Generated noise under cone movement in all likelihood is a more significant contribution to lowering perceived driver quality than harmonic distortion.
Perhaps we should introduce the term 'disharmonic distortion' for these kinds of deviation from the original signal. Buzz and rub measurements to chart these phenomena have been developed by Klippel.
Perhaps we should introduce the term 'disharmonic distortion' for these kinds of deviation from the original signal. Buzz and rub measurements to chart these phenomena have been developed by Klippel.
that would make more sense to me but alas not so.
i would, by my flawed reasoning make the arrow double headed
linking the models that way would allow the non linear mechanisms of the mechanical properties of the loudspeaker and non linear distortions to be a sub function of linear,no?
Last edited:
There are not only the unwanted harmonics caused by nonlinear distortion. Almost every natural sound has overtones which are also integer harmonics (exceptions intentionally omitted for easier understanding) of the fundamental. This is what allows us to distinguish instruments, voices etc from each other.
When a system with linear distortion (i.e. nonlinear frequency response) is fed with a sound consisting of fundamental and harmonics - the amplitude ratios between them are altered and therefore also the way how that input signal sounds to us. I.e. it sounds coloured.
Regards
Charles
silverprout well, isn't it true? Every component will generate distortion. The worst distortion will be frequency bumps and dips, phase shifts, these are more prominent than the mechanical distortions and result from integration.
Drivers are optimized for a given speed or frequency range and damping. Finding a low distortion driver is useless if it cant be integrated or the frequency response / shifts is too problematic to design a good sounding XO.
Sound coloured, this is driver material distortion, paper/polymer/metal. Astonishingly polymers sound the less colored to the ear, who would have thought a violin sound could be reproduced by a plastic cone?
Drivers are optimized for a given speed or frequency range and damping. Finding a low distortion driver is useless if it cant be integrated or the frequency response / shifts is too problematic to design a good sounding XO.
Sound coloured, this is driver material distortion, paper/polymer/metal. Astonishingly polymers sound the less colored to the ear, who would have thought a violin sound could be reproduced by a plastic cone?
some more sharing of flawed reasoning.
i conceived that a loudspeaker driver's non linear attributes where what caused the problems to begin with.
in my flawed understanding everything from geometry to mechanical properties along with it's electrical characteristics made it what it was.
looking at it this way for me allows explanations as to why frequency response varies with amplitude (and i'm not talking perception here)
i conceived that a loudspeaker driver's non linear attributes where what caused the problems to begin with.
in my flawed understanding everything from geometry to mechanical properties along with it's electrical characteristics made it what it was.
looking at it this way for me allows explanations as to why frequency response varies with amplitude (and i'm not talking perception here)
The purpose of modeling as we do is to simplify things enough to make the math reasonably manageable, yet retain enough accuracy to make the model useful for many or most practical purposes.
So, there are design trade offs in choosing models.
What you are suggesting in the quote above might complicate the math a lot without a whole lot of corresponding gain in practical benefit for designing speaker systems. But, it would probably depend on the exact details beyond what is shown in the diagram.
Depending on what you want to focus on, you might want to come up with your own models for your own design and analysis purposes. But, before you do something like that, probably wise to use the existing tools, get some experience applying them, and see how close you can come to doing what you need with them.
Generally that is believed to not be the case. By far the most notable problems with loudspeakers are the linear stuff.
I don't see this as "flawed", it's quite true.
For the most part frequency response does not change with amplitude. What changes there are, are mostly the result of thermal compression which is altogether different than any thing that we are talking about here. There can be huge frequency response changes in passive systems as the drivers heat up and the coil resistances change. I have measured several dB of changes from this, especially in smallish home systems. Not so much in a system like mine. But short of the thermal aspects, the frequency response will not change with level.
On the perception side things are quite different as it appears that humans have a significant perceptual change with level even if the underlying system is not changing.
It's a good job the FR response doesn't change with level, our perception that it does is problem enough for me, I tend to listen at unrealistically low levels and I'm still looking for a simple practical solution that is adjustable with level.
I'd like to know why I've noticed that with dome based systems, since moving to CD/waveguide, response seems much more consistent with level.
Personally i only take care about integration, all other problems are specialized engineer work, i've downloaded all the Gedlee available documents on the website but i read it when i have some time to spend on theory.
I've no problem with plastic cones, if the loudspeaker driver datasheet is corresponding to my application, i use it.
For my last design there was only two loudspeakers drivers (on earth) that were corresponding to my acoustical specification sheet.
And only one corresponding to mechanical specification sheet...
Last edited:
Yes. I have long suspected it is a diffraction issue (or lack of it) as diffraction perception is level dependent. The difference is greater IMO than any hearing response curves.. I believe this confounds conversations like this. Fade to black is nowhere near as distorted as I used to believe was normal.
Perception doesn't change the measurement results. But perception has an influence on the relevance of the measured results.
If we knew exactly how much THD of which order at what frequency and SPL and/or what amont of IMD (and the best test signal used to measure it) were cleary below perception then it would be possible to build speakers that are not perfect but keep all imperfections below the threshold(s) of audibility.
Regards
Charles
If we knew exactly how much THD of which order at what frequency and SPL and/or what amont of IMD (and the best test signal used to measure it) were cleary below perception then it would be possible to build speakers that are not perfect but keep all imperfections below the threshold(s) of audibility.
Regards
Charles
Except that isn't always guaranteed to work since human threshold of audibility is a nonlinear process. The presence of other frequencies can make a sound that was inaudible in isolation become audible while the other frequencies are also present.
here is some research on this subject, which was recently discussed in one of the other threads around the forum.
Measurements are the domain of the engineer and perceptions are the domain of the customer. What everyone wants to know is how these two domains are linked together. In some cases this link is well known and other not so well known. In some cases the perceptions are real acoustic events and in others they have little to do with reality - they are just perceptions.
I have tried for my whole life to accurately determine the links between these two. The first step is to determine what are real perceptions of the underlying engineering and what are perceptions brought about by factors that are not acoustically related - like appearance or brand or personal bias. Once one gets down to the real parameters that matter, they can be correlated to the measurements to find metrics that tell us what product attributes are likely perceptual problems and what are not. This last stage is partly done, but not completely, and is notoriously difficult to do.
Isn't the other way around the more common, i.e. masking. Masking accounts for all theories of data compression, etc. But the way that you have stated it I have not heard of before.