I found exactly the same for these two specific phenonema, with my own tests which I believe have justified significance *for me* as that is what I do precieve reliably in *controlled* testing.
Who did the same, apart from BarleyWater, so that these "indicative data points" might accumulate towards fact and common beliefs some day? Do we need an AES (read: industry consortium) paper for that?
Point is, excellent phase (square) response can be had for free and with zero ill effects using any decent speaker, there really is a free lunch.
OTOH, sacrificing important design options in favor of getting minimum phase with passive XOs sure may lead to failing speakers... unless done as clever as in the Synergy horns....
OTOH, sacrificing important design options in favor of getting minimum phase with passive XOs sure may lead to failing speakers... unless done as clever as in the Synergy horns....
I agree with most all the points Barleywater makes in #77, flat frequency and phase response sound realistic, and harmonic and intermodulation distortion increase loudness with a false sense of dynamics.
I disagree with his implication that compression drivers have increased harmonic distortion and attendant intermodulation distortion compared to any other type of driver at similar SPL levels, in general the reverse is true.
I have measured both harmonic and IM distortion with a number of different compression drivers and found distortion quite low at domestic listening levels.
As far as a "free phase lunch", it takes time and proper design application to get it right, whatever way you achieve it.
Danley and Geddes primarily have chosen passive designs and achieved good phase response, others like Dave Gunness use active means, the Gunness tuned Mackie HD 1502 has amazingly flat phase and frequency response, especially considering the rather common 2 way HF Horn/15" woofer and low price point.
At any rate, I don't need another AES paper to verify all that, when you make A/B comparisons and measurements accurate reproduction (or lack of) becomes apparent.
Art
Attachments
Last edited:
Flat frequency response as a "good thing" is pretty much a given and has been for a long time. Phase response remains an unproven hypothesis. The nonlinear distortion would have to be extremely high before it had any effect on loudness. A few % THD is not going to be audible. In the home environment nonlinear distortion is not an issue, as I think that you alluded to. We did tests on nonlinearity in compression drivers up at 110 - 120 db and found that it was not statistically independent.
Art;
We've been here a bit before; at typical home listening levels, most compression drivers have higher distortion levels than their non compression counterparts. Inherently compression drivers compress air beyond linear range in the compression cavities, and at intended operating levels in the horn throat too. For most listeners at domestic levels, the distortion levels of compression drivers is completely masked. For others with trained ears, harmonic distortion levels >1.5% are readily discernible, and a little bit higher, annoying. For PA, direct radiators don't cut it, compression drivers rule, and the tender eared need not attend if pristine sound outweighs real live performers and social contact with humans.
And yes, no shortage of dynamic drivers with horrid performance, and no shortage of improper application.
Earl;
Have you ever done testing with broadband burst signals?
I certainly have. Broadband bursts with steep high pass and steep low pass filtering at crossover point are complemented with chosen crossover filters. For example a burst signal with flat response 20Hz-1kHz with 200dB/octave low pass behavior is then filtered with 24dB/octave 1kHz high pass filter. The resultant signal is played through tweeter or other driver. Frequency response of captured result reveals complex distortion products. In the case with my little active speaker the results correlate very strongly with how much better the speaker sounds with steep crossover compared to with LR4, or linear phase version of LR4:
The shelf of frequency components above 1kHz, jutting out at about -30dB from the peak response, under the LR4 conditions accompanies every broadband transient in program material. Obviously these are not just simple harmonic components. The LR4 attenuation tail flattens out at about 250Hz due to the presence of intermodulation difference components. The results indicate the formation of a dynamic noise floor that is only about -30dB.
When the broadband burst signal is narrowed by the super steep high pass crossover filter the response results show essentially simple harmonic distortion peaks more than 40dB below peak against a very low noise floor. This result is consistent both with distortion as measured with pure sine signal or measured with swept sine technique.
Listening to the speaker system under the two different crossover conditions is night and day.
Here is results of narrow band burst centered about 1700Hz:
The peak drive voltage was somewhat higher than for above results. Once again response shows mostly simple harmonic structure, some intermodultion structure, and a very low apparent noise floor.
Results of dual peak bursts with centers about 1700Hz and 6kHz with same peak drive voltage as the single 1700Hz burst:
Intermodulation components dominate. Energy for the harmonics of the 1700Hz burst component have been tapped and taken up in formation of intermodulation components.
The apparent noise floor has risen. Harmonic distortion is always attended by intermodulation. The wider the bandwidth the greater the potential for harmonic content.
Inject continuous white noise into a loud environment and the perception is often a quieting effect. The damping muscles of the middle ear are brought into a near constant level of action. Inject noise modulated to the environmental noise sources and perception is of a louder environment with grainy texture.
Rephrased this reads: We did tests on nonlinearity in compression drivers up at 110 - 120 db and found that it was statistically dependent.
What kind of testing was this? Subjective evaluation of music v simple harmonic distortion tests?
I've never heard a compression driver that I cared for, thus don't own any, but I would be very interested in seeing broadband burst test results for some.
I read all white papers with interest and innocent trust. Right up until a particular speaker is named. I do not think the last word on distortion and phase audibility have been spoken. More truly independent studies are needed. To me its not so much that speakers are primitive, its that they were a great an advanced design to start with that quickly, much like the propeller driven airplane, perfected. To this day we see great planes from the 40s in air races competing against new design and winning as well. For the same reason.
I strongly believe that Electrostatic speakers have strengths unmatched by moving coil speakers, I also strongly believe that moving coil speakers have strengths unmatched by Electrostatic speakers. Many have criticized my comments that a square wave response of an electrostatic speaker showed it was better at some things than a moving coil speaker. I accept this method may not have been the best way to compare speakers, but it does clearly show worse performance than say amplifiers or DAC's.
I strongly suspect that the brain creates an illusion of sound, and overcomes a lot of the distortion due to wanting to produce a coherent interpretation. This in my opinion is why measurements on their own do not give the full picture, and the combination of ears and brain makes some distortion far worse than other areas for peoples opinion of relative sound quality.
This suggests that a primitive and bad speaker can be believable, even if it is primitive.
Is the poor measured performance intrinsic to the physics of transducers that could ever exist?
Is the poor measured performance intrinsic to the "primitive" nature of loudspeaker tech?
Both the above questions are more interesting than arguing that one objective measure of speakers is invalid or another. I'm sure we all would like to have the perfect transducer. Maybe this is impossible, and maybe its a function of the primitive technology.
Given the vastly different performance I perceive between electrostatic speakers and moving coil speakers in sound strengths and weaknesses, I suspect until a mathematician or physicists proves no transducer can be significantly better than what we have, I subscribe to the view the speaker technology is the problem and its current application can be regarded as "primitive".
I strongly suspect that the brain creates an illusion of sound, and overcomes a lot of the distortion due to wanting to produce a coherent interpretation. This in my opinion is why measurements on their own do not give the full picture, and the combination of ears and brain makes some distortion far worse than other areas for peoples opinion of relative sound quality.
This suggests that a primitive and bad speaker can be believable, even if it is primitive.
Is the poor measured performance intrinsic to the physics of transducers that could ever exist?
Is the poor measured performance intrinsic to the "primitive" nature of loudspeaker tech?
Both the above questions are more interesting than arguing that one objective measure of speakers is invalid or another. I'm sure we all would like to have the perfect transducer. Maybe this is impossible, and maybe its a function of the primitive technology.
Given the vastly different performance I perceive between electrostatic speakers and moving coil speakers in sound strengths and weaknesses, I suspect until a mathematician or physicists proves no transducer can be significantly better than what we have, I subscribe to the view the speaker technology is the problem and its current application can be regarded as "primitive".
Setting aside the world of 2013 and looking forward to, say, 2063, many things will be different. DSP will be essentially free. Materials technology will have finally broken through the limitations of the 19th and 20th century, nano and bio-grown materials will be common and abundant, and I strongly suspect aluminum and steel will be on their way out.
This means that wall-sized "intelligent" speakers, with millions of tiny deposited-film segments driven by tiny integrated amplification modules, will be cheaper than steam-loudspeaker technology based on early 20th-century metallurgy and magnetics. We already see this in plasma and LCD televisions and digicam sensors, with integrated per-pixel electronics as an integral part of the display and sensor. Delta-sigma DACs take advantage of cheap DSP to replace large-area precision resistor arrays of the older R2R converters.
All it takes is for DSP, nano, and bio-grown tech to pass critical thresholds, and the old tech is pushed aside in only a few years. Once solar is cheaper than coal-fired electricity, the old power stations will go off-line faster than you would believe. The same applies to loudspeaker tech; once wall-sized DSP loudspeakers with first thousands, then millions of micro-emitters is actually cheaper than the bulky stuff we have now, the transition will be pretty quick.
There was a long transition period for digital cameras, since there was about a decade where digital had many advantages over film, but it wasn't really better esthetically and visually. About three to five years ago, that changed, and digital pulled ahead in all respects, with better color, more dynamic range, and astonishing low-light performance, never seen before with any film. That pretty much finished off film.
Once the wall-emitter tech passes a certain threshold, it will replace everything else. Entire walls will be a visual and auditory emitters, with enormous processing part as an integral portion of the device. This will be commonplace, along with a whole new array of construction materials.
When? 2063 is as good a guess as any. There are really strong economic pressures for superior construction materials with less destructive impact on the environment (remember, 10 billion people with higher living standards than now), and the relentless pressure for more efficiency, lighter, and more intelligent machines ... what I call the virtualization of the machine world.
Look at a late 19th-century steam engine. Despite the enormous size, weight, and water and fuel requirements, the power output of this massive machine was only about 600 to 1000 hp ... and it had to pull very heavy trains up and down slopes. Nowadays, we have hybrid and turbodiesel cars with 130 to 250 hp output that can dramatically outperform that steam engine, and at far less weight and fuel consumption. These trends will only continue, resulting in less weight, and more intelligence embodied in every technical device. I don't think loudspeakers will be exempt from that trend.
This may sound like a strange world, with everything from houses to vehicles made of featherweight nano and bio-grown materials, and steel, aluminum, and concrete on their way out. But look back 200 years, when houses were made of heavy timbers and brick, and everything was far heavier (and lower performance) than we have now.
Iron, steel, and concrete have been used for hundreds to thousands of years, but that's only because they're cheap and abundant. That will change in 50 or 100 years, with raw materials at a serious premium, and computing power at unimaginable levels essentially free. All man-made items will have far more "intelligence" at the design, fabrication, and finished-item stages; less labor and weight, and more intelligence. We see this now, and we're nowhere close to the end of line towards virtualization and weightlessness.
I strongly suspect that 100 years from now, barring a massive technological collapse, there will be no mass production factories anywhere in the world, and all man-made objects will be grown in place using a convergence of bio and nano technology. Why? Cheaper and better. The old materials will have a nostalgia value, but will be quite expensive, and used more for trim than structural elements.
Last edited:
There is still that little device that permits to deviate the panorama to
a little square, being it a photo-sensible celluloid or CCD.
Yes, the good old lens!
Adaptive optics are coming to consumer devices pretty soon. What used to be exotic military (Star Wars) technology is now standard in modern astronomical telescopes and is working its way towards consumer devices. We already have really good optical stabilization (I have an Olympus OM-D EM-5, and the 5-axis stabilization is astonishing), and the next step are adaptive optics.
There's a big difference between crude overall feedback stabilization (as used in subwoofers) and intelligent DSP at every point of a large array; that's what used in modern telescopes, with each mirror segment computer-controlled to optimize the overall image, despite the warps and gyrations of the air-path, and done at the precision of a fraction of an optical wavelength.
The array-DSP will be applied to micro-emitter loudspeakers, probably starting with very small mobile-phone and portable-device virtual 3D soundfield emitters. Remember, with all that DSP power, the DSP can track the listener, observing head position and ear location, and dynamically synthesize a 3D soundfield for that exact location in space. Hard to do now, very easy and cheap in the future.
As the listener-tracking 3D emitter array comes down in cost and improves in performance, it will replace headphones, and then loudspeakers, creating an exact virtual 3D soundfield for the individual listener. It might seem creepy that the micro-emitter will be watching you, but modern cameras already have surprisingly good face detection, and offer the photographer the choice of autofocussing on the near or more distant eye ... and can do all of this in a fraction of a second.
What about different outer-ear (pinna) shapes? Not a problem. The next-gen Xbox has such good 3D infrared detection it can watch your heartbeat across the room. Definitely creepy, particularly since Microsoft says the new Xbox won't work without the Kinect camera connected ... and a live Internet connection. NSA, here we come. Wave at the nice man observing your living room!
There's a big difference between crude overall feedback stabilization (as used in subwoofers) and intelligent DSP at every point of a large array; that's what used in modern telescopes, with each mirror segment computer-controlled to optimize the overall image, despite the warps and gyrations of the air-path, and done at the precision of a fraction of an optical wavelength.
The array-DSP will be applied to micro-emitter loudspeakers, probably starting with very small mobile-phone and portable-device virtual 3D soundfield emitters. Remember, with all that DSP power, the DSP can track the listener, observing head position and ear location, and dynamically synthesize a 3D soundfield for that exact location in space. Hard to do now, very easy and cheap in the future.
As the listener-tracking 3D emitter array comes down in cost and improves in performance, it will replace headphones, and then loudspeakers, creating an exact virtual 3D soundfield for the individual listener. It might seem creepy that the micro-emitter will be watching you, but modern cameras already have surprisingly good face detection, and offer the photographer the choice of autofocussing on the near or more distant eye ... and can do all of this in a fraction of a second.
What about different outer-ear (pinna) shapes? Not a problem. The next-gen Xbox has such good 3D infrared detection it can watch your heartbeat across the room. Definitely creepy, particularly since Microsoft says the new Xbox won't work without the Kinect camera connected ... and a live Internet connection. NSA, here we come. Wave at the nice man observing your living room!
Last edited:
This is brilliant stuff. Many thanks. Much to think about.
First on my list, and related to Owenhamburg's (also brilliant) post is the question whether in principle DSP can exactly reproduce Renee Fleming at our ear/hearing reference.... even assuming all the computing power in the world to "undo" your room acoustics.
For sure, today it is a "fools errand" to imagine using Blumlein-like math (and even recording Fleming in outer space) you could put anything much like her wave-front at your chair with stereo speakers or even, pinna influence considered, with headphones.
Great stuff. Glad to see this thread has left egotistical system comparisons behind.
BTW, Lynn Olson's model obliterates "down the hall" as any measure of system quality since the sound is highly personalized for the hearing of one listener. And systems have to be tested/verified by hearing tests, not mics.
Ben
Funny thing, I "hear" the most perfectly realistic performance of Renee Fleming singing with no speakers playing.
Last edited:
I did say that wrong, it should be: The nonlinear distortion was not "statistically significant" from a completely undistorted signal in compression drivers up to 110-120 dB. In other words, it was inaudible.
This work was peer reviewed and published in the AES journal some ten years ago. The test was blind with several dozen subjects under controlled conditions. Your results are all anecdotal.
Air nonlinearity in compression drivers and horns is insignificant since it is extremely low order and requires very high SPLs. 1.5% "distortion" does not mean a thing as all of my other papers on the subject can attest to.
You seem to be completely in the dark regarding the work done in this area.
Instead we can now all hypothesize about our imaginary ultimate sound systems - no need to have to deal with reality.
The comments were about "what might be" 50 years from now. It is a refreshing way to think.
I work in a neuroscience dept. The things folks are doing now, and capable of doing, is mind blowing compared what was state of the art 50 years ago.
It is difficult to even imagine what the kinds of solutions might look like.
These are exciting times!
Markus - have you actually heard one of those systems? I have (different brand, same concept). The designers were so impressed with how it sent beams in the right locations that they seemed to completely miss the sound quality. It sounded like a cheap table top radio - but yes, this radio did have separate beams in different directions. But would I ever consider such a thing for serious stereo music listening, or HT? No way.
I think that it will be very difficult to bring the sound quality of these multi-element designs up to the standards that most of us (well at least myself) have come to expect. Signal dynamics for small devices will always be an issue, because these things don't scale well. 20 small speakers do not have the same output capacity as one speaker 20 times the area.
Its not that it can't be done - in 2060, I am sure that we will have such things, but the technology has a long way to go yet before it will satisfy the more demanding listeners.
Personally, I don't think we'll have to wait 50 years, at least I hope not.
For instance: Nanotube Loudspeakers | MIT Technology Review
What if the wall coverings in the listening room were made entirely of sound emitters, processed with DSP, using motion tracking and infrared tracking?
It seems like a lot of the technology is already in place, or at least nearly in place to come up with some incredible solutions.
For instance: Nanotube Loudspeakers | MIT Technology Review
What if the wall coverings in the listening room were made entirely of sound emitters, processed with DSP, using motion tracking and infrared tracking?
It seems like a lot of the technology is already in place, or at least nearly in place to come up with some incredible solutions.
Nothing personal and just a minor R&D note, but you are promoting a fallacy that every student has to memorize in Stats 101.
Present company excepted, there are lots of ways of screwing up an experiment and therefore to being unable to "find" statistical significance. Any bit of sloppiness or testees that are sleepy or too few participants or any number of inadvertent pitfalls....
It never proves anything to fail to achieve statistical significance, contrary to your assertion. If establishing a "no difference" conclusion was your purpose, it sure doesn't do it.
(And lots of peer-reviewed articles that never should have seen the light of day.)
Ben
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.