the main issue with Sd modulation is that it directly modulates the amplitude of the signal reproduced for a broad band of frequencies (AM modulation). This is just as bad as modulation of the Bl factor in the motor. Both cause severe intermoduation where e.g. the bas is modulating the midrange. Cheers, Lars@Purifi
I guess its not showing in the simulation as I'm using ideal drivers that seem to have constant sensitivity no matter the Sd used.
edit.
Yeah amplitude modulation seems quite audible and something to avoid https://purifi-audio.com/2019/12/07/amfm/
Is it correct to assume that the directivity is affected as well? Seems like it happens, off-axis is amplitude modulated even more? Or is it? When Sd gets smaller, pattern widens where wavelength is similar to cone diameter but also SPL reduces so perhaps the amplitude modulation effect is more pronounced on-axis after all. Anyway, something to avoid altogether.
edit.
Yeah amplitude modulation seems quite audible and something to avoid https://purifi-audio.com/2019/12/07/amfm/
Is it correct to assume that the directivity is affected as well? Seems like it happens, off-axis is amplitude modulated even more? Or is it? When Sd gets smaller, pattern widens where wavelength is similar to cone diameter but also SPL reduces so perhaps the amplitude modulation effect is more pronounced on-axis after all. Anyway, something to avoid altogether.
Last edited:
Here is GIF with SPL adjustment, -3db when SD is 55cm2 and +3db when Sd is 105cm2, is this more correct what would happen? If it is, crazy stuff.
Even the system group delay modulates with excursion, crossover varies. Conclusions can be drawn in haste that audibility of group delay also depends on SPL, excursion happening in the system, like diffraction perhaps. And something whose audibility cannot be tested in headphones for example unless whole listening environment was somehow constructed with sources doing the modulation and reflections auralized.
Even the system group delay modulates with excursion, crossover varies. Conclusions can be drawn in haste that audibility of group delay also depends on SPL, excursion happening in the system, like diffraction perhaps. And something whose audibility cannot be tested in headphones for example unless whole listening environment was somehow constructed with sources doing the modulation and reflections auralized.
Last edited:
Yeah I assume it would be very hard to simulate all the effects, but already this is quite shocking to watch and kind of demonstrates why big speakers tend to sound better than small even if they weren't perfect. Keeping excursion low (/ bandwidths narrow) seems paramount. And of course good drivers would help, especially for small systems 
Yeah I assume it would be very hard to simulate all the effects, but already this is quite shocking to watch and kind of demonstrates why big speakers tend to sound better than small even if they weren't perfect. Keeping excursion low (/ bandwidths narrow) seems paramount. And of course good drivers would help, especially for small systems
Funny how this has been witnessed/observed for decades, before we had the data.
Big cones moving minimally beats smaller cones moving maximally, despite equal Vd.
Finally we have a dynamic representation of frequency response and dispersion, not just static visualisation.
Interesting your animated GIF doesn’t go fast enough. Imagine real music with 100-150bpm…
VituixCAD2’s six-pack now becomes all wiggly jiggly.
Woo hoo look at that blubber fly!
Last edited:
Looks to me like the behavior is entirely driven by the off axis response changing due to the diaphragm diameter changing, that's why the on axis doesn't change, you've already defined it (idealized flat probably). The sim won't change the on axis, just off due to the diffraction calculation. So we should at least ask the question of whether we are just playing with models and not reality.
To the extent the sim is accurate of realworld behavior, a larger driver is one solution, but so is a 3-way. The midrange won't move much even when crossed fairly low, and as long as the woofer is crossed at frequency where it is still omni-radiation, there shouldn't be much issue.
To the extent the sim is accurate of realworld behavior, a larger driver is one solution, but so is a 3-way. The midrange won't move much even when crossed fairly low, and as long as the woofer is crossed at frequency where it is still omni-radiation, there shouldn't be much issue.
its just diffraction tool generated responses, used the three Sd values mentioned in purifi article and made data for those, on same baffle. Then just swapped in one set of responses at a time, exported the sixpack, combined them as GIF to kind of see how response changes with excursion cycle. So its not real sim as such, just attempt to visualize what the changing Sd roughly means.There will still be issues with the woofer when its radiating area is modulated. This would lead to excursion-dependant efficiency. As already mentioned this will lead to a behaviour similar to nonlinear BL.
Regards
Charles
Edit: I was a bit slow with typing. My response is refering to post 2028
Regards
Charles
Edit: I was a bit slow with typing. My response is refering to post 2028
Yep, the amplitude modulation seems severe though and affects the whole bandwidth of the driver making the excursion. And in addition directivity changes with Sd, which is also kind of roughly here giving some hint of magnitude of things. Real world response would be more complex especially on the top end. Also, this is quite radical example, 5" driver taken to limits. Anyway, good food for thought.
Edit, 5" driver in example has 8mm surround and it makes the Sd vary guite much. My 15" driver has ~15mm surround and Sd varies much less, roughly between 760-960cm2 if considering surround fully exluded or included to Sd. Only about 25% increase, whereas the 5" driver was almost 100%, 55-105cm2
Last edited:
Thought experiment about distortion and directivity/stereo imaging: If sound containing low frequencies, like a synth or drums, is hard panned to either side it would now make the amplitude modulation, hysteresis, all sortsa non-linear distortion effects happen only that side of stereo speaker setup. This would affect imaging, right? Phantom images would now be smeared / modulated as the sides of stereo pair would differ. Is it correct to assume non-linear distortion is not just about bad sound as in distortion sounding distortion, but effects extend to break down imaging as well?
If its so I wonder how audible is this, anyone made tests with it? This kind of perceived distortion would be dependent of program material I think. It has been tradition to mix low frequency sources on the center, which would make both sides distort roughly the same and imaging would not suffer as bad perhaps hiding the problem some. Anyway, imaging would be improved with reduced system distortion if above is true.
If its so I wonder how audible is this, anyone made tests with it? This kind of perceived distortion would be dependent of program material I think. It has been tradition to mix low frequency sources on the center, which would make both sides distort roughly the same and imaging would not suffer as bad perhaps hiding the problem some. Anyway, imaging would be improved with reduced system distortion if above is true.
^ All nonlinearity is "program" specific.
No one has ever tested your hypothesis to my knowledge.
In my estimation this nonlinear image destruction would be very slight. Nonlinearity is a small effect - on the order of a few %. Since imaging is dominantly a HF effect, this small % change in content would not highly affect the image - not until its level became intolerable.
Image is highly affected by timing of room reflections and diffraction, which are linear effects.
No one has ever tested your hypothesis to my knowledge.
In my estimation this nonlinear image destruction would be very slight. Nonlinearity is a small effect - on the order of a few %. Since imaging is dominantly a HF effect, this small % change in content would not highly affect the image - not until its level became intolerable.
Image is highly affected by timing of room reflections and diffraction, which are linear effects.
Changes in Sd on this order is new to me. How was it determined - how reliable are these estimates?
Hi,
numbers were taken directly from Purifi article https://purifi-audio.com/2021/10/14/some-speaker-problems-that-needed-solving/
I understood its extreme worst case scenario example of the issue. Numbers come from making driver surround area compeletely included or excluded from Sd. Nominal Sd found in driver datasheets is usually calculated taking half of the surround into account.
I'm sure lrisbo can comment more accurately about the issue.
numbers were taken directly from Purifi article https://purifi-audio.com/2021/10/14/some-speaker-problems-that-needed-solving/
I understood its extreme worst case scenario example of the issue. Numbers come from making driver surround area compeletely included or excluded from Sd. Nominal Sd found in driver datasheets is usually calculated taking half of the surround into account.
I'm sure lrisbo can comment more accurately about the issue.
A quick glance at the post indicates to me that this Sd change is simply a guess.
Why does Sd change with a forward thrust versus a backward one? (as shown in his figures.) It seems nonsensical to me. Sure there is some change, but nothing even close to 100%. Maybe 10% worst case.
Why does Sd change with a forward thrust versus a backward one? (as shown in his figures.) It seems nonsensical to me. Sure there is some change, but nothing even close to 100%. Maybe 10% worst case.
hi gedlee,
In a normal half roll when moving out of the gap: more of the rubber near the cone moves compared to the parts closer to the frame and vice vera when the coil moves into the motor. This is caused by the tangential forces in a revolved geometry. The phenomenon can be simulated in FEA rooms such as Comsol.
We still lack a good measurement procedure to capture Sd(x). I did one method where I had both laser on the cone and a mic inside the box. This was accurate enough to clearly show the difference between inverted and non inverted half rolls. A high resolution air flow sensor would be great to have.
Simulations show that peak to peak Sd variation across the stroke can be up to 25-30% for small long stroke drivers. Having back pressure from a box makes the variation worse. The Sd modulation is reduced for large diameter drivers with narrow surrounds.
The most convincing moment was that when we finally had our Sd linear surround and we saw the distortion drop 20dB compared to using a normal half roll.
cheers,
Lars/Purifi
Lars
Thanks for your response. It was the 100% claim that I objected to. Worse case of 25-30% is more believable (and this true only for very small speakers at large excursions.) But, still, some measured confirmation would be best.
The axial displacement of the cone and surround will always vary approaching zero at the clamp. This is simple geometry. But I don't see how the "tangential forces in a revolved geometry" (hoop stress), that results changes the displacement profile. They are orthogonal.
But, as you say, this can/should all be designed for. If that's what you did then congrats.
Thanks for your response. It was the 100% claim that I objected to. Worse case of 25-30% is more believable (and this true only for very small speakers at large excursions.) But, still, some measured confirmation would be best.
The axial displacement of the cone and surround will always vary approaching zero at the clamp. This is simple geometry. But I don't see how the "tangential forces in a revolved geometry" (hoop stress), that results changes the displacement profile. They are orthogonal.
But, as you say, this can/should all be designed for. If that's what you did then congrats.
yes, thank you we designed a surround geometry that has constant Sd. It looks rather strange but wokrs very well.
not sure where the 100% claim comes from - hopefully not from our write up?
Another way to observe a half roll: the crest point is moving ro a lower radius for the cone moving out and higher radiius when moving in. The crest point is a proxy for the effective piston radius. if you take a half roll in a linear extruded geometry then it would not move as much. The tangential forces are of course othogonal but they stiffen the object in the axial direction (geometric stiffening). This is pretty much why we use cones and not flat disks. When the surround diameter is large compared to the roll width then the stiffening gets more uniform and the Sd change is reduced
not sure where the 100% claim comes from - hopefully not from our write up?
Another way to observe a half roll: the crest point is moving ro a lower radius for the cone moving out and higher radiius when moving in. The crest point is a proxy for the effective piston radius. if you take a half roll in a linear extruded geometry then it would not move as much. The tangential forces are of course othogonal but they stiffen the object in the axial direction (geometric stiffening). This is pretty much why we use cones and not flat disks. When the surround diameter is large compared to the roll width then the stiffening gets more uniform and the Sd change is reduced
I think this impression comes from my posts here on this thread. Thats how I understood it, and that it was extreme example scenario.
Very useful for the illustrations to get magnified look what kind of effects varying Sd makes in system response.
Last edited: