Distortion in loudspeakers

[oivavoi]

Here's a discussion:

https://www.bksv.com/media/doc/BO0385.pdf

Good stuff. Reading now.

 

[watchnerd]

When I simplistically compared my electrostatic (ML reQuest) and cone (JBL LSR 308) for distortion. The low frequency hash is not coming from the speakers... Room noise and USB mic noise and such. The results vary with frequency tested, too.

Once again showing that electrostats have significantly lower distortion than dynamic speakers.

And, yet, this doesn't seem to lead to them being preferred in listening tests.

I don't know if this is a conditioning issue (i.e. they sound "weird" to people) or that speaker distortion just isn't that important for some reason, or at least not until it reaches egregious levels.

I will say one thing, though, just based on personal experience:

The subjective assessment of "loud" seems to be related to distortion.

When I've heard low distortion monitors like ATC playing loud, I've been surprised by the SPL meter reading much higher than I would have subjectively guessed.

 

[watchnerd]

Checked out their website now. These seem like seriously good figures. Here's the distortion on the small two-way SC205, with a 5-woofer:

And here's the larger four-way (or technically I would call it a 3 or 3.5-way) SC407:

Seems pretty impressive to me.

Yeah, they also publish polar dispersion plots for all their speakers, too!

 

[fas42]

Here's an idea for how to do it in a different way: Shouldn't it be possible to just play a piece a music through a loudspeaker in a anechoic chamber, and record it using a microphone with known properties? And then compare the original digital file with the recorded file using audiodiffmaker or something similar? The resulting difference would be some kind of distortion, as far as I can understand. I'm not sure how it could be quantified, though. The difficult thing to know would be if the measured differences were due to the microphone or the speaker. But what one could do in this way, was to compare several speakers, and see which one had the least distortion. So it could be a relative indicator of distortion, at least. And one could repeat it with different kinds of musical tracks. Or am I missing something here?

There are ways of calibrating the test setup to differentiate microphone from speaker distortion - say, moving the microphone further away while keeping the level being output from the speaker constant, and seeing if there is variation in the spectrum, etc.

The concept of using something like DiffMaker in that way is excellent, but the reality is that these tools haven't been developed enough. Diffmaker itself is too flawed, essentially unusable to get sensible results - it's wide open for another version of that approach, which is sufficiently refined to get useful information.

 

[fas42]

I don't know if this is a conditioning issue (i.e. they sound "weird" to people) or that speaker distortion just isn't that important for some reason, or at least not until it reaches egregious levels.

I will say one thing, though, just based on personal experience:

The subjective assessment of "loud" seems to be related to distortion.

When I've heard low distortion monitors like ATC playing loud, I've been surprised by the SPL meter reading much higher than I would have subjectively guessed.

Very much agree with this. Unpleasant distortion will make a system sound much louder than what meter readings show; "acceptable", or low level distortion is easily accommodated by our hearing system, and gives the subjective impression of the sound being intense, or powerful, at high volumes - just like natural sounds. This is part of the package of "convincing" sound - the distortion of the speaker "doesn't matter" to our ear/brain; what registers, overwhelmingly, is the impact of the "musical message".

 

[oivavoi]

Once again showing that electrostats have significantly lower distortion than dynamic speakers.

And, yet, this doesn't seem to lead to them being preferred in listening tests.

I don't know if this is a conditioning issue (i.e. they sound "weird" to people) or that speaker distortion just isn't that important for some reason, or at least not until it reaches egregious levels.

I will say one thing, though, just based on personal experience:

The subjective assessment of "loud" seems to be related to distortion.

When I've heard low distortion monitors like ATC playing loud, I've been surprised by the SPL meter reading much higher than I would have subjectively guessed.

This goes back to the thread about blind testing procedures: I'm not sure how much importance we should attach to the preferences that show up in such listener tests, after statistical tests have been applied to the responses and some slight statistical tendency can be found in one direction or the other. Some recent listening tests show that people have started to prefer 128-bits over CD quality for example, since that's what they're used to from youtube. I would assume that if you re-run these listening tests on people who had electrostats at home, the results would have been different. Such tests, also, rarely incoporate actual usage at home - typical loudspeaker placement in the living room, listening outside the sweetspot, different kinds of volume and musical material, etc.

In short: I don't see the work of Toole and Olive on listener preferences as the final answer. I think the goal of hifi should be to reproduce the signal as accurately as possible - when it comes to frequency, time and non-linear distortion. When it comes to preferences, each and everyone of us can change the result with tone controls to suit our particular tastes. I don't think there's a "correct" answer on dispersion patterns either.

 

[RayDunzl]

Recent versions of REW include the ability to add specified levels of harmonic distortion to the sine wave generator for your listening pleasure:

 

[watchnerd]

In short: I don't see the work of Toole and Olive on listener preferences as the final answer.

Sadly, I think Toole, Olive, and Geddes will be the pinnacle of loudspeaker research, the last great experimenters in 2-channel stereo from external speakers, the last hurrah of the hi fi paradigm that started in the 1950s.

Most of the new research money is going into VR/AR and HRTF.

We're probably in the "peak speaker" epoch right now. 20 years from now it will all be personalized to your head, ear, etc.

 

[oivavoi]

Sadly, I think Toole, Olive, and Geddes will be the pinnacle of loudspeaker research, the last great experimenters in 2-channel stereo from external speakers, the last hurrah of the hi fi paradigm that started in the 1950s.

Most of the new research money is going into VR/AR and HRTF.

We're probably in the "peak speaker" epoch right now. 20 years from now it will all be personalized to your head, ear, etc.

You may be right. Which makes it rational to find the absolutely best speaker money can be around this time, and pass it on as a heirloom... As for me, I will never really warm to headphones. Yes, I suppose that the implementation of head tracking etc can create convincing phantom images. But it will never beat the experience of listening to actual physical soundwaves hitting your entire body. Not for me at least.

Btw, funny that you mention Toole, Olive AND Geddes. They have all done technical research and pyschoacoustic tests. Neverhteless, their design philosophies couldn't be more different. Toole and Olive aim for wide dispersion. Geddes aims for the opposite. Is one of these approaches clearly misguided? I would say no. Both approaches have their merits. Which again shows that it may prove elusive to use listening tests to find the "right" way of building a loudspeaker.

 

[Blumlein 88]

Once again showing that electrostats have significantly lower distortion than dynamic speakers.

And, yet, this doesn't seem to lead to them being preferred in listening tests.

I don't know if this is a conditioning issue (i.e. they sound "weird" to people) or that speaker distortion just isn't that important for some reason, or at least not until it reaches egregious levels.

I will say one thing, though, just based on personal experience:

The subjective assessment of "loud" seems to be related to distortion.

When I've heard low distortion monitors like ATC playing loud, I've been surprised by the SPL meter reading much higher than I would have subjectively guessed.

I think one needs to be careful of over-generalizing here. To my knowledge at least of publicly available info, Harman only used one particular ML speaker. It also had a less than great natural frequency response along with naturally different directional characteristics. If there is additional info I would love to know about it. I don't know of other comparitive listening tests done blind where electrostats did poorly in the test.

Now Harman has an algorithm which has a high correlation of predicting how well a speaker will score in their listening tests. Presumably they could do the spin-o-rama measurements of a number of ESLs and see what the predictions were. We know for certain some ESLs will do better than others. I wonder if any of those available would be predicted to do extremely well? Some ESL's are naturally rather good with tight near level FR.

I can see why further investigating this is of little benefit to Harman as a commercial company. ESL's are not inexpensive to make, not easy to make in large numbers, and not attractive to nearly so many people as conventional quality speakers.

I have experienced what you describe in some speakers that sound clean so you are listening louder than you would have guessed. For some reason ESL's are different in the other direction. At quite low levels you hear so much detail you normally would have to turn things up much louder to hear as much.

They wouldn't it to us I don't guess, but it would be nice if we had Harman's algorithm for predicting listener satisfaction from spin-o-rama testing.

 

[oivavoi]

Harman/Toole/Olive are not the only one who has done such studies, also. Here's a nice review from 2009: https://www.researchgate.net/public..._sound_quality_-_a_review_of_existing_studies

 

[DonH56]

Maybe, but not so pure as the beer/kahlua bottle.

Actually, I'm a bit surprised to see it.

http://hyperphysics.phy-astr.gsu.edu/hbase/Music/flutew.html

View attachment 5323

Phooey, had a long reply and somehow lost it, network glitch...

The tone of an instrument depends upon the player as well as the design of the instrument (bell flare, shape of the tubing, parameters of the strings, body cavity and venting, etc.) Thinking back, it may have been wind instruments I had in mind; not sure about strings (insert stereotypical brasshole joke about strings here). I remember the studies clearly showing a nearly pure sine wave from a flute but have no hope of finding those old papers now. Brass, and to some extent reeds, have much more complex signal spectra and are not purely harmonic.

My orchestra does not allow beer or kahlua on stage, and few pieces call for that particular instrument...

Very generally/loosely speaking horns and panels exhibit lower distortion than conventional designs, with horns providing the edge in efficiency (sensitivity, dB/W) and maximum output. Panels have limited excursion and tend to distort very badly very rapidly when you crank the volume, one reason for adding subs and hybrid systems like ML. Horns (and related compression drivers) can too but are so much more sensitive (dB/W) that it is generally less an issue for them. Panels also suffer from modes in the panels, and dispersion that varies with frequency.

I remember seeing distortion plots for B&W 800-series speakers and being impressed but am not sure where I saw them... Way back in the 1980's B&W was showing off their laser interferometry tester to show cone modes and breakup, I got to see it in action, very cool.

 

[Blumlein 88]

Harman/Toole/Olive are not the only one who has done such studies, also. Here's a nice review from 2009: https://www.researchgate.net/public..._sound_quality_-_a_review_of_existing_studies

I am aware they are not the only people doing research on directivity. I notice in one mentioned in this paper that dipoles did poorly less than 1 meter from the wall, but very well more than 1 meter. They suggest it has to do with reflection of the rear wave, and it may. The other factor is the low end response of dipoles is effected significantly between such positions while that of a box speaker is effected to a much smaller degree. You can get a big dip or add a good half octave extension with careful placement of dipoles relative to the rear wall.

 

[watchnerd]

As for me, I will never really warm to headphones. Yes, I suppose that the implementation of head tracking etc can create convincing phantom images. But it will never beat the experience of listening to actual physical soundwaves hitting your entire body. Not for me at least.

It's interesting that you say that because many of the millenials that work for me feel the opposite -- they're so accustomed to listening to personal audio from smart phones since their tweens, and playing video games with headphones on, that they think speakers sound "weird".

 

[watchnerd]

I have experienced what you describe in some speakers that sound clean so you are listening louder than you would have guessed. For some reason ESL's are different in the other direction. At quite low levels you hear so much detail you normally would have to turn things up much louder to hear as much.

This is true. But ESLs also never seem to induce the same somatic sensation as dynamics, no matter how loud I crank them.

 

[fas42]

I have experienced what you describe in some speakers that sound clean so you are listening louder than you would have guessed. For some reason ESL's are different in the other direction. At quite low levels you hear so much detail you normally would have to turn things up much louder to hear as much.

A competent system delivers both. For different reasons. If you wind up the volume the electronics maintain their composure, and the subjective sense if that you're getting closer to the sound, that it's getting "stronger"'; then if you reduce the volume dramatically, to background levels or less, everything is "still there", no part of the "information" disappears - the reason for the latter is that low level noise issues, or lack speaker suspension flexibility are not masking the detail. The downside with normal quality dynamic drivers is that their suspensions always have to be brought up to a stable equilibrium, from cold - my technique is to hammer them hard; I don't bother listening to them seriously at start up, because I know they will not deliver the low volume clarity until given a decent workout.

 

[Fitzcaraldo215]

I am aware they are not the only people doing research on directivity. I notice in one mentioned in this paper that dipoles did poorly less than 1 meter from the wall, but very well more than 1 meter. They suggest it has to do with reflection of the rear wave, and it may. The other factor is the low end response of dipoles is effected significantly between such positions while that of a box speaker is effected to a much smaller degree. You can get a big dip or add a good half octave extension with careful placement of dipoles relative to the rear wall.

All true in my experience with ESL dipoles since the mid-80's. My rule of thumb for placement has been at least 5' from the wall behind. I violate that, but only for my surround and back channels in 7.1. My main fronts are just over 5' from the front wall, i.e., the wall behind them.

The 5' idea was based on some reading way back about the ear's ability to separate direct from reflected sound. That distance caused a delay of about 10 milliseconds, which was thought an adequate threshold, as I recall.

I know what you mean in your last sentence, but I think you are referring to the wall behind the dipoles, which is actually the front wall of the room from the listening position. People, especially reviewers in magazines, do that front wall/rear wall thing all the time, and it slips past the editors regularly.

I do not think dipoles are ever a good solution in the deep bass.

 

[Blumlein 88]

This is true. But ESLs also never seem to induce the same somatic sensation as dynamics, no matter how loud I crank them.

They are limited in dynamics. Best impactful ESLs I heard were some of the M1 Soundlabs driven by a pair of true 1 kilowatt amplifiers. Didn't have the jump factor of horns, but it had a potential jump factor normally absent in panels. So I think the limitation is power handling of the panels and power to make them work.

 

[Fitzcaraldo215]

This is true. But ESLs also never seem to induce the same somatic sensation as dynamics, no matter how loud I crank them.

Well, I am using all ML dipole hybrids and a really good JL fathom f113 sub. It is mainly Mch listening I do, so many more drivers and amps are involved. The sub and DSP room correction are huge contributors to awesome dynamics. My Spectron Class D amp for the front channels also delivers over 900 watts into 4 ohms. I listen to mainly classical, mind you. But, I do swing to rock on occasion. I have never felt the slightest sense of any dynamic limitation, even in comparison to other systems using direct radiators, horns, etc. Yes, horns will play louder, but I have zero interest in going there and ruining my hearing.

 

[watchnerd]

Well, I am using all ML dipole hybrids and a really good JL fathom f113 sub. It is mainly Mch listening I do, so many more drivers and amps are involved. The sub and DSP room correction are huge contributors to awesome dynamics. My Spectron Class D amp for the front channels also delivers over 900 watts into 4 ohms. I listen to mainly classical, mind you. But, I do swing to rock on occasion. I have never felt the slightest sense of any dynamic limitation, even in comparison to other systems using direct radiators, horns, etc. Yes, horns will play louder, but I have zero interest in going there and ruining my hearing.

I have 2 sets of Martin Logans, and 3 different dynamic speakers lying around (2 sets of Dynaudios, both active and passive, one set of little JBLs actives), so this is not coming from a lack of comparisons.

I didn't say dynamic limitation (somebody else did), I said somatic feel.

I'll pick a classical example, since that's your genre:

If I play Dvorak's 9, New World, 4th movement, I get more of a somatic effect from dynamic speakers than I do from my ML's.

The ML's have a bigger, more lifelike soundstage, but the dynamic speakers have more impact in the orchestral power zone (upper bass / lower mid).