Audibility of distortion in horns!

[Pan]

How do you then explain the complete lake of interest in this subject by people like Tool and Olive?

I have not spoke to either of them so I have no idea how they are reasoning about it.

The scientific studies say that this is not correct for the "better" and higher data rate coders. They found no audible artifacts for any listeners. This was published in the AES.

The discussion was about 90% data reduction which means 128kbit/s or thereabout. You claimed transparency for that amount of data reduction and I question it.
I did adress higher bitrates as being good for many situations. There are listeners who can pick audible distortion at the higher rates as well but I'd say that for the average listener and music style/production on a good rig 192-256 is starting to get hard to hear. Don't remember if I ABX'd 320 or 256 with detection. The specific tune was a well produced dynamic electro/pop/jazz thing with lots of ambience. Compressed rock and similar was harder and stopped at lower rates.

None the less this is of course really cool, that such huge amount of information can be removed with so little (and something no) audible effects.

When listening to things like this and performing tests, it's one thing to throw in a bunch of students and let them listen on their own, and another thing to use more experienced listeners and also to guide them. Meaning if someone hears an artefact it use to be possible to help other hear it as well, once you point it out to them.

So when setting up studies, you need to be specific about what you are searching for. Listening capabilities of average joe or the limits of human perception period.

 

[PMA]

That is a fascinating plot with multiple 1/2 orders. I suspect strong thermal modulation going on there. What kind of tweeter is that? I'd like to do those tests myself.

There are "oil-canning" diaphragm effects that can do this, but that takes an especially weak cone to do.

By the way - Prague is my favorite city in the EU. A beautiful historic city.

It is Beyma T2030 tweeter. I have two pieces and both have similar behavior. These subharmonic frequencies and their multiples occur quite suddenly at certain amplitude of speaker terminals voltage (not very high, BTW).

Thank you for your kind words regarding Prague.

 

[Pan]

The results of my test of distortion did not indicate any difference in symmetric versus non-symmetric nonlinearities. There is no scientific data that says that this is the case and no reason to believe that it would be the case based on the hearing mechanism.

And I did not say so either. If you read posts no17 and no18 again you'll see.

What I was saying was basically that it's flawed thinking, this thing with pleasing 2nd order HD. As you know any nonlinearity will result in modulation distortion.

 

[Pan]

Linear distortion = alteration to spectrum or waveform/time domain.

Nonlinear distortion = any alteration to the signal that results in new spectral components.

Maybe I don't understand the words, but those statements don't make sense to me.

Why?

sorry that really doesn't clear it up...

Read it again then. 🙂

It's obviously a bit simplified but pretty much that's the way it is.

Post #34 by gedlee expands on it if that helps.

 

[Pan]

One cannot expect to use a loudspeaker beyond its intended application and not have it sound bad. If your position is that one can always play a speaker loud enough that it has audible NLD then yes, that's true, but that is a completely absurd position.

My position is that I can always design a speaker such that it has no audible NLD at intended listening levels. In my room that's about 110 - 120 dB - easily enough headroom for any playback level that I am likely to reach.

Of course but it doesn't come cheap and that's my point. You need to make an effort to reduce distortion to problem free levels and I question anyone claiming a speaker totally free from audible distortion.

 

[gedlee]

One of the most surprising things is that what sounded horrific to my ears during the actual playback was far less objectionable in the recording listening back over headphones at a moderate level.
Art

Very interesting observation. Doesn't this imply that there are likely things that get unmasked at higher SPLs that are not as troublesome at lower SPLs? These "things" could be linear, you just don't know, but the implication is there in what you are saying.

 

[gedlee]

I question anyone claiming a speaker totally free from audible distortion.

I'd claim that for my speakers (at everything but painful SPLs). Not for all speakers of course, but many at mid volumes, a few at high volumes. Its the volume level that this departure happens that is the real criteria.

 

[gedlee]

It is Beyma T2030 tweeter. I have two pieces and both have similar behavior. These subharmonic frequencies and their multiples occur quite suddenly at certain amplitude of speaker terminals voltage (not very high, BTW).

Thank you for your kind words regarding Prague.

That implies an instability, the oil-can model. This will happen suddenly once the system jumps over from stable to unstable. A thermal modulation would happen at all levels, but likely increase with increasing power. It would not come on quickly like that.

I still think this kind of thing is very interesting from the standpoint of having absolutely no knowledge of any work having been done on audibility of these kind of distortion. The thing is that almost all masking is upward in frequency, which is also the case with nonlinearities. But a sub-harmonic would not be makes, probably not at all.

 

[gedlee]

The discussion was about 90% data reduction which means 128kbit/s or thereabout. You claimed transparency for that amount of data reduction and I question it.

The number 90% was just chosen to imply that huge data reductions are possible. Make it 70-90% then!

 

[PMA]

With test sine signal, that subharmonic distortion is clearly audible. But I have not done any rigorous test with music, however I expect it will be audible as well.

 

[turk 182]

Turk,

If the recording and playback were removed, that is you heard what the mic heard through the original playback system, the only way to listen at consistent levels would be to ascend twice as far in the sky for each 6 dB increase in drive level. However, high frequency air absorption would then progressively attenuate the HF, making the harmonic distortion harder to hear at greater distances.

the measurement mic is mono and doesn't move i have two ears and i can move my head which i believe is what aids in detecting distortion( via phase discrimination)

From what you wrote above, it appears you can't make much distinction between the drivers when using the music source.

i can hear differences with the drivers i just can't say that distortion level is objectionable.

One of the most surprising things is that what sounded horrific to my ears during the actual playback was far less objectionable in the recording listening back over headphones at a moderate level.

exactly my point a large part of the original stimulus is modified/removed

Edit: Moderation fixed broken quotes..

 

[turk 182]

that didn't work like i thought. i really gotta figure out the multi quote thingy.

 

[wesayso]

I have not spoke to either of them so I have no idea how they are reasoning about it.



The discussion was about 90% data reduction which means 128kbit/s or thereabout. You claimed transparency for that amount of data reduction and I question it.
I did adress higher bitrates as being good for many situations. There are listeners who can pick audible distortion at the higher rates as well but I'd say that for the average listener and music style/production on a good rig 192-256 is starting to get hard to hear. Don't remember if I ABX'd 320 or 256 with detection. The specific tune was a well produced dynamic electro/pop/jazz thing with lots of ambience. Compressed rock and similar was harder and stopped at lower rates.

None the less this is of course really cool, that such huge amount of information can be removed with so little (and something no) audible effects.

When listening to things like this and performing tests, it's one thing to throw in a bunch of students and let them listen on their own, and another thing to use more experienced listeners and also to guide them. Meaning if someone hears an artefact it use to be possible to help other hear it as well, once you point it out to them.

So when setting up studies, you need to be specific about what you are searching for. Listening capabilities of average joe or the limits of human perception period.

There's quite a bit of difference between codecs and their settings. I followed AAC development from Nero for a while long ago.
An interesting group of tests can be found here at different bitrates:
Audio quality of encoders at 320 kbit/s - SoundExpert

 

[turk 182]

no offense wesayso but what does that have to audibility of distortion in horns?

audio encoding is topic well worth discussing and deserves it's own thread.

i started this one to prevent the off topic in an other thread.

 

[wesayso]

Sorry for that, no intention to derail.

 

[turk 182]

Read it again then. 🙂

that's a little asinine and does not offer any clarification.

 

[pos]

Here is a harmonic distortion comparison between two compression drivers, taken from Voishvillo's AES paper about the JBL D2 driver (convention paper 8502).

The comparison is between a 3" VC driver with a titanium dome diaphragm (and possibly ferrofluid in the gap) and the D2 which is a dual 3" VC driver with polymer rings.
It is interesting to note that the 1/2 harmonic mainly occur in the region where the diaphragm is most probably undergoing heavy breakup modes.

JBL's own tube measurements of drivers equipped with 4" aquaplassed magnesium and beryllium diaphragms do not show such 1/2 harmonics rises :
476Mg
476Be

 

[dewardh]

What I was saying was basically that it's flawed thinking, this thing with pleasing 2nd order HD. As you know any nonlinearity will result in modulation distortion.

That, and the too often ignored reality that we're talking about mechanical systems that do not, in fact, behave like our oversimplified theoretical or electronic models. Cone breakup is commonly modeled as a simple resonance and we look at it with a single tone sweep, ignoring what "breakup" actually means . . . that different parts of the cone are moving in different directions or with different velocity at the same time, and not tracking the voice coil at all. The question becomes not just what that "sounds like" at the frequency at which it is occurring, but what the "rest of the music" sounds like when that is happening.

 

[kevinkr]

that didn't work like i thought. i really gotta figure out the multi quote thingy.

I fixed your quotes, you just forgot that the close quote is [/QUOTE] FWIW I do this all the time too! 😀

 

[scholl]

Thinking about the sub-harmonic issue, this is well worth studying. I can see how it could change a lot of things. It is a wide open field as far as I can tell. The Beyma paper didn't get much traction, but I am not sure that means anything.

My take away was the subharmonics were due to mechanical interactions and could be minimized by the device designers making different choices in materials, shapes and construction. IT would be interesting to see how BMS and the JBL push pull drivers stacked up.