plane.

 

 

[Chief bonehead]

1 hour ago, captainbeefheart said:

 

Ideally both.

yep. In the “ideal” world. I’m still looking. 

1 hour ago, captainbeefheart said:

 

The transfer function comes down to each different type of filters cutoff characteristics, and is mathematically defined.

and how are they “typically” defined?  Isn’t it into a consistent load?  Do speakers represent a consistent load?

1 hour ago, captainbeefheart said:

 

First order Butterworth (-6db/octave slope) is phase coherent, the outputs combine perfectly.

yep with constant loads. Not varying loads. 

1 hour ago, captainbeefheart said:

 

But it's very demanding on the drivers, they need to be able to handle and faithfully reproduce frequencies slightly outside their sweet spot

 

which means at higher voltages the drivers just might be outside their linear range and begin to distort. A large overlap will occur but you might be able to get a single point in space where the curve just might look very good. But again, is a good looking curve all that is needed?

1 hour ago, captainbeefheart said:

. A second order Butterworth steepens the slope up to -12db/octave but the outputs are 180° out of phase which is why you see the tweeter leads flipped around. Most Klipsch crossovers are third order Butterworth which has fast rollof of -18db/octave with good phase characteristics. The outputs phase sum together well in either polarity, and the phase response is a gradually changing shift over the audio range. The pitfall of this design is it can not be corrected using time delay for the case where the loudspeakers do not radiate from the same vertical plane. I kinda like the fourth order linkwitz-Riley Crossover. The outputs sum with flat frequency response, the outputs are in phase at the crossover frequency, and the phase relationship of the output allows time correction for drivers that are not in the same acoustic plane.

I like whatever order is needed to provide the things in need in order to have constant amplitude, similar phase at acoustic crossover, constant coverage, similar distortion characteristics and very close coverage patterns at the crossover freq, etc etc that meets those needs. The crossover is a tool and should not be limited to textbook designs that are derived using a constant load because in the real world, typically not the ideal world, you don’t have a constant load as it will vary over a frequency. And typical, hornload speaker is even worse because of their efficiency. And to make matters worse you have to adjust gain between the drivers and provide eq along with crossovers thus the name networks. The network is a tool that is adjusted to make the drivers blend and become coherent cause after all, we don’t want to hear a good voltage transfer curve……if we did our job correctly, we want not only a good freq curve but also one that meets our other criteria. Wouldn’t that make a network very specialized for a set of specific drivers with specific specs?  Just wondering. Maybe being wholeheartedly concerned what the acoustic slope is a better goal?

1 hour ago, captainbeefheart said:

 

 

 

[captainbeefheart]

5 hours ago, Chief bonehead said:

yep. In the “ideal” world. I’m still looking. 

and how are they “typically” defined?  Isn’t it into a consistent load?  Do speakers represent a consistent load?

yep with constant loads. Not varying loads. 

which means at higher voltages the drivers just might be outside their linear range and begin to distort. A large overlap will occur but you might be able to get a single point in space where the curve just might look very good. But again, is a good looking curve all that is needed?

I like whatever order is needed to provide the things in need in order to have constant amplitude, similar phase at acoustic crossover, constant coverage, similar distortion characteristics and very close coverage patterns at the crossover freq, etc etc that meets those needs. The crossover is a tool and should not be limited to textbook designs that are derived using a constant load because in the real world, typically not the ideal world, you don’t have a constant load as it will vary over a frequency. And typical, hornload speaker is even worse because of their efficiency. And to make matters worse you have to adjust gain between the drivers and provide eq along with crossovers thus the name networks. The network is a tool that is adjusted to make the drivers blend and become coherent cause after all, we don’t want to hear a good voltage transfer curve……if we did our job correctly, we want not only a good freq curve but also one that meets our other criteria. Wouldn’t that make a network very specialized for a set of specific drivers with specific specs?  Just wondering. Maybe being wholeheartedly concerned what the acoustic slope is a better goal?

 

 

As a speaker engineer you can choose whatever filter you feel is best but they are still strictly defined by specific mathematical equations with very specific transfer characteristics. The transfer characteristics or cutoff characteristics is where we enter the stop band part of the filter, not the pass band. In a crossover network these points happen to be where the crossover frequency is. It seems complicated to have to worry about the changing of impedance over the entire frequency range but you don't have to really. Take the woofer circuit, the inductive reactance from the voice coil will increase at high frequencies changing the load impedance but it's moot because we are now in the stop band part of the curves where it's attenuated as it's the woofer we are discussing, you only have to worry about the frequencies coming through the woofer which is in the pass band part of the filter and flat. It just happened to have a resonance via the capacitance chosen just at the knee of the transfer curve, all that was needed was to be damped to keep the filter curve flat, it had nothing to do with the type of filter that was chosen. For the tweeter, the voice coil reactance is so small compared to the woofer that it will only come into play at an even higher frequency which is moot because we are now in the flat pass band part of the tweeter network anyway. The reactive elements of the drivers shouldn't create a problem for the filter network designs.

 

The way I view it is I would want a flat transfer for all crossover points then move onto making the acoustic frequency response flat. As noted it wasn't the actual choice of filter type that gave us that resonance it was the 245mH inductor and 80uF capacitor that happened to put a resonance right at the worst spot, right at the end of the pass band just before the stop band "knee" making a +1.6db boost. All the important stuff is only happening right where it goes from pass band to stop band which is a specific frequency, so all we as engineers have to do is make sure it's a flat transfer and it's all good. The rest of the changing impedance vs frequency from the driver coil  is kinda moot because it will fall either in the pass band or stop band part of the filter.

[henry4841]

Bought an Onkyo TX-5000 for refurb and recap. Pulled output board to recap and the electrolytic Nichicon caps tested as good or better then the replacement caps I use. Since it works fine I decided to leave complete recap off. So much for all 40 year old electrolytic caps as being bad or going bad. Sure they are old and may fail but so will the sun one day. Probably has more to do with the quality of the Nichicon caps of that era. 

[Chief bonehead]

4 hours ago, captainbeefheart said:

 

As a speaker engineer you can choose whatever filter you feel is best but they are still strictly defined by specific mathematical equations with very specific transfer characteristics. The transfer characteristics or cutoff characteristics is where we enter the stop band part of the filter, not the pass band. In a crossover network these points happen to be where the crossover frequency is. It seems complicated to have to worry about the changing of impedance over the entire frequency range but you don't have to really. Take the woofer circuit, the inductive reactance from the voice coil will increase at high frequencies changing the load impedance but it's moot because we are now in the stop band part of the curves where it's attenuated as it's the woofer we are discussing, you only have to worry about the frequencies coming through the woofer which is in the pass band part of the filter and flat. It just happened to have a resonance via the capacitance chosen just at the knee of the transfer curve, all that was needed was to be damped to keep the filter curve flat, it had nothing to do with the type of filter that was chosen. For the tweeter, the voice coil reactance is so small compared to the woofer that it will only come into play at an even higher frequency which is moot because we are now in the flat pass band part of the tweeter network anyway. The reactive elements of the drivers shouldn't create a problem for the filter network designs.

 

The way I view it is I would want a flat transfer for all crossover points then move onto making the acoustic frequency response flat. As noted it wasn't the actual choice of filter type that gave us that resonance it was the 245mH inductor and 80uF capacitor that happened to put a resonance right at the worst spot, right at the end of the pass band just before the stop band "knee" making a +1.6db boost. All the important stuff is only happening right where it goes from pass band to stop band which is a specific frequency, so all we as engineers have to do is make sure it's a flat transfer and it's all good. The rest of the changing impedance vs frequency from the driver coil  is kinda moot because it will fall either in the pass band or stop band part of the filter.

Well I tried……🙂

[captainbeefheart]

2 hours ago, Chief bonehead said:

Well I tried……🙂

 

I apologize if I didn't explain it well.

 

You seem to be hung up on load so lets deal with that. The application of a Butterworth filter as we use in these crossovers is known as a Cauer Topology and the maths depend upon whether you have even or odd number of filters and load isn't even a variable.

 

I tried to make it clear that when discussing the transfer characteristics they only occur at a specific frequency, in our case the crossover frequency, this is where the filter goes from pass band to stop band. So worrying about the driver impedance vs frequency isn't necessary as I will try and clarify. For the woofer, any anomalies in regard to changing impedance with frequency will be dealt with by the amplifier or not important. For frequencies below the cutoff point we will be in the pass band of the filter (view it as a straight wire because impedance of the series filter networks are low since we are in the flat part of the filter transfer curve. This means the amplifier will see it directly. For frequencies above the cutoff point we are now in the stop band part of the filter where impedance is high for the series element and the shunt element has a low impedance shunting signal to ground so the drivers change in impedance vs frequency is also moot here. The only place of importance is where the corner frequency is. Let's look at the tweeter. For frequencies above the cutoff frequency we will be in the pass band part of the filter network (low series impedance, high shunt impedance) and you can view the filter network as a wire because of the low series impedance, the amplifier will now see the tweeter directly and deal with it's anomalies directly. If we go below the cutoff frequency we are now in the stop band part of the filter (high serial impedance and low shunt resistance), and this means that the amplifier isn't even seeing the tweeter because the shunt element is of low impedance and dumping the frequencies before getting to the driver. So the only places for us to have to dial in is where the cutoff (crossover) frequencies are.

 

Is that a little clearer?

[Edgar]

13 minutes ago, captainbeefheart said:

For the woofer, any anomalies in regard to changing impedance with frequency will be dealt with by the amplifier or not important. For frequencies below the cutoff point we will be in the pass band of the filter (view it as a straight wire because impedance of the series filter networks are low since we are in the flat part of the filter transfer curve. This means the amplifier will see it directly. For frequencies above the cutoff point we are now in the stop band part of the filter where impedance is high for the series element and the shunt element has a low impedance shunting signal to ground so the drivers change in impedance vs frequency is also moot here. The only place of importance is where the corner frequency is. Let's look at the tweeter. For frequencies above the cutoff frequency we will be in the pass band part of the filter network (low series impedance, high shunt impedance) and you can view the filter network as a wire because of the low series impedance, the amplifier will now see the tweeter directly and deal with it's anomalies directly. If we go below the cutoff frequency we are now in the stop band part of the filter (high serial impedance and low shunt resistance), and this means that the amplifier isn't even seeing the tweeter because the shunt element is of low impedance and dumping the frequencies before getting to the driver. So the only places for us to have to dial in is where the cutoff (crossover) frequencies are.

 

If I'm reading this correctly, you're saying that the impedance of the load does not matter except for its value at exactly the cutoff frequency. I recommend that you model a driver or two in SPICE and observe the results. Classical crossover design assumes a constant resistive load. When the load is complex, the reactive load components interact with the reactive filter components and change the filter characteristics. When the real part of the load differs from the assumed value, the cutoff frequency and filter Q change.

 

Furthermore, a textbook approach to filter design ignores the fact that the driver's acoustic response may vary wildly, even with constant signal input. The output of the driver needs to be factored into the crossover design, too.

[Chief bonehead]

1 hour ago, captainbeefheart said:

 

I apologize if I didn't explain it well.

 

You seem to be hung up on load so lets deal with that. The application of a Butterworth filter as we use in these crossovers is known as a Cauer Topology and the maths depend upon whether you have even or odd number of filters and load isn't even a variable.

 

I tried to make it clear that when discussing the transfer characteristics they only occur at a specific frequency, in our case the crossover frequency, this is where the filter goes from pass band to stop band. So worrying about the driver impedance vs frequency isn't necessary as I will try and clarify. For the woofer, any anomalies in regard to changing impedance with frequency will be dealt with by the amplifier or not important. For frequencies below the cutoff point we will be in the pass band of the filter (view it as a straight wire because impedance of the series filter networks are low since we are in the flat part of the filter transfer curve. This means the amplifier will see it directly. For frequencies above the cutoff point we are now in the stop band part of the filter where impedance is high for the series element and the shunt element has a low impedance shunting signal to ground so the drivers change in impedance vs frequency is also moot here. The only place of importance is where the corner frequency is. Let's look at the tweeter. For frequencies above the cutoff frequency we will be in the pass band part of the filter network (low series impedance, high shunt impedance) and you can view the filter network as a wire because of the low series impedance, the amplifier will now see the tweeter directly and deal with it's anomalies directly. If we go below the cutoff frequency we are now in the stop band part of the filter (high serial impedance and low shunt resistance), and this means that the amplifier isn't even seeing the tweeter because the shunt element is of low impedance and dumping the frequencies before getting to the driver. So the only places for us to have to dial in is where the cutoff (crossover) frequencies are.

 

Is that a little clearer?

Actually it was me that didn’t explain it clearly. Looking at crossovers independently without realizing that it is part of closed system is the only way to get to the acoustic response that we are designing for…..ultimately that is what your ears hear and that determines the quality of the sound. 

[jjptkd]

1 hour ago, captainbeefheart said:

 

I apologize if I didn't explain it well.

 

You seem to be hung up on load so lets deal with that.

 

Is that a little clearer?

This reminds me of the saying there is no correlation between intelligence and wisdom-- not implying that you are not wise just that creating speakers or stereo equipment for that matter seems to be somewhat of an art along with science otherwise anyone could just pull out a book and a calculator and make a top shelf product. 

 

Also reminds me of an old speaker review of the original Chorus where a guy thought they sounded good, measured them and found out just how awful they actually were, redesigned the crossover to make them near "perfect" and then they sounded like crap, ultimately leading him to return them back to stock.

[captainbeefheart]

1 hour ago, jjptkd said:

This reminds me of the saying there is no correlation between intelligence and wisdom-- not implying that you are not wise just that creating speakers or stereo equipment for that matter seems to be somewhat of an art along with science otherwise anyone could just pull out a book and a calculator and make a top shelf product. 

 

Also reminds me of an old speaker review of the original Chorus where a guy thought they sounded good, measured them and found out just how awful they actually were, redesigned the crossover to make them near "perfect" and then they sounded like crap, ultimately leading him to return them back to stock.

 

I never once stated acoustic response was not important. I am only saying choosing a specific filter will have it's pro's and cons and as the engineer you need to pick your trade offs. For example the third order filters often chosen because of their steeper slope BUT because of it's phase response it's impossible to correct so it's a trade off. Shouldn't one strive for a flat transfer and then look at the acoustic response? Who wants a resonance smack dab at the crossover frequency? I was specifically asked by Chief Bonehead this question and I stated that I feel both the transfer properties and the acoustic response are important.

 

1 hour ago, Chief bonehead said:

Actually it was me that didn’t explain it clearly. Looking at crossovers independently without realizing that it is part of closed system is the only way to get to the acoustic response that we are designing for…..ultimately that is what your ears hear and that determines the quality of the sound. 

 

As stated above, I am not disregarding acoustic response and never said it's not important. I know it's a complete system I just tend to solve one problem before solving the next.

 

2 hours ago, Edgar said:

If I'm reading this correctly, you're saying that the impedance of the load does not matter except for its value at exactly the cutoff frequency. I recommend that you model a driver or two in SPICE and observe the results. Classical crossover design assumes a constant resistive load. When the load is complex, the reactive load components interact with the reactive filter components and change the filter characteristics. When the real part of the load differs from the assumed value, the cutoff frequency and filter Q change.

 

Furthermore, a textbook approach to filter design ignores the fact that the driver's acoustic response may vary wildly, even with constant signal input. The output of the driver needs to be factored into the crossover design, too.

 

Let's put numbers to it then. Eminence Kappa has a voice coil inductance of 1.4mH and the drivers resonant frequency is 35Hz. Ask yourself how will the voice coil reactance effect the network? Well, the resonant frequency of the 1.4mH and 80uF capacitor is 475Hz, as I already stated the problems arise near the crossover frequency. The 35Hz resonant frequency is in the networks pass band where it's flat with a 0° phase shift. How will this effect the network? The 1.4mH voice coil inductance will start to increase in impedance with frequency in the stop band part of the network. Again how will this effect the network? I have added the 1.4mH voice coil inductance to the simulation and it changes nothing as expected which is why I left it out to begin with, crunching numbers quickly led me to this conclusion and the simulation confirms it.

[captainbeefheart]

I was told to look at the bigger picture, which is exactly my point. You cannot ignore the nature of the filter you decide to choose because ultimately it will have an effect on the acoustic response. Example, choosing an even number of filter elements vs odd numbered, they will have different phase characteristics you cannot ignore. It's not always about amplitude and slope steepness.

[Edgar]

2 minutes ago, captainbeefheart said:

Let's put numbers to it then. Eminence Kappa has a voice coil inductance of 1.4mH and the drivers resonant frequency is 35Hz. Ask yourself how will the voice coil reactance effect the network?

 

It's not clear to me exactly what point you are trying to make. The effects of the inductance are entirely predictable. Whether those effects can be neglected depends upon the configuration of the network and the frequency band of interest, and is a judgment to be made by the designer, hopefully in an enlightened fashion. But it cannot just be summarily dismissed.

 

Quote

Well, the resonant frequency of the 1.4mH and 80uF capacitor is 475Hz,

 

That would be exact if they were considered in isolation. But they're not.

 

Quote

as I already stated the problems arise near the crossover frequency.

 

By my reading, you said at the crossover frequency, and argued that anywhere else it can be "view[ed] as a straight wire". Asymptotically true, but again, the decision as to where the effects can be considered negligible is a judgment call.

 

You are equating piecewise linear approximations with actual performance. Even a little bit of optimization requires better approximations than that.

 

Quote

The 35Hz resonant frequency is in the networks pass band where it's flat with a 0° phase shift.

 

That cited 0° phase shift is an approximation. No passive network has 0° phase shift over any significant bandwidth.

 

Quote

The 1.4mH voice coil inductance will start to increase in impedance with frequency in the stop band part of the network.

 

No, the impedance will start to increase at anything above DC. Again, the point where it becomes significant is a designer's judgment call.

 

Quote

I have added the 1.4mH voice coil inductance to the simulation and it changes nothing as expected which is why I left it out to begin with, crunching numbers quickly led me to this conclusion and the simulation confirms it.

 

Then you have made that judgment call. It may be entirely appropriate in this portion of this network. (In fact you have provided evidence that it is.) But it will not be true in all cases. Context is vitally important.

[Edgar]

5 minutes ago, captainbeefheart said:

Example, choosing an even number of filter elements vs odd numbered, they will have different phase characteristics you cannot ignore. It's not always about amplitude and slope steepness.

 

Now that is something that I can agree with wholeheartedly!

[captainbeefheart]

5 minutes ago, Edgar said:

 

It's not clear to me exactly what point you are trying to make. The effects of the inductance are entirely predictable. Whether those effects can be neglected depends upon the configuration of the network and the frequency band of interest, and is a judgment to be made by the designer, hopefully in an enlightened fashion. But it cannot just be summarily dismissed.

 

 

That would be exact if they were considered in isolation. But they're not.

 

 

By my reading, you said at the crossover frequency, and argued that anywhere else it can be "view[ed] as a straight wire". Asymptotically true, but again, the decision as to where the effects can be considered negligible is a judgment call.

 

You are equating piecewise linear approximations with actual performance. Even a little bit of optimization requires better approximations than that.

 

 

That cited 0° phase shift is an approximation. No passive network has 0° phase shift over any significant bandwidth.

 

 

No, the impedance will start to increase at anything above DC. Again, the point where it becomes significant is a designer's judgment call.

 

 

Then you have made that judgment call. It may be entirely appropriate in this portion of this network. (In fact you have provided evidence that it is.) But it will not be true in all cases. Context is vitally important.

 

Judgement call? How?

 

These are defined by the crossover points which makes my assumptions of frequencies below it will be in the pass band and above it in the stop band correct. It has to work out this way as it's the nature of the filter network's application. Moving either left or right from the "chosen" crossover frequencies this assumption rings true.

 

Can you provide an example where the voice coil inductance will create an issue with the network? I think it makes complete sense that the reactance of the voice coil will only be an issue when near the crossover frequency because if it's not then it's either in the pass band or in the stop band part of the network. So for the woofer the reactance increase will be in the stop band, there is no getting around this. For the tweeter it will be in the pass band, there is no avoiding this either.

 

7 minutes ago, Edgar said:

That cited 0° phase shift is an approximation. No passive network has 0° phase shift over any significant bandwidth.

 

In context to the 35Hz resonance we are -7° phase shift, insignificant if you ask me.

[captainbeefheart]

Also, of course I know that the voice coil reactance will increase with frequency from DC up, but again in the context of the discussion it won't in a meaningful way until you are up into the stop band part of the woofer network.

[Edgar]

37 minutes ago, captainbeefheart said:

Judgement call? How?

 

In that the designer observes or calculates the effects, and in the context of the desired characteristics decides that those effects can be neglected.

 

Quote

These are defined by the crossover points which makes my assumptions of frequencies below it will be in the pass band and above it in the stop band correct. It has to work out this way as it's the nature of the filter network's application.

 

That is exactly the context to which I refer.

 

Quote

Can you provide an example where the voice coil inductance will create an issue with the network?

 

I have some old woofer SPICE models ... somewhere. If I can find them, then I'll experiment a bit and report back.

 

Quote

So for the woofer the reactance increase will be in the stop band, there is no getting around this. 

 

Exactly. And if the voice coil inductance is large enough then, for example, it can conceivably turn a 2nd-order LPF into a 3rd-order LPF, along with the attendant phase shift.

 

Quote

In context to the 35Hz resonance we are -7° phase shift, insignificant if you ask me.

 

And if you are the designer, then you have just made exactly the judgment call to which I referred.

[Edgar]

33 minutes ago, Edgar said:

I have some old woofer SPICE models ... somewhere. If I can find them, then I'll experiment a bit and report back.

Turns out I had a ready-made example sitting right on my hard drive. A while back I designed a 1200 Hz 3rd-order LPF for a Philips AD12240W8 that I had measured. In the graphic below, the circuit on the left side, feeding VM2, is the woofer model and the filter constructed of real capacitors and inductors that I had on-hand. The circuit on the right side, feeding VM1, is the textbook 3rd-order Butterworth filter using ideal components. The effects upon the frequency and phase response are obvious, and are not limited to the crossover frequency (green is actual, red is textbook).

 

The second graphic shows the same, with the woofer inductance compensated with a Zobel network. The improvement is significant, in my judgment.

[Deang]

That peak is crazy, you sure you didn’t pencil that in. 

[Edgar]

9 minutes ago, Crankysoldermeister said:

That peak is crazy, you sure you didn’t pencil that in. 

All of the component values are shown. You are welcome to simulate it for yourself. https://www.ti.com/tool/TINA-TI