My results on the equal loudness test
KeithHandy posted in Tools on February 3rd, 2008
I’ve got a neat little application I wrote, which, for simplicity’s sake, I call “louden”. It’s like a cross between hard limiting and distortion. Unlike hard limiting, there is no “attack” or “release”; the effect is virtually instantaneous, working on half a wave cycle at a time (if you consider a “wave cycle” to be a period of time where the wave goes below and above the center line exactly once, the part it works on is from the lowest point below the line to the highest point above it — and, alternately, vice-versa — and only does anything if one or both go out of range). Unlike ordinary distortion, which is what you wind up with when you set extremely low attack and release times on an ordinary limiter, “louden” doesn’t flatten the tops or bottoms; it maintains whatever curved shape was between them.
All this surely sounds like I’m full of shit, and in a way I’m sure I am. There’s no reason why the above idea/algorithm should have any merit. But this is one of those lucky times when it actually kind of does. When I push audio through this code, and test how far I can mangle it, the end result to my ears is like tape saturation; a kind of “gentlest of distortions”, which, despite having this non-digital character, happens to be extremely compatible with digital, because of how it reins more sound into a smaller space (yes, loudness war, I’m a bad, bad boy) and keeps those tops and bottoms inside the lines like a good kindergartener.
Despite its usefulness in ambushing the Red Hot Chili Peppers, I think there’s also a way to mold this into a useful limiter and de-esser for vocals. I read somewhere online (scroll down to “The Highs Have It”) that half the reason we need such a thing as “de-essers” in the first place is because the compressors we’re using aren’t responsive enough to the higher sibilance sounds. A typical old-skool compressor or limiter responds to the sound as it measures it… not as our ears would measure it.
So the key to making compression sound more like it’s working right, to us, would be to trick it into hearing more similarly to the way we hear. A lot of testing has been done to find what are called “equal loudness curves”, first by Fletcher and Munson in 1933, and then by Robinson and Dadson in 1956 (not sure why it always takes two guys to do the research). Apparently testing on this has continued since then, and today the current standard is called ISO 226. There’s no absolute way to determine how “accurate” any of this is, because it’s essentially a test of human perception.
With so much conflicting information — and of course, all the data changes drastically when you test people at different volumes — I decided to just test myself, using a simple online tool, and come up with my own damn curve. The results are undoubtedly biased by the type of headphones I use, the volume I tested at, and various childhood traumas percolating in my amygdala — but I seriously think this is better than looking at a bunch of pre-existing charts and trying to decide which one to rely on.
Although there are exceptions, I see that large parts of my graph look like a downward slope at 6 dB per octave. That means, twice as high = half as loud. Which brings me to a question: with all this supposedly being about “perception”, isn’t there also a reality that higher pitched sounds at the same amplitude would naturally have more energy? I know when I mix music, I’m more than happy to see kick drums and bass guitars with waveforms that nearly hit the top and bottom of the track, but I wouldn’t want to see a flute doing that. Yet, if we hear the flute as being loud enough to compete with the bass guitar, doesn’t that mean the flute is putting out just as much “energy”, only doing so by vibrating faster instead of wider? Why is our only measurement of sound energy — decibels — based on width (amplitude) alone? Why don’t we have a unit for width and frequency combined?
I’m not saying that our senses aren’t distorted; I’m just asking, shouldn’t the “distortion of our senses” be judged relative to that diagonal slope instead of a horizontal line? Shouldn’t “higher is louder” be accepted as an objective reality? I’m not joking.
By the way, I originally put the 12K and 16K where they are in the above screenshot because I simply couldn’t hear them. As I check again now, I realize I can hear the 12K, but I have to really crank the bejeezus out of it, so I’m not going to take the chance of over-interpreting that. I am, after all, approaching 40. I suggest you try the test yourself, but be sure to heed Warning #1 in large text at the top of the screen.
The point of all this, for me: if I make two EQ curves — first, one to simulate the exaggerations that our own ears/minds apply to sound, and then one to reverse it — and stick my “louden” effect between them, I should wind up with a nice “instantaneous” vocal limiter that is not as scientifically correct as it is psychoacoustically correct…
…which is what we want, because, we’re not scientists… we’re, uh, psychos.
