Interesting you brought this up. I was thinking about the design of the Ariel and the new speaker the other day.
I designed them for an old friend in Portland, Oregon who owned a pair of stacked Quad ESL57's. My goal when designing the Ariels was a time-decay performance as good as the ESL57 (single, not stacked) with 8 to 10 dB more efficiency, 8 to 10 dB more headroom, and much better imaging by using diffraction reduction techniques.
That goal was met: Quad ESL57 owners like the Ariels, and the two very different speakers sound remarkably similar to each other. The Ariels were good enough they replaced the LO-2's I had designed at Audionics in 1979.
The second prototypes that were built in 1993 are still in my living room, on the second set of drivers (with a replacement set of 6 drivers in reserve). Although they're 20 years old, I still prefer them to most direct-radiator speakers on the market, and they work very, very well with most vacuum-tube amplifiers.
The goal for the new speaker is similar: another 8 to 10 dB gain in efficiency and headroom while retaining the time-decay performance of a Quad ESL57. This has been a far harder goal, has taken much longer than the six-month-long Ariel project (more like 5 years), and needed additional talent and skill from France, Norway, and Australia.
But the new speakers are about 95% done. The first prototypes sound better than the first prototypes of the Ariels, and are closer to the design goal. The second version of the prototypes will be in my living room; it's a good question whether they will last for twenty years or not.
Here's quick answers to Soongsc's questions:
The measured efficiency is between 98 and 100 dB/meter/watt, and the intended listening distance is between 3 and 8 meters. The crossover and horn mounting "sled" will have distance compensation so first-arrival response is flat at the listening position.
Time-decay is similar to the Ariels, less than 0.5 mSec, and will be further optimized.
I am not sure what this question means. The rise time is set by the tweeter, as it is in all loudspeakers. Thermal compression due to VC heating will be 10 dB less than the Ariel due to higher efficiency and larger VC assemblies, and 15 dB less than commercial high-end loudspeakers with efficiencies in the 85 to 88 dB/meter/watt range.
In subjective terms, this results in an impression of much greater headroom, noticeably different dynamic rendering, and more vivid tone colors. "Pano" is probably the expert on this aspect of subjective sound; I've never heard the world-famous Paris setup, but I've heard enough "hints of greatness" from Altec in the past to keep me going on this project.
This is set by the drivers, although the crossover plays a significant role in reducing IM distortion. Manufacturer's specs show very typical figures of harmonics 40 to 50 dB below the fundamental, with the important difference that higher efficiency means that 1 watt input results in 8 to 15 dB louder sound than typical audiophile speakers.
Last edited:
Just out of curiosity, what amp(s) was driving the stacked 57s? I found the stacked 57s to be very responsive to changes in amplification. A difficult load to drive, as you know. They gave me my first "Oh! Amps really do make a difference" moment.
That was a long time ago. Vacuum-tube for sure; the ESL57's are pretty much unlistenable on transistor amps. Stereo 70's, Scott, Fisher, old Heathkits, I dunno. Although direct-heated triodes were first in mind when I designed the Ariels, my Portland friend didn't actually own any DHT amplifiers, just restored 50's stuff (push-pull pentode with moderate feedback).
My previous speakers were linear-phase satellite speakers with an efficiency in the 86 dB/meter/watt range with a subwoofer. I thought they were pretty good until I heard the second prototypes of the Ariels. It was then that I became very aware of the differences between amplifiers; with the previous speakers, differences were not that easy to notice ... everything sounded pretty much the same, no matter which amp I used.
My previous speakers were linear-phase satellite speakers with an efficiency in the 86 dB/meter/watt range with a subwoofer. I thought they were pretty good until I heard the second prototypes of the Ariels. It was then that I became very aware of the differences between amplifiers; with the previous speakers, differences were not that easy to notice ... everything sounded pretty much the same, no matter which amp I used.
I don't think I have ever heard two amplifiers that sound the same, but I have not tried to specifically collect statistics either.
I think the main advantage of most tube amplifiers is that the high current paths are nearly isolated from the low current paths. This gives them more advantage, especially for speakers with high resolution. Some problems with solid state amplifiers can be masked by using soft diaphragm based drivers, but that would also be a compromise. My previous examination of various horn configuration seem to point towards revealing amplifier problems making the sound a bit on the harsh side. The only way to solve it is fixing the amp. I have not played around with tube amps ever since I got shocked by 500V release from caps while working on one.
When you say linear phase, is it actually flat phase throughout the audio range? Or is it minimum phase?
"...I have not played around with tube amps ever since I got shocked by 500V release from caps while working on one..."
I began my tube electrician career attempting to strip wire attached o "add on" PSU capacitor with my teeth, effectively discharging 400V + cap in my mouth
I became a fan of High End audio shortly after
I began my tube electrician career attempting to strip wire attached o "add on" PSU capacitor with my teeth, effectively discharging 400V + cap in my mouth
I became a fan of High End audio shortly after
Class A amps are quite interesting, lots of people say it's the biasing that reduces distortion, but I think it is in fact the high bias current and low wattage that helps stabilize ground currents, thus reducing it's interaction with the amp circuit, combined with careful layout, you do get an amplifier with good resolution.
Hello Pano,
There is something special with the Hiraga amplifier it is, due to the topology of the output stage, the output impedance is not negligeible.
This can explain some analogy in the results obtain on a given loudspeaker between the Hiraga and a tube amplifier.
Best regards from Paris, France
Jean-Michel Le Cléac'h
Hello,
You should read what Mills and Hawksford wrote on the subject.
http://www.essex.ac.uk/csee/researc...J12 Distortion reduction MC current drive.pdf
My answer will only concern high frequency loudspeakers and specially compression drivers :
With a high output impedance many HF non linarities are reduced. I used to show real mesurements made on a TAD TD2001 driver on a Le CLéc'h horn that show the reducing of such distortion.
Give a look to :
http://www.melaudia.net/zfoto/rueil1010/TD2001_Melaudia_J321-694x377.gif
For bass loudspeakers the benefit should be more related to damping. With low Qts loudspeakers, using high impedance in series (or current drive) we can obtain an optimal damping. (My own tube amplifer used on the TD2001 possess an output impedance as high as 51ohms).
Nelson Pass wrote a paper on that:
http://www.firstwatt.com/pdf/art_cs_amps.pdf
Best regards,
Jean-Michel Le Cléac'h
You should read what Mills and Hawksford wrote on the subject.
http://www.essex.ac.uk/csee/researc...J12 Distortion reduction MC current drive.pdf
My answer will only concern high frequency loudspeakers and specially compression drivers :
With a high output impedance many HF non linarities are reduced. I used to show real mesurements made on a TAD TD2001 driver on a Le CLéc'h horn that show the reducing of such distortion.
Give a look to :
http://www.melaudia.net/zfoto/rueil1010/TD2001_Melaudia_J321-694x377.gif
For bass loudspeakers the benefit should be more related to damping. With low Qts loudspeakers, using high impedance in series (or current drive) we can obtain an optimal damping. (My own tube amplifer used on the TD2001 possess an output impedance as high as 51ohms).
Nelson Pass wrote a paper on that:
http://www.firstwatt.com/pdf/art_cs_amps.pdf
Best regards,
Jean-Michel Le Cléac'h
Jean-Michel, yes the Hiraga has a higher output impedance than a typical SS amp. I don't know the actual number, tho - never measured it. I think that, along with some other parts of the design, kept the Hiraga amp stable on the double electrostatic panels where other amps went into oscillation. At least it sounded that way, I never did scope it.
Soongsc: I never really thought about ground currents in Class-A, but that's an interesting idea. I'm sure that Lynn "Mr. Current Loops" Olson can fill in some gaps there.
For me, the fact that the output stage is almost always pulling more current than the musical signal seems to stabilize things. The variations in the musical current are swamped by the idle current.
Soongsc: I never really thought about ground currents in Class-A, but that's an interesting idea. I'm sure that Lynn "Mr. Current Loops" Olson can fill in some gaps there.
For me, the fact that the output stage is almost always pulling more current than the musical signal seems to stabilize things. The variations in the musical current are swamped by the idle current.
Hi Pano
One can see why a horn driver can benefit from a higher impedance source, ideally a matched one, it is just like or similar to driving an antenna.
If one were delivering power (like into a radiation resistance) with a fixed Voltage source, then if you half and double the load impedance, one see’s the delivered power changes by + and – 3dB, a 6dB variation.
On the other hand, if driven by the nominal impedance as load, then one sees the same change in load impedance causes a -3dB change in each case or a 3dB variation. In the crossovers at work, that extends farther to where some issues are gone with the right conjugating source impedance.
I think I can explain the amplifier issue. In the mid 70’s I worked at an amplifier company named Grommes that made both SS and Tube amplifiers. I had found the old mac60’s (tube amps that were traded in for the modern SS gear) I had gotten from the TV store I had worked at in High school sounded much better driving my RTR ESS panels and my home made ESS speakers than the SS amps I had or worked on.
At intersonics, I ran into a more extreme case of the same issue with piezo electric transducers which were mostly capacitive when off tune.
The problem is when you have a reactive load, the current is not in phase with the Voltage. For a capacitive reactance, the current leads the voltage in a sine wave. With the levitation source off tune, they were so extreme that the current is displaced nearly 90 degrees in phase instead of “in phase” like a resistor and much worse than any electrostatic speaker I had measured..
Transistors have a “safe operating area” and when plotted out in a push pull output stage, one finds they are normally limited to a bowtie shape in two of the four quadrants. For a resistive load, the load line is simple, it’s a proportional slope X voltage = X current.
If you imagine a load line where current is maximum when the output voltage in a sine wave is zero and current is zero when voltage out is max, that ellipse or circle shape extends greatly outside the safe area.
Since transistors fail in an instant, protection circuits prevent the output stage drive from leaving the safe area and this means that one may well find loads which the amplifier will not drive above some level. Again, an oscilloscope will show waveshape weirdness.
A class A amplifier has half of it’s maximum current flowing when there is no signal and for the same output power, requires MUCH larger plate or collector dissipations than push pull ab or b.
For a transistor, that means a single ended or push pull 10 W class A amplifier output device would need to have a much larger safe area than it would have if push pull ab or b. For driving the levitation sources, it meant using more transistors to where it could deliver rated current into a short safely, then it laughed at the reactive load.
Part B is a leading phase angle can tax bode criteria, ss amps usually have a small decoupling choke in the output stage to disconnect it from such dangerous loads "up high". .
Jean-Michel you are fortunate indeed to only be dealing with one or a few listeners, constant directivity is a necessary if an audience is to hear as close to the same thing as possible or the reverberant spectrum is to be similar to the direct field.
Best,
Tom
One can see why a horn driver can benefit from a higher impedance source, ideally a matched one, it is just like or similar to driving an antenna.
If one were delivering power (like into a radiation resistance) with a fixed Voltage source, then if you half and double the load impedance, one see’s the delivered power changes by + and – 3dB, a 6dB variation.
On the other hand, if driven by the nominal impedance as load, then one sees the same change in load impedance causes a -3dB change in each case or a 3dB variation. In the crossovers at work, that extends farther to where some issues are gone with the right conjugating source impedance.
I think I can explain the amplifier issue. In the mid 70’s I worked at an amplifier company named Grommes that made both SS and Tube amplifiers. I had found the old mac60’s (tube amps that were traded in for the modern SS gear) I had gotten from the TV store I had worked at in High school sounded much better driving my RTR ESS panels and my home made ESS speakers than the SS amps I had or worked on.
At intersonics, I ran into a more extreme case of the same issue with piezo electric transducers which were mostly capacitive when off tune.
The problem is when you have a reactive load, the current is not in phase with the Voltage. For a capacitive reactance, the current leads the voltage in a sine wave. With the levitation source off tune, they were so extreme that the current is displaced nearly 90 degrees in phase instead of “in phase” like a resistor and much worse than any electrostatic speaker I had measured..
Transistors have a “safe operating area” and when plotted out in a push pull output stage, one finds they are normally limited to a bowtie shape in two of the four quadrants. For a resistive load, the load line is simple, it’s a proportional slope X voltage = X current.
If you imagine a load line where current is maximum when the output voltage in a sine wave is zero and current is zero when voltage out is max, that ellipse or circle shape extends greatly outside the safe area.
Since transistors fail in an instant, protection circuits prevent the output stage drive from leaving the safe area and this means that one may well find loads which the amplifier will not drive above some level. Again, an oscilloscope will show waveshape weirdness.
A class A amplifier has half of it’s maximum current flowing when there is no signal and for the same output power, requires MUCH larger plate or collector dissipations than push pull ab or b.
For a transistor, that means a single ended or push pull 10 W class A amplifier output device would need to have a much larger safe area than it would have if push pull ab or b. For driving the levitation sources, it meant using more transistors to where it could deliver rated current into a short safely, then it laughed at the reactive load.
Part B is a leading phase angle can tax bode criteria, ss amps usually have a small decoupling choke in the output stage to disconnect it from such dangerous loads "up high". .
Jean-Michel you are fortunate indeed to only be dealing with one or a few listeners, constant directivity is a necessary if an audience is to hear as close to the same thing as possible or the reverberant spectrum is to be similar to the direct field.
Best,
Tom
I can't think of any way to isolate the AC current-return path in direct-coupled circuits, whether tube or transistor (which I why I don't think Loftin-White or stacked power supplies are a good idea for tube amps).
See Part One of the 2004 ETF presentation or Part Two. The key point of the presentation is to be aware that distortion comes in two forms: voltage distortion, and current distortion.
To find the sources of current distortion, you have to look at the complete current loop, which includes the AC current return-path. These are usually all summed together in direct-coupled and RC-coupled amplifiers. In transformer-coupled amplifiers, there is an option to isolate the return paths, and the Amity and the Karna amplifiers take advantage of interstage ground isolation.
What I like about triode amplification is inherently linear operation, low proportion of high-order distortion terms (thus low IM distortion), and very linear Miller capacitance. The requirement for a +300V to +500V B+ supply is awkward, true, which is why I recruit experienced tube-amp builders to confect my amplifiers. I do the concept, circuit, current and voltage analysis and parts selection, and recruit a friend for the actual build. Same thing for loudspeakers. I'm no good at woodworking, either, but I'm open to their suggestions what's practical and what's not.
What the electronics and loudspeakers have in common is intrinsically linear operation and avoidance of poorly defined, unstable regions where modeling breaks down. In amplifiers, this is Class AB operation, proximity to slewing (which can start 5x lower than hard slewing), and inadequate linear current delivery for the final power stage. In loudspeakers, diffraction causes trouble that is mostly beyond the scope of equalization, and operating tweeters close to nonlinear regions (at the lower part of the working range) introduces program-driven IM distortion. None of these are exotic problems, are measurable if you know where to look, and are audible as unnatural sound.
Many designers think Class AB operation is well-characterized. I don't agree with them: I've worked on a production line back in my Audionics days, and conventional Class AB transistor amplifiers have bias that wanders all over the place. Sure, if you average over several minutes, the bias eventually settles down, but under dynamic conditions, the transistor dies are heating up and cooling down over several seconds to as long as a minute, and during that time, the bias is wandering all over the place. The thermal feedback loop is much longer than the 10 millisecond to 500 millisecond pulse of music, and frankly, sounds like it. The sliding-bias approach used in deluxe amplifiers conceals the hard 0.7V on-off transition of the power transistor, but does not changes the underlying change in current gain as an array of devices are brought in and out of circuit.
Class AB tube amplifiers spread the switching transition over 20V to 50V instead of the transistor 0.7V on-off region, and don't suffer from the thermal-bias instabilities of transistor amplifiers. (Vacuum-tube characteristics are almost entirely independent of operating temperature, unlike transistors or MOSFETs.) However ... the great majority of Class AB pentode amplifiers use RC-coupling between the driver and power stage, and the coupling cap "blocks" when the power tube grid draws current. Recovery from the blocking typically takes 200 mS to 2 seconds, depending on the size of the coupling cap, and this draws attention to what would otherwise be a very brief interval of clipping.
Last edited:
Put another way, I design from the device upward, while other designers design from the overall concept downward, treating the actual drivers, diaphragms, magnets, etc. as "black boxes" to assemble the overall concept, akin to using op-amps to build a preamp or linestage.
I'm interested in the physical limits of the devices: what do triodes, pentodes, bipolar transistors, MOSFETs, compression drivers, horns, direct-radiators, etc. do as they approach the outer boundaries of their operating region ... and what techniques are available to linearize them in the intended operating region, and what's available as they approach the limits of the operating region, go beyond it, and return.
Thus, settling time, which tells you a lot about a transducer or amplifying element as it recovers from a transient. The actual musical signal may never have the dV/dT or dI/dT to fully stimulate the device, but it's interesting to see how the device reacts to the signal.
When I started digging into horn design, I was curious about how impulse response correlated to the edge of the coverage pattern. Old-school constant-directivity horns like the Altec Manta-Ray and the JBL Bi-Radial have diffraction "pinches" in the throat to create the intended coverage pattern ... but the downside is reflections in the time domain and a somewhat ragged edge to the pattern.
In RF antennas, the time, frequency, and polar domains are linked, and cannot be separated. An antenna that is highly directional with sidelobes at the sides and rear will also have reflections in the time domain and ripples in the frequency domain, while an antenna with a smoothly-defined polar pattern that is free of sidelobes is also well-behaved in the time and frequency domain. Once again, the edge condition also defines other aspects of performance, with the usual no-free-lunch tradeoff we see in audio.
You want the most linear low-level operation, Class A. Sorry about the cost and inefficiency. You want the most compact and efficient? Class D, with a lot of signal-processing to randomize the PWM switching spectrum that drives the power devices. In between? Class AB, with clever helper circuits to broaden out and soften the AB switching transition. Note there is no "best" approach; it's a question of what's most important to you.
I'm interested in the physical limits of the devices: what do triodes, pentodes, bipolar transistors, MOSFETs, compression drivers, horns, direct-radiators, etc. do as they approach the outer boundaries of their operating region ... and what techniques are available to linearize them in the intended operating region, and what's available as they approach the limits of the operating region, go beyond it, and return.
Thus, settling time, which tells you a lot about a transducer or amplifying element as it recovers from a transient. The actual musical signal may never have the dV/dT or dI/dT to fully stimulate the device, but it's interesting to see how the device reacts to the signal.
When I started digging into horn design, I was curious about how impulse response correlated to the edge of the coverage pattern. Old-school constant-directivity horns like the Altec Manta-Ray and the JBL Bi-Radial have diffraction "pinches" in the throat to create the intended coverage pattern ... but the downside is reflections in the time domain and a somewhat ragged edge to the pattern.
In RF antennas, the time, frequency, and polar domains are linked, and cannot be separated. An antenna that is highly directional with sidelobes at the sides and rear will also have reflections in the time domain and ripples in the frequency domain, while an antenna with a smoothly-defined polar pattern that is free of sidelobes is also well-behaved in the time and frequency domain. Once again, the edge condition also defines other aspects of performance, with the usual no-free-lunch tradeoff we see in audio.
You want the most linear low-level operation, Class A. Sorry about the cost and inefficiency. You want the most compact and efficient? Class D, with a lot of signal-processing to randomize the PWM switching spectrum that drives the power devices. In between? Class AB, with clever helper circuits to broaden out and soften the AB switching transition. Note there is no "best" approach; it's a question of what's most important to you.
Last edited:
An oversight reading the Safe Operating Area curves nearly bankrupted Audionics in the early Seventies. Before they hired Bob Sickler, the company's main product was a 100wpc Class AB transistor amplifier. The engineer was an old-school Tektronix guy, but we were having a return rate of more than 50%. Some days more amps came back then we shipped out. This situation went on for six months, and we all started to wonder how much longer we'd be keeping our jobs.
Bob was the new-hire, and he noticed something about the SOA specs of the driver transistors. Like just about all power transistors, the SOA curves were plotted in log units, with a nice straight load-line approaching the forbidden region. He pulled me aside and quietly mentioned that load-lines for loudspeakers are not straight, are they? I said no, speakers are pretty much reactive everywhere. 90 degrees (a perfect circle) might be rare in all but electrostats, but 20 to 30 degrees (an ellipse) is plenty common, especially above and below the box frequency or if the speaker has an unfriendly crossover (which is very common).
But these are driver transistors, right? They don't see the full load, right? Well, that's true, but all bipolar power transistors do is amplify current, not isolate it. The speaker load still appears at the driver transistors, just with less current.
Uh-oh. So if I get this right, then all it takes is a long bass passage, with peak energy around 50 to 100 Hz, then the SOA of the driver transistors will be exceeded ... in fact, exceed it for 10 milliseconds, and bye-bye driver stage. When the driver stages goes, since it is direct-coupled, the output stage goes too, a few milliseconds later. Net result? All driver and output transistors failed, burned emitter resistors, and a burned circuit board. We were lucky no amplifiers had outright caught on fire; some of the returns had completely charred circuit boards, with gray smoke lines around the ventilations holes in the top plate.
The discovery of this oversight caused a major sensation at the company, since the old-school Tek engineer was one of the founders and major shareholders, and didn't appreciate the new-hire demonstrating his oversight had come close to bankrupting the company. Bob almost lost his job, but was saved by the President demanding an immediate fix before we shipped one more amplifier ... and recalling all the amplifiers out in the field (which was nearly a hundred at the time).
The fix? Not that pretty. Instead of one driver transistor per side, we used three, with separate emitter-resistors ballasting each one so they wouldn't current-hog and destroy each other. The circuit required a fast driver, so we couldn't just willy-nilly swap in another part ... and transistors were pretty slow back then.
The fix took care of the reliability problem. The ultimate fix was a complete brand-new amplifier by Bob Sickler (note the new amplifier retained the paralleled drivers of the old amplifier). In an ironic turn so typical of the audio business, the new amplifier (which sold more than 2000 units) was named after the old-school engineer who fought it every inch of the way.
Last edited:
I remember asking when I was shown the rather small, and rather cryptic SOA curves why the transistor vendor used log units, when that made the critically important curves harder to read and understand?
All I got was a significant look and a brief "why do you think?"
Oh. So much for my idealism about device vendors.
(To put this in context, this was only a few years after Watergate, Nixon's resignation, and the ignominious helicopter evacuation of Saigon.)
All I got was a significant look and a brief "why do you think?"
Oh. So much for my idealism about device vendors.
(To put this in context, this was only a few years after Watergate, Nixon's resignation, and the ignominious helicopter evacuation of Saigon.)
Last edited:
Hi Lynn
Hey thanks for the flash back / story, I like classic circuitry and miss working on that stuff. The circuit sort of reminded me of the old SWTP tiger amp, another “cutting edge” design I had experience changing expensive parts on regularly.
Yeah it’s funny too back when everyone had an oscilloscope, it was pretty routine to put current on one axis and voltage on the other to see reactance in a load (or the R on one and L on the other and look at the “stereo” fuzz ball). Nowadays, who even has one?
It’s funny how things were when transistors were so new. When I was in high school I saw an article in popular electronics for a 100W SS amplifier (an unheard of power at the time).
I convinced my vocational electronics teacher to let me make one as a project. He helped procure the parts and even with help it was a good bit of cash for me to come up with.
It used an interstage transformer , a Triad ty-160x and a pair of Germanium Delco DTG-110b’s (expensive numbers I will never forget).
Anyway, on power up, it instantly smoked the emitter resistors and blew the devices (probably the opposite order). Cutting to the end after 3 more pairs of devices, my teacher set up scopes and meters, replaced the outputs and powered it on a variac, everything cool and then ZZZZZsmoke in front of the whole class.
Since he couldn’t get it to work either, he gave me credit for it and near the end of the year he got a letter back from Delco who said the design was withdrawn because of stability problems. I think we had established that already.
An antenna array can be similar to an array of acoustic sources and an issue with an array of sources is that while we think of the Huygens wavefront acoustic produce, they also radiate as independent sources with their own individual directivity. The individual radiations produce an interference pattern which show up as side lobes or regions of addition and cancellation depending on the angle / distances to the sources.
A horn can produce strong lobes too, but it may not also.
When they don’t have strong side lobes, they can project less sound outside the pattern than the array approach of similar size.
If you have facebook, the measured very directional speaker shown at the top, has very low level side lobes and very narrow pattern.
this 10X30 degree horn is an extreme case (made for projecting high frequencies to the far side of a large stadium to help overcome hf air absorption), it does show a very homogenous beam (a single acoustic source that has 64 hf drivers). Oddly enough just getting up to 8-10 Khz out to 800-1000 feet is very hard
https://www.facebook.com/DanleySoundLabs?ref=ts
While the array design programs many of the array folks usually provide look nice, they often do not include the individual radiations in the display, cannot show phase cancellation (shows power sum) and so what one measures is often very different than what is predicted.
In large room acoustics, the sound that excites the reverberant field is a big part of what kills intelligibility and musical quality and so when you investigate to find out why the arrays do not work that well or as well as our smaller full range horn systems, you find they actually radiate much more like arrays of sources here, there is a lot more sound going places it shouldn’t be and it sounds/ measures differently in every location;
http://www.danleysoundlabs.com/danley/wp-content/uploads/2012/01/line-array-paper.pdf
I guess this makes sense too in our market we have to harp on / educate / demonstrate this stuff to people.
Like any speaker flaw, like pattern flip say, it is not something that is discussed much if ever in trade mags or in print as it doesn’t sell or create the illusion that leads to a sale. Too bad we can't see sound more easily haha
Best,
Tom
Hey thanks for the flash back / story, I like classic circuitry and miss working on that stuff. The circuit sort of reminded me of the old SWTP tiger amp, another “cutting edge” design I had experience changing expensive parts on regularly.
Yeah it’s funny too back when everyone had an oscilloscope, it was pretty routine to put current on one axis and voltage on the other to see reactance in a load (or the R on one and L on the other and look at the “stereo” fuzz ball). Nowadays, who even has one?
It’s funny how things were when transistors were so new. When I was in high school I saw an article in popular electronics for a 100W SS amplifier (an unheard of power at the time).
I convinced my vocational electronics teacher to let me make one as a project. He helped procure the parts and even with help it was a good bit of cash for me to come up with.
It used an interstage transformer , a Triad ty-160x and a pair of Germanium Delco DTG-110b’s (expensive numbers I will never forget).
Anyway, on power up, it instantly smoked the emitter resistors and blew the devices (probably the opposite order). Cutting to the end after 3 more pairs of devices, my teacher set up scopes and meters, replaced the outputs and powered it on a variac, everything cool and then ZZZZZsmoke in front of the whole class.
Since he couldn’t get it to work either, he gave me credit for it and near the end of the year he got a letter back from Delco who said the design was withdrawn because of stability problems. I think we had established that already.
An antenna array can be similar to an array of acoustic sources and an issue with an array of sources is that while we think of the Huygens wavefront acoustic produce, they also radiate as independent sources with their own individual directivity. The individual radiations produce an interference pattern which show up as side lobes or regions of addition and cancellation depending on the angle / distances to the sources.
A horn can produce strong lobes too, but it may not also.
When they don’t have strong side lobes, they can project less sound outside the pattern than the array approach of similar size.
If you have facebook, the measured very directional speaker shown at the top, has very low level side lobes and very narrow pattern.
this 10X30 degree horn is an extreme case (made for projecting high frequencies to the far side of a large stadium to help overcome hf air absorption), it does show a very homogenous beam (a single acoustic source that has 64 hf drivers). Oddly enough just getting up to 8-10 Khz out to 800-1000 feet is very hard
https://www.facebook.com/DanleySoundLabs?ref=ts
While the array design programs many of the array folks usually provide look nice, they often do not include the individual radiations in the display, cannot show phase cancellation (shows power sum) and so what one measures is often very different than what is predicted.
In large room acoustics, the sound that excites the reverberant field is a big part of what kills intelligibility and musical quality and so when you investigate to find out why the arrays do not work that well or as well as our smaller full range horn systems, you find they actually radiate much more like arrays of sources here, there is a lot more sound going places it shouldn’t be and it sounds/ measures differently in every location;
http://www.danleysoundlabs.com/danley/wp-content/uploads/2012/01/line-array-paper.pdf
I guess this makes sense too in our market we have to harp on / educate / demonstrate this stuff to people.
Like any speaker flaw, like pattern flip say, it is not something that is discussed much if ever in trade mags or in print as it doesn’t sell or create the illusion that leads to a sale. Too bad we can't see sound more easily haha
Best,
Tom
I was wondering about those array-design programs, and frankly, the rationale for line-array speakers in domestic audio. If the phase vectors are conveniently omitted from the calculations, no wonder the summation looks so pretty ... unlike real life, where phase addition/subtraction (at the listening position) matters a lot.
I jump into the time domain if summation looks questionable. That's what I think when I see a line array that's being promoted for domestic use in a high-end context. Yes, very pretty, but the arrival times for all these little drivers will be different, and worse, if the array is curved or electronically delayed to approximate curvature, what about distance compensation, eh?
Noncoincident arrival time would certainly account for the mixture of blur and glare in commercial PA systems ... it's loud, all right, but you can't understand a word, thanks to a hash of non-coincident arrivals from the large arrays. Since the arrivals in the 0-3 mSec range are particularly critical for perception of intelligibility and tonal coloration, that's right where the line arrays will fall down, while still measuring fairly decent in the frequency domain. The reader of the FR graph will not realize all those innocent-looking little ripples are the result of a large number of random-phase additions and subtractions from the noncoincident arrival times of the arrayed drivers.
Maybe not as bad as misreading a log-log SOA curve from the transistor vendor (all power devices fail, smoke comes out, charred circuit board, fire in the house), but bad sound for thousands of people. The audience is used to bad PA ... a proud tradition going back to the 1920's ... so they probably cut the venue some slack. It must be frustrating for the musicians, though.
P.S. I got a request from GregOH1 to discuss the Gary Pimm open-back boxes. Gary uses the Eminence Beta 8's, a driver I'm not a fan of, but I admit he gets way better sound than I expect from a Beta 8. Credit must go to the box, because the driver isn't that wonderful.
The box is really simple; it's about 15" wide, 12" high, 18" deep, and open on the back. It's almost completely filled with a big wodge of Bonded Logic recycled-cotton, leaving about 2~3" of air-space between the stamped basket of the Beta 8 and the wodge of Bonded Logic cotton filling. It pretty much touches the back of the ceramic magnet, if my memory serves.
The Bonded Logic has a way of working itself out when loud bass is played, so I've suggested vertical or horizontal slats to Gary as a way of stiffening the top and/or sides of the box and also preventing the Bonded Logic from walking out. You'd have to insert the BL from the front through the hole for the driver instead of simply pushing it into the open back, but it compresses just fine and is easy to work with, unlike horrible fiberglass batting, which gets all over the room and is a skin (and lung) irritant. Both Gary and I also think that Bonded Logic sounds a lot better, too.
Last time I was at Gary's, he had a FFT measuring rig as part of his computer-controlled crossover system, so I could see the response of the open-back box. Really good measurements. The box is flat to about 100 Hz, with maybe a tiny, tiny rise of 0.5 dB or less at the nominal "baffle peak", but subjectively and by measurement, much better performance than the u-box or open-baffle equivalent. Unlike an open-baffle system, there is no in-band equalization, which takes a large burden off driver excursion and amplifier power requirements.
That's why I'm still considering a dual-15 woofer system with the upper woofer in a Gary Pimm style box (scaled up for a 15" driver) and the lower woofer in a vented enclosure, with the vent on the floor so it can take advantage of floor gain.
Potential issues? Well, if the filling material gets too close to the cone, it starts to mass-couple, and driver efficiency goes down, along with a subjective degradation of transients (although this does not show in the impulse measurements). So the filling material needs to be spaced at least 3 inches (or more) away from the cone, and kept at that location. The Bonded Logic filling is reasonably cohesive and does not break apart, but constraining its movement is a good idea ... internal strings inside the enclosure can do this pretty easily.
I jump into the time domain if summation looks questionable. That's what I think when I see a line array that's being promoted for domestic use in a high-end context. Yes, very pretty, but the arrival times for all these little drivers will be different, and worse, if the array is curved or electronically delayed to approximate curvature, what about distance compensation, eh?
Noncoincident arrival time would certainly account for the mixture of blur and glare in commercial PA systems ... it's loud, all right, but you can't understand a word, thanks to a hash of non-coincident arrivals from the large arrays. Since the arrivals in the 0-3 mSec range are particularly critical for perception of intelligibility and tonal coloration, that's right where the line arrays will fall down, while still measuring fairly decent in the frequency domain. The reader of the FR graph will not realize all those innocent-looking little ripples are the result of a large number of random-phase additions and subtractions from the noncoincident arrival times of the arrayed drivers.
Maybe not as bad as misreading a log-log SOA curve from the transistor vendor (all power devices fail, smoke comes out, charred circuit board, fire in the house), but bad sound for thousands of people. The audience is used to bad PA ... a proud tradition going back to the 1920's ... so they probably cut the venue some slack. It must be frustrating for the musicians, though.
P.S. I got a request from GregOH1 to discuss the Gary Pimm open-back boxes. Gary uses the Eminence Beta 8's, a driver I'm not a fan of, but I admit he gets way better sound than I expect from a Beta 8. Credit must go to the box, because the driver isn't that wonderful.
The box is really simple; it's about 15" wide, 12" high, 18" deep, and open on the back. It's almost completely filled with a big wodge of Bonded Logic recycled-cotton, leaving about 2~3" of air-space between the stamped basket of the Beta 8 and the wodge of Bonded Logic cotton filling. It pretty much touches the back of the ceramic magnet, if my memory serves.
The Bonded Logic has a way of working itself out when loud bass is played, so I've suggested vertical or horizontal slats to Gary as a way of stiffening the top and/or sides of the box and also preventing the Bonded Logic from walking out. You'd have to insert the BL from the front through the hole for the driver instead of simply pushing it into the open back, but it compresses just fine and is easy to work with, unlike horrible fiberglass batting, which gets all over the room and is a skin (and lung) irritant. Both Gary and I also think that Bonded Logic sounds a lot better, too.
Last time I was at Gary's, he had a FFT measuring rig as part of his computer-controlled crossover system, so I could see the response of the open-back box. Really good measurements. The box is flat to about 100 Hz, with maybe a tiny, tiny rise of 0.5 dB or less at the nominal "baffle peak", but subjectively and by measurement, much better performance than the u-box or open-baffle equivalent. Unlike an open-baffle system, there is no in-band equalization, which takes a large burden off driver excursion and amplifier power requirements.
That's why I'm still considering a dual-15 woofer system with the upper woofer in a Gary Pimm style box (scaled up for a 15" driver) and the lower woofer in a vented enclosure, with the vent on the floor so it can take advantage of floor gain.
Potential issues? Well, if the filling material gets too close to the cone, it starts to mass-couple, and driver efficiency goes down, along with a subjective degradation of transients (although this does not show in the impulse measurements). So the filling material needs to be spaced at least 3 inches (or more) away from the cone, and kept at that location. The Bonded Logic filling is reasonably cohesive and does not break apart, but constraining its movement is a good idea ... internal strings inside the enclosure can do this pretty easily.
Last edited: