Thank to had confirmed this by your own experience. yes, it is in the dynamic domain that everything happens in a way you cannot achieve by an equalizer. And, yes, i believe like-you that:" the real proof is in the listening".
Curing The Problem, Cures The Symptoms....
Christophe, I second all of your points above.
I might further add that IME when the drivers are compensated correctly, power handling goes up dramatically, live recordings become surround with sounds coming from way behind the speakers, from way behind the listener, and from way out past the sides of the speakers (provided that absolute polarity is correct).
Another consequence is that close mic'd sounds are revealed as exactly that...a narrow depth band of sound between the speakers....which is as it should be actually.
Yamaha beryllium dome tweeters (NS-10, NS-20T etc) are notorious for being ear bleeding, tissssy etc.
With zobel and resonance compensation these tweeters become beautifully behaved with clean, enjoyable response way out past hearing limits....no tissssss, no harshness, just fantastic HF extension and great detail.
Ditto the Yamaha NS-20T woofer then becomes useful as full range driver, with a hint of the tweeter (series 1.1uF) on top.
Typical GNFB amplifiers measure well into resistive load, sound good into resistive load, but IME don't expect GNFB amplifiers to sound optimal into reactive loads.
Dan.
Christophe, I second all of your points above.
I might further add that IME when the drivers are compensated correctly, power handling goes up dramatically, live recordings become surround with sounds coming from way behind the speakers, from way behind the listener, and from way out past the sides of the speakers (provided that absolute polarity is correct).
Another consequence is that close mic'd sounds are revealed as exactly that...a narrow depth band of sound between the speakers....which is as it should be actually.
Yamaha beryllium dome tweeters (NS-10, NS-20T etc) are notorious for being ear bleeding, tissssy etc.
With zobel and resonance compensation these tweeters become beautifully behaved with clean, enjoyable response way out past hearing limits....no tissssss, no harshness, just fantastic HF extension and great detail.
Ditto the Yamaha NS-20T woofer then becomes useful as full range driver, with a hint of the tweeter (series 1.1uF) on top.
Typical GNFB amplifiers measure well into resistive load, sound good into resistive load, but IME don't expect GNFB amplifiers to sound optimal into reactive loads.
Dan.
Hmmm, if the reason for a Zobel is to modify frequency response, how does that work? A Zobel network (or general impedance flatenning), if properly implemented will not effect frequency response. It adds another current path that flattens impedance and therefore current flow.
I know that nobody that has done it and "heard a difference" is going to be dissuaded from that. I just want to hear a good theoretical argument , maybe see some measurements backing up the premise.
If frequency response is key then why don't we concentrate on that in our crossover networks? As I advocate above, decide what response you need your filter to provide and put in the topology and values that achieve that.
In my experience that seldom includes impedance flattening.
David
Hello David, yes indeed, the zobel provides another current path.....but it also provides damping, and I think this is the key (must be) to the change in dynamic behaviour....drivers become cleaner and tweeters lose that unnatural 'tisssiness', much complained about regarding metallic dome tweeters.
Compensated drivers ought to allow a simpler preceding filter stage, plus the benefit of additional damping applied directly across the driver terminals.
I fully agree the maths of your techniques are correct, but in practice zobel networks provide worthwhile benefit.
Dan.
Well, i'm afraid the theoretical arguments cannot convince-you. There is so many thing we don't know about the way our ears+brain work. By example the benefit to have ultra high slew rate margins in closed loop amplifiers. Stupid, isn't it, when the measurable difference is after 1Mhz ? And yet ....
The measurement you will have will be acoustic levels. They will just show differences in response curves. Nothing about dynamic behavior, that i don't know how to measure.
Man, it is just like your carbon damping material (or the "in middle" damping position). You can get the same acoustic levels measurements with a bigger closed enclosure (as an example), or to empty all the volume with fiberglass in a more little one. But, did-you agree there will be a difference in music reproduction between all those enclosures ?
http://users.ece.gatech.edu/mleach/ece4445/downloads/zobel.pdf
Nice article by Marshall Leach. He does full compensation of a driver (motional impedance and inductance rise) and also uses a more accurate two part compensation for the less than 6dB per Octave L rise.
Nice article by Marshall Leach. He does full compensation of a driver (motional impedance and inductance rise) and also uses a more accurate two part compensation for the less than 6dB per Octave L rise.
Nice one. Thanks a lot David. That is exactly the schematic i published as an example.
Lets agree together, as a start, that using flat impedance load after a passive filter cannot present negative effects. OK ?
If you look at the filter i published earlier, you will see that, when it is not necessary, we don't hesitate to simplify the networks. I don't disagree with your inputs on this point.
Now, you have 3 people (including Dominique Petoin site, if you can understand French or use Google translation), with no connection between them, not totally unwashed in electronic, with nothing to sell, not believers in snake oil, who reports the same listening experience with each their own words.
Don't it teases you curiosity to try and experiment, as i did the first time ?
I'm sure that, it you confirm positive results, with all your knowledge, you will find a better way to explain why and how. If not, you will be in the same uncomfortable position: "I have seen an OVNI". 😉
Last edited:
The TB W2 800 are 8 Ohm IIRC and have no ferrofluid. At a guess Id think the Bandors, like the old Mark Audios have ferrofluid.
I hope it is true, so many tweeters with the coil glued after few years !
I hope it will be the same with foam surrounds 🙂
yes i think so, tbh i dont know if the Bandors ever used it. Without knowing id suspect they do (or did at least) to help stabilise the rocking modes when using no rear spider. Saying that Bandor have a different device so perhaps never required ferrofluid, unlike the alpairs I have.It could explain the moderate impedance peak though it could be due to something else in their design beyond my understanding. It is my understanding that higher excursion gives a higher peak also. Maybe Qes has an influence too, the W2s are very low Qts which would also imply high back EMF and i think that results in a lower impedance. Tho i may have that backwards...maybe someone will correct me if thats not entirely accurate (wouldnt want to mislead)
Last edited:
There's a spider on the Bandors, and quite a hefty looking motor etc. I'll post a pic tomorrow -
Monacor do a black anodised version of the W2-800/Omnes 2.01, I think that one is 8 ohm
Test Lautsprecherchassis Breitbänder - Monacor SPX-20TB -
Monacor do a black anodised version of the W2-800/Omnes 2.01, I think that one is 8 ohm
Test Lautsprecherchassis Breitbänder - Monacor SPX-20TB -
Here are my thoughts on Zobel networks:
I am mostly on Dave's side. In most cases they are needed simply to make the xover network behave into the complex load, to give the intended target response, ie the target electro-acoustic response divided by the driver response. Without a zobel, a second order low pass will not always give you the desired second order response, and most certainly with low pass filters on three-way vented systems, or especially on couple cavity bass loading, will never give the desired responses without some form of zobel.
However, the perfect zobel will never be derived from the Thiele-Small parameters, but must be calculated from an accurate equivalent circuit derived from a measurement of the impedance curve of the driver/box.
Even after doing this, once you optimize the network to meet your target xover response, you can nearly always get better results by letting the zobel network be re-optimized away from the perfect values, in essence they become additional reactive elements giving greater degrees of freedom to the optimizer to reach the intended goal.
Now, as to the idea of the zobel "damping " out the driver, then this is rarely required. The amplifier provides this function over nearly the whole frequency range, even when there is a network in the way, at least for even order networks. If you look at the source impedance looking back through the network towards the amplifier, all second order networks are low impedance at both low and high frequencies, and only have a high source impedance at the cutoff frequency. This frequency should not be at the resonance frequency of the driver, so therefore damping where it is needed is generally maintained.
This is not of course the case with an odd order network due to the last component in the network being a series reactive element. This is why a third order high pass on
the tweeter will always interact with the tweeter impedance at its resonant frequency and give you an odd looking response. This is where a zobel might be needed.
However, Zobel or not, the reasons for any change in sound is purely contained in the subsequent changes in frequency response, and not due to some special feature associated with zobel networks in isolation.
The idea of trying to design a network from values calculated from theoretical classic shapes, and hence needing to flatten the impedance curve of the driver to represent a resistive load is both quaint and very backward in so many ways.
Now, with regard to flattening of the impedance presented to the amplifier, this of course can have benefits to the amplifier if the speaker otherwise would present a highly reactive load to the amp. If the real part of the impedance is low at the same time as the imaginary component is high, there will be high peak current demands placed on the amplifier, and depending on the amplifier design this can affect the sound quality of the amplifier. There are many speakers on the market that suffer from this, and mostly due to poor network design.
As to the benefits of further flattening the impedance curve to be almost purely resistive, as we did with the 104.2, this has two potential benefits:
1/ because the current is always in phase with the voltage, amplifier dissipation can be dramatically lower, with lower instantaneous dissipation in the output devices. Also there are less demands on the power supply, further benefitting the amplifier sound.
2/ a resistive impedance for the speaker eliminates any interaction with the amplifier source impedance which would otherwise alter the frequency response. This is especially crucial for most tube amplifiers, which have typically quite high source impedances that can result in response modification of the order of several dB. No wonder tubes sound so different with different speakers!
Andrew
I am mostly on Dave's side. In most cases they are needed simply to make the xover network behave into the complex load, to give the intended target response, ie the target electro-acoustic response divided by the driver response. Without a zobel, a second order low pass will not always give you the desired second order response, and most certainly with low pass filters on three-way vented systems, or especially on couple cavity bass loading, will never give the desired responses without some form of zobel.
However, the perfect zobel will never be derived from the Thiele-Small parameters, but must be calculated from an accurate equivalent circuit derived from a measurement of the impedance curve of the driver/box.
Even after doing this, once you optimize the network to meet your target xover response, you can nearly always get better results by letting the zobel network be re-optimized away from the perfect values, in essence they become additional reactive elements giving greater degrees of freedom to the optimizer to reach the intended goal.
Now, as to the idea of the zobel "damping " out the driver, then this is rarely required. The amplifier provides this function over nearly the whole frequency range, even when there is a network in the way, at least for even order networks. If you look at the source impedance looking back through the network towards the amplifier, all second order networks are low impedance at both low and high frequencies, and only have a high source impedance at the cutoff frequency. This frequency should not be at the resonance frequency of the driver, so therefore damping where it is needed is generally maintained.
This is not of course the case with an odd order network due to the last component in the network being a series reactive element. This is why a third order high pass on
the tweeter will always interact with the tweeter impedance at its resonant frequency and give you an odd looking response. This is where a zobel might be needed.
However, Zobel or not, the reasons for any change in sound is purely contained in the subsequent changes in frequency response, and not due to some special feature associated with zobel networks in isolation.
The idea of trying to design a network from values calculated from theoretical classic shapes, and hence needing to flatten the impedance curve of the driver to represent a resistive load is both quaint and very backward in so many ways.
Now, with regard to flattening of the impedance presented to the amplifier, this of course can have benefits to the amplifier if the speaker otherwise would present a highly reactive load to the amp. If the real part of the impedance is low at the same time as the imaginary component is high, there will be high peak current demands placed on the amplifier, and depending on the amplifier design this can affect the sound quality of the amplifier. There are many speakers on the market that suffer from this, and mostly due to poor network design.
As to the benefits of further flattening the impedance curve to be almost purely resistive, as we did with the 104.2, this has two potential benefits:
1/ because the current is always in phase with the voltage, amplifier dissipation can be dramatically lower, with lower instantaneous dissipation in the output devices. Also there are less demands on the power supply, further benefitting the amplifier sound.
2/ a resistive impedance for the speaker eliminates any interaction with the amplifier source impedance which would otherwise alter the frequency response. This is especially crucial for most tube amplifiers, which have typically quite high source impedances that can result in response modification of the order of several dB. No wonder tubes sound so different with different speakers!
Andrew
Hi, Andrej. Thanks for you input. I agree on most of it.
On point 1, i don't fully agree. First, The serial resistance of the serial coil are high, compared to the amplifier one. second, considering the high impedance of the load at the two picks of impedance around the resonance in bass reflex, there is nothing witch can explain this ~2db of change in levels due to the Motional impedance compensation where we don't see the two peaks and the resonator's charge.
This bring-me to the point 2. We made, of course comparisons of motional Vs non motional, with equalized responses curves of the non compensated version, to fit exactly the compensated one.
The difference was obvious. It is not the frequency response witch give this effect of "better defined" bass reproduction when compensated.
I'm curious about the exact experiences you made in real life with motional compensation for bass drivers. and what brought-you to your conclusions. I had no time to make a full investigation to figure out all the effects in action (and we had no waterfall tools at this time). But i bet there is something in action about damping.
I'm inclined to be thinking it is how the amp deals with it.
I once considered bass damping when I used valve amps more often as it was critical with a high Zo and limited OPT inductance and limited power out. Even using simple amplifier circuitry I was finding that it didn't matter much whether I used a conjugate or simply adjusted the feedback ratio when it was within range, the main result was the desired response.
With regard to the need for compensation between a driver and the filters there is rarely a case where the layout can't be reduced to something more mundane looking. Technically it can be the same thing.
Old News....
2/ Agreed 100%, but not just tube amps. IME any system benefits by clean resistive load....less NFB correction required benefits sound all round.
Dan.
1/ Yes, my experience also, and I stated that too. Also amps typically go harder/louder without/before getting nasty...win/win....clipping becomes relatively benign.
2/ Agreed 100%, but not just tube amps. IME any system benefits by clean resistive load....less NFB correction required benefits sound all round.
Dan.
Christophe,
I am still reading the entire thread and trying to get past the objections so I can see how you are measuring and picking values. I will have to use a translator for the French page you put up but can see what some of the sections are saying even without speaking the language. I have had this same discussion with a very good friend of mine who is in complete agreement with you and does actively design speakers for professional applications as of today. I am interested in the inverse conjugate circuit design correcting the impedance rise and also the minimization of the back emf output of the voice-coil. How I understand this from my long conversations about this with my friend this circuit should be intimately mounted as close to the terminals as possible, no cabling between it and the driver. I plan on using this circuit with an active crossover so I am not worried about doing that section passively. This will be a two way device with two amplifiers and the active xo before the amp. We have had long discussion of the improvement in the action of the amplifier and the improvement in the feedback loops function due to the elimination of the reactive component outside of the feedback loop. I will continue to read the rest of the thread and catch up to where you are today. Thank you for continuing in the face of the rejection of the concept I have read so far. My friend has a complete B&K laboratory and fft measurement system and he does has measured response that shows this method does work. I am sorry that I know he would not want to share his personal work as it is ongoing and he has a financial interest in what he is doing.
ps One of the important factors that my friend has talked about with the inverse conjugate network is that it is very important to match the "Q" of this network with the "Q" of the device.
Steven
I am still reading the entire thread and trying to get past the objections so I can see how you are measuring and picking values. I will have to use a translator for the French page you put up but can see what some of the sections are saying even without speaking the language. I have had this same discussion with a very good friend of mine who is in complete agreement with you and does actively design speakers for professional applications as of today. I am interested in the inverse conjugate circuit design correcting the impedance rise and also the minimization of the back emf output of the voice-coil. How I understand this from my long conversations about this with my friend this circuit should be intimately mounted as close to the terminals as possible, no cabling between it and the driver. I plan on using this circuit with an active crossover so I am not worried about doing that section passively. This will be a two way device with two amplifiers and the active xo before the amp. We have had long discussion of the improvement in the action of the amplifier and the improvement in the feedback loops function due to the elimination of the reactive component outside of the feedback loop. I will continue to read the rest of the thread and catch up to where you are today. Thank you for continuing in the face of the rejection of the concept I have read so far. My friend has a complete B&K laboratory and fft measurement system and he does has measured response that shows this method does work. I am sorry that I know he would not want to share his personal work as it is ongoing and he has a financial interest in what he is doing.
ps One of the important factors that my friend has talked about with the inverse conjugate network is that it is very important to match the "Q" of this network with the "Q" of the device.
Steven
Last edited:
- Status
- Not open for further replies.