John K, many years ago i made an experiment. I took a midrange and connected it directly with an amplifier. I then put a cap in front of the amplifier to make a first order xover. In a second experiment i put a cap after the amplifier and used an LCR filter parallel to the midrange. Tuned to the same acoustic frequency response the resulting Step response looked the same so i concluded that both versions are equivalent in the time domain. I can do that again now and see what happens. I will now build the midrange into the sphere but i decided to use a smaller 3 liter variety. Then i will build the woofer section. I hope that i have something playing soon.
I can not 100% understand that question. Do you want to make a step response from the incomming ( amplifier ) signal and compare to the outcomming signal ( acoustic response ) ? Before we had the CSD, Cepstrum, Wavelet we measured the step response with an analog generator and scope. Yes, i think the step response is still a very good way to visualise the time domain behaviour. It is not easy though to interpret small changes. What we found most interesting is how steep the first pulse is and how it overshoots. Another interesting observation is how the step response goes "through zero" because you can see the behaviour of the cabinet there. The sound hits the cabinet walls and comes back through the cone. We know that since the late 70th but i swear Migeo has a better explanation.
If you achieve the same acoustic transfer function regardless of where you placed the cap then I would basically expect to see the same behavior. But that is not what you said. Your comment was that the LRC network made the amplifier's damping unnecessary. The damping arises dues the the current generated in the loop around the driver by the back EMF produced but the driver's motor. If the impedance of the loop, including the VC impedance, is Z then the damping force is -BL x (Vb/Z) where Vb is the back EMF. If a driver is connected directly to an amplifier then adding an LRC network across the driver will have negligible influence on the loop Z unless the amplifiers output Z is on the order of the voice coil DC resistance.
The LRC network you used was important when you had the cap after the amplifier because if flattened the impedance seen by the cap to look more like a pure resistance and thus yield a 1st order HP response. Without the LRC the cap would resonate with the driver impedance peak and you would have had a very different looking response. But, when you moved the cap to before the amp you did two things: you significantly changed the loop impedance seen by the driver, thus the damping of the amplifier/driver system, and then you altered (dare I say equalized) the input to the amp such that the output of the midrange driver remained the same. The same basic response would be expected in such a case, given a predominantly linear system. But you should have added one more leg to the experiment; with the cap before the amp remove the LRC network. If the magnitude of the output impedance of the amplifier was much less than the resistance in the LRC shunt you would have observed the same result demonstrating that when a driver is directly connected to an amplifier such an LRC network has no affect on damping because it dozen't affect the loop Z seen by the driver, unless the amp's output Z is high.
But the experiment you performed was reasonable. It goes to dispel the often quoted advantage sighted by active crossover proponents that direct connection of a driver to the amp means the amp has better control over the driver. (I actually have a very different take on that statement which I may introduce tot he discussion at some point.) That is not the case as if the voltage across the driver's terminals is the same then the driver will behave the same. The problem is that the voltage must be the same. A driver's impedance is subject to variation from things like voice coil heating and variation of Le with excursion. The voltage delivered across a driver's terminals by a passive crossover is typically much more sensitive to such changes in driver Z than a driver connected directly to the amplifier with crossover before the amp (be it active or line level passive). A number of years ago I performed a quick study of this effect using a Vifa MG14 driver. I measured the voltage across the driver's terminals at different levels with the driver connected to a passive 2nd order low pass crossover. The results were as below:
I had initially suspected this change in response was due to VC heating and increased Re, but as revealed through simulations the effect was more a result of changes in Le with excursion. My thinking was that such behavior could explain why some passive speakers tend to sound harsh as the volume was increased. Obviously drivers which do not suffer from such Le modulation will not exhibit this effect.
ANyway, I understand what you were trying to get across, but you did the wrong experiment to show the effect of an LRC shunt on damping.
What I'm saying, back in those days, the amplitude of ringing, for example -20db from the main input, might just have been ignored because it was considered not of significance. Especially the low end of the used spectrum of the driver.
Actually, Joachim is quite correct and you seem to be out in left field. Joachim may not have explained himself in the utmost of detail so that any layman could understand, but he was still on the mark. The "damping" the amplifier was doing was essentially in amplifying only the energy in the frequency band dictated by the cap ahead of it. If the cap is placed after the amp, it is not capable of effectively absorbing/dissipating the energy stored in the driver near resonance that the amplifier is producing. The LCR trap, as a frequency selective current divider with a transfer function that is the inverse of the driver's resonance curve, effectively reduces stored energy in the driver by drawing current limited by the cap away from the driver in the band of resonance. Very simply, without the LCR trap, the single cap placed after the amp is not capable of dissipating the energy produced by the amp that is being stored in the driver at resonance. The LCR trap facilitates that dissipation where you effectively create a current source and divert the current away from the driver. This is all really basic stuff, no? 
As for damping factor in general, passive networks consisting of loudspeaker drivers, coils, and caps are energy storage devices. Low output impedance amplifiers are the opposite - acting like energy drains. Again, this is very basic stuff and lack of damping exhibited by the cap without an LCR trap is Exhibit A evidence of the energy storage aspect of passive networks. Are you an electrical engineer?
As for damping factor in general, passive networks consisting of loudspeaker drivers, coils, and caps are energy storage devices. Low output impedance amplifiers are the opposite - acting like energy drains. Again, this is very basic stuff and lack of damping exhibited by the cap without an LCR trap is Exhibit A evidence of the energy storage aspect of passive networks. Are you an electrical engineer?
Still fntn, i find John K.s experiment intesting and i did not say that both varieties sound absolutely identical but when you whould only see the step response in the linear range it whould be hard to tell if it is an active or passive system. Yes, when the physical parameters of the drivers are changing over level a direct connection ( active speaker ) whould be better. I whould then maybe go to current drive to take advantage of that.
The intesting thing is that in the absolute top High End ( except Professional Monitors ) not a single active system has a reputation. Speakers from Wilson, Magico, Dynaudio, just to mention a few are all passive. Why ? Did we come so far with passive that the result is better or do the customers not trust the internal technology ? I have some thoughts about that but it is hard to put it in a nutshell.
The intesting thing is that in the absolute top High End ( except Professional Monitors ) not a single active system has a reputation. Speakers from Wilson, Magico, Dynaudio, just to mention a few are all passive. Why ? Did we come so far with passive that the result is better or do the customers not trust the internal technology ? I have some thoughts about that but it is hard to put it in a nutshell.
Yes, it's pretty obvious that passive networks are more substantively impacted by slight changes in voice coil impedance. That's nothing new or shocking. The real value in active speaker systems is the isolation (damping) provided by the amplifier between the area of high energy (output side of amp) and the area of low energy (input side where signal shaping takes place).
As for Magico, Wilson, Dynaudio and so forth, I don't think the exclusive use of passive components is performance based but rather a conscious decision to fulfill the demands by the target audience for simplicity. Frankly, the highest performing loudspeakers we've ever produced were active - no contest. And in some ways, with active subs, a number of "high end" manufacturers have already been going the active route. Speaking of sub bass, frankly, I've heard the Magicos and their bass performance I find to be quite lacking for a "high end" manufacturer. That assessment has been hinted at with major magazines who've accepted advertising from Alon but have noted "softer dynamics" and "dryness" at the fundamental frequencies. Yes, at the moment it seems a lot of the high end makes are passive but I don't expect it to stay that way.
As for Magico, Wilson, Dynaudio and so forth, I don't think the exclusive use of passive components is performance based but rather a conscious decision to fulfill the demands by the target audience for simplicity. Frankly, the highest performing loudspeakers we've ever produced were active - no contest. And in some ways, with active subs, a number of "high end" manufacturers have already been going the active route. Speaking of sub bass, frankly, I've heard the Magicos and their bass performance I find to be quite lacking for a "high end" manufacturer. That assessment has been hinted at with major magazines who've accepted advertising from Alon but have noted "softer dynamics" and "dryness" at the fundamental frequencies. Yes, at the moment it seems a lot of the high end makes are passive but I don't expect it to stay that way.
As I said back on page 50, flattening FR does reduce distortion as non-flat frequency response is distortion, you should know that. Unless, of course, you think that non-flat FR is not distortion in and of itself. One does not (at least should not) need elaborate distortion measurements to recognize this aspect of distortion.
You said "...in my experience, distortion rises dramatically when applying "EQ" in the form of even a modest 4-5db high Q boost or cut to an actively driven loudspeaker". John demonstrated measurements shortly thereafter that debunks your claim only to have you make assumptions on the conditions to try to make your claim again by way of equivocation. John had explicitly stated "Eq is a linear function and can not introduce distortion of its own", contrary to your claim that "distortion rises dramatically when applying "EQ" in the form of even a modest 4-5db high Q boost or cut to an actively driven loudspeaker". It does not just as John proved with his measurement with fundamental centered on the dip that you claimed (at the time) would inevitably add distortion. That was a 5db high Q boost (in addition to other areas of boost/cut to flatten the FR). Your claim is simply without merit. Were there any to it, there would have been at least some modicum of increased distortion. There was none.
This doesn't even address the fact that the EQ of the driver's response, nearly perfect on the measured axis and including the breakup area, shows that 1) it can be full equalized on-axis as demonstrated in the FR measurement and 2) that there was no change in harmonic distortion due to the equalization as well.
Dave
I find the complete lack of understanding of linear systems remarkable here. This stuff is all covered in a first course in dynamic systems. Look at the basic model of driver motion, valid from DC throughout the piston range:
F = MX",
F = BL x i -Rms x X' - X/Cms
I = Ua/Zd - Ub / ZL
Ub = BL x X',
where X" = acceleration, X' = velocity, X = position, I is the current flowing through the voice coil, Ua is the potential applied across the driver terminals, Ub is the back EMF generated by the driver and is equal to BL x X',
Zd is the driver VC impedance, Re + j Le, and ZL is the loop impedance seen by the driver looking back at the amplifier.
Now given that, here is what ZL looks like for, first the case where the cap is after the amp and the LRC shunt is in place. The LRC shunt has a resistance of 5 ohms at resonance.
Then what ZL would look like when the cap is moved to before the amp,
And last, what ZL would look like if the the LRC shunt is removed. I assumed the amplifier's out put impedance is a constant 0.5 ohms for a nominal damping factor of 16. At the driver's resonance ZL drops from 5 ohms to 4.5454 ohms (5 in parallel with .5). With a good solid state amp the DF would be on the order of 100 so the amp's output Z would be more like 0.08 ohms and at resonance ZL would drop from 0.08 ohms to 0.078 ohms (5 in parallel with 0.08). As I said, unless the amplifier's output Z is large the LRC shunt has minimal effect on damping when the cap is before the amp.
As I said, unless the amplifier's output Z is large the LRC shunt has minimal effect on damping when the cap is before the amp. The next question should be, how can the system with cap before the amp (with or without LRC shunt) have the same impulse and step response as the system with the cap and LRC shunt after the amp when it is apparent that ZL, hence the electromotive damping of the driver, is so different? Tell you what, you guys can figure it out for yourselves. I'm not of this planet and the cosmos is calling.
Probably the reason fully active systems do not have as high a reputation is just because in a passive system there is a lot more going on with amplifier/ crossover interaction, along with fredom of choice of amplifiers. Also, recall that my plots show the voltage across the driver. But it is current flowing through the VC that is directly related to the motive force. With an active crossover the voltage will scale linearly but the driver impedance will still modulate. In the case I showed, the VC Z actually increases due to increased Le at higher levels. So, with increased Z at 2k HZ and the a linearly scaled voltage across the driver the output would actually be compressed because the current would not scale linearly being V/Z. But in the passive case the effect of increased Le makes the voltage increase super linearly and thus it tends to maintain higher current through the VC with result there is less compression and perhaps even expansion. Depending on the specific case the passive system may seem harsh or it could seem more dynamic or how knows. What it does and what it sounds like are not so simple as we know.
F = MX",
F = BL x i -Rms x X' - X/Cms
I = Ua/Zd - Ub / ZL
Ub = BL x X',
where X" = acceleration, X' = velocity, X = position, I is the current flowing through the voice coil, Ua is the potential applied across the driver terminals, Ub is the back EMF generated by the driver and is equal to BL x X',
Zd is the driver VC impedance, Re + j Le, and ZL is the loop impedance seen by the driver looking back at the amplifier.
Now given that, here is what ZL looks like for, first the case where the cap is after the amp and the LRC shunt is in place. The LRC shunt has a resistance of 5 ohms at resonance.
An externally hosted image should be here but it was not working when we last tested it.
Then what ZL would look like when the cap is moved to before the amp,
An externally hosted image should be here but it was not working when we last tested it.
And last, what ZL would look like if the the LRC shunt is removed. I assumed the amplifier's out put impedance is a constant 0.5 ohms for a nominal damping factor of 16. At the driver's resonance ZL drops from 5 ohms to 4.5454 ohms (5 in parallel with .5). With a good solid state amp the DF would be on the order of 100 so the amp's output Z would be more like 0.08 ohms and at resonance ZL would drop from 0.08 ohms to 0.078 ohms (5 in parallel with 0.08). As I said, unless the amplifier's output Z is large the LRC shunt has minimal effect on damping when the cap is before the amp.
An externally hosted image should be here but it was not working when we last tested it.
As I said, unless the amplifier's output Z is large the LRC shunt has minimal effect on damping when the cap is before the amp. The next question should be, how can the system with cap before the amp (with or without LRC shunt) have the same impulse and step response as the system with the cap and LRC shunt after the amp when it is apparent that ZL, hence the electromotive damping of the driver, is so different? Tell you what, you guys can figure it out for yourselves. I'm not of this planet and the cosmos is calling.
Probably the reason fully active systems do not have as high a reputation is just because in a passive system there is a lot more going on with amplifier/ crossover interaction, along with fredom of choice of amplifiers. Also, recall that my plots show the voltage across the driver. But it is current flowing through the VC that is directly related to the motive force. With an active crossover the voltage will scale linearly but the driver impedance will still modulate. In the case I showed, the VC Z actually increases due to increased Le at higher levels. So, with increased Z at 2k HZ and the a linearly scaled voltage across the driver the output would actually be compressed because the current would not scale linearly being V/Z. But in the passive case the effect of increased Le makes the voltage increase super linearly and thus it tends to maintain higher current through the VC with result there is less compression and perhaps even expansion. Depending on the specific case the passive system may seem harsh or it could seem more dynamic or how knows. What it does and what it sounds like are not so simple as we know.
Linkwitz stated clearly that the application of EQ to amplitude response anomalies is effective only when non linear distortion is not present. I didn't say that in all instances of application of EQ that distortion rises dramatically. In my experience, I recall applying EQ and seeing distortion rising dramatically - that is essentially anecdotal not empirical. What you are claiming is empirical and it clearly disagrees with my experience and what Linkwitz was saying as I posted earlier a reference to his website. You can continue to beat this horse to death but my guess is most people reading this are going to believe Linkwitz and those who share his view over you unless you post a lot of data showing the presence of significant levels of non linear distortion accompanying an amplitude response dip which were reduced by EQ that simultaneously fixed the amplitude problem. So far, a very specific case of amplitude anomaly (again, most likely caused by something external to the driver) which was not accompanied by non linear distortion was corrected by EQ. A linear problem was fixed by a linear process. When a non linear problem results in an amplitude response anomaly and you are able to correct it via EQ - then you will have actually proven something. And I'm sure pigs will fly when that happens.
Last edited:
I will quote Linkwitz.Linkwitz stated clearly that the application of EQ to amplitude response anomalies is effectively only when non linear distortion is not present. I didn't say that in all instances of application of EQ that distortion rises dramatically. In my experience, I recall applying EQ and seeing distortion rising dramatically - that is essentially anecdotal not empirical. What you are claiming is empirical and it clearly disagrees with my experience and what Linkwitz was saying as I posted earlier a reference to his website. You can continue to beat this horse to death but my guess is most people reading this are going to believe Linkwitz and those who share his view over you unless you post a lot of data showing the presence of significant levels of non linear distortion accompanying an amplitude response dip which were reduced by EQ that simultaneously fixed the amplitude problem. So far, a very specific case of amplitude anomaly (again, most likely caused by something external to the driver) which was not accompanied by non linear distortion was corrected by EQ. A linear problem was fixed by a linear process. When a non linear problem results in an amplitude response anomaly and you are able to correct it via EQ - then you will have actually proven something. And I'm sure pigs will fly when that happens.
No need to try to expound on this.
He follows with:
This is moot with DSP EQ. There is no difficulty nor impracticality other than cost. It is not an issue of impossibility.
The next quote from SL addresses your current misinterpretation:
He says nothing to the effect that EQ is ineffective in the presence of non-linear distortion. Quite the contrary, one should certainly investigate the non-linear distortion of any driver, but that has nothing to do with whether nor not one applies EQ to linear distortion in drivers. The same applies whether one is applying passive EQ or DSP. That's simply good design practice.
A much better debate would likely have ensued had you not very early on (Post 533, pg. 54) started with insults such as "For anyone thinking of addressing the lion's share of response anomalies with EQ who might be buying into this continuous stream of rubbish - I will present once again the counterpoint, fact checking view." Nothing we said disagrees with Linkwitz, quite the contrary. The most unfortunate part is that your "facts" have often been anything but that, while the attitude certainly did not invite congenial debate. This is just one case in point.
So far your "facts" have been nothing but anecdotes and when confronted with empirical evidence that could easily be re-created if one cared to, your response has been simply equivocation and misinterpretation. You continue that with your most recent post as if anyone here had even implied that EQ corrects non-linear distortion. You really need to read more carefully and be a bit more circumspect in your inferences and accusations. And you are once again making the same flawed interpretation of what has been said explicitly with regard to non-linear distortion. I can only assume that since this has been addressed previously, you are using it in an ingenuous attempt to support your statements.
Dave
Last edited:
fntn, in all fairness to you the argument was not about what the LRC shunt does when the cap is present, it was about what it does when the cap is not present. When the cap is not there the LRC is superfluous unless the amp output Z is large.
If you want know about my back ground go to google scholar and do a search on Kreskovsky.
If you want know about my back ground go to google scholar and do a search on Kreskovsky.
We're experimenting with transmission line loading of the Scan 4624G. It kind of resembles the exit of a jet engine. Hope to have the driver mounted and tested soon. LCRs usually require large expensive caps but the rear chamber we have for this driver so far is pretty expensive to build so the only aspect that is likely to sway the final decision one way or the other is power handling and distortion.
Do you use a software package to design loudspeaker crossovers or do you design them using standard engineering calculations? - Because I've never seen a crossover with an LCR resonance trap used alone as the signal shaping component. And "the argument" was not "about what it does when the cap is not present" - that is clearly something you invented in your head. This is precisely what Joachim said:
It seems that you are intent on having an argument - in this case - with yourself since no one else is talking about shaping the response of a loudspeaker with an LCR trap alone.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.