Linkwitz Orions beaten by Behringer.... what!!?

So now we get to vote on scientific facts? That's a new approach. Kind of modern actually.

Show me a single scientific study that finds nonlinear distortion significant. I'll show you several on the other side of the argument.
Heavens above! MP3 compression is nonlinear distortion of the original signal - the encoded signal varies from the the original in a precisely measurable way - subtract one from the other, there is a remainder, this is called "distortion"!

Now that we definitely have "distortion", of the nonlinear variety, there are vast numbers of 'scientific' studies precisely delineating how audible this is, depending on all sorts of factors ...
 
How do you know it's not the timing errors the mp3 compression algorithm creates that makes it unlistenable ;).
Any "error" is a distortion, no matter how one terms the particular type of artifact. I've done some tests using the highest quality levels of encoding, and once one learns what to listen for, the slightest fine tuning of the parameters of the encoding, the type of "distortion" generated in other words, can be heard to audibly vary as a result.
 
Very few can distinguish MP3 in 256 kbps vs lossless in a blindtest.

With what music? Commerically produced "Pop" with female vocals. Not suprising - it's even not "real music" (ie. uncompressed two microphone live event recording) and it's what MP3 was designed to do.

I have some recordings in vinyl, CD and mp3 where the MP3 could almost be the master tape. (eg. Autumn "Red")

Try something more difficult and/or complex and well recorded. Like Lynn's poster child: mass choral works (perhaps with a little organ thrown in behind the orchestra) . Eg. Chesky's recording of Psalm 24 on their set up disk.

Or, something even a nine year old can spot: Australia Art Orchestra's "Ringing the bell backwards" CD vs MP3

On middle of the road music the MP3 is simply ess pleasing to listen to than the uncompressed version: something is missing. Or amiss

And you don't need anything fancier than an ithingy with a FLAC app and cheapish headphones (eg. Grado 60s) to hear that.
 
The point is not that one cannot create enough nonlinear distortion to be audible, of course they can. The point is that there is no valid measure of nonlinear distortion that correlates to perception so you cannot say that this is what you hear because there is no way to determine that as "fact". It's just an opinion.

Low bit-rate MP3 may sound terrible, but that is beside the point. Most loudspeakers have nonlinearities that are sufficiently well controlled that they are not audible. But again, I could make a loudspeaker that did have audible nonlinearity, but why would I want to do that? It is easy enough to design out nonlinearity and anyone who does not do that is not a very good designer.

Hence, either your loudspeakers are poorly designed, or you are not hearing nonlinear distortion. It is one or the other.
 
Very few can distinguish MP3 in 256 kbps vs lossless in a blindtest.
Contrary to popular misconception, classical music is the easiest to compress, while click-based music is hardest (castanets, weird electronic music (Aphex Twin etc). There is even a technical reason for it.
And coders designed from ground-up to be perceptually indistinguishable do a remarkable job with classical music at ~130kb/s range (Musepack). MP3 has known limitations that precludes it from being transparent on some samples even at the highest bitrate.
Non of this is on topic of course.
No, the cause of the audible change is nonlinear distortion, vs. linear distortion [---]
One anecdotal evidence suggests that (simulated) loudspeaker nonlinearity, even Le variation, sounds relatively innocent, compared to the gross/shouty character some speakers posess while turned up (diffraction generators mostly).

So now we get to vote [---]
My vote goes to this
 
Its neither. We do hear nonlinear distortion and its not generated by the speakers because they're decently designed. That only leaves electronics.

Yes, it is possible to have audible nonlinear distortion in electronics, but that certainly does not mean that all electronics have this problem. Most, if not all, studies have shown that in general nonlinear distortion in electronics is not a significant problem. Of course you may think that it is, but that's a long ways from proving that it is.

There is a metric that has good correlation to perception.
Why no-one in the industry uses it? Obscure licensing scheme or just plain laziness with a dose of ignorance?

Ignorance and laziness.

It's like the old joke about a drunk looking for his keys under the streetlight: 'cause that's where the light is. Everyone measures THD because it is so easy. No one bothers to ask if it means anything.
 
Contrary to popular misconception, classical music is the easiest to compress, while click-based music is hardest (castanets, weird electronic music (Aphex Twin etc). There is even a technical reason for it.
And coders designed from ground-up to be perceptually indistinguishable do a remarkable job with classical music at ~130kb/s range (Musepack). MP3 has known limitations that precludes it from being transparent on some samples even at the highest bitrate.

Daniel, that's interesting what you write about kinds of music causing difficulties for for MP3. Do you have references you could point me to?
 
AX tech editor
Joined 2002
Paid Member
Only slightly off-topic: Siegfried Linkwitz just received an award from the professional speakers association ALMA.
His keynote speech is an interesting overview and State of The Speaker presentation.
Several letters betweem him and Laurie Fincham of KEF, and his speech 'Whatever happened to the quality of reproduced sound in the Home'.
Recommended reading.

Jan
 
His first slide highlights the problems that I have with Sigfried: "How close have we come to creating the illusion of a live acoustic event in the home?"

Why is that the criteria? Most recordings do not seek this illusion. They seek to create an entirely constructed two channel illusion at the playback. Is the vast majority of recorded music to be excluded from his consideration? Well, yes actually. I have talked with him about this and he does not consider anything but live recorded classical music to be worthy of his investigations. Oh well.

The reason that I did not participate in the OP topic here is that it was based on exactly the question posed above. I do not accept that criteria as all encompassing, but simply one (small?) aspect of sound reproduction. I do not exclude live venue recording as relavent and I certainly do not condone others excluding studio recordings either.
 
Last edited:
Now guys, the audio industry deserves slightly more credit that that, or maybe not but “how we got here” is worth examining.
RF communications have always been a vastly larger and more important area of research and in design of transmitting gear, confining the output to the “fundamental” or the assigned frequency is the object and stuff radiated outside that at it’s harmonics etc are “bad”. At that time the total energy in out of band in it’s harmonics made a logical measure.

In audio, in the days of “hot bulbs”, all of the active devices had somewhat similar distortion characteristics and at an engineering level, THD was still something of an indicator of the non-linearity.
The problem was outside engineering, they used THD or it’s absence as a measure of quality and “sell product”.
The problem is, we don’t hear all harmonics equally, the 2nd harmonic can be monstrously large by modern standards and still be undetectable, nearly undetectable in music even at large values. The reason is perceptual / frequency masking and that music is already harmonic in nature and so even harmonics fit in musically.

http://upload.wikimedia.org/wikipedia/en/d/d9/Maskingpatterns_sp11.jpg

On the other hand, I had a pre-amp once, a big name SS unit. Passing a 20Hz sine (who would do that?) , I noticed “a sound” up high. I killed the subwoofer amp and it was still there, like a whirring noise. I eliminated the other gear by substitution and finally I hooked the preamp up to my old HP3562a and used the vector subtraction to measure the “distortion”. I forget the number but indeed, the THD was VERY low.
It was VERY low but a “feature” of high negative feedback circuits in general is that while they suppress the level of harmonic distortion, they push it’s distortion spectrum higher and higher.
Here, with a 20Hz sine input, the 17th harmonic was one of the largest (still very small) BUT if you look at an Equal loudness curve, you can see that at the 17th harmonic of 20Hz is 3400Hz AND your ear is about 80dB (100,000,000 times) more sensitive at 3400Hz than at 20Hz. If one had a pure tone just above the threshold of audibility at 20Hz and silent background, if the 17th harmonic had a power level of just .0001%, it would have an equal perceived loudness as the fundamental at 20Hz. Hence, at 20Hz sine wave, my preamp made a whirring noise as a result of it’s distortion spectrum extending that high.
One of the things about tubes and low feedback generally is that the distortion spectrum does not extend very high. That single feature alone could account for why in the day a 50W tube amplifier sounded like it was more powerful than a 50W SS amp that had measured much lower distortion THD. Back then, many assumed the tube had “headroom” but while clipping was a bit more gradual, a pp amp flat tops just above rated power (so usually has a TINY amount of headroom).

How about a different more all encompassing view of distortion, maybe from Star Trek.
ZZzzzzssss Distortion is what prevents Veeger from performing its mission. Veegers mission is to replicate the input signal into radiated sound . Distortion is what prevents Veeger from completing mission replicating input signal without alteration as sampled by measurement microphone.
Veeger logic suggests that since recordings are an unknown element and mostly contrived in the studio, a better subjective evaluation of accuracy or the presence or absence of a loudspeakers flaws might be conducted using the generation loss test, ideally semi anechoic conditions like outdoors or in a large room mid height.

This way you hear the original reference (unlike the original performance and recording process) and each generation gives you a more and more amplified caricature of everything that prevents the speaker from faithfully reproducing the original signal. Get to even 3 generations and you have a very good sounding loudspeaker.
Anyway, some thoughts from deep audio space and now back to a meeting.
Best,
Tom
Danley Sound Labs
 
Now guys, the audio industry deserves slightly more credit that that, or maybe not but “how we got here” is worth examining.
RF communications have always been a vastly larger and more important area of research and in design of transmitting gear, confining the output to the “fundamental” or the assigned frequency is the object and stuff radiated outside that at it’s harmonics etc are “bad”. At that time the total energy in out of band in it’s harmonics made a logical measure.

In audio, in the days of “hot bulbs”, all of the active devices had somewhat similar distortion characteristics and at an engineering level, THD was still something of an indicator of the non-linearity.
The problem was outside engineering, they used THD or it’s absence as a measure of quality and “sell product”.
The problem is, we don’t hear all harmonics equally, the 2nd harmonic can be monstrously large by modern standards and still be undetectable, nearly undetectable in music even at large values. The reason is perceptual / frequency masking and that music is already harmonic in nature and so even harmonics fit in musically.

http://upload.wikimedia.org/wikipedia/en/d/d9/Maskingpatterns_sp11.jpg

On the other hand, I had a pre-amp once, a big name SS unit. Passing a 20Hz sine (who would do that?) , I noticed “a sound” up high. I killed the subwoofer amp and it was still there, like a whirring noise. I eliminated the other gear by substitution and finally I hooked the preamp up to my old HP3562a and used the vector subtraction to measure the “distortion”. I forget the number but indeed, the THD was VERY low.
It was VERY low but a “feature” of high negative feedback circuits in general is that while they suppress the level of harmonic distortion, they push it’s distortion spectrum higher and higher.
Here, with a 20Hz sine input, the 17th harmonic was one of the largest (still very small) BUT if you look at an Equal loudness curve, you can see that at the 17th harmonic of 20Hz is 3400Hz AND your ear is about 80dB (100,000,000 times) more sensitive at 3400Hz than at 20Hz. If one had a pure tone just above the threshold of audibility at 20Hz and silent background, if the 17th harmonic had a power level of just .0001%, it would have an equal perceived loudness as the fundamental at 20Hz. Hence, at 20Hz sine wave, my preamp made a whirring noise as a result of it’s distortion spectrum extending that high.
raises hand, similar results too ;)

One of the things about tubes and low feedback generally is that the distortion spectrum does not extend very high. That single feature alone could account for why in the day a 50W tube amplifier sounded like it was more powerful than a 50W SS amp that had measured much lower distortion THD. Back then, many assumed the tube had “headroom” but while clipping was a bit more gradual, a pp amp flat tops just above rated power (so usually has a TINY amount of headroom).
So true, Never considered any amp at it's rated power, Always derated it by 6dB for personal use, 3dB in Pro use.
How about a different more all encompassing view of distortion, maybe from Star Trek.
ZZzzzzssss Distortion is what prevents Veeger from performing its mission. Veegers mission is to replicate the input signal into radiated sound . Distortion is what prevents Veeger from completing mission replicating input signal without alteration as sampled by measurement microphone.
Veeger logic suggests that since recordings are an unknown element and mostly contrived in the studio, a better subjective evaluation of accuracy or the presence or absence of a loudspeakers flaws might be conducted using the generation loss test, ideally semi anechoic conditions like outdoors or in a large room mid height.

This way you hear the original reference (unlike the original performance and recording process) and each generation gives you a more and more amplified caricature of everything that prevents the speaker from faithfully reproducing the original signal. Get to even 3 generations and you have a very good sounding loudspeaker.
Anyway, some thoughts from deep audio space and now back to a meeting.
Best,
Tom
Danley Sound Labs
Veeger made me laugh, good analogy

Don't forget Harry Potter placing the egg in the bath. ;)
 
Its neither. We do hear nonlinear distortion and its not generated by the speakers because they're decently designed. That only leaves electronics.

Electronics with perfectly resistive loads show very low levels of distortion. Reactive loads are another story.

Here are results using very nice 5.24" mid-woofers. Two woofers are setup in separate enclosures, and placed for center to center distance of 12". 73Hz and 440Hz sine waves are used for testing. Microphone is placed equidistant from centers of each driver.

Response for each woofer to 73Hz signal looks like:

image001.gif

And response to 440Hz signal looks like:

image003.gif

Response when one speaker has 73Hz signal, and other speaker has 440Hz signal:

image004.gif

No IMD is visible. The microphone circuit is not introducing any obvious IMD.

Here is response when single driver is fed two tone signal of 73Hz and 440Hz signal combined:

image005.gif

IMD components are clearly visible. In this test setup the difference is audible too.

With bigger FFT, IMD difference components are resolved from HD components of 73Hz signal:

image006.gif

When 146Hz and 440Hz signals are combined, IMD is greatly reduced:

image007.gif

At these levels of IMD, I have great difficulty in reliably hearing the difference when 146Hz and 440Hz are reproduced through two different drivers and when they are reproduced by a single driver.

In typical application, this woofer can be used below 73Hz, but when driven hard audible sound quality decays.

"Decently designed" covers a lot of territory.

In above testing, signals were about 3V p-p.

No real fault is attributable to the electronics. Results only point to design choices that must be made contingent to application demands.

More food for thought. Here are results using 2" wide band driver. A broadband burst signal is used, depicted by white trace. Electronic 1kHz high pass filters of 24dB/oct and >100dB/oct are used:

image010.gif

When bandwidth is added to burst signal around 6kHz:

image011.gif


Clearly the path in speaker design is use of electronics in constraining driver bandwidth.

From last two pictures, it is also clear that what is passed off as noise floor, is floor of modulation distortion that rides order of magnitude above that seen with electronics tested with purely resistive loads.

Electronics capability far exceeds transducer performance, and advances in audio are dependent on understanding transducer behavior.
 
Daniel, that's interesting what you write about kinds of music causing difficulties for for MP3. Do you have references you could point me to?
Well, not relevant anymore, aeons ago there was a problem with storing your music collection in digital form and groups of people looked for a (transparent) solution. Turns out, that MP3 has a problem with percussive sounds, commonly called 'preecho'. No matter how many bits are thrown at the problem, how the psychoacoustic model is perfected, some problems still remain. It turns out that MP3 has Fourier-based transform (MDCT to be excact) that has a too large window to 'contain' impulsive sounds after quantization, so pre- and post-(mostly inaudible) ringing occurs - preringing being highly audible, no masking mechanism in the hearing system. Modern 'transform' based codecs don't have this problem, mainly due shorter block size.

Ignorance and laziness.

It's like the old joke about a drunk looking for his keys under the streetlight: 'cause that's where the light is. Everyone measures THD because it is so easy. No one bothers to ask if it means anything.
You should be proud that Moore-s 'Rnonlin' metric is vastly more obscure than yours. Simplicity and all.

the audio industry deserves slightly more credit that that
Without 'it' we wouldn't have sound in our homes/theaters, wouldn't we?
 
Using images from a CRO to demonstrate whether there is audible distortion or not is completely useless - the ear is vastly more sensitive than the hopeless dynamic range of a typical waveform image. Only the output of a simulator, which can be zoomed in to high levels of magnification if the processing settings are done right will allow the viewing of the sort of glitches which look, are visually innocuous, but which could be quite disturbing aurally ...