No one knows what are the "objectives". We only know subjective things. By agreement we can choose some things over other things and call them "objectives", but they are bound to remain subjective as they are set by us.
A simple answer is that bigger speakers tend to be more efficient, or the user is driving them with a more powerful amp - because big speakers need big amps, right, 😉 ? - hence the amps driving them are more "comfortable" at a particular SPL. The behaviour of the power amp is all important here, I've been able to get ridiculously small speakers sounding huge simply because the electronics driving them are behaving themselves ...
In one respect - larger images, this is another issue with sound as a gradient.
Ex. 2" compression driver 400-2 khz operation without horn (eq'ed) vs. with horn.
The bounding of the freq.s in the horn effectively "enlarge" and move forward what we might think of as acoustic center for those freq.s..
From an acoustic perspective try this out by "cupping" your hands to your mouth and speaking to others - your voice sounds physically larger as a result. The point where the intensity starts to drop-off rapidly is effectively from a larger source.
Well my big multicell horns certainly sound big. Bigger than most direct radiator midrange drivers. When I used to build little 3" fullrage speakers they always sounded small, even at low volumes. So of that's bass, but some of it is something else.
That said, there are exceptions. The little JohnBlue JB3 is a tiny speaker that sounds big. I was amazed the first time I heard it. Just a little 3" fullrange in a ported box, simple as can be. How can it sound so big? With the size of the box/baffle, it has to be near omni for much of its range. But that shouldn't make a difference, right?
That said, there are exceptions. The little JohnBlue JB3 is a tiny speaker that sounds big. I was amazed the first time I heard it. Just a little 3" fullrange in a ported box, simple as can be. How can it sound so big? With the size of the box/baffle, it has to be near omni for much of its range. But that shouldn't make a difference, right?
You've got it, 😀 ! That's exactly the sort of thing I'm talking about - I suspect the whole setup when you heard that was very carefully optimised.
Even my crazy UULE gets pretty good: I get big depth of stage way behind the screen on orchestral, etc; and I done a recent AC/DC album as loud as the plastic surround can take quite a few times - good cymbals, quite reasonable drive, loud enough so impossible to hear anyone talking ...
Edit: Just looked at the JB3, yes, should be excellent: mount and couple it tightly to a very heavy structure, with decent amp and it should be frightening ...
Last edited:
My path to higher quality sound took me towards that "big" sound too. I realized I liked high-efficiency designs even before I realized I liked uniform directivity.
My earliest examples of impressive speakers were Klipschorns and large Infinities with arrays of EMIT tweeters and EMIM midranges. I wanted fhe delicacy of the ribbons with the power and something else I couldn't put my finger on but liked in the Klipsches. But the Klipsch sounded a little artificial and the Infinity did too, but in a different way I thought if I could combine the best of each, I'd really have something.
Not long after that, I found JBL 22xx series drivers and Altec VOTTs. The VOTT sounded powerful like the Klipsch, but could be too strident. Of course, the Klipsch stuff could be pretty squawky too. The JBL drivers impressed me the most, because they sounded very natural, and were able to remain clear even when pushed harder than any of them. So that all began to come together for me.
I suspected that special thing I liked about the Klipschorns might have been their corner placement, that they acted sort of like giant constant directivity horns. So I began to study that, and it formed many opinions I still hold to this day.
I built my first constant directivity cornerhorns using JBL 2205 woofers in the bass bins, 2115 drivers for mids and 2370 tweeter horns. Experimented with 2405 slot tweeters too, to get that sparkle I liked in the EMITs.
My current models have evolved and are more refined than the first ones, but even the first ones were damn good, in spite of the fact they couldn't remain acoustically close to the corner through the midrange. I attribute their good sound largely a result of thier driver complement. They really sounded fantastic.
As I'm sure many of you would agree, a large, high-quality, high-efficiency speaker just sounds better than anything else. All mid-efficiency speakers sound artificial in comparison, even if played at the same volume. I always assumed it was dynamics or some kind of micro-compression, but never cared to investigate and find out why because the difference was so stark. I just stick with high-efficiency designs because they sound night-and-day better.
My earliest examples of impressive speakers were Klipschorns and large Infinities with arrays of EMIT tweeters and EMIM midranges. I wanted fhe delicacy of the ribbons with the power and something else I couldn't put my finger on but liked in the Klipsches. But the Klipsch sounded a little artificial and the Infinity did too, but in a different way I thought if I could combine the best of each, I'd really have something.
Not long after that, I found JBL 22xx series drivers and Altec VOTTs. The VOTT sounded powerful like the Klipsch, but could be too strident. Of course, the Klipsch stuff could be pretty squawky too. The JBL drivers impressed me the most, because they sounded very natural, and were able to remain clear even when pushed harder than any of them. So that all began to come together for me.
I suspected that special thing I liked about the Klipschorns might have been their corner placement, that they acted sort of like giant constant directivity horns. So I began to study that, and it formed many opinions I still hold to this day.
I built my first constant directivity cornerhorns using JBL 2205 woofers in the bass bins, 2115 drivers for mids and 2370 tweeter horns. Experimented with 2405 slot tweeters too, to get that sparkle I liked in the EMITs.
My current models have evolved and are more refined than the first ones, but even the first ones were damn good, in spite of the fact they couldn't remain acoustically close to the corner through the midrange. I attribute their good sound largely a result of thier driver complement. They really sounded fantastic.
As I'm sure many of you would agree, a large, high-quality, high-efficiency speaker just sounds better than anything else. All mid-efficiency speakers sound artificial in comparison, even if played at the same volume. I always assumed it was dynamics or some kind of micro-compression, but never cared to investigate and find out why because the difference was so stark. I just stick with high-efficiency designs because they sound night-and-day better.
And the answer is simple: the amplifier is having to work much, much harder to push the mid-efficiency speakers to the same SPLs. And it distorts - the amplifier, that is. If the amplifier had enough true grunt then there wouldn't be that difference ...
There's no "must" to it, I'm sure someone will be able to point to at least one counterexample, 🙂. However, in my experience it's always been that way, and my reading of other people's experiences have always confirmed that.
There's a relatively simple way to test this: add high power dummy load resistors in series and parallel to the drivers in the high efficiency speaker, to effectively reduce their sensitivity to that of conventional drivers. This will require the amp to work as per a mid-efficiency speaker - would be interesting to see what people think then ...
There's a relatively simple way to test this: add high power dummy load resistors in series and parallel to the drivers in the high efficiency speaker, to effectively reduce their sensitivity to that of conventional drivers. This will require the amp to work as per a mid-efficiency speaker - would be interesting to see what people think then ...
Alternatively, if you were crazy enough, and could afford or organise it, do the series and paralleling with complete, identical loudspeakers. This would ensure that the impedance characteristics as seen by the amp were sufficiently correct - then stick all but one speaker behind a soundproof partition ... and turn up the wick!
The voice coil changes temperature immediately. The time constants refers to the time to a 90% level. But still, the voice coil can change significantly in under 1 ms. This is too slow to affect the waveform, but not too slow to affect the signals envelope. The thermal effects therefor do modulate the envelope of the signal even if they are too slow to affect the actual waveform. This is commonly misunderstood. Griesinger claims that it is the signals envelope that the ear uses to detect most of the more significant aspects of music - mostly in the range of 700 Hz -6 kHz.
Note that this modulation is completely independent of what people call "thermal compression". The two are not the same thing. One could have high modulation and low thermal compression or low modulation with high compression. One depends on thermal capacity and the other on thermal resistance.
And if the difference in level of modulation between a 1" direct radiating tweeter and a compression driver is more than 1000:1 then it could well be that even though this is admittedly a small effect, it cannot be put down as insignificant without some data. I have taken some data and it does show a huge difference between a small speaker and a big one but it is hard to isolate the effect to say for sure if it is audible, or it is something else?
That's why I say that the data all "indicates" that this is an issue. I do completely agree with Pano on this that small speakers never sound as good to me as bigger ones - dynamics, something. Efficiency does matter.
When one considers that almost no test has shown amplifier differences to be significant I have to conclude that it isn't the amps. In fact, most audible distortion in an amp is crossover distortion. This is made much worse in a high efficiency speaker than a low efficiency one. No, I am sorry, amps just aren't the reason for high efficiency speakers sounding better.
Not a valid test since this will change the frequency response of the system. The parallel resistor is OK, but the series one is a problem.
Go ahead and put a parallel resistor across your speaker and see if you can even hear a difference let alone it explaining the vast difference being heard between a high efficiency speaker and a low efficiency one.
We used to measure voice coil temp at KEF. We would inject a little current through each voice coil of a system and calibrate it to run a meter and read directly in degrees C. Afterwards we were able to simulate it for a given speaker. If you run the input signal through an RMS detector (rectify and log convert), then feed it to one or two time constants, then you can get a voltage proportional to temp to a fairly high degree, for any arbitrary music input.
The most important time constants are first the voice coil, and secondly the magnet structure. My recollection was that VC time constants were from 2-3 seconds for tweets, up to 10 or so for woofers. Time constants for a magnet structure would be around a half hour to an hour.
Watching the temperature was interesting. You could play music at a pretty good level and it would stay at an ambient 20 degrees. Turn it up a little and no change. Turn it up a little more and see the meter just start to twitch a little. Turn it up some more and Wham! It runs sky high and you jump for the volume control to save the tweeters. This is typical of the relationship between power applied, and your perception of playback volume (it also explains why the tweeters always go during a drunken party. "I only turned it up a little bit.")
There may be an "instantaneous change" to temperature, but the time constant really does slow down the change.
Subjectively, think in terms of a recording engineer adding a compressor to a recording with a slow response, a very soft corner and ultimately only 3 to 6 dB of total compression (before system burnout, probably only 2 or 3 dB for loud listening at below headbanging levels).
Compared to all the other hard limiting that goes on in the recording chain, do you think this would be readily noticable?
David
David
But you see I fundamentally disagree with the above claim and I would say that theoretically one cannot change the input signal power and NOT see a change in voice coil temperature. Hence, if you did not see a change then your test must not have had enough resolution. I suspect that your RMS detection had a time constant that was too long to see the rapid changes.
The instantaneous rate of change of VC temperature depends on its thermal capacity and not its thermal time constant. It is a capacitor/resistor sort of thing. The instantaneous changes depend on the capacitor value while the long term changes depend on the resistor and capacitor values as the time constant. Hence rapid changes in the signal will cause rapid changes in the temperature inversely proportional to the thermal capacity - the more copper the less the instantaneous change - while the long term temperature will depend on the resistance (or heat conduction) away from the voice coil. Regardless what the current steady state temperature is, the instantaneous temperature modulations about the steady state will be the same - inversely proportional to the thermal capacity. More thermal capacity less modulation.
Small tweeters have little copper and are less efficient than say a compression driver so they require more current into less copper for a given SPL. By my calculations this difference can be 1000 times.
All this was bothering me and so I asked my physicist friend about it. He said that Yes, the VC temperature will change instantaneously depending on the thermal capacity. That's the definition of thermal capacity - degree change for a given power input. Power is proportional to the square of the current. The current IS a motion of the electrons in the wire and this motion is the same thing as temperature. As soon as the electrons move the local temperature rises. AT the quantum level they are actually the same thing. The longer term average of this temperature then depends on how quickly this change in entropy can be dissipated.
Think of it as an AC signal on top of a DC one. The thermal modulation is the AC signal and the DC term is the steady state temperature. You cannot see the AC signal if your detector has too long a time constant, all you will see is the DC. The DC depends on the thermal time constant and the AC on thermal capacity.
Not all recordings are heavily compressed and some not at all. But then not all recordings sound as different between a small speaker and a large one. At any rate compressors in studios do NOT act instantaneously - they all have inherent time constants.
Last edited:
When I was testing cooling systems for woofers, I ran both destructive and non-destructive tests, naturally pushing them very hard. During one of the tests, with power just under the thermal limits, I noticed a peculiar but not unexpected symptom. I could actually hear the SPL drop at high power levels, at the rate of about a decibel a second.
In this test, I ran a 40Hz sine wave cycled on for fifteen seconds, then off for fifteen seconds, continuous power but with a 50% duty cycle.
The local ambient temperature was pretty stable, because the thermal mass of the magnet structure was large. Local ambient was high, and this sort of "biased" the voice coil temperature, with instantaneous power spikes rising the voice coil temperature above that point.
What I found - the peculiar but not unexpected symptom - was that the SPL was highest at the beginning of the 40Hz burst, and steadily declined as the voice coil heated. The time constant of the voice coil rise was pretty close to what Dave said, fairly slow, increasing resistance enough to decrease SPL about 1dB/second in that particular woofer under that particular test. When the 40Hz burst stopped, the voice coil began to cool, approaching the local ambient temperature.
This is in line with observations/speculations made by both Earl and Dave. I definitely saw compression with a time constant that brought audible change within just a few seconds - within the "signal envelope" Earl described - even in a woofer that has large thermal mass. But this was most noticeable only after the time constants Dave described - the heat soaking of the motor structure - had already occured. It would characterize it as an SPL modulation from power/thermal change. Since I saw it in woofers occuring within just a few seconds, I would expect the same kind of behavior in tweeters in tenths of seconds.
One other thing, I didn't notice it hapenning in a cool woofer. It only seemed to happen in a warm woofer. I have no doubt it was happening at a smaller scale when the magnet was cool, but it was significant as the magnet warmed, and local ambient approached boiling, around 200 degrees Fahrenheit. This shifts the whole operating range of the voice coil up, so thermal spikes increased the voice coil temperature further still. A thermal spike of 200 degrees on a cool voice coil only brings temperature up to maybe 250 degrees. But the same 200 degree thermal spike on a warm woofer brings it up to 400 degrees, and that's getting close to the point where voice coil resistance doubles.
That's what the first link shows. It was a valve that would nudge the airflow into a unidirectional motion so there would be a hot air outlet and a cool air inlet, something that could be run through an intercooler. My goal was to create unidirectional flow without creating asymmetry, to avoid having more pneumatic resistance to cone movement one direction or the other.
Some proposed cooling systems use an external air pump flowing through the vents across the gap, but this tends to move the cone slightly in the direction of air flow, creating asymmetry. I wanted to avoid that, and so spent a lot of effort designing a system that would introduce flow without causing asymmetry.
The valving mechanism I created did a great job of creating unidirectional flow without introducing cone motion asymmetry, so that part was a complete success. But it was a total failure at reducing heat. It seems the existing venting system used in most loudspeakers works pretty well already, and doesn't need much help from an external pump or heat exchanger.
What I found was lacking was cooling for the rest of the motor structure. The magnet and pole pieces get super hot, mostly from magnetic eddy currents and also from heat radiated from the voice coil. This heat was stored in the magnet, raising the local ambient temperature. The thermal time constant was huge - this is the slowest moving part of the thermal system - which makes it easy to overlook in terms of the things Earl and Dave are talking about. Since it moves so slowly, one would expect it to have little effect on transients. It is a big deal for reliability though, 'cause the magnet is like a Thermos bottle, metal surrounded by ceramic, and it holds that heat in and bakes the adhesive holding the voice coil on the former.
I do wonder though if it is also part of the transients we're talking about, in a backwards sort of way. Remember that I said I found the (10 second) compression modulation didn't happen at low temperatures, only at high temperatures. I think this is because at low temperatures, the voice coil temperature spikes (riding on the local ambient) aren't rising to the resistance-doubling point. But increase local ambient and they do. This make a big difference in the SPL, which is what causes audible compression. So if the time constants of smaller drivers, midranges and tweeters are less lengthy - as they surely must be - then a driver might warm fairly quickly and enter this state sooner than woofers do.
In this test, I ran a 40Hz sine wave cycled on for fifteen seconds, then off for fifteen seconds, continuous power but with a 50% duty cycle.
The local ambient temperature was pretty stable, because the thermal mass of the magnet structure was large. Local ambient was high, and this sort of "biased" the voice coil temperature, with instantaneous power spikes rising the voice coil temperature above that point.
What I found - the peculiar but not unexpected symptom - was that the SPL was highest at the beginning of the 40Hz burst, and steadily declined as the voice coil heated. The time constant of the voice coil rise was pretty close to what Dave said, fairly slow, increasing resistance enough to decrease SPL about 1dB/second in that particular woofer under that particular test. When the 40Hz burst stopped, the voice coil began to cool, approaching the local ambient temperature.
This is in line with observations/speculations made by both Earl and Dave. I definitely saw compression with a time constant that brought audible change within just a few seconds - within the "signal envelope" Earl described - even in a woofer that has large thermal mass. But this was most noticeable only after the time constants Dave described - the heat soaking of the motor structure - had already occured. It would characterize it as an SPL modulation from power/thermal change. Since I saw it in woofers occuring within just a few seconds, I would expect the same kind of behavior in tweeters in tenths of seconds.
One other thing, I didn't notice it hapenning in a cool woofer. It only seemed to happen in a warm woofer. I have no doubt it was happening at a smaller scale when the magnet was cool, but it was significant as the magnet warmed, and local ambient approached boiling, around 200 degrees Fahrenheit. This shifts the whole operating range of the voice coil up, so thermal spikes increased the voice coil temperature further still. A thermal spike of 200 degrees on a cool voice coil only brings temperature up to maybe 250 degrees. But the same 200 degree thermal spike on a warm woofer brings it up to 400 degrees, and that's getting close to the point where voice coil resistance doubles.
- Speaker Voice Coil Cooling System - Forced air intercooler system
- Speaker Voice Coil Cooling System - Heat Sink
- Woofer cooling device - Test Cycle
- Speaker Voice Coil Cooling System Patent
That's what the first link shows. It was a valve that would nudge the airflow into a unidirectional motion so there would be a hot air outlet and a cool air inlet, something that could be run through an intercooler. My goal was to create unidirectional flow without creating asymmetry, to avoid having more pneumatic resistance to cone movement one direction or the other.
Some proposed cooling systems use an external air pump flowing through the vents across the gap, but this tends to move the cone slightly in the direction of air flow, creating asymmetry. I wanted to avoid that, and so spent a lot of effort designing a system that would introduce flow without causing asymmetry.
The valving mechanism I created did a great job of creating unidirectional flow without introducing cone motion asymmetry, so that part was a complete success. But it was a total failure at reducing heat. It seems the existing venting system used in most loudspeakers works pretty well already, and doesn't need much help from an external pump or heat exchanger.
What I found was lacking was cooling for the rest of the motor structure. The magnet and pole pieces get super hot, mostly from magnetic eddy currents and also from heat radiated from the voice coil. This heat was stored in the magnet, raising the local ambient temperature. The thermal time constant was huge - this is the slowest moving part of the thermal system - which makes it easy to overlook in terms of the things Earl and Dave are talking about. Since it moves so slowly, one would expect it to have little effect on transients. It is a big deal for reliability though, 'cause the magnet is like a Thermos bottle, metal surrounded by ceramic, and it holds that heat in and bakes the adhesive holding the voice coil on the former.
I do wonder though if it is also part of the transients we're talking about, in a backwards sort of way. Remember that I said I found the (10 second) compression modulation didn't happen at low temperatures, only at high temperatures. I think this is because at low temperatures, the voice coil temperature spikes (riding on the local ambient) aren't rising to the resistance-doubling point. But increase local ambient and they do. This make a big difference in the SPL, which is what causes audible compression. So if the time constants of smaller drivers, midranges and tweeters are less lengthy - as they surely must be - then a driver might warm fairly quickly and enter this state sooner than woofers do.
Last edited:
Sorry but this is wrong. If we apply heat to a thermal mass then it may instantaneously start to rise in temperature, but it will rise at a rate totally determined by its time constant. Don't you remember those plots where time constant of an exponential rise was graphically shown? (an e to the tau type curve) The initial slope is extended to the asymptotic level. The time constant is defined as the time for the initial slope to intersect the asymptotic level. Yes it is an RC type thing. Apply DC to the RC circuit and the initial slope of changing voltage at the capacitor is running at a rate totally set by the time constant.
For the KM1 project I spent days measuring 2 part time constants: Apply 10 Watts to a tweeter and plot temperature rise of the voice coil with a B&K strip chart recorder. (Temperature rise was proportional to voltage across the coil resulting from a fixed DC current injection. The resistance of copper changes about 0.4% per degree C for calibration). It is a classical exponential rise to the asymptotic temperature. Since the magnet structure would also heat up (much more slowly) then I would repeat the same test at 1/100 or 1/1000 of the chart speed and then see the gradual rise of the structure temp. This was necessary for the KM1 as the studios would run them hard all day and the temperature would contnue to creep up. (per the structure's time constant)
Now the ultimate temperature rise does depend on thermal conductivity out of the element. A stasis is reached between the power applied and the power dissipating and the temperature will reach its final value (after many time constants). In our driver example the thermal mass and the air gap are the first two elements. Temperature of the coil rises, increasing the bleed out rate until the heat output matches the heat input and temperature stabilizes.
Search the web for KEF S-Stop info, I think a white paper was published. It was a protection circuit that did what I said earlier. Input signal to a driver when to a full wave rectifier, then to a log converter (getting power input from the voltage squared) that went to an RC time constant used to represent the driver time constant. When the capacitor charged to sufficient voltage, set by the maximum temperature we thought the driver was good for, a relay clicked and turned off the driver. Voltage at the relay was made highly proportional to voice coil temperature because the time constants were taken into effect. Without proper regard to the time constants then the circuit would trip too early or too late, depending on the dynamics of the music.
In the end, S Stop was a simple analog computer converting signal input to a voltage proportional to real time temperature. Proper time constants were a necessary element for accuracy.
David
Last edited:
Very intelligent design imo, though apparently it wasn't much a commercial success (probabaly difficult for the typical audiophile to wrap his head around, since it departed significantly from generally accepted paradigms). Never heard a pair, unfortunately.
Yes, I think what you describe here works very well. Do you have a ballpark rule of thumb for differentiating between the early field and the later field? I have been using 10 milliseconds as a target.
Sounds like an excellent approach to me. I think the spectral content of this later field is imporant - if the spectrum is tilted downward too much, it doesn't sound as good (this conclusion supported by 100% agreement among my listening panel of one). Obviously the approaches you have found to work well - Kantor's and your dedicated reverb speakers - would do a fine job in this regard.
In "Sound Reproduction", Toole cites studies indicating that speech intelligibilty is improved by reflections (under some conditions), and proposes that the ear/brain system is better able to decipher complex sounds if given multiple "looks" via reflections. So increased density of room reflections may not be necessarily detrimental to clarity.
In between directional monople and full-bore omni, one possible choice is the directional bipole. With aggressive toe-in and sufficient distance from adjacent walls, the later field becomes more desirably busy (assuming the room isn't overdamped) while leaving the early field suitably dry.
Last edited:
- Status
- Not open for further replies.