Some argue that regarding distortion, the distribution of harmonics is as important, if not more important, than the total level. For instance, the 2nd and 3rd harmonics are meant to be the least harmful, with some disagreement about the winner. Others have suggested that odd harmonics tend to sound harsher compared to a series that's dominated by even ones. Some also suggest that our hearing listens for odd harmonics as loudness cues, and that it can be fooled by distorting audio systems. Meanwhile, others say that audio systems with a slight exponential curve (mimicking but also exaggerating natural air distortion) can sound more 'dynamic'.
All I know is that when it comes to music, a square wave synth and a half-rectified sine synth certainly differ in character, with the 'evens' having a softer choral/pipe organ sound to my ears. Square waves can be a little harsh, but the closely related triangle waves (different phase angles, and more damped at higher frequencies) can also sound nice. OTOH, a sawtooth wave can have a piercing trumpet-like brightness to it.
It's hard to say off-hand what distribution of harmonics would be ideal. I'd have to do subjective tests, while remembering to adjust the distortion loudness for equal energy.
Besides that, there's IMD. I'd have to investigate further, but IMD could potentially induce even higher 'secondary' harmonic distortion.
As a thought experiment, a speaker produces a 1kHz sine wave at an amplitude of 0.1mm. In a THD test, the result is quite good, say 0.1% THD. However, by adding a DC offset, the voice-coil is pushed to a non-optimal position and THD grows to 1%. Now, instead of a DC offset, apply a bass signal. Not only will IMD side-bands be created, but the THD will also be modulated.
All I know is that when it comes to music, a square wave synth and a half-rectified sine synth certainly differ in character, with the 'evens' having a softer choral/pipe organ sound to my ears. Square waves can be a little harsh, but the closely related triangle waves (different phase angles, and more damped at higher frequencies) can also sound nice. OTOH, a sawtooth wave can have a piercing trumpet-like brightness to it.
It's hard to say off-hand what distribution of harmonics would be ideal. I'd have to do subjective tests, while remembering to adjust the distortion loudness for equal energy.
Besides that, there's IMD. I'd have to investigate further, but IMD could potentially induce even higher 'secondary' harmonic distortion.
As a thought experiment, a speaker produces a 1kHz sine wave at an amplitude of 0.1mm. In a THD test, the result is quite good, say 0.1% THD. However, by adding a DC offset, the voice-coil is pushed to a non-optimal position and THD grows to 1%. Now, instead of a DC offset, apply a bass signal. Not only will IMD side-bands be created, but the THD will also be modulated.
Not if it used to calculate a frequency response it cannot.
Not sure if that was a response to my post, but if it is then the word distribution has been confused with dispersed. I am not talking about distortion profiles.
Consider a given magnitude at say 6kHz. How can you tell if it is 6kHz fundamental or some 2nd harmonic of 3kHz or 3rd harmonic of 2kHz or 4th harmonic of 1500Hz and so on and so on.
In an FFT approximation to a magnitude response, the distortion energy is distributed throughout the measure and indiscernible in such a display because of that.
Sure it can.
It seems like you are focused on a simple RTA implementation of FFT, which agreed, cannot separate distortion from response.
But there are many implementations of FFT, for different purposes.
For instance, a sine sweep that REW uses, like Juhazi mentions.
Other FFT implementations...MLS, dual channel...etc
And even if there was no way to discern between response anomalies and harmonic distortion (but there is
) , wouldn't we still want to use FFT to measure ?hi Sonce, thanks for your kind advice...but this thread is nothing about speaker design as i mentioned before.
speaker driver design maybe better, but just marginally.
please see how our experts going now
could you go more detail on what being nonsense? so why are you still here?The first post is so full of pedantic nonsense it hardly matters
obsessed: so you were aware that the quoted statement is not a widely accepted conclusion? whats the point?
and just found the long running thread by Earl Geddes, still reading it but it does not seem to be conclusive to everyone either.
Geddes on Distortion perception
is it better if you can provide some facts or information for my further "education"? is this kind of response helpful to anyone?
ok its you again, so whats up?
sure i could be incorrect and you are the only expert here, why not "educate" me instead?
Last edited:
It's been shown there is very little deviation from linearity. Are you still concerned with how small details are reproduced in the presence of larger signals?How did the levels scale with power input?
Any measurements of that?
Sure it can.
It seems like you are focused on a simple RTA implementation of FFT, which agreed, cannot separate distortion from response.
But there are many implementations of FFT, for different purposes.
For instance, a sine sweep that REW uses, like Juhazi mentions.
Other FFT implementations...MLS, dual channel...etc
And even if there was no way to discern between response anomalies and harmonic distortion (but there is
The discussion was not of spectrometry but far more general. MLS measures are a perfect example of how phase randomisation in the measurement process prevents distortions being identified. The FFT can be useful, but it is only as useful as an understanding of it limitations. There seems to be a belief that a simple magnitude response represents what we hear. It is not and is use in this application such as via MLS analysis is fundamentally flawed.
i have some thoughs regarding the rather dated test by Earl Geddes: http://gedlee.azurewebsites.net/Papers/The Perception of Distortion.pdf
please forgive me if it has been addressed somewhere else, without all these i really have hard times to validate or assess any implications.
- participants should be audiophiles instead of random individuals (critical)
- source material (CD) is not specified (critical)
- test files or processing method should be open to public (extremely critical)
- D/A converter is not specified (critical)
- amplifier is not specified (critical)
- transducer should use well known multi-way high-end speakers instead of earbuds (critical)
- test methodology is not specified (critical)
please forgive me if it has been addressed somewhere else, without all these i really have hard times to validate or assess any implications.
- participants should be audiophiles instead of random individuals (critical)
- source material (CD) is not specified (critical)
- test files or processing method should be open to public (extremely critical)
- D/A converter is not specified (critical)
- amplifier is not specified (critical)
- transducer should use well known multi-way high-end speakers instead of earbuds (critical)
- test methodology is not specified (critical)
I disagree. Distortion in loudspeakers with increasing power input or power output is well known and easily measured. What has been shown in the method adopted here is how distortion energy has been dispersed within the measurement analysis and has therefore been rendered indiscernible.
- Status
- Not open for further replies.