Hearing Notches, HyperRecruitment, and IMD

I should re-read my own posts more often… I wrote about this topic in an earlier post, but I failed to connect my own dots until I chance reread the post just moments ago.

I have a hearing notch in my right ear around 1250-1500 Hz, or D#5-F#5, just more than 2 octaves above middle-C, which extends for roughly a full critical band. I hear absolutely nothing in my right ear in that range, even when sound levels have been amplified to lethal levels. Completely gone.

That means there must be substantial damage to the inner hair cells in that region of the cochlea – producing no sensation to the brain. Correspondingly, it would be reasonable to conclude that the outer hair cells – responsible for controlling the amplitude of vibration of the basilar membrane at those frequencies – must also be badly damaged. Which further implies that we should expect larger than normal vibrations of the basilar membrane in that region.

It often happens that severe hearing notches are accompanied by hyper-recruitment hearing in bands just above the notch in frequency. It seems a reasonable hypothesis, that if outer hair cells were severely damaged in the hearing notch zone, then the next higher frequency band may have some amount of damage too, and it is responded to by hair cells even closer to the oval window where sound enters the cochlea.

Also, the unconstrained vibrations of the basilar membrane will also respond more than they should even when frequencies are above the hearing notch range. And that contributes to excessive vibration at the next higher frequency band, making it more difficult to control with the extant outer hair cells in that band.

And indeed, I seem to have some hyper-recruitment in the band from 1500 Hz to around 2200 Hz, centered mostly around 1800 Hz.

With Crescendo it is relatively easy to treat hyper-recruitment. You just need to place a notch or dipping EQ just ahead of the Crescendo engine. That causes Crescendo to believe that the hearing corrections correspond to a much more severe recruitment slope, and causes the corrections to be more gently applied so that we never exceed the comfort level in those frequencies.

But it is also the case that even after this dipping EQ for hyper-recruitment correction, I sometimes hear a crunchy sort of sound on high female chorus and high string section sounds in music, in the broad region of 1.7-2.5 kHz. But it can drive you crazy just trying to isolate the precise frequencies. They seem to float around in that upper band. And it happens that if I sometimes dip the frequency band around 750 Hz, then the “crunchies” become considerably lessened.

I attribute this behavior to IMD, or Intermodulation Distortion. The ear is naturally nonlinear, and when two tones are presented at loud enough level, you can often hear pseudo tones – tones that really aren’t in the presentation, yet you can clearly hear them anyway. Those tones are being generated by the nonlinearity of our hearing. And they are real to our perception, even though they aren’t present in the signal at all.

At weaker presentation levels, those tones are still there, but it becomes harder to hear them. You can use probe tones to find them, where a beat sensation becomes evident as the probe tone nears the IMD product tone.

The rate of fall off in intensity of these IMD product tones is often greater than the actual reduction in overall volume, but not always.

If we present two tones at frequencies F1 and F2, with F2 > F1, then the IMD product tones will be located at (2 F1 – F2) and at (2 F2 – F1). These both fall outside of the band between F1 and F2.

In the case of (2 F2 – F1), these IMD product tones will fall to the high side of the higher frequency, F2. And as F1 ranges from zero up to F2, the IMD product tone moves from 2 F2 down to F2 itself.

And if you notch out the tone at F1, say at 750 Hz, then the IMD product tone due to a higher tone for F2 will fall away linearly with the depth of the notching. But if you instead notched out F2 then the IMD product would decrease at twice the rate of the notch depth.

So we know that I have a profound notch in my hearing between 1250 Hz and 1500 Hz. And we suspect that any stimulation at those frequencies will be met with unrestrained vibration in the basilar membrane, hence encouraging the development of IMD product tones with respect to incoming tones at lower frequencies.

We also know that nearly all music has more acoustic power in the lower frequencies, hence those low frequencies will more likely encourage IMD product tone formation against higher tones. But in this case, even though the musical spectrum falls away starting just below the notch frequencies, the unconstrained vibration in the basilar membrane overcomes that fall-off. And so the most likely production of IMD tones will occur between that notch frequency region and all lower frequency tones which carry appreciable power.

And so to be most effective, we should notch out those frequencies in the region of the hearing notch to produce a fall-off of IMD tone intensity at twice the rate of the spectral notching depth. And it is nature’s serendipitous coincidence that the tones we are suppressing can’t be heard anyway due to my hearing notch.

So instead of notching out 750 Hz for occasional relief against the “crunchies”, I should actually be notching out 1250-1500 Hz, with a relatively mild dipping EQ. That offers the chance to kill off as much IMD production as possible, against all the lower frequencies, and with substantially more effectiveness due to the squared effect of notching the higher frequency by some amount.

And sure enough, when I do this – notch out 1.37 kHz to a depth of 3 dB and a bandwidth of 0.26 octaves, covering the band from 1250-1500 Hz, the listening experience has no crunchies overall. I still need to dip the frequency band around 1800 Hz due to my hyper-recruitment. Real musical tones at that frequency are still responded to with excessiveness due to real hyper-recruitment hearing.

But the shallow notch at 1.37 kHz, by only 3 dB, effectively prevents the formation of audible IMD products from 1.4 kHz to 2.7 kHz – the very region that broadly exhibited occasional crunchy sounds. The attenuation of IMD product tones is as deep as 6 dB.

Everything always seems so clear in retrospect, but discovery is anything but a straight-line process…

  • DM

[Note: Whereas the notching in the 1250-1500 Hz zone actually reduces the sound levels in that band, but I can’t hear them anyway, the EQ dip at 1800 Hz does not reduce the level of the apparent sound to me. Even though those frequencies around 1800 Hz really are attenuated, they don’t sound like it to my hyper-recruitment hearing, with the help from Crescendo. They just sound like they should at those frequencies. I can’t detect any decrease in amplitude after Crescendo correction. And Crescendo keeps them from ever appearing at uncomfortably loud levels.]