Very much what I've found with the horn systems I've worked with (we've discussed this before, recently). It needs to be acoustical 4th order, Bessel or L-R. That can be achieved with 3rd order electrical, but real 4th order electrical often works and sounds better. And yes, it can be frustrating getting the driver positions right. Small shift can matter.
I takes a lot of time, trial and error to fully learn this.
hi
is anybody aware of a mid bass horn that would work well with the ah425 radian playing up to 700hz ?
and off course as low as possible to make integration with a sub easier .
Perhaps the 100hz horn plans at this site ( see link below) would suit your needs? I don't know about recommending a sub that would go up to 100hz easily however.
I am following this thread with interest but I always get caught when thinking about mid base horns for myself in that you need some distance between you and them ( i.e. a large size room) and a large room is not what I have. To get good integration it is my understanding that you need one wavelength distance which I guess would be approx 10-15ft away for 100hz. So I think for my needs the Onken would be the better fit.
inlowsound.com
I am following this thread with interest but I always get caught when thinking about mid base horns for myself in that you need some distance between you and them ( i.e. a large size room) and a large room is not what I have. To get good integration it is my understanding that you need one wavelength distance which I guess would be approx 10-15ft away for 100hz. So I think for my needs the Onken would be the better fit.
inlowsound.com
Well, partial horns then. The first prototypes in Dallas have a 700 Hz crossover, and are horns from there on up. (The supertweeter ribbons used in the listening sessions are back here in Colorado.) Dr. Geddes' Summa (which I've heard in the limited context of the RMAF) is a waveguide from 950 Hz on up. The Altec Model 19 has an 800 Hz crossover, if I remember right. All of them use combine a 15" paper-cone woofer with a horn MF/HF system. The first system of this type was the Lansing Iconic in the mid-Thirties; field-coil drivers all around, for the simple reason that Alnico magnets for loudspeakers did not exist.
Paul Klipsch, hallowed be thy name, was comfortable with combining a direct-radiator with a MF horn; the Cornwall goes back to 1959, and some Klipsch fans consider it one of PWK's best loudspeakers. The ones I've heard were much, much better than the Heresy, which I suspect was PWK's idea of a joke (the Heresy MF and HF horns are 3 to 5 dB hotter than the bass radiator, unlike the Cornwall, which is balanced flat).
There aren't many commercial all-horn systems; the Germans seem to be the only ones to go out on a limb and build these for the high-end market.
Rather than engage in religious arguments, which never go anywhere, I'm curious why the sonics in the bass region are so different. This isn't metaphysics; there's a good reason, right here in the physical world, why sonic differences exist. PWK would hold that it's the low IM distortion of horns that gives them their distinctive sound (and dynamics). Based on previous posts, Dr. Geddes would strongly disagree.
Me? I'm not sure. I'm still wondering why two studio-grade 15" drivers, both operated in their most linear regions, can sound different ... at all. At input levels of less than 1 watt, the difference in IM distortion should be pretty small, and there's no difference in flatness at all, both are well within the piston range.
But then again, I'm the zero-feedback Class A DHT-triode guy, focussed on mechanisms of very low-level distortion, and interactions between power supply and amplifying devices.
Maybe it's low-level stiction effects in the spider, which indirectly affect mechanical resistance (which is a bulk specification with no reference to linearity). Maybe. But if that were the case, horn loading would make it worse, since the cone is now moving less, not more, so stiction effects would be more dominant. But the air-load is greater, which transfers some of the cone damping from the BL/amplifier system to the air-load itself. (The air-load in direct-radiators is so trivial that nearly all the damping comes from the amplifier, which is seen through the magnetic system, which in turn is nonlinear.)
Last edited:
Lynn,
Stiction, maybe..
Were your tests to determine that the frequency response of the different 15"s were identical (if they were not identical, they should sound different) done using the same amp where the sonic differences show up?
The frequency response of a zero-feedback Class A DHT-triode amp is modulated by irregularities in loudspeaker impedance.
Did you test the impedance of the different drivers to see if they were identical ?
Art
This covers much of it, same as it always was 
The coupling of the air with the driver is more efficient and in turn gives better dynamics across the board with a more alive sound and presence. It's hard to go back to anything once you live with a good bass horn.
http://invalid.ed.unit.no/~dunker/why.html
http://kolbrek.hoyttalerdesign.no/index.php/horns/why-horns
You don't need to sit a wavelength away to get the benefit of a bass horn. Some listen to them pretty close. I like to listen from 14' feet out or more myself with my new 40 hz front firing low bass horns and my down firing front wall placed subwoofer bass horn. Of course the horn paths are several feet long. For midbass i like straight short hypex horns loaded with a single 15 (listen at around 11 feet) although you can get great bass out of a good 8, 10 or 12 but the horn is longer.
The coupling of the air with the driver is more efficient and in turn gives better dynamics across the board with a more alive sound and presence. It's hard to go back to anything once you live with a good bass horn.http://invalid.ed.unit.no/~dunker/why.html
http://kolbrek.hoyttalerdesign.no/index.php/horns/why-horns
You don't need to sit a wavelength away to get the benefit of a bass horn. Some listen to them pretty close. I like to listen from 14' feet out or more myself with my new 40 hz front firing low bass horns and my down firing front wall placed subwoofer bass horn. Of course the horn paths are several feet long. For midbass i like straight short hypex horns loaded with a single 15 (listen at around 11 feet) although you can get great bass out of a good 8, 10 or 12 but the horn is longer.
Last edited:
I would go along with such a thing, even though I have never 'properly' investigated such. Most drivers are pretty unpleasant, 'off', cold, and the cheaper they are the worse this aspect usually is - the only solution I've found so far is to drive them hard for an hour or so, to 'free them up' - ignore what they sound like in the meantime, and only seriously listen to them once they've 'broken in' for the day ...
There's a pretty radical solution if the problem is really stiction: modulate the bass channel with triangular-spectrum noise between 30 and 60 kHz. In effect, the stiction is "dithered", just like the nonlinear BH curve of tape modulation is straightened out with HF bias between 60 and 100 kHz.
Not that I'm going to do that: just throwing it out there for the multi-amp folks that are using computer crossovers. Just add SACD/DSD levels of noise for the octave just above the audio band ... who knows, maybe that's why SACD/DSD is known for the "smooth" sound. Maybe the HF noise is really performing a useful function: smoothing out Class AB output-transistor switching gremlins, dithering away correlated HF distortion in the amplifier, and improving the transfer function of the loudspeaker driver.
I avoid amplifier elements with Class AB transitions, including op-amps (most operate in Class AB to avoid overheating). But there are plenty of op-amps in the signal chain in modern recordings, so I can't pretend it's not still there ... although dithering at the 16th, 20th, or 24th bit in the original A->D conversion probably goes a long way to mask the problem.
Returning to horns (in any part of the spectrum), the hallmark sound is most apparent at low levels. There's a wealth of detail, tone color, and just sheer music at background-music levels that is not there with typical low-efficiency audiophile speakers. The low-level qualities are shared with electrostats, most notably the Quad ESL57's. I've always liked the sound of the old JansZen Z-130 4-panel tweeter; in fact, I owned a pair of them in college, sitting on top of AR-6's.
The dynamic mismatch I referred to in previous posts is most apparent at low levels: if a system is misdesigned, there's an odd effect of some of the drivers seeming to "drop out" at low levels, leading to an impression of imbalance. If a misdesigned system is subjectively balanced at an 80 dB level, it will sound odd at 50 dB. At 100 dB, you don't notice as much since everything is so damn loud, you're not paying attention to the finer points of spectral balance. But at low levels you do notice spectral balance, particularly if it's the bass that seems to be dropping out.
Electrostats, like horns (and direct-radiator arrays), are dominated by the air-load. That may be part of the reason for the "direct" sound they're known for. Many years ago when I visited Hiroshi Ito in Hawaii, we visited one of his friends in Honolulu, who had a built a really unusual loudspeaker. (I regret to say both Hiroshi and his friend have since passed on, so I have no further information about the system.)
It was a big fiberglass Tractrix, about 1 meter/40 inches across, with a single JansZen electrostat at the rear, powered by an offbeat single-MOSFET amplifier (in Class A of course). It sounded, really, really good, with striking stereo image, floating in space in front of the horns. It didn't sound like a horn at all, but it also didn't sound like the big panel ESL's, either. Very dynamic and colorful sound, remarkably different than anything I've heard before or since.
Was it flat? It certainly sounded smoother than any horn I'd heard before, and with much better spatial qualities ... almost like an MBL (but with better tonality and resolution). The custom electronics could have had built-in EQ, I didn't think to ask at the time.
Not that I'm going to do that: just throwing it out there for the multi-amp folks that are using computer crossovers. Just add SACD/DSD levels of noise for the octave just above the audio band ... who knows, maybe that's why SACD/DSD is known for the "smooth" sound. Maybe the HF noise is really performing a useful function: smoothing out Class AB output-transistor switching gremlins, dithering away correlated HF distortion in the amplifier, and improving the transfer function of the loudspeaker driver.
I avoid amplifier elements with Class AB transitions, including op-amps (most operate in Class AB to avoid overheating). But there are plenty of op-amps in the signal chain in modern recordings, so I can't pretend it's not still there ... although dithering at the 16th, 20th, or 24th bit in the original A->D conversion probably goes a long way to mask the problem.
Returning to horns (in any part of the spectrum), the hallmark sound is most apparent at low levels. There's a wealth of detail, tone color, and just sheer music at background-music levels that is not there with typical low-efficiency audiophile speakers. The low-level qualities are shared with electrostats, most notably the Quad ESL57's. I've always liked the sound of the old JansZen Z-130 4-panel tweeter; in fact, I owned a pair of them in college, sitting on top of AR-6's.
The dynamic mismatch I referred to in previous posts is most apparent at low levels: if a system is misdesigned, there's an odd effect of some of the drivers seeming to "drop out" at low levels, leading to an impression of imbalance. If a misdesigned system is subjectively balanced at an 80 dB level, it will sound odd at 50 dB. At 100 dB, you don't notice as much since everything is so damn loud, you're not paying attention to the finer points of spectral balance. But at low levels you do notice spectral balance, particularly if it's the bass that seems to be dropping out.
Electrostats, like horns (and direct-radiator arrays), are dominated by the air-load. That may be part of the reason for the "direct" sound they're known for. Many years ago when I visited Hiroshi Ito in Hawaii, we visited one of his friends in Honolulu, who had a built a really unusual loudspeaker. (I regret to say both Hiroshi and his friend have since passed on, so I have no further information about the system.)
It was a big fiberglass Tractrix, about 1 meter/40 inches across, with a single JansZen electrostat at the rear, powered by an offbeat single-MOSFET amplifier (in Class A of course). It sounded, really, really good, with striking stereo image, floating in space in front of the horns. It didn't sound like a horn at all, but it also didn't sound like the big panel ESL's, either. Very dynamic and colorful sound, remarkably different than anything I've heard before or since.
Was it flat? It certainly sounded smoother than any horn I'd heard before, and with much better spatial qualities ... almost like an MBL (but with better tonality and resolution). The custom electronics could have had built-in EQ, I didn't think to ask at the time.
Last edited:
There is perhaps something in there .. if the signal is very 'busy' then aspects of the system are being exercised which aren't otherwise, or perhaps some DC offset or related behaviour isn't being allowed to build up, which 'corrupts' the sound.
The 'stiction' or whatever problem it "really" is, is that which separates hifi, from 'musical' playback. As an extreme example, I use a cheap TV driven by Blu-ray player as a 'test bed'. Fired it up again yesterday, and the initial sound is quite atrocious, ghetto blaster quality - but, keep it firing away as loud as one dares ... and slowly, steadily, the quality builds and builds, detail starts to appear, the treble becomes sweeter and builds in impact, the soundstage increases and depth comes into the picture. By the final stage the volume is run at the limit of the electronics, and one wishes there was far more travel on the level control,
.And with sorted out systems, using any sort of speakers. Years ago I had one of many "I'll be damned!!" moments, when the system was running so softly that I had to put my ears to the speaker drivers to hear anything, the level was equivalent to headphones being used for background 'muzak' - and there was all the sound, still: rich, full, detailed, deep, all the usual, positive adjectives could be used ...
Under Klippel testing the driver's resistance vs. excursion is expressed as Kms(x) seen here:
http://www.klippel.de/uploads/media/Klippel_Nonlinearity_Poster.pdf
A few drivers have been measured with the klippel system here, but it requires their free membership to view them:
Klippel Reviews & Driver Specs - Car Audio | DiyMobileAudio.com | Car Stereo Forum
-while that resistance is likely part of the reason a particular driver has better lower spl detail (..particularly as freq.s decrease), my *thinking more and more points to the surround's damping of the diaphragm (as freq.s increase).
This is more than the internal loss of the surround itself - but also the amount of resistance and lossy bending the surround applies to the outer edge of the diaphragm (which would other-wise freely vibrate like a "whizer").
I'd bet that this amount of damping is also more (or less) effected depending on the cone geometry and surface area.
Finally I would also think that it depends on how much the surround becomes a termination to the diaphragm's outer edge, and how much the surround overlaps the diaphragm's surface.
*I've heard this driver (or its 8 ohm derivative) in several designs and it has fantastic low suspension resistance from most of its linear excursion:
http://www.madisoundspeakerstore.co...eak-illuminator-18wu/4741t-00-7-woofer-4-ohm/
(it's also on the klippel testing on the DIY Mobile Audio website)
-and yet, the sound always seems over-damped to me, with a loss in low level harmonics and venue effects. Now this could be related to the cabinet (and particularly any "stuffing" in those cabinets - though the crystal cable speakers don't have "stuffing"), and I've not heard the driver on an open baffle.
Last edited:
I'm a little at a loss getting my head around how additional electrical excitation WAY above the Woofer's passband could affect its mechanical resistance within its passband.
Not saying that it can't - just that I don't understand it.
Would you care to elaborate a bit further?
Thanks!
Marco
This is also something I have been wondering about, namely: how can people using non-negligible output impedance amplifiers (such as DHTs and to a lesser extent virtually all tube amps) know the extent to which the differences they are hearing between loudspeakers X and Y are due to the 'speakers themselves and not to the different amounts of damping by the amp's output impedance?
( A simple web page illustrating the phenomenon: Loudspeaker Enclosure )
Marco
They have the right intentions but wrong design integration.
It makes no sense to have the BL and Km so linear if it is going to be impedance limited.
The leads to the coil seem a bit tight, I wonder how long this will last before it starts to break if you really let it drive to those limits.
I like the way they allocate those vent holes in the former, wonder if they also have a vent in the center of the core.
Last edited:
Let's take a look at what stiction really is: granular, nonlinear movement. Push a rock across a concrete pavement, and it will not move smoothly. It'll jump from high point to high point, not smoothly at all.
What happens when we move a spider made of impregnated fabric? It's only smooth when you slowly average the motion over time. In reality, the fibers grind against each other with a slip-strike motion. Try listening to a loudspeaker as a microphone. Is it linear? No. The spider, and to some extent the surround, are optimized for large motions (for long excursions), but no real attention is paid to small motions.
The spider could be made of Mylar or similar plastics, but then we'd be dealing with annoying issues of VC sag over time, as well as less overall excursion. Impregnated fabrics are the solution we're stuck with, for the time being, since they work well enough for most applications, and are durable.
Would a microphone diaphragm be suspended with impregnated fabric? Of course not. It's a terrible material for the application, and there's no requirement for long excursions or concerns about sag, since the working diaphragm is so light.
Adding linear resistance to a mechanical system is kind of tricky; many of the resistors are prone to stick-slip effects (unlike electrical resistors) and exhibit low-level nonlinearity. Look at all the trouble turntable vendors go to avoid stick-slip in the main bearing, or similar problems with bearing chatter in tonearms.
An analogy can be made with power amplifiers: in the aggregate, and ignoring low-level distortion, the Zout is 0.008 ohms, or something like that. That's averaged over the entire sine wave, and distortion products are ignored. When we break out of the black-box model and look at what the output transistors are really doing, we discover they are switching on and off with a 0.7V transition, combined with charge-storage effects. This leads to instantaneous transitions in the output impedance as arrays of current-gain devices switch on and off; the transition can be smoothed with dynamic-bias circuits, but not entirely removed. You could switch arrays of batteries with relays, and the Zout measurement would be similar. The fact the output would consist entirely of square waves would not appear in the Zout measurement.
Direct-radiators are constant-acceleration devices. This means at higher frequencies, cone motion is very small, and low-level problems in spider and surround will appear as noise in the 2nd and 3rd harmonic distortion trace. Since it has low correlation with the input signal, it is not cyclic, and will tend to be rejected by the distortion analyzer, especially if the sweep speed is slow (averaging over many cycles).
I've worked in test instrumentation. It's a truism in the business that the instruments do not see what they not designed to look for. When you use a swept spectrum analyzer, you have to be aware the instrument is not designed to see single events, and will give very distorted-looking traces. FFT measurements have their own set of limitations; averaging is useful for extending the S/N ratio of the display, but it rejects non-correlated data as noise. That "noise" may not be noise; it can be granular, non-linear motion, with partial correlation to the input signal. If the test measurement rejects the low-level nonlinearity, that does not mean it is not there.
Rather than treat the loudspeaker as an idealized black box, we have to look at how each element actually performs, and see whether existing measurements capture the behavior or not.
What happens when we move a spider made of impregnated fabric? It's only smooth when you slowly average the motion over time. In reality, the fibers grind against each other with a slip-strike motion. Try listening to a loudspeaker as a microphone. Is it linear? No. The spider, and to some extent the surround, are optimized for large motions (for long excursions), but no real attention is paid to small motions.
The spider could be made of Mylar or similar plastics, but then we'd be dealing with annoying issues of VC sag over time, as well as less overall excursion. Impregnated fabrics are the solution we're stuck with, for the time being, since they work well enough for most applications, and are durable.
Would a microphone diaphragm be suspended with impregnated fabric? Of course not. It's a terrible material for the application, and there's no requirement for long excursions or concerns about sag, since the working diaphragm is so light.
Adding linear resistance to a mechanical system is kind of tricky; many of the resistors are prone to stick-slip effects (unlike electrical resistors) and exhibit low-level nonlinearity. Look at all the trouble turntable vendors go to avoid stick-slip in the main bearing, or similar problems with bearing chatter in tonearms.
An analogy can be made with power amplifiers: in the aggregate, and ignoring low-level distortion, the Zout is 0.008 ohms, or something like that. That's averaged over the entire sine wave, and distortion products are ignored. When we break out of the black-box model and look at what the output transistors are really doing, we discover they are switching on and off with a 0.7V transition, combined with charge-storage effects. This leads to instantaneous transitions in the output impedance as arrays of current-gain devices switch on and off; the transition can be smoothed with dynamic-bias circuits, but not entirely removed. You could switch arrays of batteries with relays, and the Zout measurement would be similar. The fact the output would consist entirely of square waves would not appear in the Zout measurement.
Direct-radiators are constant-acceleration devices. This means at higher frequencies, cone motion is very small, and low-level problems in spider and surround will appear as noise in the 2nd and 3rd harmonic distortion trace. Since it has low correlation with the input signal, it is not cyclic, and will tend to be rejected by the distortion analyzer, especially if the sweep speed is slow (averaging over many cycles).
I've worked in test instrumentation. It's a truism in the business that the instruments do not see what they not designed to look for. When you use a swept spectrum analyzer, you have to be aware the instrument is not designed to see single events, and will give very distorted-looking traces. FFT measurements have their own set of limitations; averaging is useful for extending the S/N ratio of the display, but it rejects non-correlated data as noise. That "noise" may not be noise; it can be granular, non-linear motion, with partial correlation to the input signal. If the test measurement rejects the low-level nonlinearity, that does not mean it is not there.
Rather than treat the loudspeaker as an idealized black box, we have to look at how each element actually performs, and see whether existing measurements capture the behavior or not.
Last edited:
Ran out of time on the previous post.
Loudspeakers are very complex electrodynamic transducers that radiate in 3 dimensions; the most common measurements only capture a small portion of what they are really doing.
Simple example: what does the impulse response look like in 3-space? You'd think something as basic as that would be common, but it's actually really hard to do. I've seen one paper published decades ago in the AES Journal, but very little since. And impulse response is about as fundamental as it gets: it's what every scope displays as voltage over time.
Now imagine that 3-D impulse display (probably looking like an expanding wavefront with lots of wrinkles in it) with nonlinear distortion somehow also displayed, maybe as ripples of color. Never mind capturing the data in 3-D, how do we display it? We're running out of dimensions.
The ear easily detects resonant energy tails; how do we display that in 3-D? Any takers?
These measurements only cover the most basic problems of loudspeakers; there are plenty of others. We can argue which are audible, and which aren't, but the departures from an ideal pulsating sphere are very gross, and not captured by the measurements you usually see in published reports.
Loudspeakers are very complex electrodynamic transducers that radiate in 3 dimensions; the most common measurements only capture a small portion of what they are really doing.
Simple example: what does the impulse response look like in 3-space? You'd think something as basic as that would be common, but it's actually really hard to do. I've seen one paper published decades ago in the AES Journal, but very little since. And impulse response is about as fundamental as it gets: it's what every scope displays as voltage over time.
Now imagine that 3-D impulse display (probably looking like an expanding wavefront with lots of wrinkles in it) with nonlinear distortion somehow also displayed, maybe as ripples of color. Never mind capturing the data in 3-D, how do we display it? We're running out of dimensions.
The ear easily detects resonant energy tails; how do we display that in 3-D? Any takers?
These measurements only cover the most basic problems of loudspeakers; there are plenty of others. We can argue which are audible, and which aren't, but the departures from an ideal pulsating sphere are very gross, and not captured by the measurements you usually see in published reports.
Last edited:
Very interesting thoughts, Lynn. I have noted another behaviour that supports the possible influence of stiction in the spider, corresponding to the lessening of its impact when the impregnated fabric is "conditioned" by vigorous movement -- normally one album is immediately followed by another, but if some reason a major delay occurs between the end of one recording and the the start of the next, then there is a very noticeable loss of quality, a reduction of the detail being reproduced when the music starts. This may in fact be due to the spider material quite rapidly reverting to its 'cold' behaviour ...
Last edited:
I really am suspicious of the spider, and to a lesser degree, the surround. The terrible part of a loudspeaker is you need at least 80 dB of linear dynamic range. The cone is well-behaved at low levels, and the magnetic system looks pretty good too (setting aside arcane issues like domain noise in ceramic magnets).
The long-lasting sticky goo on the Altec/GPA fabric surround helps a lot at, or the selection of high-loss, vibration-absorbing rubber seen in the best Skaaning drivers. I'm not so sure about foam surrounds, though ... I've seen not seen much evidence it's quiet at low levels.
The spider, though ... that's tough. It has to suppress the spider mode around 500 Hz, act as the primary restoring force on the cone and voice coil, keep the whole moving system from tipping off-axis for the life of the driver ... and oh yes, keep noise generation to a minimum. That's a lot of contradictory requirements. Usually, something has to give when you ask any material to do several things at once, and particularly if "noise generation" doesn't appear on any published measurement. If you were designing a pro driver intended to work all day at 100-watt levels, which would you throw out?
The long-lasting sticky goo on the Altec/GPA fabric surround helps a lot at, or the selection of high-loss, vibration-absorbing rubber seen in the best Skaaning drivers. I'm not so sure about foam surrounds, though ... I've seen not seen much evidence it's quiet at low levels.
The spider, though ... that's tough. It has to suppress the spider mode around 500 Hz, act as the primary restoring force on the cone and voice coil, keep the whole moving system from tipping off-axis for the life of the driver ... and oh yes, keep noise generation to a minimum. That's a lot of contradictory requirements. Usually, something has to give when you ask any material to do several things at once, and particularly if "noise generation" doesn't appear on any published measurement. If you were designing a pro driver intended to work all day at 100-watt levels, which would you throw out?
Last edited: