Bit depth and what that means to the home recordist.
In the digital audio medium, few things are as aurally unpleasant as hard clipping. Unfortunately, I sometimes receive files to mix that have been clipped because of outdated information, hearsay, and a general misunderstanding of digital audio.
Digital clipping occurs when a signal surpasses the headroom of the system, resulting in distortion as the waveform is limited and squared off. This is referred to as hard clipping. In some analog systems, such as with magnetic tape or vacuum tubes, clipping is less dramatic and may even be desired (guitar amplifier distortion, for example). This is referred to as soft clipping and will be ignored for the remainder of this post.
The headroom of a system is defined as difference above normal operating level and the onset of clipping. Dynamic range is defined as the ratio between the loudest undistorted amplitude of a system to the RMS amplitude of noise in the system.
Simplified somewhat, digital audio is made up of zeroes and ones comprising binary words, each representing the amplitude of an analog signal at a particular point in time. The sample rate for CD audio is 44.1kHz. When converting an analog signal to digital for CD, 44.1 thousand words are written each second to represent that signal. The greater the bit depth, the longer the word and the higher the possible resolution. For example, a 16-bit signal (such as CD audio) has 216, or 65,536 possible values. For each added bit of information, the resolution doubles. So a 24-bit signal has 16,777,216 possible values. The theoretical maximum dynamic range for a 16-bit signal is 96dB and 144dB for a 24-bit signal, limited by quantization error. In reality, limitations in the analog components of digital audio converters can reduce a system’s dynamic range and dither can increase it, making these figures approximations.
Now, what does that mean to someone making a recording at home?
In the days of analog tape, noise was a real concern and engineers employed a number of techniques to maintain an optimum signal-to-noise ratio. Performances could be lost if the signal was not loud enough. When digital systems first started to appear, 16-bit was adopted as the standard, and engineers continued to track with many of the same techniques used with analog mediums. With the limited dynamic range afforded by a 16-bit system (human hearing has approximately 140dB of dynamic range), noise was still a concern and “using all the bits” by pushing the signal level was important.
Unfortunately, this way of working presents some problems. Recording at the edge of a system’s operating maximum can be risky – a clipped signal is often irreparable. Also, many processes in recording and mixing can add gain to the signal. If there is no headroom left, the engineer will be forced to make adjustments to allow for the extra gain. Possibly most important of all, audio processors, both analog and digital, are designed to operate at an ideal level. Working at too-high a level in the digital realm can cause the input stage of these processors to overload, introducing distortion.
These days, however, every digital system should offer an option to record at 24-bit. If the reason for doing so hasn’t become clear by now, just follow blindly and do it anyway. Recording and mixing at 24-bit allows one to adjust a signal to peak at a lower level with no loss in resolution compared to a 16-bit signal.
As far as signal level, just keep it low. Tracking with peaks in the -18dB to -12dB range has absolutely no negative impact on the resolution of a signal. In fact, those 18dB equate to about 3 bits of resolution. By keeping your signal at a reasonable level, you’re avoiding clipping, even if the source being recorded becomes considerably louder before you can reach the input gain knob. In addition, you’re not overdriving your preamp, your processors, or your master bus. You may find your workflow is much more relaxed when you’re not constantly having to adjust your signal level to account for small changes in the performance and signal processing. Depending on your equipment, you may also notice a dramatic improvement in the quality of your recordings.
Still concerned with “using all the bits”?