Positive Feedback ISSUE 66
march/april 2013



The (Un)Broken Audiophile - Part 2
by David McClain, Acudora, Inc.


david mcclain

(Part 1 can be found here http://www.positive-feedback.com/Issue65/hearing.htm)

The Harmonic Structure of Sound

Years ago my uncle, the celebrated Musicologist, Earnest McClain, taught me how to use a wine glass as a Helmholtz Resonator to pick out the harmonics of piano tones. By varying the opening of the glass cupped over an ear you could very definitely pick out the second and third, and higher, harmonics as you strike a piano key.

Years ago, Master Piano Tuners would use this trick to keep the octaves across the keyboard in tune with one another, as well as finding the perfect 5th's and 4th's. Today, my piano tuner comes armed with a laptop computer and a fancy tuning gauge that listens to the built-in microphone of his laptop.

My uncle also taught me how to excite almost any note in a long tube, any long tube, with a trumpet mouthpiece. Bugle players must lean this skill as well.

All interesting sounds have a harmonic structure, beyond their fundamental pitch, that gives rise to their timbre—their unique sound. This is why a trumpet sounds different from a clarinet. Music developed chord structures that mimic the harmonic structure of musical instruments, and this became the basis for harmony.

But the harmonics produce by an instrument is not the whole story. The individual amplitudes of the harmonics help shape that sound's timbre as well. Take a cello, and a broomstick. Stretch another cello string across the length of that broomstick and play it with the bow.

It doesn't sound very much like a cello. And that's because, even though both are stringed instruments, and both have the same pitch and multitude of harmonics, the cello's body selectively amplifies some of those harmonics, giving rise to what we call formants. Human voices have formants too. Say "ooaahaayee" and you can hear the formant structure of your voice becoming evermore emphatic of higher frequencies.


When you apply equalization to a sound you can, if too much EQ is applied, radically change the personality of the instrument. This is akin to changing the formants, or the shape of the cello's body. Overemphasizing the higher harmonics of an oboe can turn it into a muted trumpet sound.

There is a place for creative equalization. Each of us chooses our speaker systems and headphones according to personal taste. A spectrally flat monitor, while accurate, can become very tiresome to listen to for any length of time.

But the manner in which we perceive the effects of creative EQ varies, depending on our hearing acuity. People with normal hearing experience something very different from those of us with hearing impairment.

EQ applied in an effort to restore damaged hearing can only work correctly at one particular loudness level. Below that level, it isn't enough boost, and above that level it becomes too much.

Figure 1 – Static EQ vs Nonlinear Anti-Recruitment Compression. Static EQ can only be correct at one specific loudness level, where it crosses the anti-recruitment compression curve. Below that it offers too little gain, and above that it offers too much gain.

If you have hearing impairment, then you must apply nonlinear compression to the sounds before they reach your ears—complementary to the recruitment hearing that you normally experience. And then, if you apply shaping EQ ahead of that correction you can hear what people with normal hearing experience.

Linear Compression

Linear compression might also be tried for overcoming hearing impairment. And it works better than static EQ. But it can only sound correct at two particular loudness levels. Below the first, it still doesn't offer enough gain. And between those two levels it is still too much gain.

Figure 2 – Linear Compression vs Nonlinear Anti-Recruitment Compression. Can only be correct at two specific loudness levels where the two compression curves cross each other. Below the lowest crossing, linear compression offers too little gain. Between the two crossings linear compression still offers too much gain.

A DIY System – NYC Compression

There is a simple trick that you can try, that approximates the needed nonlinear compression across a few wide frequency bands. Multiband linear compressors are commonly used items for Mastering sessions—typically 3 or 4 bands span the entire audible range. The trick relies on a technique, long known among mixing studios in New York City, called NYC Compression, or Parallel Compression.

NYC Compression splits the audio signal into two paths. One goes untreated to the final mixer, while the other is routed through a linear compressor with a fairly steep compression ratio, then added back to the dry signal. If you try this, you will "beef-up" your bass, and highlight the highs.

At loud levels, the dry signal dominates and the compressed signal adds very little noticeable effect. But as the dry signal drops in level, the compressed chain adds more and more. It should be clear that when the signal level in the dry chain equals the compressed level, there would be a 6dB boost through their sum.

If you plot out the effective compression curve of this parallel compression, its shape starts to mimic the nonlinear compression needed to overcome recruitment hearing. In fact, a very specific amount of compression ratio, compression threshold, and makeup gain will lead to a very good approximation of the anti-recruitment compression needed.

While our Crescendo corrects 100 Bark bands at quarter-Bark spacing, this technique can only provide 3 or 4 very broad bands of correction. Crescendo also applies its corrections in Phon space, while this NYC Compression technique works in dBSPL, or sound intensity, space. But it sort of works, it is simple to do, and can offer some improvement. It gives just a hint at what the more sophisticated Crescendo could provide.

Figure 3 illustrates the procedure using the Mastering Compressor plugin packaged with PreSonus's Studio One 2. It is a 5-band (linear) mastering compressor with good dynamic range on the compression threshold.

Figure 3 - Showing the PreSonus mastering compressor arranged as an insert in the output master channel. This compressor allows its processing to be mixed with the dry signal directly. I set the mix to 50% Wet/Dry, and backed off the output gain by 6 dB to allow headroom for the compression mix.

My bands are arranged as DC-750Hz, 750Hz – 1.5kHz, 1.5kHz – 3kHz, 3kHz – 6kHz, and 6kHz – 20kHz. I personally don't really need anything beyond 10kHz. (See text discussion).

My threshold elevations were guessed at being None in the bass region, then 30dB, 50dB, 60dB, and 65dB in the ascending bands. These translate to -96dBFS in the bass region, then -67 BFS, -47dBFS, -37dBFS, and -32dBFS.

All bands, except the bass region, have compression ratio of 4.5, and makeup gain close to 21dB. The bass was given a compression ratio of 20:1 and a makeup gain of -36dB to keep it out of the way.

When I listen through it, I have to say it sounds pretty good… perhaps a bit too bright sounding, but not bad… Definitely much better than nothing!

For this purpose, if your hearing threshold elevation at some frequency is 20dB or less, just consider that to be normal hearing. And so select the bass region to be wide enough to cover all frequencies up to the point where your threshold elevation exceeds 20dB.

There will be no real NYC Compression applied to this bass band. We'll set the compression ratio to 10:1 or higher, and set its compression threshold as low as it will go, and ask for no makeup gain for this channel.

In effect, it will still compress, but it will be acting at such low volume levels that your music will likely never reach the point where compression dominates. And besides, most music is very bass heavy, and so will be unlikely to invoke the parallel compression.

Now you have 2 or 3 bands remaining. Divide the remaining frequency space among those bands. Bearing in mind that most music exhibits a roll-off above 1kHz of around 6dB/octave, you want these bands to be in roughly logarithmically equal sizes, so that the higher bands retain as much audio power as possible.

So, for example, if your bass band occupies all the frequencies from DC to 1kHz, then you have from 1kHz to 20kHz to divvy up among the remaining bands. Taking the log10 of these two frequency limits, we see that the logarithmic span is 1.3 in log10kHz.

If we have 3 bands remaining, set them so that each has a logarithmic width defined by the ratio 1.3 / 3 = 0.43 which corresponds to a frequency ratio of around 2.7:1, or about an octave and a fourth.

Figure 4 – Closeup view of the multiband mastering compressor.

Now, in each of the 3 bands, you need to estimate your threshold elevation near the middle of the band. And whatever that value is, call it Pt, set your compressor threshold to (Pt – 20) dB.

But wait… those threshold elevations are numbers like 30dB, or 70dB. We are inside a digital processor where the values can never exceed digital max, or 0dBFS.

So, yes, we have to convert from real-live listening levels to dBFS measure. For that you need a sound pressure meter. Make a sinewave oscillator with peak amplitude around -17dBFS and play it through your speakers, taking note of what the sound meter shows as the physical loudness level in dBSPL.

In my studio/lab, I normally calibrate my playback system so that an RMS level of -20dBFS produces a sound level of 77dBSPL. This is about 6dB down from THX Theater Standard levels, and seems to offer comfortable but loud levels of playback. Going all the way to THX Standard levels would seem too loud for my taste. But feel free to select your own preferred levels, and just note what they are.

Now whenever I speak of levels like 60dBSPL, you know exactly how to represent it in the digital dBFS system. My 77 dBSPL at -17dBFS peak sine amplitude leaves 20dB of headroom above my 0dBVU level of -20dBFS RMS. That's plenty for nearly all classical and wide dynamic range music reproduction. My conversion from dBSPL to dBFS requires that I subtract 97dB. So a level like 60 dBSPL RMS becomes -37dBFS RMS.

[As an aside, most all of these multiband linear compressors are peak-level compressors, not RMS compressors like those used in Crescendo.]

So, going back to setting the compression threshold, if your Pt were 50dB, then you'd set the compression threshold to (50 – 20 = 30) dB in SPL space, which becomes -67dBFS. Repeat this procedure for each band.

The hardest part is now behind us. For compression ratio, choose 4.5:1, and for makeup gain choose 21dB. Do this for each of the active bands. And then mix the output of the multiband compressor back into the dry chain.

Depending on how severe your corrections need to be, you may need to back off on the output gain to allow headroom for mixing in the corrections. You can make up for this attenuation with your analog volume control on your amplifier.

Now you have, approximately, a 3-4 band Crescendo, sort of… It won't sound as good as a real Crescendo, but I'll bet it sounds a whole lot better than nothing to you.

If you are unsure about what your threshold elevations are, try varying the compression thresholds in each band until it sounds best to you.


I've just made it even easier for everyone to try out Crescendo on their Mac computers. There are 3 files to download:

http://refined-audiometrics.com/tekram/CrescendoLive-Installer.pkg  (12 MB)

http://refined-audiometrics.com/tekram/Acudora-Save-Shares.app.zip (5 MB)

http://refined-audiometrics.com/tekram/eval-license-03012013.txt (1.4 KB)

Install with the installer. Either before or after that, but before attempting to run CrescendoLive.app, be sure to unzip and run the Acudora-Save-Shares.app. Copy the entire contents of the eval license text file and paste into the Acudora-Save-Shares text window and hit the Save License button.

This permits the evaluation copy of CrescendoLive to run on any Intel Mac computer (Mac-Mini, MacBook, tower systems, etc.), until 1 March 2013. There are no limitations in this version, except to state that it will cease functioning after March 1 of this year.

Nobody needs to contact me unless they wish to do so. If people feel they need longer to evaluate the system, they can contact me for a different license.

To reiterate the settings I use on my MacBook, I listen to WAV uncompressed audio files through iTunes, piped through our AcudoraPipe to CrescendoLive for processing, then out to a Lavry DA11 and into either KRK Nearfield Monitors, or my Beyerdynamic DT-880 or Sennheiser HD650 headphones for listening.