In The Acoustic Realm And At The Microphone

In general, we can expect to find noise floors in quiet recording rooms to be about 40 dB SPL (not A-weighted – more about that later) or greater. Such a noise floor usually sounds very quiet, unless there is a lot of high-frequency noise present (usually there isn’t, unless you have computer disk drives running, snicker, snicker). Meanwhile, acoustic music as heard by the listener generally exists across a maximum amplitude range from about 50 dB SPL (pianissimo solo vocalist) to 110 dB SPL (fortissimo brass section), with occasional louder excursions in amplified music. Levels much above 120 dB SPL are difficult to achieve simply because of their power requirements and that’s just as well, because above that level the rarefaction portions of the sound wave become non-linear as they begin to approach a vacuum! That’s probably why such levels hurt!

So, anyway, the dynamic range of the typical acoustical environment is around 90 dB, about the same as a CD. And the actual encountered dynamic range of music to noise floor will hardly ever be more than 60 dB.

Obviously, the closer the microphone is to the sound source, the better its acoustic signal-to-noise ratio is going to be. But that ratio is always going to be limited by how loud the source is vs. how loud the noise floor in the room is. If you have a singer singing at 90 dB SPL over a noise floor of 40 dB SPL, then the signal-to-noise ratio is only 50 dB! And that’s quite good!

All of this should lead you to the insight that the dynamic range of 16-bit digital recorders is probably adequate. In fact, none of us make 16-bit recordings, because by the time the signal hits the mike, it typically needs only ten bits (60 dB) to store it.

So, in a perfect world, it would appear that we shouldn’t really need 16 bits, or 20 bits, or whatever. But the world is neither perfect nor simple. So let’s look at some issues here: the nature of noise floors, the nature of headroom and the nature of music signals.

About Noise floors

When we think about noise floors, it helps to be a little more specific. Neither acoustical nor electrical noise floors have a flat spectrum, and to make matters more confusing, they have different spectra.

Acoustical noise floors have predominantly low-frequency energy. This works out well for us, because we don’t hear low frequencies very well and aren’t annoyed by the mid-level low frequency stuff. For this reason, we often use so-called A-weighting when we make low-level measurements, because A-weighting approximately represents the response curve of our hearing at a level of 40 dB SPL. The important thing to note about A-weighting is that the bottom octave (32 Hz.) is attenuated by approximately 33 dB, while the second octave (63 Hz.) is attenuated by about 21 dB and the third octave (125 Hz.) is attenuated by 11 dB.

This is a meaningful and useful adjustment to make, and it means that our 40 dB SPL environment may very well “sound” like a 30 dB SPL environment if there is no high-frequency noise in the room, such as those pesky computer fans. At the same time, virtually all of the acoustical energy in the room will probably be low frequency.

What this means is that acoustical noise may be less of a problem than it appears, and that so long as the spectrum of the acoustical noise doesn’t contain high-frequency components, you can often capture desired audio signals (i.e. music) from your microphone that appear to have as much as 70 dB dynamic range.

On the other hand, electrical noise is usually random “white noise.” Such noise has a spectrum with a 3 dB/octave slope rising toward higher frequencies. Such noise is, by comparison with acoustical noise, quite audible and annoying. This will be of considerable significance when we get to the microphone.

About Headroom

When we make a recording, we cannot reasonably predict with any precision what the loudest level will actually be, so we leave a safety margin. This margin – headroom – is the slack we cut ourselves to keep from getting in trouble with distortion. Much of recording engineering craft has to do with tip-toeing between noise and distortion, keeping just enough headroom to be safe while not bringing up the noise too far. Remember, as we increase headroom, we lose signal-to-noise. It is also worth noting that in the post-production process we usually get rid of the headroom, and bring the audio peaks up to 0 dBFS.

About Music Signals

One final thing to keep in mind is the difference between the signal-to-noise ratio and what we might call the signal-to-signal ratio – the range between the loudest and softest signals. Take a look at the following Noise Level Analysis of Billie Jean, coming from the console.

	Michael Jackson’s Billie Jean

This graph shows the level over time of Billie Jean. Notice that the music is quite tightly controlled, with a very limited range except for the first 2 bars of kick drum/snare solo. The actual dynamic range of the music is about 12 dB and the average level (Leq) is only 8 dB below Lmax. This is the result of extensive compression and very strict level management.

The insight to be gained from this is that often the music itself (particularly multitrack pop and rock music) has a very narrow dynamic range and you need comparatively little headroom or noise floor. To be cynical and crass, you may need nothing more than a decent noise gate or quick hand with a fader at the beginning and end of the selection to get rid of the remaining noise!