Hezz said:
This is very insightful and right in line with a concept that I have been working on for the last several months. Cone breakup is actually NOT chaotic, but perhaps it would be better if it was. In every case that I have looked at in audio the idea of increasing the "entropy" (randomness or disorder) improves the subjective perception. The idea that an audio system would sound better if all its coherent artifacts were incoherent - increased entropy - is actually quite profound.
I am actually doing a major project which is not in audio on this exact concept. Making something noisey that is highly coherent and annoying into something that is incoherent, but not noticable. No reduction in sound level need occur at all, its just that the ear doesn't notice it anymore. Our hearing system is actually active and it takes energy to run that active feature. The hair cells lock into signals and actively increase the vibrations on the cochlea which results in the huge dynamic range that we have. If the signal is coherent then this mechanism is in full force, but if it is incoherent then it doesn't function at all. The coherent case become fatiging and the incoherent case could actually be pleasant, and yet there is no change in the actually level of the sound energy.
There is much to this concept that is appealing. I'm almost ready to file some patents.
Can we really be sure about that, given that even pistonic movement can be chaotic, showing jump resonance phenomena, when not electrically damped, arising from nonlinearites (there is an AES paper from Tymphany on that). Can we safely assume that the damping mechanism in cone breakup are linear enough to not cause any chaotic transient effects?gedlee said:
- Klaus
KSTR said:
I am not familiar with the Tympany paper. Can you reference it. and what does it mean "when not electrically damped". What you are saying just doesn't sit right with me and I have studied nonlinear dynamics intensly.
Chaos REQUIRES very strong nonlinearity and the cone breakup is linear as far as I have ever seen. And even if its not linear it is not nonlinear enough to go into chaotic motion. I don't think that there is such a thing as "chaotic transient effects"
Tymphany have removed it from their site, it seems. So I've sent it to you via Email. It's about speakers under current drive conditions (not what normal drivers are build for, hence)
Official AES-link is:
http://www.aes.org/e-lib/browse.cfm?elib=12904
- Klaus
Official AES-link is:
http://www.aes.org/e-lib/browse.cfm?elib=12904
- Klaus
gedlee said:
Dr. Geddes is making a very important point here. If cone breakup were truly random, like modulation noise on a tape deck, it would be more benign than the resonant behavior we actually get. The ear/brain/mind notices the degree of coherence in the breakup, and attention is drawn to its resonant character - the ear/brain/mind is trying to decipher what materials are physically made of, something we do unconsciously all the time.
Is that "thump" we just heard at night made of wood? Was it a tree branch? Is there something rustling in the grass? Millions of years of evolution have attuned the ear, brain, mind, and most of all, emotions to assessing what's going on in the immediate environment. Music builds on this perception, and tickles it with all kinds of interesting changing tonal characters. Density is not a problem - if anything, it is musically stimulating, and gives the ear/brain/mind a fun puzzle, like the double-track effect of understanding the poetry of words in a song at the same time you appreciate the musical harmonies.
Coherence in the noise is very readily detected - this is a system that can readily perceive a few sour notes in a large ensemble of symphonic or choral performers, something well beyond the ability of our spectrum analyzers and MLS gizmos. If the loudspeaker is playing its own little mindless tune, following along with the music, that interferes with the complexity of a large-scale ensemble work - like some joker playing along with a kazoo in the next room. Once you become aware of the "joker", you will be very annoyed, and will find it hard to pay attention to main performance.
Well, that's the situation that audiophiles are in. Once you become aware of the colorations of loudspeakers, they are annoying - and not easy to "forget" once you've learned what a certain coloration sounds like. That's why audio is so fad-driven - it's a new set of colorations that fools us - for awhile. This "fooling" process may last for several years, until yet another "improvement" comes out, and then makes the colorations of the previous generation immediately audible. Then everyone jumps off the old bandwagon and on to the new one, and pats themselves on the back about how clever and smart they are.
Remember "Perfect Sound Forever" and all the industry leaders who insisted that 44.1/16 PCM was audibly indistinguishable from a live mike-feed - and the analog guys were a bunch fuddy-duds who didn't want to move with the times? A decade later, 96/24 PCM becomes the new standard in the recording studio, and guess what, when you flip the switch between 44.1/16 and 96/24 PCM, the much higher resolution format sounds noticeably more - you guessed it - "analog". So who's right and who's wrong?
One of the most interesting points in the Newell and Holland book is the studio practice of assessing studio-monitor loudspeakers by their ability to resolve between different ADC/DAC combinations. The lowest-quality loudspeakers render 16 and 24-bit PCM the same; better loudspeakers can resolve the difference between 16 and 20-bit; and the best loudspeakers can reliably and consistently resolve the difference between 16, 20, and 24-bit sources. I was surprised to find this method of assessing loudspeakers has become common practice in high-end studios (at least in the UK).
Lynne
Your last paragraoh is what I have been saying fro years. Loudspeaker popularity is Fad driven.
To prove a point being made here to your self (maybe I'll do this and post it, who knows, but anyways) add a very low level tone that is below the RMS value of the signal. Lets say that its always below the signal so the signal to noise is some - x dB. Listen to this and when your ear locks onto this tone it will stand out like a sore thumb. This is what cone modes do. They are a constant mechanism for a pure tone of fixed frequency that is always present.
I agree with Lynne on the point about never using a driver once it starts to break up. I can deal with the classic edge hole, because its not really a "mode" per see, but once there is a resonance, the driver is no longer useable. I believe that randomizing these resonance would make them far more palatable. (Of course the driver manufacturers aren't interested.)
And to Soongsc, you are incorrect. Cone shape will not makes the modes nonlinear. You seem to be confused about nonlinearity in this context.
Your last paragraoh is what I have been saying fro years. Loudspeaker popularity is Fad driven.
To prove a point being made here to your self (maybe I'll do this and post it, who knows, but anyways) add a very low level tone that is below the RMS value of the signal. Lets say that its always below the signal so the signal to noise is some - x dB. Listen to this and when your ear locks onto this tone it will stand out like a sore thumb. This is what cone modes do. They are a constant mechanism for a pure tone of fixed frequency that is always present.
I agree with Lynne on the point about never using a driver once it starts to break up. I can deal with the classic edge hole, because its not really a "mode" per see, but once there is a resonance, the driver is no longer useable. I believe that randomizing these resonance would make them far more palatable. (Of course the driver manufacturers aren't interested.)
And to Soongsc, you are incorrect. Cone shape will not makes the modes nonlinear. You seem to be confused about nonlinearity in this context.
gedlee said:
I found this out the hard way on my first loudspeaker for Audionics in 1976. I was stuck with the driver array since the cabinets and drivers were already on hand, so I had to make-do with what I had. Even though the instrumentation was very crude - a pink-noise generator and an Altec 1/3 octave RTA (ugh) - it became evident the B139 woofer had a huge resonance around 1.5 kHz. It was obvious the Q was a lot higher than the RTA's filters would show - it was only a few dB high, but was very audible on pink-noise.
Twiddling around with a notch filter made it possible to drop the resonance in and out of audibility. Having already learned a few pointers from Laurie Fincham of KEF and the BBC Lab, I knew that pink-noise was the most sensitive subjective test stimulus, so the pink-noise aligned notch filter was certainly going to work on music - which it did.
But for reasons that are lost in the mists of time, it wasn't practical to use that particular notch filter in the production crossover (although there were plenty of others in the system). Since the B139 crossover was around 150~200 Hz (yes, big iron-core inductors), I figured 20~30 dB of rejection (below the level of the main response) would get rid of the coloration. No, it didn't. The threshold for rejection was somewhere around 40~50 dB. It took a lot of low-pass filtering to get rid of it - several octaves and a minimum of an 18 dB/octave slope. I really had to hit it with a hammer to make it go away.
Frankly, I was both astonished and dismayed. That meant that many, if not most, commercial loudspeakers had plenty of "buried resonance" as the BBC described it. There was absolutely no way a conventional frequency-response measurement would show the difference between 20 and 50 dB of rejection of the 1.5 kHz B139 resonance - but it was immediately audible on pink-noise.
After time and some ear-training, you could hear the resonance on music as well. With music, the coloration would come and go in an odd, unexpected fashion - if you didn't know it was there, waiting to come out and bite you, it would be easy to blame the recording - it sounded a lot like bad microphone technique. But it was nothing of the sort - it was a high-Q resonance in the rigid expanded-foam "cone" of the B139 driver.
This is where I have a bone to pick with magazine reviewers. I have actually heard the editors of the highest-profile, largest-circulation review magazine say that they consider reviewers to be "more objective" if they don't know about the technology of what they're writing about. Worse, this kind of idiocy has spread to mainstream newspapers and TV news, where knowledge of the subject is actually considered a detriment to "objectivity".
How on Earth ignorance has become re-defined in the USA as "objectivity" I don't know, but it speaks volumes about the quality of news coverage we get. If this trend continues, we'll start seeing "even-handed" coverage of the Flat-Earth Society versus a "designated spokesman" of the AAAS, with the newscasters accusing the AAAS spokesman of being an "elitist" when they use "big words". I wouldn't put it past the newscasters to start sniggering about "AAAS" itself.
Since very few magazine reviewers have actually designed a loudspeaker, they have no idea of what raw-loudspeaker colorations actually sound like. Thus, the off-the-wall magazine descriptions of the sound - "chocolate midrange" and similar absurdities. If more audiophiles could hear for themselves the colorations of raw drivers with pink-noise, comparing the DUT to a known-good headphone, the silly stuff would get chased off the market pretty quickly.
By the way, the words "breakup" and "resonance" are sometime used interchangeably, and that can lead to misunderstandings. I sometimes do this myself, and it probably confuses people.
A better way to think of resonance is a small-signal phenomenon. "Small-signal" has a precise meaning - it means present at all signal levels, and most importantly, present right down into the noise threshold. Small-signal resonances are typically modelled by an imaginary array of RLC filters in parallel with a "perfect" transducer, but it should be kept in mind this is a very simplified model. Real-world resonances can be quite directional at mid and higher frequencies, which is where the RLC equivalent starts to break down. Resonance is usually characterized as a "linear" distortion, since it is level-independent.
Breakup is strongly level-dependent, and with many diaphragm materials, has a very sudden onset, like clipping in an amplifier. Diaphragm materials that are soft and lossy are more like soft-clipping amplifiers - no sudden onset, but the tradeoff is degraded performance at levels just below breakup. Going to the performance of the diaphragm in more detail, infinitely rigid materials do not exist, and all diaphragms are flexing to some degree. So the distinction between sudden and gradual breakup is not hard and fast, but more a matter of degree. Breakup is considered an aspect of "nonlinear" distortion, since it is strongly level-dependent.
Since we're talking about real-world loudspeakers, the two problems can interact. Both arise from diaphragm flexing - if the diaphragm were infinitely rigid, neither would arise. If the diaphragm is pushed hard enough, not only will it go completely out of control, it will mechanically fail, typically near the surround, spider, or voice-coil attachment, wherever the G-forces are greatest.
In amplifier design, the areas to watch for are where "linear" and "nonlinear" effects interact. These are hard to measure, but an experienced designer will know which parts of the amplifier are under the greatest stress under dynamic conditions, and build-in a reserve of headroom, even if the specs never reveal the improvement. But the amplifier will sound better and also be more reliable.
The same applies to loudspeakers. It is important to know the regions where the diaphragm is reaching the limits of its performance envelope, and steer energy away from these areas. This not only improves the sound, but has a favorable effect on long-term stability and reliability.
This may sound like an obvious thing to do, but driver vendors do not want to draw attention to the weakest areas of their products - they want you to look elsewhere, and that's the exact area that is most important to the system designer. The same applies to amplifiers - the specs you need to know, the ones that affect critical areas of design, are frequently not available from the parts vendors.
This is part of the reason I like to talk to the actual engineers that have designed the transducer, or the active and passive elements in an amplifier. When you bypass the marketing department, you get closer to the truth - which is why big companies go to considerable trouble to isolate the engineering staff from the public. The engineers are keenly aware of all the weak points (they wouldn't be good engineers if they didn't), and have the bad habit of telling the truth!
A better way to think of resonance is a small-signal phenomenon. "Small-signal" has a precise meaning - it means present at all signal levels, and most importantly, present right down into the noise threshold. Small-signal resonances are typically modelled by an imaginary array of RLC filters in parallel with a "perfect" transducer, but it should be kept in mind this is a very simplified model. Real-world resonances can be quite directional at mid and higher frequencies, which is where the RLC equivalent starts to break down. Resonance is usually characterized as a "linear" distortion, since it is level-independent.
Breakup is strongly level-dependent, and with many diaphragm materials, has a very sudden onset, like clipping in an amplifier. Diaphragm materials that are soft and lossy are more like soft-clipping amplifiers - no sudden onset, but the tradeoff is degraded performance at levels just below breakup. Going to the performance of the diaphragm in more detail, infinitely rigid materials do not exist, and all diaphragms are flexing to some degree. So the distinction between sudden and gradual breakup is not hard and fast, but more a matter of degree. Breakup is considered an aspect of "nonlinear" distortion, since it is strongly level-dependent.
Since we're talking about real-world loudspeakers, the two problems can interact. Both arise from diaphragm flexing - if the diaphragm were infinitely rigid, neither would arise. If the diaphragm is pushed hard enough, not only will it go completely out of control, it will mechanically fail, typically near the surround, spider, or voice-coil attachment, wherever the G-forces are greatest.
In amplifier design, the areas to watch for are where "linear" and "nonlinear" effects interact. These are hard to measure, but an experienced designer will know which parts of the amplifier are under the greatest stress under dynamic conditions, and build-in a reserve of headroom, even if the specs never reveal the improvement. But the amplifier will sound better and also be more reliable.
The same applies to loudspeakers. It is important to know the regions where the diaphragm is reaching the limits of its performance envelope, and steer energy away from these areas. This not only improves the sound, but has a favorable effect on long-term stability and reliability.
This may sound like an obvious thing to do, but driver vendors do not want to draw attention to the weakest areas of their products - they want you to look elsewhere, and that's the exact area that is most important to the system designer. The same applies to amplifiers - the specs you need to know, the ones that affect critical areas of design, are frequently not available from the parts vendors.
This is part of the reason I like to talk to the actual engineers that have designed the transducer, or the active and passive elements in an amplifier. When you bypass the marketing department, you get closer to the truth - which is why big companies go to considerable trouble to isolate the engineering staff from the public. The engineers are keenly aware of all the weak points (they wouldn't be good engineers if they didn't), and have the bad habit of telling the truth!
Soongsc and Lynne
Unless you have data to the contrary, cone resonance modes or "breakup" is a "linear" phenomina. This does not mean that they are "flat" it means that they are independent of level. I have not seen a loudspeaker whose "resonances" were very dependent on level. Sure they will change slightly, for a whole array of reasons, but for all practical purposes you will get the same ferequency responce at .1 volt, 1 volt or 10 volt - i.e. it is linear in level.
The shape of the cone or the material is of little consequence because the vibration displacements of the "modes" is very small. This does not mean that the "effect" is small, it simply means that at high frequencies, where the breakups occur, the displacements are small.
You would have to show me real data to change my mind on this. I have had this argument for nearly 30 years, and I am still 100% confident in my position. If you disagree then "Show Me The Data".
Unless you have data to the contrary, cone resonance modes or "breakup" is a "linear" phenomina. This does not mean that they are "flat" it means that they are independent of level. I have not seen a loudspeaker whose "resonances" were very dependent on level. Sure they will change slightly, for a whole array of reasons, but for all practical purposes you will get the same ferequency responce at .1 volt, 1 volt or 10 volt - i.e. it is linear in level.
The shape of the cone or the material is of little consequence because the vibration displacements of the "modes" is very small. This does not mean that the "effect" is small, it simply means that at high frequencies, where the breakups occur, the displacements are small.
You would have to show me real data to change my mind on this. I have had this argument for nearly 30 years, and I am still 100% confident in my position. If you disagree then "Show Me The Data".
gedlee said:
Earl, you are about ten years too late to that party. That is exactly what the whole "NXT" thing is about. It was developed in the UK. The idea is to excite a membrane with (what I presume is) a semi-conventional voice coil. But the location of the voice coil, the shape of the membrane, the thickness of the membrane (may even be variable), and other factors are designed specifically to *distribute* the resonances.
I think they used to call this technology "distributed mode resonances" for a while. For years it was hailed as the biggest breakthrough in speakers since Rice & Kellogg. They claimed it was going to revolutionize speaker design.
Now, ten years later, there are a handful of commercial examples. Mostly seems to be used where a thin form-factor is an advantage. Like having fold-out "wings" on a laptop that sound better than the 1" speakers they typically use.
But I've never heard anyone say that they actually sounded *better* than conventional speakers (although that was their stated goal in the beginning). And besides, it isn't really all that hard to make a conventional loudspeaker that operates essentially pistonically. In other words, the cone break up modes are suppressed by 40, 60, or even 80 dB.
But you can't do that by running a 15" paper cone woofer up to 1000 Hz, sorry.
gedlee said:
I am in complete agreement with you here. When I first started investigating cone resonances 25 years ago, I assumed that higher drive levels would induce cone flexure at lower frequencies. I was quite surprised to find out that this was not the case.
Like you pointed out, at any reasonable drive level, the frequency response of a given driver is essentially the same (except for compression effects due to heating of the voice coil).