..possible, anything's possible.
-pretty sure it wasn't a freq. alteration due to impedance though, always seemed to be used with solid-state gear.
No, I think it goes deeper than that. The reviewers who are giving the glowing reviews are listening to the audiophile-flavored, low-efficiency TAD on 200-watt transistor amplifiers aimed at the audiophile market (and priced to match).
This has become a red flag for me; if the reviewer only listens to solid-state, their tonal preferences are almost certainly going to be different than mine. This basic preference affects all of their choices: what they describe as "accurate" (I've heard these reviewer's systems for myself) I hear as hard and metallic, a coloration quite unlike live, acoustic music. I've only heard three transistor amps that don't have that annoying metallic coloration, and two of them are not commercially available. (Gary Pimm's amp and the no-longer-manufactured LNPA.)
My yardstick is acoustic music; I really enjoy German and London techno dance music, but I wouldn't pretend it was suitable for making tonal assessments of a system. And I never listen to audiophile-oriented jazz or blues recordings - ever. That stuff drives me out of the room.
The top-of-the-line TAD I heard in a private demo (not at a show) sounded really dreadful. The RCA Living Presence recordings, which have a romantic warm, glowing sound (partly due to RCA ribbon mikes), sounded bright and shrill. That's just plain wrong. The Mercury Living Presence recordings, which don't forgive systems with HF problems, nearly took the top of my head off. Yet on my system with Ariels, Karnas, and the Monarchy DAC, the same recordings sound you-are-there in their liveness and sense of presence.
But ... an audiophile that has their system optimized for all-solid-state amplification would criticize these late-Fifties recordings as "old-fashioned" and "inaccurate", while the latest Diana Krall is, of course, "accurate". Except I don't listen to Diana Krall, or any of the other made-for-audiophile recordings you hear at shows.
If a system is only enjoyable on the latest audiophile recordings, something is terribly wrong.
Last edited:
Poor Diana Krall gets so much flack. But I think the engineering and production is brilliant. They have managed to make, package and sell a very specific sound -over and over. It takes a lot of work and manipulation to get that "sound". It's a sound that sells.
Her earliest recordings are not like that, they are more natural and make an interesting contrast to the current sound.
As for the TAD stuff, I can't really judge - it's been too long. I just remember the 15" woofers and the compression drivers sounding pretty good in custom systems. I've heard GOTO sound awful, too. But it was not the drivers, it was the crossover.
There is certainly a widespread taste for a rather hash and brittle sound. It sounds fake to me. As you say, un-ampified music just doesn't sound like that.
But I think the engineering and production is brilliant. They have managed to make, package and sell a very specific sound -over and over. It takes a lot of work and manipulation to get that "sound". It's a sound that sells.Her earliest recordings are not like that, they are more natural and make an interesting contrast to the current sound.
As for the TAD stuff, I can't really judge - it's been too long. I just remember the 15" woofers and the compression drivers sounding pretty good in custom systems. I've heard GOTO sound awful, too. But it was not the drivers, it was the crossover.
There is certainly a widespread taste for a rather hash and brittle sound. It sounds fake to me. As you say, un-ampified music just doesn't sound like that.
That's right. The current line has no apparent connection to the previous line of products except price and beautiful cabinetry. The sound is completely different; I understand the current designer used to work for KEF and was part of the Uni-Q design team. Think of a Uni-Q at TAD price points.
Here's one of Bjorn Kolbrek's simulations comparing the power factor of a T=0.707 LeCleac'h horn that is exit-angle-matched to the Altec 288 (8 degrees) compared to a conical horn. (Both horns are simulated in free air, not mounted in a cabinet.)
The thin black trace shows the resistance term seen by the 288 driver with the LeCleac'h horn, and the thin red trace shows the reactance (or reflection) of the LeCleac'h horn. Like an antenna at RF frequencies, the goal is to get signal out of the antenna and radiated into free space, not reflected back to the transmitter. For an RF system, the wasted power heats up the power tube and could potentially damage it, and for an audio system, the reflected power bounces off the diaphragm and phase plug and re-emerges 2 milliseconds or so later.
Reflections in horns are quite undesirable, since they result in significant reflections in the time domain. Since they are reflections and not resonances, conventional frequency-domain equalization does not improve the time-domain response - it makes it worse.
The heavy black trace shows the resistance term seen by the 288 driver with a conical horn, and the heavy red trace shows the reactance (or reflection) of a conical horn. (The simulated conical uses a smoothed Quadratic Throat transition region between the compression driver and the horn expansion.)
Note much poorer diaphragm loading in the critical 1 to 5 kHz region, despite the conical being the same size as the LeCleac'h horn. More significantly, the reflections are much stronger as well. The LeCleac'h has reflected energy down to 10% at 700 Hz, while the conical only drops into the 10~20% region at frequencies above 5 kHz. The "lower efficiency" of the conical in the 1 to 5 kHz region is actually energy bouncing around in the horn. Sure, you can solve the energy-loss problem with more input power, but what does that do for all that reflected energy bouncing around inside?
Why am I not a fan of conical and near-conical horns? This is why. The combination of Newell and Holland's direct measurements (shown in previous posts) combined with Bjorn's simulations, showed the problems with uniform diaphragm loading are quite severe, and not responsive to equalization. Yes, the gross FR aberrations can be corrected with notch filters, and/or active EQ. But that has no effect on a reflection series, and coloration in the 1 to 5 kHz range is the worst possible spectral place for problems to happen in a loudspeaker. If you've wondered why commercial constant-directivity loudspeakers used in PA systems and theaters have an unpleasant fatiguing quality in the midrange, this is the reason.
The thin black trace shows the resistance term seen by the 288 driver with the LeCleac'h horn, and the thin red trace shows the reactance (or reflection) of the LeCleac'h horn. Like an antenna at RF frequencies, the goal is to get signal out of the antenna and radiated into free space, not reflected back to the transmitter. For an RF system, the wasted power heats up the power tube and could potentially damage it, and for an audio system, the reflected power bounces off the diaphragm and phase plug and re-emerges 2 milliseconds or so later.
Reflections in horns are quite undesirable, since they result in significant reflections in the time domain. Since they are reflections and not resonances, conventional frequency-domain equalization does not improve the time-domain response - it makes it worse.
The heavy black trace shows the resistance term seen by the 288 driver with a conical horn, and the heavy red trace shows the reactance (or reflection) of a conical horn. (The simulated conical uses a smoothed Quadratic Throat transition region between the compression driver and the horn expansion.)
Note much poorer diaphragm loading in the critical 1 to 5 kHz region, despite the conical being the same size as the LeCleac'h horn. More significantly, the reflections are much stronger as well. The LeCleac'h has reflected energy down to 10% at 700 Hz, while the conical only drops into the 10~20% region at frequencies above 5 kHz. The "lower efficiency" of the conical in the 1 to 5 kHz region is actually energy bouncing around in the horn. Sure, you can solve the energy-loss problem with more input power, but what does that do for all that reflected energy bouncing around inside?
Why am I not a fan of conical and near-conical horns? This is why. The combination of Newell and Holland's direct measurements (shown in previous posts) combined with Bjorn's simulations, showed the problems with uniform diaphragm loading are quite severe, and not responsive to equalization. Yes, the gross FR aberrations can be corrected with notch filters, and/or active EQ. But that has no effect on a reflection series, and coloration in the 1 to 5 kHz range is the worst possible spectral place for problems to happen in a loudspeaker. If you've wondered why commercial constant-directivity loudspeakers used in PA systems and theaters have an unpleasant fatiguing quality in the midrange, this is the reason.
Attachments
Last edited:
Sorry about that, guys. The referenced material in the deleted posts is on pages 108 through 113 of Newell and Holland's book, which describes the design criteria for the AX2 horn, which I used to guide the design for the AH425 horn, with Bjorn Kolbrek's substantial assistance. I very highly recommend buying the book if you are serious about horn design, or are curious about any aspect of high-efficiency loudspeaker design. The Newell and Holland book is not aimed at the Loudspeaker Design Cookbook level of reader, so if you're just getting started, stick to Vance Dickason's book. For rest of you, get it now; get the research data directly from the authors. It costs less than a single driver, and way less than a failed project.
The point made by the two authors is that horn length should be limited to 12 inches or less, so that any remaining reflections fell below perceptual thresholds, and that there is a design tradeoff between smoothness of diaphragm loading (a resistive load over the entire working range) and directivity control. In other words, directivity control by means of throat diffraction inevitably creates reflections, which then appear in several other domains - the ratio of resistance versus reactance over frequency (as shown above and in Holland and Newell's detailed measurements), impedance variations, loss of efficiency over part of the working range, and most importantly, time-domain reflections that are not correctable by standard methods of equalization.
All things I want to avoid. There are other ways to address directivity, but reflections, particularly in the critical 1 to 5 kHz range, are very undesirable and very hard to remove after-the-fact with electronics.
The point made by the two authors is that horn length should be limited to 12 inches or less, so that any remaining reflections fell below perceptual thresholds, and that there is a design tradeoff between smoothness of diaphragm loading (a resistive load over the entire working range) and directivity control. In other words, directivity control by means of throat diffraction inevitably creates reflections, which then appear in several other domains - the ratio of resistance versus reactance over frequency (as shown above and in Holland and Newell's detailed measurements), impedance variations, loss of efficiency over part of the working range, and most importantly, time-domain reflections that are not correctable by standard methods of equalization.
All things I want to avoid. There are other ways to address directivity, but reflections, particularly in the critical 1 to 5 kHz range, are very undesirable and very hard to remove after-the-fact with electronics.
Last edited:
For your interest, I have put up some of my BEM simulations on the web: BEM
There will be more coming up soon, as I have recently done some simulations of the effect of mouth rollback in LeCléac'h horns.
To explain a little more about the curves Lynn put up a few posts back: the black curves show the power factor of the horn, which is the cosine of the phase angle of the throat impedance. It directly tells you at what frequencies the horn will transmit power efficiently, as P = |p|*|U|*cos(theta), P being power, p pressure, U volume velocity, and cos(theta) the power factor. The red curves show the reflection coefficient, a reflection coefficient of one is a total reflection, zero is perfect transmission. It is calculated based on the impedance mismatch between a source with an impedance of rho*c/St (the asymptotic throat impedance) and the actual throat impedance of the horn.
Regards,
Bjørn
There will be more coming up soon, as I have recently done some simulations of the effect of mouth rollback in LeCléac'h horns.
To explain a little more about the curves Lynn put up a few posts back: the black curves show the power factor of the horn, which is the cosine of the phase angle of the throat impedance. It directly tells you at what frequencies the horn will transmit power efficiently, as P = |p|*|U|*cos(theta), P being power, p pressure, U volume velocity, and cos(theta) the power factor. The red curves show the reflection coefficient, a reflection coefficient of one is a total reflection, zero is perfect transmission. It is calculated based on the impedance mismatch between a source with an impedance of rho*c/St (the asymptotic throat impedance) and the actual throat impedance of the horn.
Regards,
Bjørn
A well designed waveguide has no internal resonances. In my designs there are no ripples in the Compression Drivers impednace curves and none in the pass band of the radiated response. Yes reflections are a very bad thing, but length of the horn has nothing to do with the problem. Its all about good design - the mouth being the critical part in this regard (assuming no internal diffraction).
"time-domain reflections" are correctable in both time, frequency and space by EQ as long as they are single mode reflections. The higher order modes that can result from reflections internal to the device are NOT correctable by electronic means however. This is a fundamental difference in the two types of modes.
I want to thank Bjorn Kolbrek for all the hard work he did on the AH425 project - without him, the AH425 would not have been possible.
I also want to thank him for publishing the BEM simulations on the AH425+288 in free air, the AH425+288 on a flat baffle, a Tractrix horn, a conical horn, an exponential horn, an OSWG, and an OSWG combined with the DE250 compression driver. This is the first time this data has been published in a public forum.
I'd like to draw the reader's attention to the differences in throat impedances of the different types of horn. The conical is the worst by far, with throat diffraction, combined with mouth reflections, as the culprit. The AH425 is one of the smoother profiles, although it's entirely possible a smoother profile could be designed.
The Newell and Holland book guided me towards the LeCleac'h profile, Bjorn Kolbrek, and Martin Seddon. Since the AX2 horn is not for sale, and the book does not describe the profile, all that was available were the Newell and Holland overall design guidelines. I agreed with the authors about the sonic virtues of the horn in the Tannoy 15" coax driver, but Tannoy does not sell their drivers to OEMs or DIYers.
I do not give great weight to directivity control. I know that Dr. Floyd Toole and Dr. Geddes place directivity control high on their list of priorities, but I must respectfully disagree with them, while agreeing on many other points. I very strongly agree about the importance of uniform and flat response at the listening position, where many high-end loudspeakers (even at absurd $100,000 price points) fall down pretty badly.
I'm in the Brit-design camp on the importance of a quick and clean time-decay characteristics, which is a traditional priority of BBC studio monitors, Laurie Fincham when he was at KEF, and Newell and Dr. Holland. This is a fairly unusual priority for a high-efficiency loudspeaker, where the primary concerns are reliable power-handling and precise directivity control (for theater and SR applications). The pro world uses digital equalization as a matter of course, so flat response is not of great interest.
My primary reason for designing the LTO Apollo (the working name for now) was a high-quality loudspeaker suitable for vacuum-tube amplifiers in the 6 to 60-watt range. There are many, many loudspeakers suitable for transistor amplifiers, but far fewer that are a good match for the modest powers of direct-heated-triode and pentode amplifiers. Even if you have a ridiculous budget of $60,000 a pair, there still aren't a lot of good choices if you want a good-sounding loudspeaker with good objective measurements that is 97 dB/meter/watt or higher efficiency - which is what low to moderate power amplifiers need.
I also want to thank him for publishing the BEM simulations on the AH425+288 in free air, the AH425+288 on a flat baffle, a Tractrix horn, a conical horn, an exponential horn, an OSWG, and an OSWG combined with the DE250 compression driver. This is the first time this data has been published in a public forum.
I'd like to draw the reader's attention to the differences in throat impedances of the different types of horn. The conical is the worst by far, with throat diffraction, combined with mouth reflections, as the culprit. The AH425 is one of the smoother profiles, although it's entirely possible a smoother profile could be designed.
The Newell and Holland book guided me towards the LeCleac'h profile, Bjorn Kolbrek, and Martin Seddon. Since the AX2 horn is not for sale, and the book does not describe the profile, all that was available were the Newell and Holland overall design guidelines. I agreed with the authors about the sonic virtues of the horn in the Tannoy 15" coax driver, but Tannoy does not sell their drivers to OEMs or DIYers.
I do not give great weight to directivity control. I know that Dr. Floyd Toole and Dr. Geddes place directivity control high on their list of priorities, but I must respectfully disagree with them, while agreeing on many other points. I very strongly agree about the importance of uniform and flat response at the listening position, where many high-end loudspeakers (even at absurd $100,000 price points) fall down pretty badly.
I'm in the Brit-design camp on the importance of a quick and clean time-decay characteristics, which is a traditional priority of BBC studio monitors, Laurie Fincham when he was at KEF, and Newell and Dr. Holland. This is a fairly unusual priority for a high-efficiency loudspeaker, where the primary concerns are reliable power-handling and precise directivity control (for theater and SR applications). The pro world uses digital equalization as a matter of course, so flat response is not of great interest.
My primary reason for designing the LTO Apollo (the working name for now) was a high-quality loudspeaker suitable for vacuum-tube amplifiers in the 6 to 60-watt range. There are many, many loudspeakers suitable for transistor amplifiers, but far fewer that are a good match for the modest powers of direct-heated-triode and pentode amplifiers. Even if you have a ridiculous budget of $60,000 a pair, there still aren't a lot of good choices if you want a good-sounding loudspeaker with good objective measurements that is 97 dB/meter/watt or higher efficiency - which is what low to moderate power amplifiers need.
Last edited:
It's your thread so that is your perogative. Lots of people still disagree with Darwin.
..apparently respectful disagreement only goes one-way.
Hmm, seems to me that at one time Earl had real problem with this from *others*. Still, it seemed to be a light-hearted "rip" (with an appropriate emoticon). Hmm.. what's more important?
I could probably argue for dispersion over that of "flat on axis", I mean on the one hand you can have flat on-axis with an extremely narrow dispersion sound - and if narrow enough the sound "appears" to be at the ear it's focused on (..like someone is whispering/shouting "in" your ear). On the other hand most everyone has the ability to "auto-correct" or "equalize" the response they hear as a learning process - so a truly flat amplitude response isn't really required.
Still, I'd have to say that within the context of normal loudspeakers and design - the highest priority is in fact flat-on axis at the listener position, or something approaching that. Of course "on-axis" is itself a bit of a misnomer because it's really several degrees for direct sound alone.. so perhaps flat on +/- 5 degrees.
I design loudspeakers for my own satisfaction, not anybody else. I'm 63 years old, retired with a moderate independent income, and don't have to answer to a marketing department, a manager, or a university department head. When I left Audionics in 1979, I made a resolution to myself to never, ever work for anyone else in the audio industry. Collaborate, sure, absolutely; I always try to give generous credit to my collaborators. Working for Tektronix taught me that small teams do the best work, because they can cross-check each other, in a friendly, collegial way. Which is pretty different than most of the high-end industry, which populated by solo designers working alone (most high-end companies have much larger marketing departments than engineers - many companies have no on-staff engineer at all, hiring out the work to a consultant).
My favorite loudspeakers are of the old Brit school: Quad ESL57, the full-size BBC monitors, and a few others. None of them are controlled directivity designs. The Altec interest started when I heard an Altec A5 system at the San Francisco Audio Club a few years back - I had to make an emergency repair on the very primitive 12 dB/octave stock crossover (which was made with the cheapest parts imaginable), but heard what I call "hints of greatness". This isn't meant to be dismissive; it means an interesting but flawed loudspeaker that still does some things better than anything else. Vividness of tone color is what stands out from that presentation, somewhat akin to the sound of the Ortofon SPU cartridge, or a Decca that actually works.
I'm not that interested in controlled directivity loudspeakers because I've heard a number of them, from different designers, and it's not a sound that engages me. I have a minority taste in audio: I don't care for solid-state electronics, don't like the sound of most delta-sigma DACs, and don't like what most magazine reviewers like. I'm not interested in what a listening panel thinks. I care about what I hear, and how to go about getting that sound.
Since I have no interest in what a listening panel thinks, or what most audiophiles think, I go my own way. I always warn people curious about building the Ariel, Amity, or Karna that they don't sound like what's on the market commercially, and in fact that's why they exist - I couldn't buy what I wanted at any price, so I design my own. From my limited experience, people hear things in different ways - there seems to a strong cultural component to perception, hearing, and emotional response to music. My own cultural experience of growing up in Japan and Hong Kong, with parents in the US Foreign Service, gave me a cultural experience very different than most Americans - there are still many things about the US that I find kind of baffling.
This makes me a cultural relativist, I guess, and not willing to say that individual A is hearing X instead of Y. I'm not them. I don't really know what they hear, nor what they like. I've met audiophiles who busily filled out a 20-point checklist as they evaluated a system. I was kind of boggled to see them do this, and more astonished to find that's how they evaluated all audio systems. Naturally, they asked me what I do - seeing that I didn't have any kind of checklist.
The answer, which surprised the questioner, is that I either like a system, don't like it, or am indifferent. Just like my reactions to the food at a restaurant, or seeing a pretty girl cross the street. Now sometimes I'm curious why something sounds a certain way, just like I'm curious how a really good meal was prepared.
Which takes us back to the preference for the Quad ESL57 and large BBC monitors. These are systems optimized for very quick, clean decays in time, as well as flat response at the listening position, free of narrowband resonances. They sound that way for a reason.
The "hints of greatness" in the Altec A5? Well, the large-format 288 on Altec's best horn, the big multicell, gets a lot of credit, along with the 416 woofer, which has an unusual old-school construction of an underhung voice coil, a light cone with a very effective edge-damping system (which provides freedom from breakup above 2 kHz, unlike most other 15" drivers), and an Alnico magnet. The crossover? Uh, no. The thin-walled plywood box? Uh, no. Those are problems, not assets, but you can't blame Altec for that. Filter design was in a very primitive state when the A5 was designed in the late Forties, and not much was known about box damping.
The weaknesses of the Quad ESL57 and BBC monitors? Dynamics - or rather, lack of them. They are the opposite of the Altecs in that respect. Throw a modern 200-watt transistor amplifier at them, and all you do is destroy a classic in a matter of seconds. Not a nice thing to do.
It has to be possible to design a high-dynamic-range system with good time response (freedom from resonance, reflection, and diffraction). Maybe not as good as electrostats, but possibly as good as low-to-moderate efficiency direct-radiators. The Ariels were designed in 1993, and were a good first attempt at this goal. I'm aiming a little higher this time, and have needed the help of Bjorn Kolbrek, Martin Seddon, the good folks at Great Plains Audio, and many others. I've heard the first prototypes, and it seems to be going in the direction I intended when I started the project. Measurements and simulations are promising so far, and 3rd to 4th-order crossovers are giving good subjective results.
My favorite loudspeakers are of the old Brit school: Quad ESL57, the full-size BBC monitors, and a few others. None of them are controlled directivity designs. The Altec interest started when I heard an Altec A5 system at the San Francisco Audio Club a few years back - I had to make an emergency repair on the very primitive 12 dB/octave stock crossover (which was made with the cheapest parts imaginable), but heard what I call "hints of greatness". This isn't meant to be dismissive; it means an interesting but flawed loudspeaker that still does some things better than anything else. Vividness of tone color is what stands out from that presentation, somewhat akin to the sound of the Ortofon SPU cartridge, or a Decca that actually works.
I'm not that interested in controlled directivity loudspeakers because I've heard a number of them, from different designers, and it's not a sound that engages me. I have a minority taste in audio: I don't care for solid-state electronics, don't like the sound of most delta-sigma DACs, and don't like what most magazine reviewers like. I'm not interested in what a listening panel thinks. I care about what I hear, and how to go about getting that sound.
Since I have no interest in what a listening panel thinks, or what most audiophiles think, I go my own way. I always warn people curious about building the Ariel, Amity, or Karna that they don't sound like what's on the market commercially, and in fact that's why they exist - I couldn't buy what I wanted at any price, so I design my own. From my limited experience, people hear things in different ways - there seems to a strong cultural component to perception, hearing, and emotional response to music. My own cultural experience of growing up in Japan and Hong Kong, with parents in the US Foreign Service, gave me a cultural experience very different than most Americans - there are still many things about the US that I find kind of baffling.
This makes me a cultural relativist, I guess, and not willing to say that individual A is hearing X instead of Y. I'm not them. I don't really know what they hear, nor what they like. I've met audiophiles who busily filled out a 20-point checklist as they evaluated a system. I was kind of boggled to see them do this, and more astonished to find that's how they evaluated all audio systems. Naturally, they asked me what I do - seeing that I didn't have any kind of checklist.
The answer, which surprised the questioner, is that I either like a system, don't like it, or am indifferent. Just like my reactions to the food at a restaurant, or seeing a pretty girl cross the street. Now sometimes I'm curious why something sounds a certain way, just like I'm curious how a really good meal was prepared.
Which takes us back to the preference for the Quad ESL57 and large BBC monitors. These are systems optimized for very quick, clean decays in time, as well as flat response at the listening position, free of narrowband resonances. They sound that way for a reason.
The "hints of greatness" in the Altec A5? Well, the large-format 288 on Altec's best horn, the big multicell, gets a lot of credit, along with the 416 woofer, which has an unusual old-school construction of an underhung voice coil, a light cone with a very effective edge-damping system (which provides freedom from breakup above 2 kHz, unlike most other 15" drivers), and an Alnico magnet. The crossover? Uh, no. The thin-walled plywood box? Uh, no. Those are problems, not assets, but you can't blame Altec for that. Filter design was in a very primitive state when the A5 was designed in the late Forties, and not much was known about box damping.
The weaknesses of the Quad ESL57 and BBC monitors? Dynamics - or rather, lack of them. They are the opposite of the Altecs in that respect. Throw a modern 200-watt transistor amplifier at them, and all you do is destroy a classic in a matter of seconds. Not a nice thing to do.
It has to be possible to design a high-dynamic-range system with good time response (freedom from resonance, reflection, and diffraction). Maybe not as good as electrostats, but possibly as good as low-to-moderate efficiency direct-radiators. The Ariels were designed in 1993, and were a good first attempt at this goal. I'm aiming a little higher this time, and have needed the help of Bjorn Kolbrek, Martin Seddon, the good folks at Great Plains Audio, and many others. I've heard the first prototypes, and it seems to be going in the direction I intended when I started the project. Measurements and simulations are promising so far, and 3rd to 4th-order crossovers are giving good subjective results.
Last edited: