In the most general sense, speakers with real Thiele/Small sensitivities in the 90~93 dB/metre range have a significant dynamic advantage over speakers in the 85~88 dB/metre range, particularly if we're talking about small midbass drivers (5 to 7-inch). Small radiating areas, small voice coils, and low efficiency do not bode well for SPL's above 100 dB.
What's most noticeable about this class of speaker is limited dynamic range in both senses - they kind of "shut off" at levels below 70 dB, and crash into the wall above 100 dB. So the user is tempted to "gain-ride" the volume control to keep them in the optimum range - if they get too quiet, the sound gets murky and congested, to the point where song lyrics become unintelligible, and they're no fun at all when they hit the dynamic limit, prompting a sudden reach for the volume to stop the screech.
As efficiency reaches the first plateau in the 90~93 dB/metre range, it's like having a car with more torque - you're not always having to change gear as road conditions change - sometimes there are two, or occasionally three, gears to choose from. So you're not always reaching for the volume control - loud parts sound fine as they are, and the speaker sounds better at lower levels, too.
With efficiency in the truly high range of 100~105 dB/metre, the most striking thing is the microdetail. The tonal character is exactly the same at whisper level as it is at full shout. Most audiophile speakers fail this test; they become disjointed and congested when you try and use them for background music - they demand to be turned up, but when you do, you have to gain-ride to avoid the crashes and breakups at high levels.
This loss of detail at low levels is something of a mystery to me: high-end speakers suffer from this, and I suspect some of the enthusiasm for single-driver systems is the desire for a speaker that can play quietly and also be musically satisfying. Altec A7's or Klipschorns, despite their size and complexity, can actually sound quite beautiful at low levels, with sparkling, crystal-clear sound. Horn-loaded Lowthers can also sound quite charming at these levels. The worst examples of speakers that just don't work at moderate levels are inefficient and complex multiway systems with 200~300 lb cabinets.
Speaker designers tend to be hypnotized by performance at the upper end of the dynamic range, while not paying enough attention to the milliwatt region. True, the mechanisms for loss of low-level detail are not well understood, with vague guesses about stiction in the driver spider and surround, or odd level-dependent nonlinearities in the cabinet damping materials. I've yet to see convincing explanations for this, though the phenomenon is common enough - table radios and high-efficiency speakers do just fine at low levels, while big-name audiophile speakers are pretty much unlistenable.
I noticed this in my own loudspeakers when I graduated from 85~88 dB speakers to the 92 dB of the Ariel. Low-level detail was much, much better - not the amazing transparency of all-horn systems, but a major improvement from where I had been. Subjectively, it was kind of a halfway house between audiophilia and the world of horns, and it was a good solution for triode-amp owners in 1993. I was very much hoping in ten year's time that Scan-Speak, Vifa, and Seas would have more efficient drivers, making a progressive update of the Ariel straightforward.
Well, it didn't happen. Driver responses get much rougher than the silky-smooth Scan-Speak/Vifa combination I chose for the Ariel, with no corresponding increase in efficiency. The Ariel drivers eventually went out of production, with only the Skaaning drivers as obvious candidates for replacement - and still no improvement in efficiency, more than 15 years later. Thus, the turn to studio-monitor drivers.
Last edited:
Yeah - I know your point in that - its quite a bunch of effects that are involved "with dynamics", thanks for putting it together again.
I guess this "mystery" will last for some more time - as its not really simple to track this down IMO.
My own concept on that is to look at the mechanical effects involved, which cause - when SPL demand is low, and thus excursion decreases - a significantly "pop up" in the mix of distortions.
There is no usual spec this behaviour can be *absolutely* related to, though it basically is burried in the Qms numbers.
"In gerneral" a high Qms tells us that those specifically friction related distortion must be low - nevertheless Dynaudio with their usual low Qms values shows that this "must not" be the case.
The problem here is that Qms by itself is related on a bunch of effects that turn out pretty different when examined more closely.
Which components in the "bunch of Qms" actually dominate, will heavily change this sonic signature.
So far I have only been able to set up the concept and have made some measurements with speakers I know.
The "results / correlation" seem to speak for my concept, but are far from anything that could create "common ground" on the subject.
Just my 2ct
Michael
I guess this "mystery" will last for some more time - as its not really simple to track this down IMO.
My own concept on that is to look at the mechanical effects involved, which cause - when SPL demand is low, and thus excursion decreases - a significantly "pop up" in the mix of distortions.
There is no usual spec this behaviour can be *absolutely* related to, though it basically is burried in the Qms numbers.
"In gerneral" a high Qms tells us that those specifically friction related distortion must be low - nevertheless Dynaudio with their usual low Qms values shows that this "must not" be the case.
The problem here is that Qms by itself is related on a bunch of effects that turn out pretty different when examined more closely.
Which components in the "bunch of Qms" actually dominate, will heavily change this sonic signature.
So far I have only been able to set up the concept and have made some measurements with speakers I know.
The "results / correlation" seem to speak for my concept, but are far from anything that could create "common ground" on the subject.
Just my 2ct
Michael
Last edited:
Yes, that has certainly been my experience. The more efficient speakers don't seem to need as much, or any, gain leveling. I like that. Very much agree with all this.
And the frequency extremes, too. But the customers seem share the mass hypnosis.....
This is a very interesting observation! When we have very sensitive speakers, I would think this would mean there is less energy wasted to heat and stored in the driver itself. If this is true, probably we would see good CSD performance, which does help reveal more of the detail at low levels. I certainly wish I had more data on this.
When converting compliance, mass, and resistance to their electrical equivalents, it's important to find out - in as much detail as possible - where the real world diverges from the model. These divergences might not have much effect on T/S computations, but they may have a big effect on the sound.
Mass, fortunately, stays pretty much the same, except for nearfield mass-loading effects from air-loads on the front surface of the baffle. But these are linear effects, so not much of a problem.
Compliance isn't as pretty. The driver breaks in and it shifts quite a bit, and the spider and surround are not really all that linear, certainly compared to a capacitor. Stiction (small-signal stop-start effects) may be a real problem, although frustratingly difficult to measure.
Resistance is worse. Electrical resistors are wonderfully linear, while mechanical and acoustical resistors are actually pretty hard to characterize, they have such complex behavior. They are quite nonlinear, small-signal characterization is a guess at best, and breaking down the real behavior of acoustic and mechanical resistors is stunningly complex, on the order of fluid dynamics.
When we measure Qms, that's really just with steady-state sine waves at a rather low level, and that provides an approximate number that satisfies the T/S machine. Measure at high enough levels to see the cone moving, and nonlinearities make the three measurements fall apart. Measure at a lower level, where stiction might start to appear, and room noise becomes difficult to remove from the measurement. It's fortunate for T/S theory that most of the damping comes from the amplifier, not resistance in the spider and surround.
Models are useful, but are dangerously seductive. They are not the real world. Talk to a chemist about the difference between H2O and what actually comes out of the tap in the kitchen - even distilled water is astonishingly complex at the atomic level. Water flow in a pipe? Check out fluid dynamics some time.
What we measure is a crude first-order approximation of what drivers are really doing. One simple example - resistive-vent enclosures, from a strict T/S viewpoint, are a poor choice because they do not use the cabinet volume as efficiently as a closed-box or vented-box. They don't offer the best tradeoff of F3 vs efficiency & cabinet size.
But - T/S theory relies on idealized perfect drivers where the RLC analogies are an exact match to the real driver. In particular, variations in BL product throws off the 4th-order vented equation, since amplifier damping is a critical part of the highpass filter. Electrical 4th-order filters require quite accurate parts tolerances - 1% caps, for example, and that's tighter than most caps you can buy off the shelf. A 1% capacitor is usually considered a precision part, and priced accordingly.
The variation in BL product cannot be assured to anything like that precision with real-world drivers under dynamic conditions - thus, most vented systems are actually mistuned from moment-to-moment, as the DC rectification effects from the vent, combined with BL variations, make the approximation to an ideal 4th-order system approximate at best.
This is where a resistive-vent system has a hidden advantage. Unlike its low-resistance vented counterpart, some of the damping is supplied by real terms (resistance in the vent), instead of all coming from the amplifier (where it is modulated by changes in BL product). In dynamic terms, the two systems will not sound the same, because the acoustic highpass filter uses different systems for filter damping.
The larger point is to look carefully for the places where the real device departs from the model, rather than just following trends in the industry. We certainly need to keep improving measurement techniques, and take out as much guesswork as we can. At the current state of the art, the model isn't all that close to the real thing.
Mass, fortunately, stays pretty much the same, except for nearfield mass-loading effects from air-loads on the front surface of the baffle. But these are linear effects, so not much of a problem.
Compliance isn't as pretty. The driver breaks in and it shifts quite a bit, and the spider and surround are not really all that linear, certainly compared to a capacitor. Stiction (small-signal stop-start effects) may be a real problem, although frustratingly difficult to measure.
Resistance is worse. Electrical resistors are wonderfully linear, while mechanical and acoustical resistors are actually pretty hard to characterize, they have such complex behavior. They are quite nonlinear, small-signal characterization is a guess at best, and breaking down the real behavior of acoustic and mechanical resistors is stunningly complex, on the order of fluid dynamics.
When we measure Qms, that's really just with steady-state sine waves at a rather low level, and that provides an approximate number that satisfies the T/S machine. Measure at high enough levels to see the cone moving, and nonlinearities make the three measurements fall apart. Measure at a lower level, where stiction might start to appear, and room noise becomes difficult to remove from the measurement. It's fortunate for T/S theory that most of the damping comes from the amplifier, not resistance in the spider and surround.
Models are useful, but are dangerously seductive. They are not the real world. Talk to a chemist about the difference between H2O and what actually comes out of the tap in the kitchen - even distilled water is astonishingly complex at the atomic level. Water flow in a pipe? Check out fluid dynamics some time.
What we measure is a crude first-order approximation of what drivers are really doing. One simple example - resistive-vent enclosures, from a strict T/S viewpoint, are a poor choice because they do not use the cabinet volume as efficiently as a closed-box or vented-box. They don't offer the best tradeoff of F3 vs efficiency & cabinet size.
But - T/S theory relies on idealized perfect drivers where the RLC analogies are an exact match to the real driver. In particular, variations in BL product throws off the 4th-order vented equation, since amplifier damping is a critical part of the highpass filter. Electrical 4th-order filters require quite accurate parts tolerances - 1% caps, for example, and that's tighter than most caps you can buy off the shelf. A 1% capacitor is usually considered a precision part, and priced accordingly.
The variation in BL product cannot be assured to anything like that precision with real-world drivers under dynamic conditions - thus, most vented systems are actually mistuned from moment-to-moment, as the DC rectification effects from the vent, combined with BL variations, make the approximation to an ideal 4th-order system approximate at best.
This is where a resistive-vent system has a hidden advantage. Unlike its low-resistance vented counterpart, some of the damping is supplied by real terms (resistance in the vent), instead of all coming from the amplifier (where it is modulated by changes in BL product). In dynamic terms, the two systems will not sound the same, because the acoustic highpass filter uses different systems for filter damping.
The larger point is to look carefully for the places where the real device departs from the model, rather than just following trends in the industry. We certainly need to keep improving measurement techniques, and take out as much guesswork as we can. At the current state of the art, the model isn't all that close to the real thing.
Last edited:
Qms as part of TS parameters is kind of "garbage bin" for anything that does not "really" count in speaker (-box) design - its just there to make math complete.
So its not a big surprise that it is not exactly helpful in the context of analysing speaker behaviour at micro details.
The basis of my concept is that there are two forms of friction involved that make all the difference.
a) velocity dependent friction
b) velocity independent friction
The latter one I see as the root cause the for loss on micro detail and hence in "dynamics". In general there is IMO more to gain in "dynamics" by extending (cleaning up) the lower SPL range than the upper range - or lets say - at least as much.
Michael
So its not a big surprise that it is not exactly helpful in the context of analysing speaker behaviour at micro details.
The basis of my concept is that there are two forms of friction involved that make all the difference.
a) velocity dependent friction
b) velocity independent friction
The latter one I see as the root cause the for loss on micro detail and hence in "dynamics". In general there is IMO more to gain in "dynamics" by extending (cleaning up) the lower SPL range than the upper range - or lets say - at least as much.
Michael
Last edited:
Hello Lynn,
I agree with most of what you say , dis-agree with your conclusion as to why. I'm hoping we can discuss this in absolute terms , no need to use bottom feeder specifications to justify one's point, it's academic that speakers built with poorly designed drivers will have poor performance.
High efficiency speakers do not have better low level resolution nor dynamics IMO and your conclusion leaves out the multiple effects for cause as well as the electronics and how they are affected by the magnitude and phase represented by the speaker and not just efficiency.
Reproduction of audio is very complex there is nothing static about it, there is always a constant juggle to keep a balance and since we have not moved from testing speakers with static normalized test, it remains difficult until one can start to measure speakers dynamically, there is no a ha moment as any wrong on the sliding scale topples the cart , regardless of how good we think the science is, hence no defining one topology for building speakers...
In absolute terms the only limit to good dynamics in a low efficiency speaker vs one that is of high efficiency , is poor design and inadequate amplification.
A large dipole line source speaker will have better dynamics and lower distortion than any high efficiency small diaphragm/chamber horn type speaker. It will not suffer from the dynamic compression that is associated with having a defined acoustic chamber as used in most box speakers and horn drivers...
High end Box speakers ( horns 2 ) or speakers like the Ariel have defined and sometimes undersized acoustic space, this will cause them to suffer from the same compression as you described as they hit this limit defined by the sq area displacement of the driver and it's limited acoustic space(rear chamber). This of course can be improved by proper rear chamber design and i would assume most of the speakers you are describing have such an issue .
This also plays into the other issue you have touched on, low level detail and intelligibility caused by the noise from poor chamber design not because of low efficiency ! Of course this is assuming proper care has already been done in the selecting and implementation of the proper drivers and x-overs, academic that mistakes here kills the day .
No designer sets out to design a low efficiency speaker unfortunately most will end up so if they want proper balance , low distortion and full bandwidth.
Having one for high efficiency alone does not make for a superior speaker IMO....
Edit: defining low efficiency as anything 86db/ 2.83V/m
Last edited:
+10 ...agree with using simplistic testing and modeling for such a complex equation adds to the difficulty of the task at hand .
Disagree: with vent resistance comment , the amplifier always have to be considered as part of the circuit ..
regards,
The test methods and soft ware a certainly capable of finding and displaying a change in micro detail, when the driver is capable of expressing it into the adjacent air.
Look here at a blink comparison that John K provided a few years ago.
Enable Tests
Same driver, same test set up. It is fairly obvious from this comparison, and you need to focus on the resonance nodes to see the changes, that the driver can provide a clear micro dynamic signal, but problems with the radiating diaphragm, rather than the other components of the driver, are the culprit in micro dynamic signal losses.
I think you may have a good point there Michael. Velocity independent friction, or coulomb friction would certainly have a level dependent effect. With coulomb friction there is a threshold level of force required before there can be motion (static coefficient of friction) but once that limit is surpassed the level of force required to sustain motion is reduced. However, I don't see that this would relate to high or low efficiency drivers. It would be associated with damping of the suspension and internal damping the cone.
I also think that before we go jumping off a bridge chasing micro detail and changes in timbre at low level we have to consider Fletcher-Munson. I won't know why one would not expect micro detail and timbre to change with playback level when how we perceive sound is dependent on level to begin with?
Seemed to me that it is not change in timbre so much as the loss of information coherence that supports timbrel comprehension, into the noise floor of the driver, that was being pointed to. That noise floor is what needs to be driven lower and you have shown that there is quite a bit more coherence available from, very likely any given driver, in that series of tests that the blink comparison was part of. No doubt what you and Michael are looking at is an aspect of the problem, but I don't think it is the main contributor here.
I thought that timbre audibly changed with volume regardless of the speaker. Like the pitch of a low frequency goes down and the pitch of high frequency audibly increases as volume goes up. Though obviously it doesn't do this the physical domain. Just inside the brain.
That said, I would say I've heard horns get screechier(for lack of a better word) at high volumes that I thought sounded revealing at low volumes.
Dan
That said, I would say I've heard horns get screechier(for lack of a better word) at high volumes that I thought sounded revealing at low volumes.
Dan
Transistors for Ariels
I've been running my (slightly modified) early model Ariels on a variety of solid state amps.
The ones I keep coming back to are orginal JLHs and a friend's more modern clone (he's been trying for a decade to improve on the original - and he's an EE PhD)
There are still a few I haven't tried - the obvious ones being something from Pass Labs or the Nat Semi based "gain clone" derivatives.
Eventually I'll finish my PP 2A3 amp and that might be it (SE 6AS4 doesn't cut it). But it took me a decade to finish the Ariels so don't hold you're breath for a followup!
I've been running my (slightly modified) early model Ariels on a variety of solid state amps.
The ones I keep coming back to are orginal JLHs and a friend's more modern clone (he's been trying for a decade to improve on the original - and he's an EE PhD)
There are still a few I haven't tried - the obvious ones being something from Pass Labs or the Nat Semi based "gain clone" derivatives.
Eventually I'll finish my PP 2A3 amp and that might be it (SE 6AS4 doesn't cut it). But it took me a decade to finish the Ariels so don't hold you're breath for a followup!
Yes they do suffer from dynamic compression especially the ones with a small rear chamber...
If you do freq testing , do at 1 watt/2/5/10 10/20/30 deg compare the graphs for consistency ...
Thats a interesting CSD comparison.
The first two resonances get lowered but also the 3rd and 4th come slightly more alive.
All in all it looks for me more balanced though and certainly a impressive change for "just a few dips of lacquer" at the right position.
Thanks for pointing there, Bud
Michael
Yes, thats the point I was focusing on.
*If* we are not able to clearly detect information at low levels as it gets masked by upcoming distortion, then this may be perceived as cut in dynamics at the low SPL end.
Michael
Agree, neither see I a direct relationship to high eff drivers.
But on the other hand, we mostly see the very high Qms drivers *not* in the low eff home audio department
In your statement (about velocity independent friction) there are two points to possibly better keep separated .
1.)Static coefficient of friction greater than dynamic coefficient of friction
2.)Static coefficient of friction equal with dynamic coefficient of friction
3.)Static coefficient of friction lower than dynamic coefficient of friction
you were pointing out 1.) - which not necessarily must be the case.
2.) & 3.) works just as well to degrade micro-dynamics.
No big deal, just to keep things sorted out...
Basically *if* we could do such precise measurement we would see (in any case) sort of clipping of a sine that gets heavier with *decreasing* SPL level.
At almost no input signal this leads to zero output signal and in case of outfading signals to a resting point of the diaphragm being slightly offset.
In addition to that there is morphing of a sine towards triangle. As a triangle has heavy high order harmonics, this is IMO easier to detect by simple (T)HD measurement.
There must be - and actually was with my measurements - an increase of higher order distortion compared to lower order distortion (2nd for example) when SPL for measurement is lowered.
Michael
Last edited:
Looking at data, the initial decay in the passband does help determine whether the system is capable of revealing detail, but there are other factors as well. However, there are lots of testimonials using the EnABL process. I have not listened to a pair personally, but under certain conditions tuned resonances do seem to sort of create some impression of airiness, which could creat impresson of more detail. I have listened in a room that put resonant devices to produce this effect, but I really have not studied this condition. SoundEasy V17 has a feature that takes recordings and generates sub harmonics added to the original recording; would the same concept apply if certain resonances were place strategically? We are getting into an area where listening impressons play an important role. Unless we are able to listening to a similar setup and compare our listening impressions with what others describe, it would be difficult to understand what's going on.
Last edited: