How much tweeter distortion is audible?

[Pan]

Headphones are very good sources for a great many tests as they take out a lot of confounding variables. Etymotic ER-4 are known as "research ear plugs" for just this reason. [/B]

If you want to exclude the room (which is not necessary for many tests, especially if the room is treated and suitable for critical work) you can use headphones but those small earplugs distorts 30-100 times more than a good headphone or speaker, very poor performance:

http://www.headphone.com/technical/...raphType=1&buttonSelection=Compare+Headphones


/Peter

 

[pjpoes]

We basically did have a small sample size. Its difficult to find subjects and good ones are extremely hard to find. There is no way that you can get anybody to do a test for more than about 30-40 minutes. You have to work with minimal numbers which seriuosly limits the anayysis that you can do.

I can relate, it's the same sorts of issues I deal with in my work. We seem to find that financial incentives of sufficient size do get people to to just about anything you want, but to get sufficient samples, I bet we are talking multi-million dollar research budget and a very understanding IRB.

None the less, that further reinforces my earlier point that you need to manually control extraneous variables. If you can't do it statistically, then they must come from the study design themselves.

I just got through a series of meetings discussing the standardization of a lab measurement in the home. We are doing roughly the same thing I would imagine going from headphones to speakers in home would be. Namely, we have to run the study in both the lab and in peoples homes, them compare the results to ensure reliability and selectivity is similar.

 

[jeroen_d]

This is the transfer curve with a -20 dB third order negative cosine. In the piano sample, no IM grunge is audible.

Thank you for that, erjee.

Indeed, we should have compression if we try to simulate the behaviour of 3rd order harmonic distortion of a loudspeaker. So I used a -26 dB third order negative cosine to get 5% distortion. The transfer function looks like this.

 

[jeroen_d]

I used AddDistortion to process this function on a 315Hz full scale sinewave. I measured the result with my soundcard and the ARTA spectrum analyzer. The original sinewave looks like this.

 

[jeroen_d]

The processed sinewave looks like this. Indeed, a 3rd order harmonic at -26dB is now apparent.

 

[jeroen_d]

And now a little experiment. I processed a short piece of Beethoven string quartet. The first half of the sample, the volume is low and distortion is well below 1%. The second half of the sample, the string quartet plays very loud and the distortion rises towards 5% 3rd harmonic distortion. As if the speaker is driven to its excursion limit.

Who can distinguish just by listening (pan, please use the superduper HD650) these files and which one sounds best? One is the undistorted sample, the other contains 5% 3rd harmonic distortion at max volume. To distuinguish, try to detect compression and lack of transparancy.

http://home.kabelfoon.nl/~dezaire/Beethoven (1).wav
http://home.kabelfoon.nl/~dezaire/Beethoven (2).wav

 

[pjpoes]

do you want us to post our guess?

 

[jeroen_d]

Well, not really. You could do it by private messaging to me, to prevent bias on the forum, and I could post the results after a few days.

But I guess anyone can draw its own (by nature limited) conclusions by participating in this little experiment. By no means I want to embarrass people or have some sort of contest between 'golden' ears.

 

[gedlee]

A couple of comments. The piece is a little short in my experience. I shoot for 15-20 sec minimum. And its easy to tell which is distorted without even listening, so this aspect makes it hard to draw any reliable conclusions. If everyone just listened that would be one thing, but surely there will be those who "need" to be right and they will simply look at the data - problem solved.

 

[pjpoes]

would you tell us if we were right in the pm?

I have Sennheiser HD600's and a discrete headphone amp, but it's all connected to a pretty noisy computer and mediocre sound card. I wasn't positive if I heard it, probably would want to make a few guesses. What if you posted the files with different names (same files) to use as second trials.

 

[breez]

Here's a tool for people trying to distinguish between the samples: http://ff123.net/abchr/abchr.html

You could also use foobar2000's ABX tool.

 

[jeroen_d]

A couple of comments. The piece is a little short in my experience. I shoot for 15-20 sec minimum. And its easy to tell which is distorted without even listening, so this aspect makes it hard to draw any reliable conclusions. If everyone just listened that would be one thing, but surely there will be those who "need" to be right and they will simply look at the data - problem solved.

Ok, I will make a piece of 20sec. And you're right, if people want to cheat it is easy to do. Just open the files in some wave editor program and you will see which one is compressed. So I stick with my comment "But I guess anyone can draw its own (by nature limited) conclusions...". If people "need" to be right, wrong, agree or disagree with your findings, well that's their own business. In the end, on a forum people can learn something if they want and people can also be very opiniated (sometimes right, sometimes completely wrong) and have big ego's.

 

[Pan]

Short samples is best if the "right"material is used but I'd prefer stereo instead of mono.

One track is obvisouly compressed and lack clearness and air.

I used HD600, I was to lazy to go get the HD650 but they have similar distortion performance allthough a different FR.

I heard the difference in both halfs but it was a little easier in the 2nd half.

As I have said numerous times before, a speaker with 0.5-1%THD even if dominantly low order HD (as measured with a sine) is not acceptable in a hifi set up.

When speakers get down to 0.1%THD at 90dB SPL things start to sound really good.


/Peter

 

[jeroen_d]

Here are two longer samples with piano music, same distortion.

http://home.kabelfoon.nl/~dezaire/Liszt_norm_1.wav
http://home.kabelfoon.nl/~dezaire/Liszt_norm_2.wav

 

[pjpoes]

well I never did better than guessing in my short time trying it here with my headphones.

 

[pjpoes]

let's make sure I used the abc/hr comparison utility right. So I load the waves into the setup, in doing this, the original can be either 1 or 2, and then I can load 1 and 2 into the later wave area's respectively? Then for setting up the abx, there is no reason to choose original, correct, A and B should be sample 1 and 2 respectively.

 

[auplater]

distortion

I hear a difference... increased "grunge" in certain sections on one of the samples but then I'm using HD 485's, not 600's or 650's so maybe that's the reason...

Single note piano solos would be an interesting sample set...

John L.

 

[JinMTVT]

Ok Guys ..i do not have 1% of the knowledge and understanding ( math, theories and experience )
that most of you here have ( probably not even 0.0001% of dr gedlee ahha )

But i was asking myself after browsing through this discussion, just trying to interest and learn about all this ...

If we hear a signal with a distortion, our brain cannot know that there is a distortion unless it can compare it with the original signal, wich is impossible in real time,
and hard unless experienced listener in AB or delay

but, i am pretty sure that our brain is used to hear some patterns and regularities in life sounds,
so it would expect to hear the same type of patterns in an audio reproduction, and this is where different type of distortion can be easily detected, while others doesn't influence the regular patterns or type of signal

i hope my msg is not too fuzzy

so without a possible comparison, distortion that leads to difference in timbre and frequency can't be detected easily

but difference in freq. response and all affecting factors
( polar and axis does affect freq. response right? )
or in phase/timing ( wich ends up being again freq. reponse ?? ) do not sound "regular" or "natural" to our brain

 

[JinMTVT]

jeroen_d:

about your 2 wave files ...
i cannot tell any particular difference even after 15 minutes of repeatitive listening

i am not also able to tell wich is playing
( differentiate the 2)

for reference,
listening on sony MDR-SA3000 headphones
through computer -> Rakk DAC+ pass output -> diy heaphone amp from our swedish friend Peter
that is supposed to have ridiculously low distortion levels

 

[JinMTVT]

i sure hope that no one is going to CHEAT on this,
as this would be serious disgrace to onself

we are supposed to be here to try and understand,
learn and grow

i'll try and listen again more tomorrow ...
but i still couldn't figure out the distorted version
and i know i have good quality sound with this setup
( but my EARs is what is lacking quality 😛 )