The Audibility of Intermodulation Distortion

In this blog post I compare a driver's intermodulation distortion (IMD) against the spectral content of music.  Unfortunately there is not a great wealth of study data to support the audibility of IMD.  However we can first begin to ask some other questions relating to it's impact on the music we hear.  I pose the following question:

Does IMD affect the reproduction of music through our speakers?  

In other words, is the low level detail contained in our music masked by IMD?  Can this be demonstrated by comparing a speaker's IMD against the music signal?  To answer this we would need to look at the spectral content of a well recorded song.  Since IMD is frequency spectrum based we would need to freeze the song at one particular point in time and analyze it's spectral content.  We could then look at the low level detail contained in the song.  If a vast majority of the low level detail is below the noise floor of the driver, then it could be argued that the driver itself is masking potential detail from your music. 

To start let's look at the IMD performance of a well regarding driver... The Lamar M2-16 Field Coil Compression driver. 

  

The above spectrum is a the Lamar M2-16 field coil compression driver on the ES-290 Biradial Horn.  To analyze the chart, we see that at 400Hz the noise is -65dB down from the test tone.  This equates to 0.056% intermodulation distortion.  However IMD rises as you move up the frequency spectrum.  At 2kHz distortion rises to 0.25% (-52dB down).   

 

Choosing a Song 

 

In order to determine if IMD is intruding on low level detail contained in our music let's look at an example. 

Below is a spectrogram image of Paul Horn's album titled 'Inside the Taj Mahal' track title  Pt. XII Bass Flute.  Paul plays his bass flute inside the Taj Mahal. 

You can hear this song on Spotify by clicking here

You can hear the sound of the instrument resonate through the corridors of the Taj Mahal.  At around half a second into the track Paul begins to play the bass flute.  At 0:01 into the track the flute is producing a steady tone with all the related side harmonics and ambient information from the Taj Mahal. 

I converted my flac file for this song into .wav so that I could import the song into Audacity software for analysis. 

Selecting the 0:01 second mark I used the spectrogram function within Audacity to produce the follow musical spectrum:

As described in my diagram, the low level detail for the song begins at -56dB down from the main note at 232Hz (-18dB down from peak zero).  In absolute terms the low level detail is a full -75dB down from peak zero.  

In my first example I showed that the Lamar M2-16 Field Coil Compression Driver side band IMD noise products are -65dB down at 400Hz.   To compare the driver's performance against the musical content I've drawn a dashed black line in the diagram below.  This highlights that the driver's noise floor is somewhere in the middle of the low level detail of the musical content.   Anything below the dashed line would likely be masked by the IMD distortion of the driver.  

How would other drivers hold up?  I tested four other drivers to see how they would fare.  I've charted their IMD performance at 2kHz (90dB SPL 1m) into the chart below for reference. 

Conclusion

It can be shown that the various driver's noise spectra has the potential to infringe on the low level musical detail reproduction.  

1 comment

James Lee

I love your stuff. IMD was in my mind when I designed a DSP based 7 way sound system.
Frequency bands were divided mathematically for the x-overs. Time domain correction was applied at the listening position for each band so all drivers would not comb filter. My goal was to say bye bye to IMD.
I also made a 4way compression driver of sorts with domes and a single exit! No horn, just a 1 inch tube.
To add to the crazy, its in a car.

Leave a comment