Beyond the Ariel

Lynn Olson said:



I surmise not all readers of this thread are aware that frequency measurements contain less information than impulse response - you can create frequency response, phase response, ETC decay, CSD waterfalls, and lots of other things from the impulse response, but the reverse is not true, since phase or time data is not present in a FR response plot.


I would not surmise that at all. I would expect that most people who frequent this or other DIY groups are more than aware that FR means both amplitude and phase. FR should be express as a phasor,
M(f) exp(jPhi(f)). I don't think there are many people who are thinking only amplitude when they think FR. As you know, there is a one to one correspondence between the time domain and the phasor frequency domain for a linear system. (Nonlinear effects THD, etc, in a loudspeaker measured at moderate level will have very minor effects on the FR obtained from the impulse, or the impulse obtained from the FR. ) Given the proper specification of the FR or impulse you can pretty much go any where you like in terms of post processing, except to extract nonlinear effects. Most DIY'ers that I know understand this.


One simple example are square and triangle waves, which have identical spectra in the frequency domain, with a diminishing series of 3rd, 5th, 7th and the rest of the odd harmonics. Although the magnitudes of the harmonics are precisely the same, the phases are different, making for a very different waveform.

I believe you are in error here Lynn. The square wave has amplitude that goes 1/n, where n indicates the harmonic, n odd. The triangle has amplitude that goes (1/n)^2



Allpass functions are invisible in the frequency domain as well -

Again, I guess you are only considering amplitude response.
 
john k... said:

I would not surmise that at all. I would expect that most people who frequent this or other DIY groups are more than aware that FR means both amplitude and phase. FR should be express as a phasor, M(f) exp(jPhi(f)).

I don't think there are many people who are thinking only amplitude when they think FR. As you know, there is a one to one correspondence between the time domain and the phasor frequency domain for a linear system. Given the proper specification of the FR or impulse you can pretty much go any where you like in terms of post processing, except to extract nonlinear effects. Most DIY'ers that I know understand this.


Yes, a complete FR - including phase - allows computation of other terms, including impulse response. But how often is the complete data set published?

When I look at FR curves going back the Fifties, where's the phase? No way can you extract phase from an old-school swept-oscillator spectrum analyzer (like the Tektronix 7L5), a swept-sinewave graph made on B&K equipment, or a Seventies-vintage 1/3 octave realtime analyzer measurement. The time information was never collected, and cannot be extrapolated from a magnitude-only graph. This was the point I was trying to make.

Going further, it's now 2007, but when you look at specs for most prosound equipment - horns especially - where's the phase or time information? Not there, sorry. The majority of published information is even worse - not only no time data, but the FR response (magnitude only) is smoothed as well, making visualization of resonances very difficult.

All I'm asking for is better, and more useful, published specs. I've been pestering driver manufacturers for any kind of time information for more than 30 years now, and most of the time, all I get is a runaround and lot of hand-waving and excuses why it's "not possible". I usually have to surreptitiously find out who the engineer is and wheedle the data out of him. Surely, after all this time, driver and speaker manufacturers should be publishing FR and impulse data, instead of excuses. Is that too much to ask?

P. S. Actually, I'm a little envious of the higher-class crowd that JohnK gets. The people I meet think they're pretty sharp if they can operate and understand a 1/3 octave RTA, and any discussion of impulse response gets lost in the swamps of "boy, that Altec A7 sure has good transient response". If you ask about the measurements, you get shown a picture of glowing RTA lights or a faded photocopy of a B&K pen graph made in 1952. Kind of hard to calculate the impulse response from that, isn't it? And Altec A4's and A7's ain't exactly portable.
 
Here's the latest thoughts about the open baffle - will construct the baffle, and measure and audition drivers after the RMAF. Curious to hear what the various EnABL'ed drivers (including Lowther) can do, and also curious if the elusive Fertin widerange field-coil driver will make an appearance.
 

Attachments

  • dipole_12+15.gif
    dipole_12+15.gif
    54.4 KB · Views: 985
Lynn Olson said:


Yes, a complete FR - including phase - allows computation of other terms, including impulse response. But how often is the complete data set published?




True, but who builds speakers based on published specks today? ;) Yes, it is 2007. We are way past the 50's. But published specs are missing a lot more than phase, like baffle and box (if used) effects. Isn't that why all these DIY CAD tools generally have at least some kind of rudimentary measurement tool packaged with them? At best published specs are useful only for identifying the general characteristics of the driver; smoothness through the operating band, breakup behavior, etc. What good would the most detailed specs in the world be if they aren’t for your configuration?

As for extracting phase form amplitude data, at least of a single driver, minimum phase reconstruction work very well if you have good amplitude data. Maybe pro sound driver manufactures don't supply sufficiently accurate amplitude data, most reprtable hifi driver manufactures do (SEAS, Vifa Peerless, ScanSpeak, etc). While I haven't looked at large panel (like Martin Logan ELS) of horns, I have yet to measure a conventional driver, cone, dome or small ribbon, that doesn't reduce to minimum phase on axis over its useful operating range. So what that leaves is what is the driver offset, a problem which is still of concern using impulse response techniques. Of course, I wouldn't consider 1/3 octave RTA a legitimate measurement of a driver's anechoic response.

I can't comment on the crowd you hang with but most of the guys I know are just average Joes building speakers as a hobby.

Altec A4's and A7's may not be very portable, but a lap top with mic, running any number of cheap (or free) software packages, certainly is.

I can certainly understand that you (or I) don't want to buy and test every driver you may contemplate using in a system, but doesn't it really have to come down to that? No matter how detailed the specs are, they won't tell you how a driver sounds.
 
Administrator
Joined 2004
Paid Member
Re: Other drivers worth considering?

mikey_audiogeek said:
have you considered Triangle drivers? Bit low on efficiency, but otherwise...

Strange.... I thought they were not selling raw drivers these days. Looks like they are.

The 1st speakers I ever built used the very first Triangle driver. I liked them a lot. The neighbors always asked me "who were the musicians you had over yesterday?" or "hey, you're getting really good on that sax." Yeah, just like Paul Desmond. ;)

Certainly worth a look/listen. My 2 cents.
 
Lynn,
looking at your OB proposal:
How are you going to deal with the energy response step at the crossover between WR driver and tweeter? Up to 1.6 kHz the WR driver will radiate with (hopefully) almost equal power to the front and rear, while the tweeter will only radiate to the front. So you will need some controlled rear wave attenuation over the passband of the WR driver or a second tweeter IMHO.
A "felt tent" behind the driver sounds too "lightweight" for the upcoming task.

In my own applications I never got that right, and I always wondered how the Orion could do without a second tweeter or a rear wave treatment of the midrange.
 
There was a discussion about impulse responses. I've collected three probably interesting responses: Summa ESP15, Manger 109 measured at Gelsenkirchen HiFi exhibition on axis and from 3m, and my Jordan JXR6 (0.15-30kHz) in steel enclosure with a lot of damping (0,5l free air and 1,5l wool) and at 0deg, 30cm. Manger is meant to be perfectly aperiodic, Jordan - well controlled modes for aluminium diaphragm.
 
Rudolf said:
Lynn,
looking at your OB proposal:

How are you going to deal with the energy response step at the crossover between WR driver and tweeter? Up to 1.6 kHz the WR driver will radiate with (hopefully) almost equal power to the front and rear, while the tweeter will only radiate to the front. So you will need some controlled rear wave attenuation over the passband of the WR driver or a second tweeter IMHO.

A "felt tent" behind the driver sounds too "lightweight" for the upcoming task.

In my own applications I never got that right, and I always wondered how the Orion could do without a second tweeter or a rear wave treatment of the midrange.

Plan to try both approaches, maybe at the same time. At 1.6 kHz, wool felt (and other damping materials) are starting work pretty well. Gary Pimm is using Bonded Logic "Ultratouch" fiber to fill his open-backed dipole woofers with great success - this recycled cotton fiber has high absorptive properties, and Pimm's system has the best, most crisp, and tonally realistic bass I've heard so far.

I'll be using the Ultratouch fiber for the open-backed Bass chamber, and possibly be trying it in small, custom-made pillows behind the other drivers - white pillows would look more than a little silly, but it's easy enough to dye them black before filling them with the Ultratouch fiber.

The whole trick with damping is keeping the fibers several inches away from the cone - this is important - and using organic-origin fibers as opposed to synthetics. I tried lots of different fillings for the Ariel, and was not at all happy with the sound of foam, fiberglass, or polyfill. Wool and cotton sounded the best.

If anything, I expect a dipole is going to be even less tolerant of odd-sounding damping materials - these can have rather weird dynamic-related colorations. If you select a damping material with suboptimum properties, you will hear a noticeable muffled or dynamically constrained quality to the sound - always compare against no damping when evaluating the sonics of damping materials. Compare speaker A against speaker B for quick comparisons - it won't so much be a tonal coloration, but more of a flattened, murky, "closed-in" sound you want to listen for. This is NOT accuracy, it's an unwanted coloration.

It's entirely possible I won't find any damping material acceptable in the midrange - that's a real possibility that affects the rest of the design.

The vertical directivity of the ribbon tweeter will bear on the decision for or against a rear tweeter - as you might expect, it's about subjective decisions of spaciousness and smooth integration with the large-area midbass driver. I also plan to have a wooden disk to fit in the 12" hole that has an 8" cutout for the 8" drivers that look interesting (Lowther, Fertin, et al).

I feel modern designs have gone astray in focussing on dispersion vs frequency compared to the basic sound of the drivers themselves. Considering how grossly colored I find contemporary "audiophile" drivers - I keep being surprised how people are able to ignore such basic colorations while looking at all those pretty-looking polar graphs. I have to be direct here, folks - I really dislike the sound and philosophy of modern high-end audio, sorry. Please look elsewhere if you find the sound of modern high-end even a little bit palatable.

I should warn readers, like I did for the Ariels, that my designs do not follow contemporary trends in speaker design. I optimize for natural sound on solo voice and choirs, followed by naturalistic qualities on other instruments (symphonic), followed by natural spatial qualities (not imaging per se) and realistic dynamics - in about that order. I make no claims for the "best" or "ultimate", but choose to address problems I see ignored in other approaches.

Everything I do is off the beaten path - look at how I design amplifiers. If all of this is just too weird for you, please, look in another direction! There are lots of talented folks here in diyAudio - you have many choices.
 
Re: Other drivers worth considering?

mikey_audiogeek said:
Hi Lynn, have you considered Triangle drivers? Bit low on efficiency, but otherwise...

Triangle drivers

Best,
Mike

Um, I listened the Triangle speakers, and they didn't leave a strong impression. But they're hardly alone - the current fad for audiophile qualities like "fast" and "slam" has led not to lower distortion and more efficiency (which sure would have been nice), but drivers with poorer self-damping, only slightly more efficiency, and ever-more-harsh sound with very rough response at the upper edge of the band.

It seems that modern designers are reacting to these problematic drivers in two ways: the "minimalists" are going for lots of excitement and thrills, using the rock-bottom simplest crossovers possible (but with very expensive parts), and letting the peaks sail right through. Unsophisticated listeners - and worse, reviewers - interpret the peaks and harshness as "speed" and "accuracy".

There's a lower-profile school that believes in extensive computer simulation and using crossovers of almost unlimited complexity. This "objective" school of designers tend to discount esoterica like audibility of capacitor coloration - or even believe it doesn't exist - so has no problem with complex op-amp circuits, multiple transistor amps, or high-parts-count crossovers with extensive notch filters and shaping networks.

I'm not in either school. I don't want to use drivers with problematic responses - too much work for too little return. I still remember the bad old days of KEF and Audax Bextrene drivers, with their characteristic qualities of lumpy midrange, and dreadfully low efficiencies (85 dB/metre typical). Now, audiophile efficiencies have crept up to 90~93 dB/metre (with a tailwind), but the drivers have gotten really peaky, and in ways that are very hard to correct - the worst peaks are typically directional, making a crossover correction useless.
 
john k... said:

As for extracting phase form amplitude data, at least of a single driver, minimum phase reconstruction work very well if you have good amplitude data. Maybe pro sound driver manufactures don't supply sufficiently accurate amplitude data, most reprtable hifi driver manufactures do (SEAS, Vifa Peerless, ScanSpeak, etc). While I haven't looked at large panel (like Martin Logan ELS) of horns, I have yet to measure a conventional driver, cone, dome or small ribbon, that doesn't reduce to minimum phase on axis over its useful operating range. So what that leaves is what is the driver offset, a problem which is still of concern using impulse response techniques. Of course, I wouldn't consider 1/3 octave RTA a legitimate measurement of a driver's anechoic response.

I think one difference between us is that I'm no longer interested in SEAS, Vifa Peerless, ScanSpeak, etc. They've had 15 years to respond to the vacuum-tube subculture and the constantly-expressed demand for substantially higher efficiency, and have done their best to ignore it and hope it would go away. Well, it hasn't, there's only been a trivial 1~2 dB change in efficiency in more than a decade, and to me, the mainstream audiophile drivers sound worse than what they were making 15 years ago. That's why I'm looking at different vendors than the usual mainstream candidates.

I'm surprised that you haven't seen drivers depart from minimum phase. This is one of the most direct indicators of cone breakup, and it's gotten much worse with the popularity of very rigid Kevlar, carbon-fiber, composite, ceramic, and metal cones. When a cone no longer moves as unit and enters the breakup region, there are multiple, asynchronous centers of radiation all over the cone. This is a clear indication of a "no-go" zone, and indirectly shows a requirement for an aggressive high-slope crossover to avoid gross coloration.

The drivers that are most interesting - to me - are the ones that don't require aggressive equalization to avoid harsh sound, and are characterized by smooth, well-controlled rolloff regions that retain their minimum-phase character to very high frequencies. Since the prosound manufacturers can't be bothered to supply either impulse or complete FR/phase data information, I'll be finding this out the hard ($$$) way - I'm not expecting any free loans from 18Sound, JBL, or Fertin.
 
Lynn Olson said:
... At 1.6 kHz, wool felt (and other damping materials) are starting work pretty well. Gary Pimm is using Bonded Logic "Ultratouch" fiber to fill his open-backed dipole woofers with great success - this recycled cotton fiber has high absorptive properties, and Pimm's system has the best, most crisp, and tonally realistic bass I've heard so far.

Ultratouch seems to have a very linear absorption capacity over a wide frequency range. I would rather imagine absorption rising with frequency, so that at 1.6 kHz the WR would be "dead" in the back. Something like 41 mm thick Basotect might achieve this:

An externally hosted image should be here but it was not working when we last tested it.
 
jzagaja said:
soongsc

Do you know any other drivers than Jordan/Bandor/Goerlich that shares similar concept of using variable thickness and stiffness? For example with all-PP composites it is possible to vary local mechanical properties in in a wide range.
Above 300Hz or so, I have not come across any as good. Bear in mind that certain material characteristics are just not documented, unless one has dedication as Ted did, it is very difficult to find the right formula without massive trial and error. The one I'm tweaking does better in the lower frequencies, but is not as good in the higher frequencies. The region from 15KHz~30KHz is the most difficult region to tweak in small metal drivers and very critical. I remember that Ted mentioned it just extented to 30KHz by chance, I wish I had that kind of luck. I understand that some people are trying some alloy with lithium that seeminly has better damping characteristics, but have not seen any on the market yet.
 
jzagaja said:
soongsc

So above 10kHz it is not possible to use polypropylene? (5-6"). It is 7 times less stiff than aluminium (oriented).

I'm wondering about Jordan's 30K ability - my friend ordered faulty units and then changed them to Vifa 10BGS. It looks much cleaner on CSD but he likes JXR6 more for its highs. For him Vifa is too dry.
If the cone is soft, the sound will not be as crystal clear because of how much vibration wave exists on the cone at the same time. The right combination of stiffness is necessary. I'm not familar with the 10BGS and could not find data on it. Soft cone material generally sounds dry because the soft cone absorbs some of the detail where the VC former meets the cone. This is the main reason alluminum is better in revealing detail.

I just went back to check my measurements, and my CSD are actually better than the Jordans.:eek: Now if I can get them to go to 30KHz, it would be nice.
 
Lynn Olson said:


I think one difference between us is that I'm no longer interested in SEAS, Vifa Peerless, ScanSpeak, etc. [...] That's why I'm looking at different vendors than the usual mainstream candidates.

I'm surprised that you haven't seen drivers depart from minimum phase. This is one of the most direct indicators of cone breakup, and it's gotten much worse with the popularity of very rigid Kevlar, carbon-fiber, composite, ceramic, and metal cones. When a cone no longer moves as unit and enters the breakup region, there are multiple, asynchronous centers of radiation all over the cone. This is a clear indication of a "no-go" zone, and indirectly shows a requirement for an aggressive high-slope crossover to avoid gross coloration.

The drivers that are most interesting - to me - are the ones that don't require aggressive equalization to avoid harsh sound, and are characterized by smooth, well-controlled rolloff regions that retain their minimum-phase character to very high frequencies. [...]

Yes, I realize that you are looking at different types of drivers. And I'll be the first to admit that I'm not part of the SET subculture though I do like and have in the past (and will in the future) owned tube amps. I had a wonderful pair of EAR 509's designed by Tim de Paravicini (100 w into 4, 8, or 16 ohms). I had to opportunity to meet and talk with Tim at the time I bought them, back in the 80's. Don't know why I sold them.

On the other issue, break up is not necessarily a departure from minimum phase. This thread went through the multiple source/minimum phase thing when we were discussing diffraction(or was it in Earl's thread) and I don't think there is any reason to go there again other to say that as far as I recall for every conventional driver I have measured it has been possible to get a near perfect minimum phase reconstruction of the measured phase, on axis, well into the breakup even with highly resonance metal cone like the Seas W18 and W22. And is it really break up or uncontrolled resonance at and above breakup that are the issue. Assume for the moment that breakup is still minimum phase, at least in some cases. If the breakup is well damped and retains minimum phase behavior, how would it be identified from either the amplitude or phase response? I think the only way to verify breakup is viewing the cone motion with a laser scan or the like. And if there is a departure from minimum phase is that necessarily an indication of breakup or just that the size of the radiating area is much greater than the wave length?

It's not a big deal though. I understand you desire for better driver data supplied from the manufacture. I think one of the issues regarding phase is that it is a relative quality. There needs to be a clear reference point. For example, assuming minimum phase for the moment, the driver is minimum phase relative to the acoustic center. But where is the AC? That isn't an easy question to answer since the only way to accurately locate the AC is by constructing the minimum phase response and then removing delay from the measured phase until they match. And while that is easy to do, there are a number of parameters in the minimum phase reconstruction that must be specified some what arbitrarily which can alter the AC position. A physical reference is a better idea and for that reason I always reference the phase to the driver mounting flange.