Audibility of non-linear distortion - are we getting carried away?

[Drew]

Hi All,

First, a bit of background: I'm trying to pick a midrange for a three way system that I'm designing for a friend. This is to be a pretty inexpensive system. This implies a pretty straightforward, non-exotic crossover. This suggests well behaved drivers with low to moderate breakup modes.

So I then started looking at non-linear distortion numbers of various drivers, and arived at a question: how much distortion is acceptable - i.e. at what point does it become audible? Now, I know that this depends on drive level, the fundamental frequency, what harmonic you're looking at, etc, etc.

So, it's a complicated question (or rather, it's many questions), but I started looking into this. What I've managed to find online (I'll have to wait until I go back to work next week before I can look through some AES journals) suggests that the threshold of perceptability is about 1% (-40 dB) in most cases and for harmonics less than or equal to 3.

Now, by this criterion, the Vifa XG18 looks to be better suited to the project than say the RS180 - ignoring the price difference. Since a friend will sell me his two XG18 for $60, I pretty much can ignore the price! Now the RS180 will prebably handle higher SPLs better, since it has some distortion headroom to spare. But at a given level of, say, 90 dB SPL, the lower distortion of the RS180 doesn't seem to matter.

So, I've said all that so I can ask: at what point can we stop looking at distortion numbers and focus on other performance critera: a smooth FR, etc?

Low distortion is great of course, but are we getting carried away and a bit myopic with this? Honest question.

[Jed]

The odd order harmonic types are quite audible in my opinion. Also, you have to set your design goals and go from there. For example, at what frequency range will the driver be implemented? If used from X-Y frequencies, does the distortion over a this frequency range have excessive distortion in bass or midrange? The ear is most sensitive in the 1k-5K area, so distortion in this area is possibly more offensive given odd order types.
Also, how low will you cross over to avoid trouble spots with distortion profile? Does your tweeter have an extended low end and low distortion and capable of crossing LR4 at 1.5K? So there is really no answer because one driver with unacceptable distortion in situation A could yield very effective results in situation B given all the variables of designing a complete system.

[cotdt]

from what i know, harmonic distortion is only a factor at high output with any decent quality driver. the various drivers all sound different due to differences in frequency response (especially local frequency response) which cannot be corrected with simple EQing. The XG18's are fuzzy drivers. The RS180 is much more pistonic but loses definition in the upper midrange.

my own blind tests with piano music suggests harmonic distortion is only audible above 3% for 3rd order. and it doesn't sound so bad either, only making the sound very slightly muddier. this is distortion added across the entire spectrum which is the worst case senario, and this is on a hi-end DAC, amp, and ModulaMT speakers (more resolving than my "hi-end" headphones). IMHO linear distortion is what matters. while a smoothed FR might look flat, a high resolution unsmoothed graph taken properly (not the usual DIY techniques) will reveal the sound signature of any driver.

in short, i agree with the thread starter that harmonic distortion testing is overblown. it's good for determining crossover points and as a diagnosis for motor linearity, but not for determining the "purity" of the sound due to the reasons i listed above. only truly pistonic drivers can have pure sound.

[Chris7]

Non-linear distortion is definitely not overblown. The non-linear distortion profiles of the drivers is the main factor that distinguishes one speaker from another, other variables being similar (crossover topology, on-axis flatness, cabinet construction, etc.). It's what accounts for the tone we stereotypically associate with poly drivers vs. paper vs. metal cone (even when their breakup is well-controlled). Just playing the devil's advocate, if anything, linear distortion is overblown. You can tweak the frequency response of a speaker with an EQ by one or two dB in various places and your ear will get used to it over extended listening, but the intrinsic tone of the speaker won't change. Honestly though, both linear and nonlinear distortion are important. With nonlinear distortion, it's not the absolute level of distortion at any one frequency, but the overall distortion profile.

[cotdt]

Non-linear distortion is definitely not overblown. The non-linear distortion profiles of the drivers is the main factor that distinguishes one speaker from another, other variables being similar (crossover topology, on-axis flatness, cabinet construction, etc.). It's what accounts for the tone we stereotypically associate with poly drivers vs. paper vs. metal cone (even when their breakup is well-controlled). Just playing the devil's advocate, if anything, linear distortion is overblown. You can tweak the frequency response of a speaker with an EQ by one or two dB in various places and your ear will get used to it over extended listening, but the intrinsic tone of the speaker won't change. Honestly though, both linear and nonlinear distortion are important. With nonlinear distortion, it's not the absolute level of distortion at any one frequency, but the overall distortion profile.

That's also what I thought at first, before I started to read the research that was being done in acoustics and psychoacoustics. Every speaker driver has its own timbre which determines its sound signature, ie. that "poly" sound, that "metal" sound. Though not all poly drivers have the poly sound.

O'Toole at Harmon did a lot of research into this stuff: Here's a link to their white papers:

http://www.harman.com/about_harman/...leadership.aspx

In short, different cone materials with different rigidity, density, speed of sound, will have a different timbre which a very high-resolution FR measurement done correctly (it pretty much never is) will reveal. A speaker may look like it has a flat (+-1dB) frequency response, but locally you get a lot of fuzziness (can be +-10dB or more!) once outside of the woofer's pistonic range, which happens at around 500-700Hz for 7"s based on nearfield measurements. Here are some excerpts:

"Why is it that resonances are so important? Because they are the fundamental building blocks of almost all of the sounds we are interested in hearing. High-Q resonances define the pitches. Medium- and low-Q resonances define the timbres, allowing us to distinguish between different voices and instruments. It is subtle differences in the resonant structure of sounds that are responsible for the nuances and shading of tone in musical sounds. Our ears are very highly attuned to the detection and evaluation of resonances, and it is therefore no surprise that listeners zero in on them as unwanted “editorializing” when they appear in loud*speakers."

"I think most people might be surprised to see that a high-Q resonance can exhibit a peak of 10 dB before it becomes clearly audible with any kind of music or sound. Similarly, why is it that the low-Q resonances, that hardly ring at all, are audible at such low levels? At least part of the answer lies in the ability of musical signals to excite the resonances. A high-Q resonance is very frequency specific, and it takes time to build up, just as it decays slowly. This means that a musical signal must hit the resonant frequency quite accurately, and stay there long enough to transfer significant energy to it. The changing nature of music, and vibrato, both conspire against this happening. In contrast, lower-Q resonances have wide footprints in the frequency domain, and so are more often and more readily excited."

"A practical problem is that many measurements these days are made with FFT, TDS, and other systems that time-window the data so that anechoic measurements can be made in normal rooms. A result of the time windowing is that the frequency response data have poor frequency resolution (≥100 Hz is common), and cannot reveal high-Q phenomena at low and middle frequencies. Resolution limitations of this kind are common among loudspeaker designers and reviewers who measure in ordinary rooms. Anechoic chambers are expensive, and measuring outdoors, in nature’s own anechoic space, has practical difficulties. It means, simply, that many commonly-used and published measurements simply cannot reveal visual evidence of certain kinds of audible problems falling within a critical portion of the frequency range – that of the human voice and below. "

"Another common measurement is one in which the audible frequency range is divided into equal fixed-percentage band*widths, such as 1/3 octaves, or in which a high-resolution measurement is heavily smoothed, or spectrally averaged, on a continuous basis. These spectral-averaging devices have extremely limited utility in the design and evaluation process."

"An interesting fact now emerges: that the conventional method of specifying frequency response, ± x dB, is useless unless the tolerance is very, very small. High-Q phenomena could be ± 5 dB, while moderate-Q resonances could be ± 3 dB and low-Q and other broadband deviations could be ± 0.5 dB, and all of them would be equally audible! Clearly, frequency response curves must be interpreted. Frequency responses must also be measured in a manner that reveals resonances as distinct from interference peaks and dips – i.e. spatial averaging. The sad conclusion, therefore, is that standard verbal descriptions of frequency response are not useful, and most published graphs are not of the right kind to allow for unambiguous interpretation. It is easy to understand how the popular belief that ‘you can’t measure what you can hear’ came to be. Bad measurements and useless specifications are responsible."

[JonMarsh]

Picking midwoofers for a speaker system is a complex task, one might say art, because of the various variables to trade off, including bass extension, Q in the systems, dispersion, non-linear disortion, line distortion, etc.

Conventional motors exhibit there shortcomings when you see a rising distortion spectra with increasing frequency- a trait that is the OPPOSITE of what should be happening physically due to the lowering cone travel, but is caused by motor inductivity modulation with increasing frequency. This doesn't help clarity of sonics.

Now, the XG18 does have falling THD in both HD2 and HD3 with increasing frequency, but the intial levels are a little higher than I'd like. The absense of upper end break up modes makes it much easier to work with, and it's a good choice for a first crossover project. It does have a flex mode with a sharp narrow dip in it's response at about 2.8 kHz, a little higher than the RS180 at 2.6 kHz, so it's not well suited to higher crossovers than the RS180 if you want to stay away from energy storage in linear distortion. BUT, it's really a pretty nice driver, I'd take it any day over the PP Seas models. I just can't justify spending $60 on it. I haven't listened to it in a finished system in the bottom end, but it should be possible to do a nice ported design. Sensitivity is a teeny bit lower than average. For the same money, I'm more inclined to work with the 830883, which has about 15 dB lower distortion at moderate (88-90 dB) playback levels, and flatter nearfield frequency response to 3 kHz.

Re nonlinear distortion, one of the first really low distortion speakers I think I ever built were the old X1 SLAMM klones. Weren't pushing the Eton 7's hard at all, a nice crossover on the very linear Focal Tc120dx2, and Audax Pro 13" and 15" in each bass cabinet. The thing about a low distortion speaker is that we have a habit in audio playback of turning up a system until it sounds like it's working at a bit of level, but not to hard; maybe just at the onset of some audible level of distoriton. If you're used to more conventional systems with conventional drivers, (midrange Seas and Scanspeak), then you go to some fairly high end stuff in a somewhat optimized configuration, it can be a little scary- ThomasW and I were listening from a Sony ES XA7ES CD player directly into an Aragon power amp into the speakers- only after a couple of hours of listening and noticing the build up of some tinitus, did we realize how loud we were listening.

Low distortion makes sense, and in my opinion directly correlates with freedom from listening fatigue, among other things, as well as being able to listen into the recording further.

Both the latest European drivers and the Dayton RS series enable very good quality speakers compared with what was possible 10 years ago, at rather reasonable prices.

~Jon

[Drew]

First, I'd like to thak everybody for their responses. Please don't get too caught up with the fact that I'm considering particular drivers for a system. I'm hoping to discuss this subject a bit more abstractly.

For the same money, I'm more inclined to work with the 830883, which has about 15 dB lower distortion at moderate (88-90 dB) playback levels, and flatter nearfield frequency response to 3 kHz.

{snip}

Low distortion makes sense, and in my opinion directly correlates with freedom from listening fatigue, among other things, as well as being able to listen into the recording further.

Ok - so at what distortion level (all other tings being equal) does listening fatigue disappear for you?

What I'm trying to get at is: at what point does it not matter anymore? As you say, driver X may have 15 dB lower distortion at a given SPL than driver Y, but the question that I am getting at is does it matter? Obviously, if driver Y is at (simplisticly) -20 dB and driver X is at - 35 dB, than it matters.

But, if driver Y is already at -40 dB, than is there any advantage in going to driver X at -55 dB? I mean, at some point it has to NOT matter. How about -100 dB vs. -115 dB?

From what I've read -40 dB (i.e. 1%) is just barely on the cusp of audibility for 3rd harmonic when the listeners are listening to 360 Hz tones at close to 90dB (can't remember the exact number). For music, it's higher. Again, I'm going to spend many lunchtimes reading some AES papers over the next few weeks, but I wanted to get a discussion rolling here.

In many things, people often assume that "bigger is always better" - I'm wondering if we're falling into an analogous situation of "less is always better". All other things being equal, then fine. Of course. But when we start having to spend $$ on crossovers to tame our low distortion $$$$ drivers, I want to know if it's justified.

This isn't intended to be a criticism of anybody, but I want to know if this design approach is justified by the data regarding audibility of distortion. As an engineer, I like to focus on things that matter. If distortion numbers of -55 dB matter vs. -40 dB, then fine. But if they don't, then there are plenty of other aspects of system design that I should spend my time and money on.

I'm not trying to defend any particular driver - those drivers mentioned in my previous post were used to as a jumping-off point for this discussion. To get caught up in those particular drivers is to miss the broader point I'm hoping to discuss.

As for Jon's point about the perception of loudness and the audibility of distortion, this is correct and well known. Perceived loudness (i.e. the point at which you want to turn it down a bit) is highly correlated with the onset of audible distortion. However, this is really a side issue to what I'm hoping to discuss. Heck, if you want to continue down this path, it would be desirable to use drivers that distort a few percent at 90-100 dB to alert us to the high SPLs and therefore help us to protect our hearing. But let's not go there.

[Dennis H]

As I posted in another thread:

.... one study I've seen that says the audibilty is a function of the square of the HD order. So the 3rd harmonic only needs to be something like half as loud as the 2nd for equal audibility. Crunching the numbers:

Order	% HD for equal audibility
2	1.00
3	0.44
4	0.25
5	0.16
6	0.11
7	0.08
8	0.06
9	0.05

Several other studies confirm the concept if not the exact numbers. So, just looking at 2nd and 3rd harmonics doesn't tell the whole story. The higher order harmonics are much more audible even if they don't contribute much to THD.

[Bri]

I'd be interested in doing some research to see if anyone has tried to quantify this, and more generally an objective measure of loudspeaker quality. Actually I'm sure people and companies have tried, it's just a question of how many have published. If someone could come up with a good utility function that takes speaker measurements and spits out a "quality" number (takes into account the various distortions, flatness, phase/delay issues, basically everything that impacts the quality of the transduction to sound waves at the ears), then uncertainty and sensitivity analysis could be performed on all the input variables. Performing such an analysis could be done with many of the "design of experiments" methods. That would help answer all of these questions like "does this kind of capacitor/inductor/wire/cone material/etc really matter?" in an objective an potentially semi-automated fashion. Does this sound like wishful thinking? I don't think that one universal "correct" utility function could ever be devised, but having a few around would be better than nothing. Food for thought at least!

[dwk]

I'd be interested in doing some research to see if anyone has tried to quantify this, and more generally an objective measure of loudspeaker quality.

Geddes has done some stuff in this direction, but as it is inherently a multi-dimensional problem, you can't expect a 'single figure of merit' to come out the end. For example, how would you define an objective trade-off between smooth power response vs HD of a specific order? They are basically orthogonal concepts, and choosing between them is likely to be highly subjective.

One thing Geddes has done is some work along the lines of quantifying the perceptability of various orders of distortion similar to the chart Dennis posted. He defined a 'GedLee metric' which was a weighted sum of the HD components, and which correlated very strongly with a perceptual ranking of the signals. This was synthetically generated, though, so it wasn't applied to measuring a system(s) and determining an objective measure. I also don't believe his actually weights were published anywhere that I've seen, so it's not terribly useful.

Another thing he did when developing the Summa was to do some studies as to what factors were most prominent in determining the subjective 'goodness' of a speaker. His result was that the main factors were:
- smoothness of on-axis response
- smoothness of power response
- flatness of on-axis
- flatness of power response.

Significantly, distortion was NOT a statistically significant factor in his tests, although I don't know the details of listening level etc. All systems were pretty decent, though, so it may simply be the case that there wasn't enough spread in the distortion values to show up. However, he's pretty convinced that distortion is simply not a factor at all in the choice of compression driver for his systems (although he then couches it by saying 'assuming the drivers are good to start with', and obviously low distortion is one of the key elements that distinguishes a good driver)

The final aspect that I discussed with him at RMAF is that he's now convinced that the perception of some signal characteristics islevel-dependent, further complicating things. His specific finding was that diffraction/early-reflection artifacts are fairly strongly level dependent, becoming more audible at higher levels; this probably also applies to his 'higher order modes' work in waveguides, as they are delayed wrt the original signal. If true, this could easily complicate the question of the audibility of distortion which also rises with level; Jon's description of fatigue etc being factors associated with higher order HD seems very similar to subjective descriptions of of HOM/diffraction.

So, the upshot is that this isn't a simple set of questions, and it involves a lot of factors that are hard to isolate and/or separate.

[Chris7]

One thing Geddes has done is some work along the lines of quantifying the perceptability of various orders of distortion similar to the chart Dennis posted. He defined a 'GedLee metric' which was a weighted sum of the HD components, and which correlated very strongly with a perceptual ranking of the signals. This was synthetically generated, though, so it wasn't applied to measuring a system(s) and determining an objective measure. I also don't believe his actually weights were published anywhere that I've seen, so it's not terribly useful.

This article compares the harmonic weighting factors in Geddes' work to those from Shorter's earlier work; they're essentially the same:

http://www.gedlee.com/downloads/THD_.pdf

What's interesting is that Geddes' harmonic scaling numbers were determined from a completely different set of starting principles -- i.e. the similarity in his metric with the previous work is more coincidental than derivative, but it reinforces the previous work.

One thing that is clear from all the research is that raw THD is a meaningless measurement; some form of harmonic scaling is necessary.

[Piotr]

I see a common misstake being repeated in this thread.

The threshold audibility of harmonics on single sine waves has "little to do" with music and loudspeakerdrivers.... or any non linear electro or electroacoustical device for that matter.

Playing complex material will produce a LOT of sideband products, not only a pure series of overtones. One should think in terms of non linearity instead of distortion. Distortion is a result of two things, a stimuli and a non linearity. Change the stimuli and the distortion changes.

People also often forget or overlook the fact that not only overtones is the result from a non linear transfer charactereistics but also subharmonics. This is intereesting with all drivers but especially tweeters.. Many people use to argue that high distortion on a tweeter in the top octave is nothing to worry about because the distortion products ("the harmonics") end up above the audible range iow. above 20kHz. This is true if we only play sinewaves on our music systems but that is rarely the situation as we all know. Low distortion in (or rather a high degree of linearity) is necessary in the whole range being excited in the device.

If the 1% limit is accepted as a level of distortion that is necessary to hear the coloration from a 2nd harmonic on a pure sinewave, then I think we need somewhere about 1/10 of that, 0.1%, on a driver for getting anywhere near transparency.

A driver with 0.5% 2nd and 3rd harmonics on pure sinewaves will be very colored when asked to reproduce music.

I don not agree that THD measuremetns are worthless because they are very indicative ( is that a word?) of the non linearity of the transducer.

/Peter

[JonMarsh]

Thank you Peter- I was thinking all morning about this topic and was going to suggest looking at it from two sides- (being schizophrenic at times, it's easy for me to also play the devil's advocate .

Some of the most illustrative work in this area regarding intermodulation distortion comes from the late Dean Jensen, who pioneered an illustrative approach to IM distortion called spectral contamination analysis, which brings up the very points you are raising here, that subharmonics result from Intermodulation distortion, and they are not at ALL related to the original frequency, and as such are quite audible.

Now, harmonic distortion testing is still useful; it shows nonlinearity in a specific frequency range, and is a guide to how a system will react- the danger arises in drawing the wrong conclusions about what is audible and what isn't. IMO, the biggest benefit in the reduction of measured harmonic distoriton in more modern, well designed drivers, is the significant reduction in intermodualtion distortion audible with music signals.

Now, myself, I think I'm maybe too easy with driver testing- certainly not pushing them nearly as hard as Sigfried Linkwitz or MarkK with multi-tone tests at fairly high SPL, but my first intent is to get a simple clear picture of basic behavior in a range common to the real world applicatoin- playback SPL around 90 dB, input voltage 2.83 V, sometimes higher for special drivers and for woofers. Even the latter I rarely find necessary to test above 10VRMS before they start showing their warts. Many 12" woofers are not acceptable even at 4 VRMS, so I don't have to do higher level testing.

So, am I too easy going, or too tough? I prefer to see close to -60 dB if possible in the normal working range at 90 dB SPL. Considering what additionally happens with IM, I think this is reasonable.

So, on that basis, I think the difference between two speakers like the XG18 and 830883 is worth considering.



















But, that's not to say you can't make a nice set of speakers with the XG18, especially for the right program material and listening conditions. And it will be easier to do, if you don't have a lot of crossover experience.

(and yes, I was driving the 830883 to higher levles, don't remember exactly how much- it's also a couple of dB more sensitive than the XG18).

Another thing a little quirky about the Vifa is the S shaped surround; though it's supposed to be anti-resonant, it appears that may not be the case, as there is an impedance reflection in the impedance curve which could be the surround, at about 350 Hz; another lesser one at 1200 Hz.

I'd suggest just trying the XG18 and see if you like it. If you do, no harm no foul.

[cotdt]

I see a common misstake being repeated in this thread.

The threshold audibility of harmonics on single sine waves has "little to do" with music and loudspeakerdrivers.... or any non linear electro or electroacoustical device for that matter.

Playing complex material will produce a LOT of sideband products, not only a pure series of overtones. One should think in terms of non linearity instead of distortion. Distortion is a result of two things, a stimuli and a non linearity. Change the stimuli and the distortion changes.

People also often forget or overlook the fact that not only overtones is the result from a non linear transfer charactereistics but also subharmonics. This is intereesting with all drivers but especially tweeters.. Many people use to argue that high distortion on a tweeter in the top octave is nothing to worry about because the distortion products ("the harmonics") end up above the audible range iow. above 20kHz. This is true if we only play sinewaves on our music systems but that is rarely the situation as we all know. Low distortion in (or rather a high degree of linearity) is necessary in the whole range being excited in the device.

If the 1% limit is accepted as a level of distortion that is necessary to hear the coloration from a 2nd harmonic on a pure sinewave, then I think we need somewhere about 1/10 of that, 0.1%, on a driver for getting anywhere near transparency.

A driver with 0.5% 2nd and 3rd harmonics on pure sinewaves will be very colored when asked to reproduce music.

I don not agree that THD measuremetns are worthless because they are very indicative ( is that a word?) of the non linearity of the transducer.

/Peter

of course distortion is higher with real music than compared to sine waves, but there are also multitone distortion tests for this purpose. And while distortion is indeed higher, it is still below audibility for any half-decent driver. So it still begs the question: what sonic difference does various levels of distortion make? based on the latest research i've read in psychoacoustics, harmonic distortion is SIMPLY NOT AN ISSUE when it comes to half-decent drivers. as i've pointed out above, sonic differences between drivers (like metal cone versus poly) comes from local frequency response differences, not harmonic distortion as long as the distortion levels are low enough.

[JonMarsh]

of course distortion is higher with real music than compared to sine waves, but there are also multitone distortion tests for this purpose. And while distortion is indeed higher, it is still below audibility for any half-decent driver. So it still begs the question: what sonic difference does various levels of distortion make? based on the latest research i've read in psychoacoustics, harmonic distortion is SIMPLY NOT AN ISSUE when it comes to half-decent drivers. as i've pointed out above, sonic differences between drivers (like metal cone versus poly) comes from local frequency response differences, not harmonic distortion as long as the distortion levels are low enough.

Now, who is deciding what half decent drivers are, and who is deciding what is audible and what isn't? And what distortion level is low enough?

This reminds me a bit of the days back in the 70's when Julian Hirsch was making his claims of what distortion was audible and what wasn't, whereas Richard Heyser was doing some rather detailed and interesting testing of speaker systems in Audio Magazine, and uncovering and identifying all sorts of problems.

There is also a matter of a trained versus untrained listener; as someone who has worked live music and recording INCLUDING classical, and is constantly amazed at the crap people think they can peddle as high end at the CES show every year, I like to think I'm at least an "experienced" listener.

I do a lot of subjective testing not just with music but also with pink noise and other test signals, running the same for comparison through a high grade set of headphones. So, I'm a little critical, and often somewhat disapointed with my best efforts.

I do a lot of driver testing, and I can hear these problems as well as measure them.

Then, there's the matter of associated equipment, listening rooms, program material, etc.

I would counsel gently but strongly against making any sweeping statements- this is a very personal matter, and what fits for one person may not for another.

Also, if you use drivers within their pistonic region of cone operation, then differences in cone materials should largely be rendered moot, and the motor is the prime determinant of sonics, along with the cabinet, crossover, etc). If you don't use drivers within their pistonic range, all bets are off- it's a case of pick your preferred coloration.

Regards,

Jon

[Dennis H]

Mark K showed that 'subharmonics' are a real problem when he popped a ribbon by feeding it a 3-tone signal that should have been well within its capabilities. Ribbons don't like low frequencies and the crossover can't filter them out because they are IM products of the tones the XO is letting through.

[Piotr]

of course distortion is higher with real music than compared to sine waves, but there are also multitone distortion tests for this purpose. And while distortion is indeed higher, it is still below audibility for any half-decent driver. So it still begs the question: what sonic difference does various levels of distortion make? based on the latest research i've read in psychoacoustics, harmonic distortion is SIMPLY NOT AN ISSUE when it comes to half-decent drivers. as i've pointed out above, sonic differences between drivers (like metal cone versus poly) comes from local frequency response differences, not harmonic distortion as long as the distortion levels are low enough.

It's not that disortion is higher (again what is "distortion"..) with music, totally new frequencies appear that is not in the recorded signal or in a simple test signal. Yes, multitone tests come closer but the wider the spectral content the more crapola is created. How can you determine that the distortion is belowe audibility? If what you say are correct then we can forget about coloration of amps and sources since they are way lower in distortion than transducers... now any experienced audiotech or audiofool know that is not the case. Dithering on 16bit RBCD is on the level around 80-90dB below the signal and is still audible. Sorry to say but you are simply wrong.

Sonic differences does not come from local Fr deviations (at least not in the first place). In such case that would show up in the summed output of the driver or in the power response of the driver. What happens at low levels on "decent" drivers is that motorgenerated distortion products end up being lower that conegenerated distortion. If a driver flex as you describe, and have "local variations" on the driver cone/membrane it will generate distortion due to non linear behaviour of the cone. I have discussed this with dlr and SL on the Madboard and with dlr in a thread here. It took some time to get them to listen though.

Now, a soft driver can be made very good, but few designers knows how to do it. A very rigid driver will be a better driver.. not counting in dispersion in the upper range of the passband, but when it comes to distortion.

/Peter

[JonMarsh]

And that occurs because the ribbon is trying to faithfully reproduce what it is fed, and the crest factor for the combined signals can be high- a conventional driver just gets to the xmax limits and gives up, more or less; ribbons have some fairly prodigous drive in the motor down to lower frequencies, and will jump the ribbon out of the gap or tear it.

I've seen some interesting things with composite materials for ribbons that shows great potential for increasing their robustness- in the meantime, one has to be careful of their inherent and very different weaknesses compared with voicecoil style drivers.

[Piotr]

You can experience subharmonics with musical instruments. An acoustic guitar will give low frequency output if two higher notes are played. An electric guitar will be even a better way to try this out on. A lot of non linear stages is in the signal chain, guitar, coil-pup's, tubes, transformer, a 12" floppy paper transducer will casue IM and harmonics/subharmonics.

/Peter

[Jim Holtz]

Mark K showed that 'subharmonics' are a real problem when he popped a ribbon by feeding it a 3-tone signal that should have been well within its capabilities. Ribbons don't like low frequencies and the crossover can't filter them out because they are IM products of the tones the XO is letting through.

Hi Dennis,

I'm not a crossover guru as you know and I've followed the distortion testing very closely with great interest. However, John K. has a different take on some of the distortion testing than many claim to be the definitive test for a driver's worthyness.

http://www.musicanddesign.com/Stored_energy_3.html

I'm curious what everyone else's thoughts are about John's conclusions.

Jim

[Dennis H]

I don't think anyone, including SL, disagrees with John's conclusion:

The primary thing to remember is that given the impulse response or frequency response of a linear system, then things like burst response, waterfall or CSD plots, etc, are all just different ways of displaying the same information. The FFt conveniently allows us to move from the frequency and time domain.

Presenting the information in different kinds of graphs makes it easier to spot different kinds of problems but any of the graphs contains all the information if you know what to look for (and have a magnifying glass. )

However, John is talking about linear distortion which is just a fancy way of saying deviations from flat in the frequency response. This thread is about nonlinear distortion which is loosely defined as any tones that aren't contained in the original signal. Two different but equally important subjects.

[JonMarsh]

This would all be pretty valid if the driver were a monotonic point source. In the real world, they aren't, and different parts will radiate in different ways off in different directions when the cone behavior get's non pistonic. "Skilled" designers manage to come up with a cone profile that has a break up/flex mode which is maked (at least partly) by radiation from other parts of the cone. Classic case is the behavior of drivers like the W22 Seas, or 8-800 Eton. Nearfield investigation shows issues with the behavior, and shaped burst testing shows big time energy storage even though there isn't an apparent on-axis bump or glitch, or very little of one.

There's an old saying... I think it was Mark Twain

Figures don't lie, but liars do figure...

One can interpret this kind of data in many ways and come to whatever conclusion you think is real... this kind of thinking led to some of the flaws in the Thor D'Apollito design, where he uses a 24 dB/octave response roll off on the midwoofer equizlied to folllow the curve nicely out beyond 6 kHz. Problem is, the driver problems are still quite audible if you disconnect the tweeter and listen to the midwoofer full range through the crossover. Scads of people of agonized over W18 crossovers, trying to get them to sound right, because of ignoring or overlooking this point. I think I recall TacoD was one of them...

[cotdt]

You can experience subharmonics with musical instruments. An acoustic guitar will give low frequency output if two higher notes are played. An electric guitar will be even a better way to try this out on. A lot of non linear stages is in the signal chain, guitar, coil-pup's, tubes, transformer, a 12" floppy paper transducer will casue IM and harmonics/subharmonics.

/Peter

Good point, but they should be very low in level so I'm skeptical that they make any difference. In your posts I think you are trying to suggest that extremely low levels of harmonic distortion like -80dB are indeed audible, but the burden of proof is on you since I could not even hear -40dB distortion. In my own blind testing, harmonic distortion did not change the sound character of the speakers. It sounded like just that: frequencies that were multiples of the fundamental that added an unpleasant edge to the sound. The sound signature itself remained the same. I think a blind test should be designed to prove your points, because otherwise we're all just speculating.

Reading the stuff written by John K., and the white papers at Harmon Kardon, I'm convinced that linear distortion is what has the largest impact on the sound, even when the FR is "flattened".

Different amps that measure similarly in simple tests do sound different. I have no idea why, but to attribute it to harmonic distortion, one will first have to show that such low levels of distortion are audible.

[Dennis H]

Nearfield investigation shows issues with the behavior, and shaped burst testing shows big time energy storage even though there isn't an apparent on-axis bump or glitch, or very little of one.

But the nearfield measurement also shows big deviations from flat response, right? So the takeway, at least for me, is where you put the mic, not what kind of graph you use to display the data -- could be a simple frequency response or an SL-style ETC at the problem frequency.

[JonMarsh]

It's not an FR response issue due to measurement technique with mic placement 3" on axis in front of the diaphragm, though eventually above 3-4 kHz you see some roll off due to phase arrival differences.

Nearfield measurements on drivers that are behaving quite pistonically don't show issues in FR- it's the one that aren't behaving pistonically that do. But the energy storage in a W22 is quite measurable at 1400 Hz in a 1 meter measurement, where the FR curve doesn't look at all suspicious.

It's also pretty clear when the cone is "gang agley" when 1/4" near field at various locations on a fairly pistonic 8" driver shows nearly identical response curves up to the breakup peak, while on something more irregular, like a SS 8" kevlar, or W22 Seas, things are not at all similar in the center or periphery of the driver, eh?

Having looked at a lot of drivers in detail this way, before having the opportunity for shaped sine burst testing, I've found it the most reliable and inexpensive way to detect non-pistonic behavior.

And whereas there are many acoustic radiating surfaces intended to work this way, like an acoustic guitar or violin, loudspeakers shouldn't, IMO.

Somewhere on here I posted ages ago a comparison of the lowly HiVi M8a and SS 21W8554 (which was used by SL in some of the Audio Artistry models). The SS has a better motor, but the HiVi has a much better cone, IMO.

And for the record, this is a free forum, and everyone should follow their own heart and path to what they believe is audio nirvanna. But if asked what I think works and why, I'll be straight about it, since I'm not trying to make money or keep trade secrets, I've got nothing to hide and no axe to grind. Other than ego jollies, I guess. :B


No question some will like a soft cone that gradually decouples and maintains a smoothish on axis behavior up to 5 kHz and beyond. But there just may be unexpected side effects to that approach - but let every one try what they think will work and please them, and let the ultimate comparison be to live acoustic sound.

After all, I designed some speakers 20 years ago with Seas PP and paper cone drivers. Just wasn't ever all that happy with them. I've made more mistakes than a lot of posters here have made projects- at last count, prior to the Isiris, the new CC three way in development, and the M12ta, over 50. Experience is a tough teacher, but some of us lunkheads don't learn any other way. :roll:

~Jon

[Mark K]

A couple of comments to clarify the issues. Some of these have been made, but are worth repeating/regurgitating.

Linear distortion is just nonflat frequency response, and in theory, can be completely corrected for a single axis. The rub, as pointed out, is that it is very difficult to know what to the true frequency response of the driver is. The typical windowed measurement has a fair bit of inaccuracy and so equalizing an inaccurate curve to flat leaves the listener with a certain amount of uncertainty regarding the true frequency response. However, the comments about our ability to know "true" frequency response are probably overly negative. By using relatively pistonic drivers below the region of cone break up, and looking at near field plots, you can certainly get a very detailed look at the frequency response quite low in frequency and developing appropriately flat system curve. Another issue in regard to linear distortion is cone break up higher up in frequency and its effects on off axis frequency response. This cannot be corrected in the equalization has all the more reason to cross a rigid piston at relatively low crust over frequency.

I agree that subtle variations in frequency response are responsible for changes in timbre and are quite key. Obviously these changes in timbre occur at all volume levels, which is quite different from nonlinear distortion, which is heavily level dependent.

So the moral of the story is that linear distortion can be mitigated by carefully controlling to a flat frequency response and observing of axis behavoir. Care must be taken in measurement, however there is no reason that a thoughtful careful DIY'er cannot achieve reasonably flat frequency response.

But what about nonlinear distortion? How important is it? I find it somewhat humorous that people point to a Dr. Geddes' work and use him as a reference for suggesting that nonlinear distortion is unimportant. Please take a look at the loudspeakers that he designs. They are clearly made for high output levels. I don't see him hawking any single driver setups on his site. The goal is of course to choose drivers and a system of drivers so that any nonlinear distortion effects are unimportant at the target listening level.

I can assure you that nonlinear distortion is quite audible. It is painfully audible in most two ways that are played loud. Once you know what you are listening to, it becomes quite annoying. Certain types of distortion are quite pleasing, hence the plethora of mediocre two ways that abound.

But there are number of problems with nonlinear distortion testing. By definition, there are plethora of tones and signals that can be tried, different multi-tones in all of these result in slightly different distortion spectra. Some distortion test signals are somewhat inherently unpleasant and some are quite pleasing.

In the bass region, distortion is often not to recognized as distortion per se.
Here's a quote from my primer (easier to cut and paste.)
--
The ear’s decreasing sensitivity at low frequencies is a significant problem and lead to the need for quite low distortion at the low end. To reiterate, if xmax is defined by a manufacturer as 10% harmonic distortion, then say, at 20 Hz, when the 10% figure is reached, the amplitude of the 40 Hz 2nd harmonic may be down only 20 dB. The problem is, the relative sensitivity of the ear has increased by 20 dB from 20 to 40 Hz. So what are heard are two tones, 20 and 40 Hz of equal magnitude. So, even though the driver is still operating in its linear range, obvious audible distortion is occurring. And this doesn’t take into account problems with the 3rd order product or higher. The 3rd order product has to be down almost 40dB just to be perceived as equal in magnitude to the 20 Hz fundamental. Wow!
--
Now this isn't necessessarily perceived as a distortion product, just a lot of 40 Hz bass. It still distortion. Try this experiment. Play music at a relatively loud level on your typical 7" woofer based system and stick your finger on the woofer. The sound actually gets a little bit richer in lower octaves. When you measure the spectrum is actually quite a bit of distortion generated. Not suprisingly, doing this lightly adds to the richness of the bass notes well before audible distortion is noted. A you can try the same trick on a tweeter, though they really isn't much a richness just a lot of painful distortion. Voice coil rub creates a lot of higher order distortion which is in general painfully obvious.

Based on what I know of the Vifa XG 18, I would not recommend it for use in a two-way. It's xmax limited distortion products are way too high, much more so than the equivalent Peerless HDS exclusive. Does this mean that in absolute terms the Vifa driver isn't good? Not all. However excursion would have to be limited somehow. This would make a nice driver as a midrange unit although it could use more sensitivity.

Finally, I think it's best if you take a step back and ask yourself to look at the problem in a somewhat different way. I'm quite convinced that there is no clear answer to the question, "at what level is distortion audible?" But instead, go with the premise that in general all things being equal, one should choose the lower distortion driver. If the Peerless HDS has superior excursion and Le distortion numbers, and has an equally flat frequency response curve, and an equally favorable off axis frequency response profile, why would you use the Vifa?

Hey, I think the Vifa xg18 is a nice driver and I have a pair sitting in my garage. But every time I look at that pair thinking to myself, "Hmmm, I should use that pair," then I will mind wanders back to the Peerless, which seems to be more than it's equal in every category.

What about Zaph's data? The Vifa has remarkably low distortion in the passband. That's an interesting question. It may be that the Vifa has very well controlled cone behavior, or that the HD products are showing us something slightly different. In any case, I wouldn't object to someone using the XG18 or a pair as a midrange pair. At a somewhat higher xover point, the excursion based distortion would be as controlled as the peerless. (I.e. the peerless crossed at 150 as the highpass in a mid filter would do the same as the XG18 crossed at 250-does that make sense?) So I think it's a bit of a wash as to which midwoofer would be better in the three-way. You could use the RS 180 and that would allow you a lower crossover point. However you would have to use a steeper filter and perhaps crossover lower. Conversely, the xg18 would allow a shallower slope and maybe a higher crossover point to the tweeter. It would be unclear to me though what's going on with cone breakup and off axis behavior.

So there, that settles it. Either the RS 180 or XG 18. Definitely one of these two. And nonlinear distortion matters :

This post edited with Dragon. Get serious.

[JonMarsh]

Very good points and exposition, Mark. And the funny thing is that I bought the Vifa XG18's thinking to use them as a midrange driver....

[Piotr]

For anyone that doubts audibility of driver non-linear behaviour, just pick a driver/loudspeaker and play sinewaves of different frequencies at different levels. Best is maybe to use two drivers of different make but similar type and size.

SL suggests using multitone signals to further push the drivers.

It´s very obvious that non linearities causes audible harmonics doing such simple listening tests.

/Peter

[Piotr]

Good point, but they should be very low in level so I'm skeptical that they make any difference. In your posts I think you are trying to suggest that extremely low levels of harmonic distortion like -80dB are indeed audible, but the burden of proof is on you since I could not even hear -40dB distortion. In my own blind testing, harmonic distortion did not change the sound character of the speakers. It sounded like just that: frequencies that were multiples of the fundamental that added an unpleasant edge to the sound. The sound signature itself remained the same. I think a blind test should be designed to prove your points, because otherwise we're all just speculating.

Reading the stuff written by John K., and the white papers at Harmon Kardon, I'm convinced that linear distortion is what has the largest impact on the sound, even when the FR is "flattened".

Different amps that measure similarly in simple tests do sound different. I have no idea why, but to attribute it to harmonic distortion, one will first have to show that such low levels of distortion are audible.

The level is high enough with instruments so you can hear a new low frecuency note.

I also do not understand when you say that the harmonics in your test did not change the sound, but that it added an unpleasant edge to it...? Sounds like coloration to me.

It´s important to understand theory and physics of music here. All musical notes has harmonics. Those harmonics is what gives different instruments (and also notes in a scale) their sound in the static portion of the note, transient and decay of course affect the sound greatly as well. BUT not all harmonics are perfectly in a mathematical sense. The non-perfekt vibration of thick strings for example (think piano and acoustic steel stringed guitar) causes the harmonics to shift slighly. Let's say a 500Hz fundamental with a 2nd harmonic at 1000Hz. In some instruments the octave may end up being 1001Hz instead. These shifts are more obvious the higher the order of the harmonic. If you play back such an instrument on a non linear transducer the 2nd harmonic of 500Hz may be a more "perfect" octave at 1000Hz. Now this 100Hz note does not match the slightly imperfect 2nd harmonic from the instrument at 1001Hz. What we get here is a "wobbling note" of 1Hz.


John K's paper is good. However there is one more mechanism in play here. If a cone is soft and give rise to peaks and dips in the Fr. response we can EQ it flat and get rid of the deviations BUT, the flexing will still be there. And flexing always means cone generated distortion. The springiness of the flexing cone (or dome) is not linear but casues HD and IMD as other non linear elements of the driver. So, when two notes are playing back on a flexing cone, modulation will be the result. A low frequency will move the whole cone, a higher frequency will more or less decouple the inner part of the cone. But when the lower Fr. moves the cone outward, the inner part will be pushed out slightly more than the outer part of the cone. This will stretch the material and increase the springiness so when the higher Fr. push on the cone the tension will be different depending on where the cone is for the moment. Operating a cone below its major break up does not mean that the cone is truly pistonic, no matter how stiff the cone is. Flexing will always be there.

The grainy flat and sometimes harsh sound one can hear from some speakers is not about small deviations in the frequency response but from HD and IMD products. These products are in the order of -40dB to -60dB or so and mask low level information in the recorded material. This affects resolution of the timbre and acoustic/ambience information. Small frequency deviations simply add or subtracts a tiny amount from the harmonics in the recorded material and tend to still sound natural.

On the other hand one could argue that all linear distortion (frequency deviations) in a transducer has elements of non linear distortion... food for thought..

/Peter

[tktran]

Imho...

If there's anything that we get carried away with, it is linear distortion. An example of that is our fanaticism of hard coned drivers, such as Dayton RS, HiVi, Accuton etc drivers...

Don't be scared of the soft cone, or non-pistonic drivers. Where is the evidence that pistonic drivers are the only path to audio nirvana? :-)
Where are all the microphones and studio monitors that used metal diaphragms/cones?

Don't be afraid of the soggy cone drivers. ;x(

Some of them have great/better/best linear and non-linear distortion,
eg. AA 6.5, 8531G, 8945A/P..

[Piotr]

True, but if everything else is equal, a stiff cone will have less distortion han a flexing cone.

I thought pretty many studios used drivers/speakers with metall drivers. B&W Tweeters Alu/diamond, TAD Beryllium, JBL has compression drivers with Beryllium I think..? Genelec uses Seas alu tweeters right?

When it comes to microphones it's my opinion that the most transparent sound comes from small capsules that push the membrane resonances as high up as possible. I have a pair of QTC1 (1/4" capsule) and no 1" capsule can match those when it comes to naturall resolution and freedom from grain. Bigger capsules often have a bump at 5-10k due to the resonance. That could be EQ'd of course but the question is, will the performance match a small capsule?

/Peter

[JonMarsh]

For anyone that doubts audibility of driver non-linear behaviour, just pick a driver/loudspeaker and play sinewaves of different frequencies at different levels. Best is maybe to use two drivers of different make but similar type and size.

SL suggests using multitone signals to further push the drivers.

It´s very obvious that non linearities causes audible harmonics doing such simple listening tests.

/Peter

Amen, brother! As personally noted above, too.

[dwk]

But what about nonlinear distortion? How important is it? I find it somewhat humorous that people point to a Dr. Geddes' work and use him as a reference for suggesting that nonlinear distortion is unimportant. Please take a look at the loudspeakers that he designs. They are clearly made for high output levels. I don't see him hawking any single driver setups on his site. The goal is of course to choose drivers and a system of drivers so that any nonlinear distortion effects are unimportant at the target listening level.

Well, since I brought Geddes in, I suppose I should defend or at least clarify my comments, since I think we're saying the same thing.

I originally referred to Geddes work as an example of something that is headed in the general direction of objectively quantifying speaker behavior, as well as mapping out the subjective space in which they are evaluated. I wasn't really intending to use it as support for the position that non-linear distortion doesn't matter, as I agree that he doesn't really say that. However, it does support the notion that it is only one consideration in speaker evaluation, and once it is 'low enough' (whatever that means) then other factors become equally or even more significant.

I think the way to interpret Geddes position is that he isn't saying that non-linear distortion is not audible. What he is saying is that 'mainstream' drivers like the B&C are now low enough in distortion that it is no longer a driving factor in speaker design. i.e. the $250 B&C compression driver is fine - no need to spend $1500 or whatever on the TAD.
So, my read on it is that his findings are that the incremental decrease in disortion going from a merely 'very good' pro driver to a 'state of the art' pro driver is unlikely to be highly audible at home levels compared to other speaker design considerations. Taking the step from an entry-level Eminence to the B&C this may not be the case, though, and I suspect that for Earl 'hi fi' drivers simply aren't in the discussion.

In this sense, your point about designing for the listening level is correct. Since Earl targets fairly loud levels, he designs a speaker that can play 10-15dB above the target 'comfortably', and by doing so he goes a long way towards ensuring that distortion isn't much of a concern.

[Mark K]

Well, I wasn't so much responding only to your post. I've seen a number of posts on other forums as well that basically use Dr. Geddes' comments to justify some rather dubious design choices. I agree that below a certain threshold, nonlinear distortion is unimportant. Still, it's difficult to know exactly where that point is. Better to err on the side of caution. Or, at least account for it in your design. I would not go so far as to say that with all or even most modern drivers, distortion is low enough to be unimportant.

DIY get togethers are great places to see this. Generally the rooms are a bit larger and speakers are driven a bit harder. Next time you go to a diy, and someone is playing something, ask yourself, is it concert level? If not, try to crank up the volume until it is. I guess my idea of loud is a bit different than others, but most 2 way systems can't cut it. There is obvious distortion and compression.

Sure, if you've got a small apartment and all you listen to is a female jazz vocalist accompanied by an acoustic trio...well, nonlinear distortion is relatively unimportant. A ribbon mated to a W15 will probably sound outstanding. The slight distortion that would exist would probably serve to enhance the perception of realism anyway. (I'm not kidding. I suspect it would.) And I've actually thought about putting together something like this for low-mod volume levels. But it's not really any kind of system to reproduce serious spl's.

[Dennis H]

Jon, sorry if you thought I was disagreeing with your nearfield measurement technique. It's one of the coolest methods I've seen and you can do it with minimal measurement gear. Heck TrueRTA could do it.

Nearfield measurements on drivers that are behaving quite pistonically don't show issues in FR- it's the one that aren't behaving pistonically that do.

Yeah, that's what I was trying to say. You said it better.

But the energy storage in a W22 is quite measurable at 1400 Hz in a 1 meter measurement, where the FR curve doesn't look at all suspicious.

I'm guessing that the ETC of a shaped 1400 Hz tone is including stuff that happens over some frequency bandwidth, not just a discrete 1400 Hz measurement. The FR dip above 1400 is probably showing up in the ETC. So back to the old rule of thumb before we had shaped bursts and ETCs: the 'perfect' driver should be well behaved, both nearfield and farfield, both on and off axis, for some amount beyond the nominal XO frequency.

[Feyz]

So it still begs the question: what sonic difference does various levels of distortion make? based on the latest research i've read in psychoacoustics, harmonic distortion is SIMPLY NOT AN ISSUE when it comes to half-decent drivers.

Who decided that, on what basis, what is a half decent deriver, what latest research? What kind of distortion are we talking about, AM distortion, FM distortion, a mixture of it? What kind of IM distortion pattern? The reality of the matter is drivers are the weakest link by a large margin both in terms of frequency amplitue/phase linearity and non-linear distoriton. It is one thing saying frequency amp/phase linearity is more important than non-linear distotion, but another thing saying non-linear distortion as happens with most half decent drivers are below audibility so they don't matter. Then I would ask why Scan-Speak drivers had been so successfull over the years, what set them apart over the others? Their frequency response even though not bad, weren't perfect either. But they had motors that had attention was paid to non-linear distortion, which IMO set them apart which made them choice of high quality speaker for many years. Now others have caught up.

[Bri]

You can experience subharmonics with musical instruments. An acoustic guitar will give low frequency output if two higher notes are played.
/Peter

Isn't this just a beat note occurring, or is there a difference between beat notes and what you are describing?

[Piotr]

I'm not sure what "beat" means..

One thing is when you tune strings, you will then hear a wobbling sound that goes away when two strings are fretted and tuned correctly. Guitarists commonly use this when tunig by ear. But then there is the subharmonic that simply ads a low frequency tone to two higher notes with a specific distance/relationship.

/Peter

[Tommythecat]

"beat" refers to "the wobbling sound"

[Bri]

I'm not sure what "beat" means..
One thing is when you tune strings, you will then hear a wobbling sound that goes away when two strings are fretted and tuned correctly. Guitarists commonly use this when tunig by ear. But then there is the subharmonic that simply ads a low frequency tone to two higher notes with a specific distance/relationship.
/Peter

"Beat frequency" is probably the more common term.

http://hyperphysics.phy-astr.gsu.edu/hbase/sound/beat.html

http://en.wikipedia.org/wiki/Beat_(acoustics)

[Piotr]

Beat it is then, thanks!

/Peter

[JonP]

Yep, works that way with RF, audio, any cyclical signal.... F1 and F2 together sums to F1+F2 and F1-F2 as well as the original signals. When you get very close in frequency, the F1-F2 product becomes a few Hz, or less, and you hear a "beat" that eventually goes to zero.

A side note on "subharmonics"... I've read, and had my wife and her flute teacher test it out.. you have two flutes play a certain two notes, and you hear a "wolf tone", a lower tone which must be the F1-F2, but high enough to sound like 200Hz... The question is, is it produced more in the air, or in the ears of the listeners? Flutes don't have a vibrating mechanical object, like other instruments, and speakers. Part of it seems to be if you are very close (i.e. it's loud) you hear it, but don't if you are a bit farther away... Some of these effects might be due to IMD, or just mixing, inside of your ears. Maybe someone knows the physio/psycoacoustics on this better...

Of course, in that case, it's not the speaker's fault!

[Dennis H]

A side note on "subharmonics"... I've read, and had my wife and her flute teacher test it out.. you have two flutes play a certain two notes, and you hear a "wolf tone", a lower tone which must be the F1-F2, but high enough to sound like 200Hz...

Not quite the same I think, although the principle is the same, because the tones will all be coming from a single tweeter in a speaker. A ribbon trying to play that 200 Hz difference tone will be in trouble if it's loud enough.

[cotdt]

To make my point, I made a frequency response measurement of an aluminum woofer, a paper woofer, and a poly woofer. Of course it's not the "true" frequency response but it will serve. They all have similarly low nonlinear distortion.

This is the unsmoothed frequency response of my Dayton RS180-RS28A bookshelf speakers (ModulaMT designed by Jon). The port response is not summed so let's just focus on the midrange frequency, which has -+1dB fuzzy peaks.



And then I measured my (treated) paper coned speakers, which also had a flat response, but instead of the +-1dB fuzzy peaks, it had +-3dB fuzzy peaks. Note that this treated paper woofer had a rigid cone using the touch test, but I guess it's not as pistonic as metal cones.

And last but not least the poly cone woofer, in this case my Ascendant Audio Poly 6.5". Low nonlinear distortion as measured by Zaph, but it had +-5dB fuzzy peaks. Note that this is a poly cone, not a composite poly whatever cone, which is much more rigid and should not be lumped into the poly cone category.

In short:
-Dayton RS180 had +-1dB "fuzz"
-Treated paper cone had +-3dB "fuzz"
-AA Poly 6.5" had +-5dB "fuzz"

Subjective:
-RS180 speaker is very detailed-sounding, spacious, good imaging, but soundstage not as wide.
-AA Poly 6.5" matched the resolution of the RS180, but not as detailed-sounding, more laid-back, wide soundstage but imaging is not as good.
-Paper is in-between the two in every aspect.

My theory is that the higher linear distortion of nonrigid cones (like poly) creates "time-smear" due to energy storage which makes it sound less detailed, more laid-back, poorer imaging, but wider soundstage. This effect is the same regardless of volume, making it less likely that nonlinear distortion is what is causing the sonic differences.

Don't get me wrong I love the AA Poly 6.5", but I wouldn't say this driver is low in distortion when linear distortion is accounted for.

[Feyz]

Ear/brain is nonlinear so generates the beat tones even though they don't exist in the sound waves.