Gain Staging in the Workstation
Gain staging is the process by which you manage the signal levels in the various sections of the signal path to minimize noise and distortion while maintaining the best signal level. The basic principle is to figure out the dynamic range of your source (whether that source is a microphone, a synthesizer, a sound module, a drum machine, or anything else) then maximize the levels of that source without adding excessive noise or distortion.
It seems that this should be a simple process; you make sure that you set an appropriate level at the first stage of your signal path (with “appropriate” being the level that ensures a good signal-to-noise ratio while leaving enough headroom that the loudest signal peaks aren’t overdriving any of the circuitry in the path, which leads to clipping). It was pretty simple in the days of analog consoles: if the signal was overdriving the channel, you could hear it. If the signal was so quiet that turning it up to the appropriate level brought up the noise floor to where it was unacceptable, you could hear that, as well.
In the world of digital recording, though, the rules can change. With high-level gear (and old-fashioned analog VU meters), there’s plenty of headroom built into the system. Zero on the VU meters is generally +4dBu, so there’ll be possibly 20dB of headroom over 0VU. Even signals with a high peak-to-average ratio (like a snare drum) might clip the circuitry, but that distortion can be relatively unobtrusive. Another thing — on a console, the signal processing (EQ and compression) is built into the console and designed to work with the levels present. Even when outboard processors were used, they used the standard calibration of 0VU = +4dB. With a system-wide standard, gain staging was a fairly straightforward and logical process.
It’s different in the workstation. To begin with, signal levels in the digital domain are measured differently than in the analog world; analog measurements quoted in dB describe the ratio between the quantity of two levels — the level being measured and a reference. To describe an absolute value, the reference point must be known. There are a number of different reference points used in various applications (dBv, dBu, dBm), but the most common in analog audio is dBu, which represents the signal level as compared to 0.775 volts RMS with an unloaded, open-circuit source (the “u” in “dBu” means “unloaded”). In the digital domain, the most common measurement is dBFS, which represents the level compared to full scale. 0dBFS is the loudest possible digital signal. Any louder, and you get distortion — and digital distortion isn’t pretty.
So, how do analog and digital scales compare? 0dBu on an analog system is equivalent to -24dBFS; so +4dB, the standard operating level, is roughly equivalent to -20dBFS. +12dBu, a nominal analog peak level, equals -12dBFS. A full-scale digital signal, 0dBFS is +24dBu. Good recording practice in the digital world is similar to the analog world; to avoid needless distortion caused by signals that are too hot, keep your standard levels around -24dBFS and peak levels at -12dBFS. You’ll still have a decent amount of headroom for unplanned peaks and more than enough resolution to avoid signal-to-noise issues in a mix.
Why It Matters
The immediate concern regarding gain is getting audio into the workstation — if you overload the converters, then you’ll have distortion that can’t be fixed. Keep your nominal signal level around -24dBFS and your peaks at -12dBFS, and you’ll be safe. But the other reason is that, once audio is in the workstation, funny things can happen when you introduce plug-ins. In theory, 32-bit floating point software has a dynamic range approaching an astounding 1,500dB; but, in practice, when summing together many high-level signals, the summing buses in different workstations may not produce the same results — especially when plug-ins are used on the buses. Even with the astounding dynamic range that’s theoretically possible with 64-bit floating point processing, you need to be aware of (and manage) levels at every stage of the mix to avoid overloads, especially when using multiple plug-ins on a channel (or on a bus). Unlike the analog world, DAWs don’t have a “nominal” operating level — at least in the sense that you’re trying to get the optimal signal-to-noise ratio while leaving enough headroom that peak signals don’t overdrive the circuitry. In the digital domain, there’s an absolute ceiling (0dBFS) — and as long as you don’t hit that ceiling, you’re fine. But, since plug-ins also have a level ceiling, even if your channel level is good, if you hit the plug-in’s ceiling, then you’ll still be clipping.
What to Do
Given the huge amount of gain available in the workstation, it’s really easy to avoid clipping. Start by turning down the input stage for every channel — you don’t need to record as loud as possible. (If you don’t think the DAW is loud enough, turn up your speakers). Be aware of the headroom you have available in each plug-in — especially when using multiple plug-ins on each track. You can have a clean recording yet, if each plug-in is adding gain, end up clipping the output of the track. Finally, keep track of the DAW’s output bus; if you’re clipping the output, turn all of the individual tracks down. With a contemporary DAW, you have a remarkable amount of headroom available; don’t be afraid to use it.
Want to learn more? Visit Sweetwater Sound’s Sweetwater’s article on this subject.