Intermodulation-distortion and Doppler-distortion ??

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Dear enthusiasts and experts of DIY, would you mind and care explaining how and why those distortions happen in certain speaker designs and how they are eliminated ??

I would also be very curious to know in simple words, what happens "sonically" as a result of those distortions ??


Thank you in advance guys
 

Good to see the Rod corrected his initial article.

[edit] I would argue that whether it is rightly called Doppler distortion or not is some what of a moot point as it would not occur if the cone was not moving at the modulation frequency.

[edit 2] It kind of interesting to look bad at the history. I wrote my initial artical back in 2001. Then, some time later Rod wrote an artical some waht at odds with mine. When I became aware of Rod's articel I wrote my revised article in response. At point SL and Art L. apparently got in the loop with Rod and the issue was straightend out. It would have been a lot simpler if Rod had contacted me before posting the his initial rersponse to my artical. I guess that's what happens when you don't belong to the "club". :)
 
Last edited:
Yes, this is a confusing topic (as far as what to call it)
It's basically this aarvin.....think of the loudspeaker driver as a spring, when a bass tone is playing, it makes a large swing. Now a treble makes a little swing, if the tones play at the same time, you will get phase distortion because the treble swing is fighting the inertia of the bass tone swing. This gets worse with greater amplitude (because the swings are larger). The remedy for this is to not let the higher frequencies swing to much by implementing a crossover.
This is the main problem with having a full range driver cover the entire FR.
I'm still confused as whether or not to call it Doppler distortion because 1 swing in itself should cause a higher or lower pitch when moving toward or away from you. I guess we could find out by comparing a driver pointing to the listener vs. one pointing upwards. I don't think it would ever be perceivable though.
 
.
I'm still confused as whether or not to call it Doppler distortion because 1 swing in itself should cause a higher or lower pitch when moving toward or away from you. I guess we could find out by comparing a driver pointing to the listener vs. one pointing upwards. I don't think it would ever be perceivable though.

I have a hard time accepting Rod's position that it is not Doppler distortion, or rather that calling it Doppler distortion is incorrect. Perhaps it might be more accurate to call it Doppler induced phase modulation distortion. It is definitely a Doppler effect since if the driver is not moving it does not occur. Rod makes the comment that "it takes time" and the "driver is not actually moving through the medium (air)". I can not agree with that. For example, if I hold my hand up and move it to the left and right, at what frequency and at what amplitude is it suddenly not moving through the air? It is its motion that disturbs the air and generates acoustic wave. Or, suppose that in the conventional Doppler explanation where the carrier is moving at constant velocity, if the carrier suddenly starts to broadcast a sound does it take time for the frequency to be shifted? No! Even when you start with the idea of phase modulation, that phase modulation is the result of the cone motion. So call it what you like, it is motion induced. And, like Doppler shift, the magnitude of the shift is dependent on the velocity, with max shift occurring at max cone velocity. Also note that in my analysis, when the velocity of the source becomes constant (carrier frequency is zero or DC) the basic Doppler result is recovered.
 
Hi,

you are wrong, when it comes to exactly describing and quantifying Doppler distortion. Doppler is frequency modulation.

Peak frequency modulation calculates as ArtificialCarrierVelocity divided by SpeedOfSoundInAir. Artificial carrier velocity is PeakExcursionOfArtificialCarrier times 2 times Pi times FrequencyOfArtificialCarrier. That is it. For 3 mm (= 6 mm peak-to-peak) and 70 Hz there is half a percent of Doppler, a tenth of a halftone step.

In any wave, equal power means, that excursion rises with the reciproc of frequency. Basses shake trousers, trebles do not. Yet two waves of different frequency only superpose and do not intermodulate.

But equal power also means, that a sound source of fixed size must rise excursion with the square of the reciproc of frequency. This is calling for trouble. It does not matter, if it is a diaphragm or the air closely in front of it, which excurs. This may be a reason, why there is often left a distance between woofer and tweeter. But if one is to lessen Doppler that way, one is to introduce destructive interference. A more thoro solution is to lower excursion by raising source size. This again introduces beaming. All ways lead to Rome, but there are only a few ways.

Uli
 
Last edited:
Intermodulation distortion is caused by non-linear movement of the cone - e.g. as the cone is pushed further and further outwards, the suspension of the driver may pose higher and higher resistance (become stiffer, resist further outward movement), therefore the cone moves smaller and smaller increments outwards as the voltage increases. The result of this is that if you are playing high frequency content (minimal excursion) and a low frequency occurs which causes the cone to move outwards, the high frequency content will drop in amplitude - i.e. the high frequency content is being amplitude modulated by the low frequency content. This is intermodulation distortion in a nut shell and if you play 1kHz and 50Hz at the same time you will get 50Hz, 1kHz tone but also weak 950Hz and 1050Hz tones from the speaker. Suspension it not the only contributor to IMD - there are electromagnetic contributions as well such as uneven magnetic flux, flux modulation, voice coil inductance modulation, etc.

Doppler distortion is caused by the cone physically having to move to create sound waves. Technically it is phase modulation since when you play a bass note which causes significant excursion, the source of high frequencies is moving forwards and backwards with respect to the listener at the frequency of the bass note. The absolute phase in degrees at a given frequency is 360*d*f/345 where d is the listening distance in meters, and f is the frequency in Hz. Ignoring the non-linear aspects of a speaker driver, Doppler distortion alone does not change the amplitude of the high frequencies therefore it is not intermodulation distortion. In my opinion, Doppler distortion alone is not audible. Since great excursion is needed to create doppler distortion, it occurs that significant IMD is also created by the speaker driver, which is audible.

Both intermodulation distortion and doppler distortion can be reduced in any speaker design by using multiple drivers* to cover different frequency bands, or reducing cone excursion - use multiple drivers or drivers with physically larger cones. That is not to say that there are disadvantages to using more/larger drivers as well.

*There is one interesting case where using multiple drivers does not solve IMD and that is coaxial speakers. Where the cone of the bass driver horn loads the tweeter, the movement of the bass driver effectively changes the geometry of the horn loading as it moves, therefore changing the amplitude of the tweeter frequencies, therefore causing IMD. This can also effect conventional multi-way speakers to some degree due to the 'cavity effect' edge-diffraction effect of placing two drivers next to each other on a speaker baffle - it's almost insignificant though.
 
Last edited:
Ignoring the non-linear aspects of a speaker driver, Doppler distortion alone does not change the amplitude of the high frequencies therefore it is not intermodulation distortion.

But it will also cause new spectral content that wasn't there before therefore its effect on the audio signal might be a little like IMD. But one woul have to dig deeper into the maths to see how strong the effect actually is.

https://en.wikipedia.org/wiki/Frequency_modulation#Bessel_functions

Regards

Charles
 
But it will also cause new spectral content that wasn't there before therefore its effect on the audio signal might be a little like IMD. But one woul have to dig deeper into the maths to see how strong the effect actually is.

https://en.wikipedia.org/wiki/Frequency_modulation#Bessel_functions

Regards

Charles
It is true that it is very similar in spectral content to IMD. Since it is proportional to excursion, my feeling is that by the time the excursion is high enough for Doppler distortion to become a problem, distortion from motor/suspension non-linearity is more of a problem. It's less objectionable than IMD in my opinion because it requires significant excursion to become audible therefore only occurs during loud bass. IMD on the other hand can occur even at low volume/excursion with simpler driver designs, and there is more IMD interaction between mid and high frequencies. IMD between mid frequencies and higher makes a grungy grating sound, while IMD/Doppler between bass and other frequencies is more of a 'warbling' sound, far less objectionable imho.

I made some test samples simulating Doppler distortion at 0mm (no distortion), +/-3mm and +/-6mm.
50Hz and 500Hz, then 50Hz and 2kHz.
https://www.mediafire.com/?m9y9lhq794wwr97
 
Last edited:
...how and why those distortions happen in certain speaker designs and how they are eliminated ??...what happens "sonically" as a result of those distortions ??
The effects on the sound can be found here in this easier-to-read article by Keith Howard on Doppler distortion. I recommend this one:

Red Shift: Doppler distortion in loudspeakers | Stereophile.com

I sense that you're not looking for the physics of what is happening, but the following Klippel papers define the sources of those nonlinearities and their measurement, and how they add to each other. They really define all the issues from their sources so that you can do the trades to minimize their effects:

https://www.klippel.de/know-how/measurements/nonlinear-distortion/intermodulation-distortion.html

https://www.klippel.de/fileadmin/kl...es/AN_10_Loudspeaker_FM_and_AM_Distortion.pdf

As far as answering your question on "how are they eliminated?":

1) reduce the peak-peak motion of the driver diaphragms using either larger diaphragms for bass frequencies or even midrange frequencies, or transform the motion of the driver to a higher pressure field using horns to load the drivers. The last approach (horn loading) is the difference in sound that you get with horn-loaded drivers vs. direct radiators: much lower modulation distortion.

2) reduce the nonlinearities in the drivers themselves--like using overhung voice coils, shorting rings, more linear electric motor designs, etc. This usually increases the cost of the drivers...rather steeply.

3) Limit the frequency range of each driver by using more loudspeaker "ways" and crossover filters (with their attendant problems).

Chris
 
Last edited:
Technically it is phase modulation since when you play a bass note which causes significant excursion, the source of high frequencies is moving forwards and backwards with respect to the listener at the frequency of the bass note. The absolute phase in degrees at a given frequency is 360*d*f/345 where d is the listening distance in meters, and f is the frequency in Hz. Ignoring the non-linear aspects of a speaker driver, Doppler distortion alone does not change the amplitude of the high frequencies therefore it is not intermodulation distortion.
No, dear TMM. Doppler effect in broad-range electroacoustics is frequency modulation. It is a sort of intermodulation, because one signal, the low-frequency one, modulates another signal, the high-frequency one. I gave simple formulas for calculating it here.

I do not believe, that the signals linked are really FM as stated by thee. 500 Hz modulated by 50 Hz with 6 mm excursion should be audibly distorted, yet i hear a pure sine.

cask05 said:
2) reduce the nonlinearities in the drivers themselves--like using overhung voice coils, shorting rings, more linear electric motor designs, etc. This usually increases the cost of the drivers...rather steeply.
... and does not reduce Doppler distortion, it actually raises it, dear Cask05. Only a reduction of excursion reduces Doppler distortion.
 
No, dear TMM. Doppler effect in broad-range electroacoustics is frequency modulation. It is a sort of intermodulation, because one signal, the low-frequency one, modulates another signal, the high-frequency one. I gave simple formulas for calculating it here.

I do not believe, that the signals linked are really FM as stated by thee. 500 Hz modulated by 50 Hz with 6 mm excursion should be audibly distorted, yet i hear a pure sine.

That just proves my point that modulation effects aren't very offensive between low frequencies and higher frequencies. Check with a spectrum analyser, they are there. I can hear it (barely) on the 500hz tone, 6mm using good headphones (audio technica ad900). The 2kHz example I can't hear the side bands at 6mm even though they are higher in amplitude relative to the 2kHz tone than in the 500hz example.
This is due to the smaller wavelength of a 2kHz tone. The reason it's not audible is probably because there is a limit in human ability to sense differences in frequency and the ratio of 1950/2000Hz is much smaller than 450/500Hz. Some people are less sensitive (more 'tone deaf' to put it bluntly!) to being able to hear changes in tone. Personally I'm known to be better than average.

FM and PM are very closely related. It is my understanding that the Doppler effect in speakers cam be considered PM as the phase of the modulated tone is directly proportional to the excursion of the cone, which is proportional to the modulating tone.

Sent from my Nexus 5 using Tapatalk
 
Last edited:
Critical band theory may help, by 2 kHz the Critical band is 400 Hz wide

the Psychoacoustic term Fastl uses for hearing small modulation of a sine is "Roughness" - we're not very sensitive, >10% modulation/sideband amplitude may be needed to clearly hear only 50 Hz offset Roughness by 2 kHz
 
TMM, are thy examples measurements of a real source, or are they simulations? If they are the latter, then please calculate Doppler distortions with my formulae, and thou will be surprised, that thou must apply more modulation. Thou may be theoretically right about the phase stuff, but Doppler really is FM and can become easily calculated as such.
 
Here is a link to some work I did back in 2001 on this. The revise version of the paper is not archived.

Copy and past the link in your browser.

"http://web.archive.org/web/20090809205449/http://www.geocities.com/kreskovs/Doppler1.html"

The page above was to my old write up. This page is to some measurements. (Note: this was done 14 years ago.)

Doppler2
 
TMM, are thy examples measurements of a real source, or are they simulations? If they are the latter, then please calculate Doppler distortions with my formulae, and thou will be surprised, that thou must apply more modulation. Thou may be theoretically right about the phase stuff, but Doppler really is FM and can become easily calculated as such.
It's a simulation, assuming 340m/s speed of sound and +/- 0/3/6mm of cone excursion modulating the phase of the higher tone.
I didn't calculate the 'amount of modulation' and apply it that way, it's inherent from the simulation because I calculated the excursion and dynamically calculated the instantaneous phase modulation to apply to the higher frequency while constructing the signal in MATLAB. If you can't hear it then the conclusion is just that - it's not audible at this excursion level.

I think you would have an extremely hard time producing audible Doppler distortion in a real loudspeaker without IMD dominating it.

Here are the spectra of the 6mm files I posted previously:
26136055174_02c297faee_o.png

26136055234_ae50cc4b5f_o.png



The MATLAB code:
Code:
    Fs=96000;
    n = 960000;
    tt = linspace(1,n,n);
    f1 = 50;
    f2 = 2000;
    xmax = 0.006;
    
    samples = sin(2*pi.*tt/Fs.*f1)';
    samples2 = 0.1*sin(2*pi.*tt/Fs.*f2+ pi.*samples'*(xmax/(340/f2)))';
    
    final_signal = 0.3*samples+samples2;
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.