Peter posted:
The low pass filter will of course deal with direct excitation of the breakups. Of course - that's quite elementary. Don't allow signals that will stimulate the breakup modes.
However, you need the high pass to deal with excitation by nonlinearities of the driver. Unless you choose to run it at low excursion (SPL) levels. The ONLY way to eliminate THD/IMD from BL/Cms/Le nonlinearities for a given driver is via a high pass filter which reduces excursion.
Nonlinearities of the driver will create their own harmonics all over the frequency range (well outside the passband of the crossover attached to the driver) which will still excite the modes. The only way to eliminate this is to eliminate the nonlinearities - any causal filter (DSP or analog) will not work.
Your other option is as you suggest - turning the volume down. Of course, that may require use of multiple drivers to reach a desired SPL level. Use of multiple drivers covering the same frequency range introduces its own set of problems to the system as a whole...
Dan Wiggins
Adire Audio
The low pass filter will of course deal with direct excitation of the breakups. Of course - that's quite elementary. Don't allow signals that will stimulate the breakup modes.
However, you need the high pass to deal with excitation by nonlinearities of the driver. Unless you choose to run it at low excursion (SPL) levels. The ONLY way to eliminate THD/IMD from BL/Cms/Le nonlinearities for a given driver is via a high pass filter which reduces excursion.
Nonlinearities of the driver will create their own harmonics all over the frequency range (well outside the passband of the crossover attached to the driver) which will still excite the modes. The only way to eliminate this is to eliminate the nonlinearities - any causal filter (DSP or analog) will not work.
Your other option is as you suggest - turning the volume down. Of course, that may require use of multiple drivers to reach a desired SPL level. Use of multiple drivers covering the same frequency range introduces its own set of problems to the system as a whole...
Dan Wiggins
Adire Audio
Hi Rick
This is based on experience from one of the most difficult situations a sound engineer can face, live accoustic folk music. I used to run the PA at large English folk festival in the main dance tent. For the evening dance sessions we would have a marquee with over 1000 people, and a 20K plus sound system. A mixture of unamplified instruments, drum kits, amplified instruments and vocals is difficult at the best of times, but then you need to get the caller, (who calls the dance moves) heard right to the back of the tent, is not the easiest job in the world!
This is based on experience from one of the most difficult situations a sound engineer can face, live accoustic folk music. I used to run the PA at large English folk festival in the main dance tent. For the evening dance sessions we would have a marquee with over 1000 people, and a 20K plus sound system. A mixture of unamplified instruments, drum kits, amplified instruments and vocals is difficult at the best of times, but then you need to get the caller, (who calls the dance moves) heard right to the back of the tent, is not the easiest job in the world!
Hennie,Hennie said:
This is actually an EXCELLENT example driver for what I'm talking about...
First off, though, you are correct in that this driver will have a LOT lower THD at 105 dB SPL than a 100, 133, 160mm Seas! It has 2-4 times the surface area, and thus requires a lot less stroke to reach a given SPL level, meaning it will operate in its linear range more, and will also tend to have less stress on the cone, surround, spider as well (less breakup issues).
Of course, a 250mm driver for a midrange is a bit extreme for most people, in that dispersion above 1 kHz or so is going to narrow, making it a speaker that will sound very good at one seat, but may not be good for sharing audio with a friend...
Anyway, about the THD plots. If THD was a linear effect, then the THD level below the SPL level would maintain a constant distance. HOWEVER, we see that at the breakup modes the THD actually INCREASES!
Take, for instance, 400 Hz (there's a peak in THD at that frequency). THD is ~18 dB down on the graph (38 dB down in real life; the THD is offset 20 dB). Now look at peak at ~3 kHz - it's 15 dB down. At 4.6 kHz it's 12 dB down.
THD will increase at breakup modes, because it's not just the raw signal that is exciting them but also nonlinearities of the driver! So you're not just getting a stimulus of V from your amp, you're getting V + d where d is the contribution of the nonlinearities.
Remember, the dominant nonlinearities of a moving driver are the BL change with excursion, the Cms change with excursion, and the Le change with excursion. They are excursion based, rather than frequency based.
So let's say you eliminate all the THD from the V contribution. And let's say that was all the THD at 400 Hz. You now have ZERO THD at 400 Hz. You will still have THD at 3 and 4.6 kHz because of the d component. And there are only three ways to eliminate that:
- Non-causal filtering (DSP IIRs with internal delays)
- Linearizing the driver (eliminating the source of the nonlinearities)
- Limiting the excursion via a high pass filter.
This is why you can filter and notch and EQ the heck out of a driver with high Q resonances in the cone, and still hear those peaks when you start to drive the speaker harder. Your EQ and filtering cannot solve the problem. Your only solution is to limit the excursion.
This is also why softer cones - which tend to have lower amplitude peaks - do not gain "harshness" as quickly as stiff cones as you push them.
Now, positive peak resonances in cones tend to amplify the contribution of d. Dip resonances will amplify them, too, but as pretty much everyone knows it's sonically much better to have a dip than a peak!
Dan Wiggins
Adire Audio
Dan,
I know I don´t write all that clear but you seem to ingore or refuse to think about what I write.
"The low pass filter will of course deal with direct excitation of the breakups. Of course - that's quite elementary. Don't allow signals that will stimulate the breakup modes."
I thought we where way above this level but ok.
Yes. A driver that peak at 6k, let´s lowpasss it at 2k 48dB/oct and yes, then we have avoided to drive the breakup directly as seen from the driving voltage. Yes this goes without saying.
HOWEVER (for the third time) some harmonic products from frequencies IN the pass band BELOW the LP will excite the breakup... as you know. Now, which one is most likely to excite this break up to a audible high level, the 3rd harmonic of a 2000Hz sine wave of the 8th harmonic at 750Hz? Or maybe the 25th(?) harminic at 100Hz?
Clarification; typical harmonic distortion products fall with increasing frecuency so the harmonic products from driver non linearitys that will coincide with the major breakup with the HIGHEST amplitude (and thereby excite the breakup resonance) will be the LOW harmonics from the frecuencies relatively high up in the pass band. For a driver with a 6k breakup this means the frequencies 2000Hz and 3000Hz for the 2nd and 3rd order that will equal 6000Hz and therefore excite the resonance.
Clear?
To avoid this harmonic products to excite the driver breakup we need to LOWPASS the driver below these frequencies. IOW a Seas Excel W22 with 4.5k breakup peak needs to be lowpassed BELOW 1500Hz which is the frequency that will have its 3rd order harmonics at 4500Hz. At 750Hz the 6th harmonic will be at 4.5k and potentially excite the resonance however now the level gets lower and lower so it´s not much to worry about anymore.
Puh!
"However, you need the high pass to deal with excitation by nonlinearities of the driver. Unless you choose to run it at low excursion (SPL) levels."
Yes, I have said that a couple of times. But it is just one of two important things to make a "pistonic" driver shine.
"The ONLY way to eliminate THD/IMD from BL/Cms/Le nonlinearities for a given driver is via a high pass filter which reduces excursion."
No, by LOWPASSING the driver as described above you will also reduce the harmonic content that can excite the breakup resonance.
"Nonlinearities of the driver will create their own harmonics all over the frequency range (well outside the passband of the crossover attached to the driver) which will still excite the modes. The only way to eliminate this is to eliminate the nonlinearities - any causal filter (DSP or analog) will not work."
This is what I have been talking about for a couple of posts now but I don´t know where the confusion lies on your part.
Digest!
/Peter
I know I don´t write all that clear but you seem to ingore or refuse to think about what I write.
"The low pass filter will of course deal with direct excitation of the breakups. Of course - that's quite elementary. Don't allow signals that will stimulate the breakup modes."
I thought we where way above this level but ok.
Yes. A driver that peak at 6k, let´s lowpasss it at 2k 48dB/oct and yes, then we have avoided to drive the breakup directly as seen from the driving voltage. Yes this goes without saying.
HOWEVER (for the third time) some harmonic products from frequencies IN the pass band BELOW the LP will excite the breakup... as you know. Now, which one is most likely to excite this break up to a audible high level, the 3rd harmonic of a 2000Hz sine wave of the 8th harmonic at 750Hz? Or maybe the 25th(?) harminic at 100Hz?
Clarification; typical harmonic distortion products fall with increasing frecuency so the harmonic products from driver non linearitys that will coincide with the major breakup with the HIGHEST amplitude (and thereby excite the breakup resonance) will be the LOW harmonics from the frecuencies relatively high up in the pass band. For a driver with a 6k breakup this means the frequencies 2000Hz and 3000Hz for the 2nd and 3rd order that will equal 6000Hz and therefore excite the resonance.
Clear?
To avoid this harmonic products to excite the driver breakup we need to LOWPASS the driver below these frequencies. IOW a Seas Excel W22 with 4.5k breakup peak needs to be lowpassed BELOW 1500Hz which is the frequency that will have its 3rd order harmonics at 4500Hz. At 750Hz the 6th harmonic will be at 4.5k and potentially excite the resonance however now the level gets lower and lower so it´s not much to worry about anymore.
Puh!
"However, you need the high pass to deal with excitation by nonlinearities of the driver. Unless you choose to run it at low excursion (SPL) levels."
Yes, I have said that a couple of times. But it is just one of two important things to make a "pistonic" driver shine.
"The ONLY way to eliminate THD/IMD from BL/Cms/Le nonlinearities for a given driver is via a high pass filter which reduces excursion."
No, by LOWPASSING the driver as described above you will also reduce the harmonic content that can excite the breakup resonance.
"Nonlinearities of the driver will create their own harmonics all over the frequency range (well outside the passband of the crossover attached to the driver) which will still excite the modes. The only way to eliminate this is to eliminate the nonlinearities - any causal filter (DSP or analog) will not work."
This is what I have been talking about for a couple of posts now but I don´t know where the confusion lies on your part.
Digest!
/Peter
OK, Dan thanks for your response. I can indeed confirm that what you say ties up with my own listening experience. This has been bugging me for a while. For example I had a Sonus Faber with a carbon reinforced and doped paper cone. With the crossover in place it measures beautifully smooth. However the raw driver response has a pronounced peak of about 10 dB. At low volumes it sounds like it measures with the crossover in place - beautifully smooth. But at moderately high volumes it becomes a bit strained. (It is a two way with quite a low f-3db and therefore it reaches high excursions rather quickly.)
But the same argument would also apply if a driver has a strongly rising response without necessarily having high Q resonances? Like some drivers with either a saturated gap or some inductance cancelling scheme? You can EQ it flat and limit its pass band, but it does nothing for those harmonics which have a fundamantal in the pass band. It is amplified by the rising response even if it does not have high Q resonances...
Another example of a rising response due to some inductance reduction method (AIC - active impedance control)
http://www.eighteensound.it/pdf/10NDA520.pdf
The reduction in inductance probably reduces Le non-linearities, but the rising response amplifies harmonic distortion? Is there a nett reduction in the end?
But the same argument would also apply if a driver has a strongly rising response without necessarily having high Q resonances? Like some drivers with either a saturated gap or some inductance cancelling scheme? You can EQ it flat and limit its pass band, but it does nothing for those harmonics which have a fundamantal in the pass band. It is amplified by the rising response even if it does not have high Q resonances...
Another example of a rising response due to some inductance reduction method (AIC - active impedance control)
http://www.eighteensound.it/pdf/10NDA520.pdf
The reduction in inductance probably reduces Le non-linearities, but the rising response amplifies harmonic distortion? Is there a nett reduction in the end?
Dan,
"First off, though, you are correct in that this driver will have a LOT lower THD at 105 dB SPL than a 100, 133, 160mm Seas! It has 2-4 times the surface area, and thus requires a lot less stroke to reach a given SPL level, meaning it will operate in its linear range more,"
Yes and my subwoofer has much less dist at 100Hz than my tweeter...
A 10" drivers should have about a little less than twice the Sd of a 8". So I would be ok with this comparison if two W22 were compared to this JBL 10". It´s not hard to reduce non-linear distortion by going towards greater surface area, but what about dispersion problems which you adressed in the last post to me reagarding using for example two 8"?
"and will also tend to have less stress on the cone, surround, spider as well (less breakup issues)."
Are you saying that cone breakup and energy storage is a non linear phenomena? SL look at it as a linear distortion and I´m somewhere in the between I´d say. Do you have any studies or papers that deals with this or do you assume it is like you describe?
"Of course, a 250mm driver for a midrange is a bit extreme for most people, in that dispersion above 1 kHz or so is going to narrow, making it a speaker that will sound very good at one seat, but may not be good for sharing audio with a friend..."
Exactly which is why you could use two 8"ers whitout any penalty instead... well other than cost that is but let´s leave that beside for now since we are talking about physics and not economy.
"Anyway, about the THD plots. If THD was a linear effect, then the THD level below the SPL level would maintain a constant distance. HOWEVER, we see that at the breakup modes the THD actually INCREASES!"
Yes, cone adds distortion the more it moves/flexes (so choose a rigid one ) and also the inductive induced distortion is a bigger factor the higher the frequency. No news for you I guess..
"Remember, the dominant nonlinearities of a moving driver are the BL change with excursion, the Cms change with excursion, and the Le change with excursion. They are excursion based, rather than frequency based."
Distortion from modulation of the inductance is frequency based. It´s less of an issue in low frequencies since the impedance is more resisitive there.
"So let's say you eliminate all the THD from the V contribution. And let's say that was all the THD at 400 Hz. You now have ZERO THD at 400 Hz. You will still have THD at 3 and 4.6 kHz because of the d component."
Zero distortion in the voltage drive doesn´t equal zero distortion in the driver. The non linearitys at 400Hz will induce some partials to the output at 800Hz, 1200Hz, 1600Hz, 2000Hz, 2400Hz, 2800Hz, 3200Hz, 3600Hz, 4000Hz, 4400Hz, 4800Hz.. and so on but no at 3 and 4.6kHz.
With multitone signals though, we will have non harmonic products due to intermodulation distortion and possibly some products would be at your proposed 3 and 4.6k.
"This is why you can filter and notch and EQ the heck out of a driver with high Q resonances in the cone, and still hear those peaks when you start to drive the speaker harder. Your EQ and filtering cannot solve the problem. Your only solution is to limit the excursion."
At the point when this happens the driver distorts so much overall so it is a minor problem.
"This is also why softer cones - which tend to have lower amplitude peaks - do not gain "harshness" as quickly as stiff cones as you push them."
Is it so? I´ve heard many soft coned system sound much more strained than sfitt cones. I would say that a stiff cones system done right actually is very good in just this regard that it can play strong but still be clean sounding.
It´s all about the implementation and as I said earlier IMO the small portion of distortion that will excite the breakup in a well implemented stiff driver will be nothing compared to the distortion that is induced of a softer cone and all it´s distortion products while playing multitone signals.
BTW, I listened to Ray Kimbers latest Isomike demo CD on my Sony DVD and Philips Matchline Widescreen tonight, damn fine sound
/Peter
"First off, though, you are correct in that this driver will have a LOT lower THD at 105 dB SPL than a 100, 133, 160mm Seas! It has 2-4 times the surface area, and thus requires a lot less stroke to reach a given SPL level, meaning it will operate in its linear range more,"
Yes and my subwoofer has much less dist at 100Hz than my tweeter...
A 10" drivers should have about a little less than twice the Sd of a 8". So I would be ok with this comparison if two W22 were compared to this JBL 10". It´s not hard to reduce non-linear distortion by going towards greater surface area, but what about dispersion problems which you adressed in the last post to me reagarding using for example two 8"?
"and will also tend to have less stress on the cone, surround, spider as well (less breakup issues)."
Are you saying that cone breakup and energy storage is a non linear phenomena? SL look at it as a linear distortion and I´m somewhere in the between I´d say. Do you have any studies or papers that deals with this or do you assume it is like you describe?
"Of course, a 250mm driver for a midrange is a bit extreme for most people, in that dispersion above 1 kHz or so is going to narrow, making it a speaker that will sound very good at one seat, but may not be good for sharing audio with a friend..."
Exactly which is why you could use two 8"ers whitout any penalty instead... well other than cost that is but let´s leave that beside for now since we are talking about physics and not economy.
"Anyway, about the THD plots. If THD was a linear effect, then the THD level below the SPL level would maintain a constant distance. HOWEVER, we see that at the breakup modes the THD actually INCREASES!"
Yes, cone adds distortion the more it moves/flexes (so choose a rigid one
) and also the inductive induced distortion is a bigger factor the higher the frequency. No news for you I guess.."Remember, the dominant nonlinearities of a moving driver are the BL change with excursion, the Cms change with excursion, and the Le change with excursion. They are excursion based, rather than frequency based."
Distortion from modulation of the inductance is frequency based. It´s less of an issue in low frequencies since the impedance is more resisitive there.
"So let's say you eliminate all the THD from the V contribution. And let's say that was all the THD at 400 Hz. You now have ZERO THD at 400 Hz. You will still have THD at 3 and 4.6 kHz because of the d component."
Zero distortion in the voltage drive doesn´t equal zero distortion in the driver. The non linearitys at 400Hz will induce some partials to the output at 800Hz, 1200Hz, 1600Hz, 2000Hz, 2400Hz, 2800Hz, 3200Hz, 3600Hz, 4000Hz, 4400Hz, 4800Hz.. and so on but no at 3 and 4.6kHz.
With multitone signals though, we will have non harmonic products due to intermodulation distortion and possibly some products would be at your proposed 3 and 4.6k.
"This is why you can filter and notch and EQ the heck out of a driver with high Q resonances in the cone, and still hear those peaks when you start to drive the speaker harder. Your EQ and filtering cannot solve the problem. Your only solution is to limit the excursion."
At the point when this happens the driver distorts so much overall so it is a minor problem.
"This is also why softer cones - which tend to have lower amplitude peaks - do not gain "harshness" as quickly as stiff cones as you push them."
Is it so? I´ve heard many soft coned system sound much more strained than sfitt cones. I would say that a stiff cones system done right actually is very good in just this regard that it can play strong but still be clean sounding.
It´s all about the implementation and as I said earlier IMO the small portion of distortion that will excite the breakup in a well implemented stiff driver will be nothing compared to the distortion that is induced of a softer cone and all it´s distortion products while playing multitone signals.
BTW, I listened to Ray Kimbers latest Isomike demo CD on my Sony DVD and Philips Matchline Widescreen tonight, damn fine sound
/Peter
Hennie said:
Hennie,
yes it will... at least to some degree. That is at least all non linarities that act on the voice coil former will also push at the cone and result in higher output than if a similar driver would have a flat response. At least Bl, spider Cms, Le will affect this. Since we are now talking about a lossy flexing cone it will be partly decoupled from surround and therefore the surround may have little significance in this regard.
/Peter
Peter,
Say you have a driver with a rated Xmax of 10mm (that is, the BL is 30% down at 10mm one way). From a distortion standpoint related to BL nonlinearities, which is worse:
1mm at 500 Hz
8mm at 50 Hz
The second - 10mm at 50 Hz - is MUCH worse. Using a low pass filter at 500 Hz will NOT HELP AT ALL. Distortion products from nonlinearities in the BL, Cms, and Le are strictly functions of excursion, NOT of frequency! And in fact, for flat SPL, we know that excursion must INCREASE by a factor of 4 for each octave you drop. Hence to eliminate the contributions of excursion-based distortion you MUST eliminate excursion. Which can only be done by high-passing the filter.
Say you are reproducing 100 dB SPL broadband (pink noise, average musical signal, etc). Which reduces more excursion for that driver: a high pass filter at 100 Hz or a low pass filter at 500 Hz?
Do you now see my point? Low-passing the signal will NOT eliminate distortions related to excursion, unless you low pass so low as to significantly reduce excursion (say a 20 Hz low pass filter). Nonlinearities in the BL, Cms, and inductance versus stroke are strictly functions of excursion, they are not related to a frequency.
And the interesting thing is that nonlinearities of the BL, Cms, and Le at 100 Hz can generate high order, high amplitude distortion components at much higher frequencies. A decade or more higher. So if you want to use a low pass filter to solve these problems take the first cone mode and divide by 10, at a bare minimum (for example, a W18EX001 with a cone mode at 5 kHz would need to be low passed hard at 500 Hz to reduce - not eliminate - ringing induced by excursion-based nonlinearities). Better to simply lop-off the low end with a high pass filter in the first place.
Here is a plot of the harmonics of a VERY good underhung driver, full copper sleeve on the pole, copper ring in the motor, copper ring on top of pole:
http://www.adireaudio.com/Files/Dan/UH_spectra.gif
Driver is running at ~67% of rated Xmax (6mm one way out of 9mm - well within the "linear" rating of the driver).
Here's the same soft parts, but with a motor that has ~30% more rated Xmax, running at the same percentage of excursion (XBL^2 motor with an equivalently flat motor, 7.1mm motion out of 11mm rated Xmax, very flat BL curve), and with NO copper:
http://www.adireaudio.com/Files/Dan/XBL_spectra.gif
This is at an actual higher SPL level - remember, same soft parts, same cone diameter, but more stroke. Yet the distortion products are LOWER! It's all because the BL curve over which the driver is operating is flatter. And that is dictated by the 70 Hz component, NOT the 735 Hz component. So the BL nonlinearities of the 70 Hz component account for big differences above 1400 Hz - note even that the 2.2 kHz component is radically different (a full 5 octaves above the source of the distortions).
Hennie,
Yes, I've heard that strain many times! We try to avoid it at all costs - it is what robs a speaker of its dynamics. You end up running at lower volume to keep the peak levels "in control", which can hurt the overall musical experience.
Rising responses aren't usually as much a problem because it usually comes from of two things:
- Overly damped motor structure (low Q driver)
- Directivity
Both of these are completely linear effects, and can therefore be dealt with EQ. Of course, with passive solutions, your only choice is to "cut" the rise, not "bump" the loss! That means that a driver that may be 95 dB efficient ends up with an effective system efficiency of 85-88 dB efficiency. But it doesn't mean the same kinds of distortions.
Dan Wiggins
Adire Audio
Say you have a driver with a rated Xmax of 10mm (that is, the BL is 30% down at 10mm one way). From a distortion standpoint related to BL nonlinearities, which is worse:
1mm at 500 Hz
8mm at 50 Hz
The second - 10mm at 50 Hz - is MUCH worse. Using a low pass filter at 500 Hz will NOT HELP AT ALL. Distortion products from nonlinearities in the BL, Cms, and Le are strictly functions of excursion, NOT of frequency! And in fact, for flat SPL, we know that excursion must INCREASE by a factor of 4 for each octave you drop. Hence to eliminate the contributions of excursion-based distortion you MUST eliminate excursion. Which can only be done by high-passing the filter.
Say you are reproducing 100 dB SPL broadband (pink noise, average musical signal, etc). Which reduces more excursion for that driver: a high pass filter at 100 Hz or a low pass filter at 500 Hz?
Do you now see my point? Low-passing the signal will NOT eliminate distortions related to excursion, unless you low pass so low as to significantly reduce excursion (say a 20 Hz low pass filter). Nonlinearities in the BL, Cms, and inductance versus stroke are strictly functions of excursion, they are not related to a frequency.
And the interesting thing is that nonlinearities of the BL, Cms, and Le at 100 Hz can generate high order, high amplitude distortion components at much higher frequencies. A decade or more higher. So if you want to use a low pass filter to solve these problems take the first cone mode and divide by 10, at a bare minimum (for example, a W18EX001 with a cone mode at 5 kHz would need to be low passed hard at 500 Hz to reduce - not eliminate - ringing induced by excursion-based nonlinearities). Better to simply lop-off the low end with a high pass filter in the first place.
Here is a plot of the harmonics of a VERY good underhung driver, full copper sleeve on the pole, copper ring in the motor, copper ring on top of pole:
http://www.adireaudio.com/Files/Dan/UH_spectra.gif
Driver is running at ~67% of rated Xmax (6mm one way out of 9mm - well within the "linear" rating of the driver).
Here's the same soft parts, but with a motor that has ~30% more rated Xmax, running at the same percentage of excursion (XBL^2 motor with an equivalently flat motor, 7.1mm motion out of 11mm rated Xmax, very flat BL curve), and with NO copper:
http://www.adireaudio.com/Files/Dan/XBL_spectra.gif
This is at an actual higher SPL level - remember, same soft parts, same cone diameter, but more stroke. Yet the distortion products are LOWER! It's all because the BL curve over which the driver is operating is flatter. And that is dictated by the 70 Hz component, NOT the 735 Hz component. So the BL nonlinearities of the 70 Hz component account for big differences above 1400 Hz - note even that the 2.2 kHz component is radically different (a full 5 octaves above the source of the distortions).
Hennie,
Yes, I've heard that strain many times!
We try to avoid it at all costs - it is what robs a speaker of its dynamics. You end up running at lower volume to keep the peak levels "in control", which can hurt the overall musical experience.Rising responses aren't usually as much a problem because it usually comes from of two things:
- Overly damped motor structure (low Q driver)
- Directivity
Both of these are completely linear effects, and can therefore be dealt with EQ. Of course, with passive solutions, your only choice is to "cut" the rise, not "bump" the loss! That means that a driver that may be 95 dB efficient ends up with an effective system efficiency of 85-88 dB efficiency. But it doesn't mean the same kinds of distortions.
Dan Wiggins
Adire Audio
Dan, Peter:
I think you are both arguing back and forth, and both making a lot of sense, but not about the same thing.
Peter, you seem mainly concerned with the passband inadvertantly exciting breakup resonance, due to _linear_ distortion (in the passband -- the resonance itself may have some nonlinear components, it doesn't have any connection to the argument).
Dan, you are worried about _nonlinear_ distortion related to excursion exciting the same (or other unrelated) resonance.
These both exist. High pass addresses Dan's concern. Low pass addresses Peter. I think you both would agree that this is not an either-or situation: we need to fix BOTH.
Does that seem reasonable to both of you?
I think you are both arguing back and forth, and both making a lot of sense, but not about the same thing.
Peter, you seem mainly concerned with the passband inadvertantly exciting breakup resonance, due to _linear_ distortion (in the passband -- the resonance itself may have some nonlinear components, it doesn't have any connection to the argument).
Dan, you are worried about _nonlinear_ distortion related to excursion exciting the same (or other unrelated) resonance.
These both exist. High pass addresses Dan's concern. Low pass addresses Peter. I think you both would agree that this is not an either-or situation: we need to fix BOTH.
Does that seem reasonable to both of you?
A typical 10" driver has an Sd of ~340 cm^2. A typical 8" driver has an Sd of ~220 cm^2. A pair of 8s are roughly equivalent to a single 12; a pair of 6.5s would be roughly equal to a single 10, in terms of surface area.
Going to towards greater surface area does solve the problem, but dispersion can be adversely affected. The solution is to eliminate distortions in the first place, by reducing nonlinearities related to stroke. There are two components of SPL: bore and stroke. Keep the bore small to help dispersion (and it also helps raise resonant modes), and increase linear stroke to make back the SPL.
It has both linear and nonlinear properties. Basic mechanics and material sciences will tell you the "acquisition" and "release" of energy in materials are not linear, and can be greatly changed based upon the strength of the applied signal. Some modes will not be excited until a given drive level is applied (in the extreme, a cone will fold and generate a VERY strong first order mode when driven extremely hard). Deformation is actually - at its very heart - a nonlinear phenomenon.
But note that I'm talking about the driver as a whole - LINEAR distortion modes can be excited from nonlinear sources! Nonlinear IMD from BL distortion can easily excite linear distortion modes of cones.
Yes, you can go to more drivers. Of course, you still do have other physics constraints - the dispersion of a pair of 6.5s will not be as good as a 10 in one dimension (say stacked 6.5s compared to a single 10; the vertical dispersion will be different).
Rather than choose a cone that will have high strength, high Q modes, I prefer one that will allow a touch more resonance but at much lower Q.
I liken it to an amp that overloads gracefully (20W rated at 0.1%, reaches 1% at 30W) to one that does not (20W rated at 0.01%, reaches 1% at 21W). I will accept an inaudible amount of distortion to reduce the level of audible distortions at higher amplitudes.
It is also HIGHLY excursion based. Check the inductance of a typical driver at 70% of Xmax forward, at rest, and at 70% Xmax rearward. You'll see a typical change of a factor of 2 or more over that range. The high frequency components riding on that will vary a LOT, because the Le of the driver is dramatically change over excursion.
The low frequencies cause the inductance to change a lot over range, which means the high frequencies suffer. Le nonlinearities are a definite excursion (and hence, low frequency) issue.
EXACTLY! Which is why I prefer to look at spectral contamination (multi-tone) plots rather than simple THD/IMD plots. The excitation is much broader, and much more real. A test at 400 Hz may show no cone modes, but when playing music the cone modes can be readily audible because there is more than just 400 Hz present!
Also you can have signals exciting the cone which are not harmonically related to the original input signal!. That is a very crucial point! Distortion products - linear and nonlinear - can interact with each other to generate new signals which bear no harmonic relation to the original signal. Meaning you can "attempt to deal" with a problem in a linear, harmonic manner and it still won't fix the problem. Because the driver ITSELF will generate non-harmonic stimulus!
Potentially. I've worked with some drivers where they sounded fine except for a really annoying ringing that was "supposed" to be 40 dB down. We tolerate distortion components in the low frequencies a LOT better than we do higher frequency components. You may in fact not perceive the THD of 4% in the bass (or 20% Doppler-based IMD) but that ringing mode will be readily excited, and if it's in the 3-6 kHz range can be audibly present.
My experience has been that a good paper or poly cone will overload much more gracefully - will change from clean to strained a lot slower and more controlled. It's not as sudden a change, and IMHO that's a good thing. Smooth changes - in frequency, power distribution, dynamics, etc - is better subjectively than sudden changes.
Potentially, it depends upon the cone. The Extremis 6.8 we sell is a soft cone, but has very low measured distortions, and really no breakup peak. It does vary.
Was that the one they distributed at CES? It is a very good recording!
Dan Wiggins
Adire Audio
tiroth said:
Absolutely. I may have focused on what happens further up but I stated that both are needed to make a stiff driver work well. I was concerned over Dan´s statement that the only way to reduce "the problem" was via a highpass.
I fully understand the excursion and non-linear relationship but after the excursion is limited "enough" via a highpass we still need to focus on the things that happen in the high part of the range and use a lowpass in order to deal with the harmonics that even under low excursion will excite the break up since.
It is the amplitude of the breakup we hear IMO not the "Q" and we need to reduce the amount of excitation by lowpass the driver according to my theorys/explanations earlier.
Thanks for a interesting discussion Dan.
/Peter
EXACTLY. It's pretty surprising what happens to drivers when you goose them with more than one or two tones... Even the best drivers out there will have 3%+ distortion at moderate (94-96 dB) output levels with music. Makes worrying about a hundredths of a percent in the amp kind of a "so what" thing...Pan said:
Dan Wiggins
Adire Audio
DanWiggins said:
Even the best drivers out there will have 3%+ distortion at moderate (94-96 dB) output levels with music. Makes worrying about a hundredths of a percent in the amp kind of a "so what" thing...
... not to mention cable directivity, Mpingo like gadgets and cryogenic treatment ...
Of course, for people with golden egos ***SORRY!*** ears ...
Relative and real distortions
Dan you said:
‘It's pretty surprising what happens to drivers when you goose them with more than one or two tones...
Even the best drivers out there will have 3%+ distortion at moderate (94-96 dB) output levels with music.
Makes worrying about a hundredths of a percent in the amp kind of a "so what" thing... ‘
I’m no expert, but believe that speaker distortion is nearly 100% low order (2nd and 3rd harmonics.)
GedLee has shown through their research that these low orders are inaudible until they are very high in level - more than 10%!!
(The same research shows high order distortions, as can be found in amplifiers, can be audible at 0.01%). Samples demonstrating here:
http://www.gedlee.com/distortion_perception.htm
Cheers
Dan you said:
‘It's pretty surprising what happens to drivers when you goose them with more than one or two tones...
Even the best drivers out there will have 3%+ distortion at moderate (94-96 dB) output levels with music.
Makes worrying about a hundredths of a percent in the amp kind of a "so what" thing... ‘
I’m no expert, but believe that speaker distortion is nearly 100% low order (2nd and 3rd harmonics.)
GedLee has shown through their research that these low orders are inaudible until they are very high in level - more than 10%!!
(The same research shows high order distortions, as can be found in amplifiers, can be audible at 0.01%). Samples demonstrating here:
http://www.gedlee.com/distortion_perception.htm
Cheers
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.