Fun ‘N Profit Time: What It All Means

Conclusion? We can hear components of sound well below the measured dynamic range of a system, and a given signal can have audible components 80 dB or more below the loudest components of the signal under some circumstances. The presence of a noise floor doesn’t mask everything else below it. At worst, stuff 10 dB below the noise floor will still be audible (i.e. you’ll be able to hear music down to 10 dB below the hiss!), and at best you’ll be able to hear specific elements as much as 60 dB below the floor.

Or, vice versa. You may be able to hear the noise floor of a recording 60 dB below the signal. Like tape hiss. Remember it?

Bob Ludwig is, as you probably figured out by now, right once again. Due to the rather amazing nature of our hearing, we can simultaneously sort out sound components that are 80 dB apart in level. We can also hear sounds that are considerably softer than the measured noise floor present. So, figure that you need to make allowances for hearing sounds 20 dB below the electronic noise floor and up to 60 dB below the acoustic noise floor.

Further, at any given moment in time you are hearing both soft and loud sounds simultaneously. With that in mind, let’s take another look at that poor long-suffering 96 dB range Compact Disc. If you record 15 dB below Full Scale, then it’s quite possible that there are audible components 80 dB down in the signal that are going to be hovering right around -95 dB Full Scale. Thanks to the miracle of dither, these signals will be audible and won’t necessarily even sound too bad.

As signal levels get softer, though, those components are going to disappear. In an analog signal trace, their disappearance will be gradual and smooth. In 16-bit digital format, once they drop below the Least Significant Bit they are going to drop out comparatively abruptly and roughly, which is part of why there is complaint about the behavior of low-level reverb tails and musical sound decays in 16-bit digital recording.

So, even though these elements are way outside the range of measured levels, they are still audible, and to some, significant. How significant? To my way of thinking, compared to issues like mic choice and placement, equalization and monitor level they aren’t terribly important. But, and this is a very big but, remember Tom Bates’ definition earlier. Low lever signals will have poorer resolution and at some point this begins to be a serious problem. Meanwhile, we are in an obsessive business and much of our work is at the edges of the envelope. So, yes, Virginia, you can hear these things, and yes, they do matter.

Happy dither!