There are a number of disconnects between current audiology practice and the restoration of accurate musical hearing. Among them are:
- Different goals,
- Inconsistent calibration standards,
- Difference between flat spectral response and subjective preferences.
Let’s briefly examine each of these as they relate to Crescendo, and provide some guidance on the proper settings for your Crescendo system.
Today’s current audiology practices are consumed primarily with the restoration of speech perception. As such, they are successful to a large degree. But their techniques are at complete odds with accurate musical hearing restoration, where the apparent ratios of harmonics must be maintained in order to preserve the timbre of musical instruments.
The compression curves offered by hearing aids are entirely incorrect for musical hearing restoration and preservation of harmonic amplitude ratios. But they work just fine for speech perception. Intentional (or otherwise) distortion of speech can often make it easier to comprehend.
Inconsistent Calibration Standards
When you receive a hearing test an audiogram is generated which represents your hearing deficits at each of half-octave frequencies, relative to their calibration standard of “normal hearing”. What does that actually mean?
ISO-226 defines the current reference for audiometric thresholds. And ANSI 3.6-1996 (or later) defines reference levels for hearing thresholds in various testing transducers for the audiometers used by audiologists. These two standards have loose (if any) connection.
ANSI 3.6 derives from earlier work performed by an Air Force researcher in 1969 who tested 25 Air Force Officer Candidates (presumably young men in their early 20’s). That work relied on a specific model of headphone to convert from electrical signals to audible sounds. The headphone was measured using a “coupler” that tries to (crudely and poorly) emulate the physical behavior of a human ear. That headphone is no longer produced.
More recent standards try to provide calibration levels for currently available models of speakers, headphones, and bone conduction transducers. Each of these standards is tied (loosely) back to the original standard produced by the Air Force study in 1969.
At best, each new standard tries to normalize their findings using “Biological References” – which means that a handful of presumably normal hearing adults (i.e., the other audiologists in the same office) listen to the threshold levels produced through the audiometer and transducers, and come to some loose agreement that the levels “sound about right…”.
Invariably there is a roughly 5-15 dB difference between the ANSI 3.6 levels produced, and the Absolute Threshold of Hearing (ATH) defined by ISO-226. At frequencies near 1 kHz the difference is around 5-10 dB, while at frequencies near 6-8 kHz the difference is around 15 dB. In each case, the ANSI thresholds used by Hearing Professionals is higher than the levels specified by the ISO standard.
ISO-226 ATH refers to the very best possible hearing, perhaps found in a 10 year old child. ANSI standards used in hearing tests are based on young adults. A dBHL unit has its zero point on the audiometry (ANSI) scale. So your audiometry is likely too low, compared to the ISO-226 ATH.
The ISO standard relates dBSPL (as measured by a microphone) to the human hearing levels in Phons. The relationship between dBHL (measured in the hearing test) and dBSPL is only loosely defined.
Crescendo is based on the “EarSpring” model for human hearing, and it clearly shows a NYC-like, ski-jump, compression compensation curve, not the linear compression curves used for speech restoration. Crescendo also works in Phon space, not dBSPL, and not dBHL.
What all this means is that if you use your reported audiometry for setting up your Crescendo, you might try adding 5 dB to the 1 kHz and 1.5 kHz levels, add 10 dB to 2 kHz, 3 kHz, and 4 kHz, and add 15 dB to the 6 kHz and 8 kHz threshold elevations. When I do this, everything really does sound more correct to me. But your mileage will possibly vary, depending on the test devices used for your own audiometry. Consistency is a poor bet between tests.
Difference between Flat Spectral Response and Subjective Preferences
Crescendo is intended to play through spectrally flat transducers. Of course, headphones and speakers are never spectrally flat. Your room affects playback through speakers. And the close proximity of the headphone speakers and the cups of your headphones interact with your pinnae to produce a wildly complicated spectral response. Absolutely crazy not flat spectral response.
Some people use software corrections, like those provided by Sonarworks Ltd., to make your speakers and headphones “spectrally flat”.
But then… do they really become Spectrally Flat? Or do they represent the ideal listener preferences? These two are not the same thing.
Dr. Sean Olive, of Harmon Research, has written extensively on the proper setup of speaker systems and related headphone measurement systems. He finds that the “ideal” listening room, as judged by listener preferences, has a -1 dB / octave rolloff. And headphones should strive for this, plus a 4 dB lift in the deep bass region. Not at all, spectrally flat.
So when Sonarworks claims to “flatten” your headphone response, they claim it is tied back to subjective ideals. Does this subjective ideal refer to the ideal listening room with -1 dB/octave rolloff? Or does it refer to subjectively sensed spectral flatness? There is no hint from them about how to interpret their words. (The operative word here is “subjective”.)
So my recommendation is to try it both ways, and choose whatever sounds best to you. Sonarworks allows you to provide a spectral tilt to the playback response. You might try using a compensating +1 dB/octave rollup to overcome any possible -1 dB/octave rolloff. That would be the +5 dB setting on the Sonarworks spectral tilt. Try using it, or not.
[Well, assuming the offsets applied to my measured audiology elevations are correct, when I try the +1 dB/octave compensation in Sonarworks and listen to female commentators on videos, the sound has a broad artificial high frequency resonance to it. A kind of ringing on her voice. So it seems that Sonarworks is likely flattening the spectrum, and not adhering to listener preferences with rolloff.]