Bias and Subsonics in Mixing and Stereo Imaging
DC Bias or Baseline Offset is one of a class of effects that are utterly unappreciable unless used in the right context or for the right purpose. Bias has no effect on the frequency content or wave shape where audible frequencies are concerned, and therefore does not change the way we perceive a sound. So it might well be asked, then, why write an article on an effect that does absolutely nothing? Well, the fact is that its effectiveness is in its handling of subsonics and DC characteristics, and that these subtle movements, while posessing no audible tone of their own, do have an effect on the way we perceive an audible sound. Subsonic patterns and DC components, in their perceptual role, form a signature of sorts that in our perception separates the audio waves with which they are associated from other waves that we hear simultaneously, even while similar frequencies or harmonics are present in the others. Without changing the wave shape, waves are perceived as being different from each other by virtue of their different crossings of the baseline which give altered patterns of compression and rarefaction. The rest of this article shall elaborate on this concept by applying it to specific uses.
The primary implementation is in mixing. If you are familiar with a Windows-based mixing application called SAW Pro, then this is an instance where DC biasing is very apparent. If you have a demo copy, you will have noticed its characteristic tone. If you load audio files into it that were recorded with other applications, then the "naturalness" of the sounds are gone, but there is remarkable "neatness" and clarity in the resulting playback. It's by no means a "plastic" sound, but it doesn't breathe. It doesn't have any background grittiness that you might have intended to emphasise. Maybe this effect can be turned off - I haven't looked that closely. This is fairly typical of the sort of mix you can get with a DC bias effect where the original biasing characteristics are flattened out and then followed by a simple block-move function. A common procedure would involve finding the average sample value within a certain segment of audio data, multiplying that value by -1 and adding the result to all samples within that segment. This should reduce all subsonic activities and DC biasing to an audio-insignificance teetering about the baseline. A common increment is then suffered by all samples in the file, to create the new bias: theoretically a constant bias.
If you have the flexibility - and the initiative - and the time, let's not forget - then you can customise your biasing according to the requirements of the project. Maybe a solo section needs to stand out without drowning the rest of the mix in its own frequency range. This is then a very useful technique. Maybe two tracks have a conflicting frequency content and filtering simply won't yield appropriate results. By making a sharp alteration to the subsonic or bias content of one or both tracks, the frequency conflict is largely overcome. You can rule the earth this way.
Everyone knows how important it is to breathe. I dare you to go without for just two minutes. In audio engineering, we sometimes like the sound of tubes (or valves, depending on which side of which ocean you're from); old fashioned coil reverbs, torn speakers, the "wrong" cables being used, etcetera. So, too, with these subsonics. It's only an audio redneck or a minidisk freak that gets critical about leaving inaudible data functions in a recording. I say remove at own risk. If you make any kind of recording from an external source, whether it is line or mic, an offset of some kind is going to result. Period. It stems from the characteristics of all the circuitry involved in the process; particularly that which precedes and constitutes the greatest gain stage of amplification.
Now, listening in mono, you won't get an audible difference, regardless what you do with it. But in stereo, that is, with effectively two tracks running, a difference can be heard. With left and right channels possessing different characteristics, the stereo image is less clear, and in a certain way, more interesting. More on that later. For now, it breathes. It has a natural quality which by flattening out, as with some DC bias functions, can only be lost. The result of this loss, with a single stereo recording, is a dull, mechanical sound. Maybe Gary Newman or Edgar Froese would adore it that way, but most of us wouldn't.
If you're aim is not to make electronic-sounding beeps and buzzes, then free-form manipulation of subsonic content is more appropriate. Creating a file of subsonic waves that are different in both channels and which have their own constant DC bias can do this quite easily. Then it must be looped to form the correct time-length, and mixed onto the original track. What you then have is a natural-sounding track that stands out from the rest of the mix with a good degree of clarity.
We mentioned stereo effects above. To elaborate, when the two channels can be most easily compared and when stereo imaging cues are well ordered, a stereo image is at its clearest. The effect of a difference in volume between left and right is hampered when a DC bias or subsonic component applied differently to both channels alters the compression and rarefaction patterns without (or yet only inaudibly) changing the wave shape. The peaks in the left and right channels are then a little closer in amplitude than are the troughs, or vice-versa. This has a confusing effect where positional perception in the horizontal plane is concerned.
It would be easily taken from the last paragraph that this effect is simply at enmity with stereo imaging. This is not so: an appreciable bias, beyond what is required for a "natural" sound, is good for rendering to an ambient quality. It can thus be applied to reverb algorithms with good effect. Moreover, the sometimes directionless sound of thunder can be due in part to a large subsonic content. A passable thunder effect can be generated with quite diverse soundwaves, even if you are equipped with just two speakers. (You really can rule the earth this way). As to the effect on other directional parameters such as HRTF filtering, interaural delay and ear canal resonance, I am not sure. The relationship between amplitude and bias amount is another area that I cannot make a judgement on.
So, above are the abilities of this simple little effect. You cannot hear it on its own because it has no effect on the wave shape, but its effect is entirely context dependent. Depending on the context, it lends or impedes definition to a sound. Where the likeness of waves is required for comparison, a bias or subsonic component works against definition. Where the difference of mixed waves is very great, it enhances the impression of separateness. The presence of a small subsonic undertone lends to a natural feeling in a stereo recording, while the absence of such diminishes it.