Print Page

Max Headroom

6/1/2007

Gain staging is the art of setting the multiple gain controls in any signal path to get the cleanest signal. Many of the rules of analog gain staging that pertain to balancing still apply to the digital realm, but when it comes to headroom and clipping, the rules of the game change completely.

A good place to begin in your pursuit of understanding how gain staging works in your DAW is to distinguish between two systems of calculation: floating point and fixed point. Most native DAWs use floating-point calculations, whereas PCM audio files and some DAWs, most notably Pro Tools, use fixed-point calculations.

FIG. 1: It''s 12 dB. Do you know where your clips are?

A Bit Too Much

Floating-point numbers use a fixed number of bits for the mantissa, or basic value; the remaining bits are used for the exponent. That amounts to a gain-ranging scheme controlled by the exponent; the mantissa is like a window of dynamic range that is moved up and down over the full dynamic range by changing the exponent.

Most commonly, 32-bit floating-point numbers with a 24-bit mantissa and an 8-bit exponent are used (exceptions include Cakewalk Sonar and Sonic Studio SoundBlade, which use 64-bit floating-point numbers). That gives the mantissa the same dynamic range as a standard 24-bit fixed-point PCM audio file. The gain ranging that is represented by the 8-bit exponent provides more than 1,500 dB of dynamic range. Plug-ins for native systems and add-on hardware like the Universal Audio UAD-1 PCI card also use 32-bit floating-point math.

In practice, it is almost impossible to clip an audio signal inside a 32-bit floating-point mixing engine — headroom isn't a problem. The problem arises when you try to send a signal with this huge dynamic range to an outside world that wants a 24-bit PCM signal — for example, when you bounce to disk or send a signal to the digital-to-audio converter (DAC) in your audio interface.

In recognition of this fact, the meters in most DAWs that have floating-point mixing engines actually reflect the dynamic range of the 24-bit PCM output. When the clip indicator lights up in your DAW, it doesn't mean the signal is clipping inside the DAW; it means that it would be clipped if it were output directly to a 24-bit fixed-point file or an audio interface. Avoiding clipping in a DAW mix is therefore usually as simple as inserting a master fader in your session and lowering it until its meter no longer shows clipping.

That being said, some attention to adding some gain staging earlier in the signal path may be useful. For instance, when you want to render individual tracks or subgroups to disk or directly to the output for external processing, it's convenient to already have them at an appropriate level. Furthermore, meters that are not pinned are more useful or at least more pleasing to look at. For one approach to gain staging in the digital domain, see “Step-by-Step Instructions” on p. 72.

On the Other Hand

With Pro Tools and other fixed-point systems, the story is quite different. Pro Tools runs on Digidesign's proprietary TDM hardware, which has a 24-bit-wide, time-domain multiplexed bus. The program uses 48-bit double-precision, fixed-point math, so each signal inside the mixer uses up two TDM channels, or multiplexing time slots, to travel over the TDM bus.

Things get pretty complicated trying to figure out exactly what Pro Tools' meters are telling you, but Digidesign reserves a number of bits above 0 dBfs for headroom. As in native systems, channel clip indicators show that the signal will clip if it is sent to a DAC (as opposed to showing that it is clipping internally). On the other hand, the clip indicator of a master-fader strip does show actual clipping. It's possible, though not easy, to clip this output — a loud film mix with a lot of channels might be able to do so.

Pro Tools LE complicates the situation further because it is a native system and uses a 32-bit floating-point engine. The metering is complicated in LE, but two key meters — the prefader-channel meter and the final master-fader output meter — are 24-bit fixed; therefore, clipping really means clipping.

To suit LE's floating-point engine, most RTAS plug-ins use 32-bit floating-point math. That makes the use of RTAS plug-ins in a TDM system a little messy, because the signals need to be converted between the two representations.

Gain staging has always been one of those skills that draw on both theory and practical experience. Although the details have changed with the migration to digital audio, that concept remains intact. Best results come from combining some understanding of what is happening under the hood with experimenting and listening on your own system.


Larry the O recently finished floating-point scuba diving in Thailand.

STEP-BY-STEP INSTRUCTIONS

1.

Record signals to peak at no higher than -6 dB on the meters.

2.

Start your mix with your faders set to an average position of half to two-thirds of the fader travel.

Start your mix with your faders set to an average position of half to two-thirds of the fader travel.

3.

Use groups. Once you've gotten a good balance of elements within a group, using a group master won't mess with that balance.

Use groups. Once you've gotten a good balance of elements within a group, using a group master won't mess with that balance.

4.

Use master faders in your sessions. When a master-fader meter clips, simply lower the fader until the clipping stops.

5.

Watch out for clipping in plug-ins. Lower the plug-in's input level if necessary.

6.

If a track-level fader is inconveniently high or low, insert a gain plug-in to adjust the level.

  Print Page