But first, let’s go over some digital-audio basics. The A/D converter in your audio interface measures or samples the incoming analog signal many thousand times a second, and it converts those samples to the binary numbers that you record on your hard drive. Figure 1 shows an analog sine wave that is sampled periodically. The digital meters in your DAW show those sample levels—not necessarily the peak signal levels that appear between samples. In your D/A converter, a lowpass filter (or reconstruction filter) creates occasional peaks between the samples that can be up to 3dB higher than the measured level at certain high frequencies. If your recorded levels are near 0dB, this can cause clipping that the meters don’t display. Similarly, some plug-ins can create larger peaks than those that went in—without any increase in volume, or audible change in the program. Even an EQ cut or roll-off can increase signal levels by producing intersample peaks due to ringing at the cut frequency.

Love the Peaks

As shown in Figure 2, a musical signal changes in level continuously as it plays. Imagine a musical passage with a low-level synth pad, but with high-level drum hits. The average level or volume of the passage is low, but the transient peak levels are high. Peak levels may be up to 24dB above average levels, depending on the type of signal. Percussive sounds have much higher peaks than continuous sounds (synth pads, organ, flute)—even if the two signals have similar average levels.

The meters in your DAW can show signal levels in two modes: RMS and peak. RMS readings correspond to the average levels, and peak readings show the level of peaks, or short transients. The average or RMS level indicates approximately how loud the sound is, and the peak level shows how close the signal is to clipping. So, as you don’t want to clip or distort the signal while recording or mixing, use peak metering—not RMS.

Reduce Recording Levels

In general, reduce your recording levels by turning down the gain in your mic preamps. Record at about –6dB maximum in peak-meter mode. One benefit of lower recording levels is that you won’t overdrive your mic preamp. The distortion of most analog gear increases as you approach maximum gain. Another benefit of reduced levels is that it creates some headroom for your plug-ins. Going for 6dB of headroom should eliminate any invisible overs. Then, you can set up your mix balances without having to adjust levels every time you insert a plug-in.

Remember, the recorded level on each track drives the plug-ins—not the track fader, which comes after the plugs. If a track was recorded at 0dB, use your DAW’s trim control (or insert a –6dB trim plug-in) to reduce the signal level going to your plugs. Do this before setting up a compressor or gate, because trim affects their gain reduction.

But doesn’t a low recording level increase noise? Yes—but in practice, it’s not a problem. A 24-bit recording has a theoretical signal-to-noise ratio of 144dB. So even if you record at -6dB, you’re still far above the noise floor, and you won’t hear any increase in hiss.

Keep Reducing

Here are some other spots in the signal path you can set your level to a maximum of –6dB:

Stems. At the output of each plug-in, check for clipping—especially in an equalizer set to boost. Find the plug’s output gain control, and turn it down until clipping stops. You might have several plugs in series, and you don’t want the output of one plug to overload the input of another.

Upsampling plug-in outputs. These DSP processes upsample the signal (increase its sample rate), do the effect, and then downsample the signal at the output. The downsampler acts like a partial reconstruction filter. So even with the effect turned off, the up and down sampling can create overs.

Mixdown master. Set the maximum peak level of the master output bus to –3dB to –6dB to allow for signal peaks that do not display on sample-reading meters. You can make up the gain later during mastering, where you boost the overall level or peak-limit/normalize to make a hot CD. Try to keep the master faders at or near 0dB. If the master faders are set low, you will turn up the channel faders to get a good mix level, which results in high-level signals that can overload the mix bus. [Note: Mix bus overload is not a serious problem if the DAW uses 48-bit or 64-bit floating-point math in the mix bus, because float processing can pass signals above 0dB without clipping. Still, it’s good practice to control the levels hitting the mix bus. Summing calculations become less accurate when signals exceed full scale.]

Compressor makeup gain. Don’t do it! Makeup gain can cause the initial uncompressed transient to clip by boosting its level. Raise the track fader instead.

A Tip For Mastering

If your mix is going to a mastering engineer, omit any stereo-bus processing. Record the file at a maximum of –6dB to –3dB so that the mix doesn’t overload the mastering engineer’s D/A converter. If you are mastering your own mixes, use an oversampling meter that displays the actual reconstructed signal, or set the limiter ceiling at –3dB to –0.3dB—not 0dB or higher. This avoids making files that sound distorted on listeners’ playback systems, as many CD players distort with samples above the clip level.

9 TIPS FOR SMOOTHER SOUNDS

• Dirty contacts can cause signal rectification and distortion. Clean connectors with agents such as Caig Labs DeoxIT (www.caig.com). 
• Use audio interfaces designed for low jitter. 
• During mastering, add dither when truncating from 24-bit to 16-bit. This eliminates quantization distortion at low-signal levels.
• Minimize the calculations your DAW has to do. Each calculation creates some error or distortion by increasing the word length beyond 24 bits. Rather than boosting the gain of a musical section and dropping it later, undo the changes, and apply only the correct amount of gain change. 
• Consider using fader moves rather than compression. Compression adds distortion because it changes the waveform. Too-fast release times cause distortion in low-frequency notes, and also cause pumping. A fader-setting change on a series of notes is less audible than a compressor working on each note.
• When you peak-limit and normalize a stereo mix to make a hot CD, try not to exceed 6dB to 7dB of peak reduction, as limiting adds distortion. Use even less limiting on mixes that don’t have loud transients. Rather than normalizing the limited signal to 0dB, normalize to –3dB to –1dB to allow for intersample peaks.
• Experiment with an analog tape plug-in to warm up a track, or record to analog multitrack tape and dump the tracks to your DAW for editing/mixing.
• To reduce graininess in a reverb plug-in, use a high-density setting, or use a convolution reverb. 
• When recording source sounds, consider using less high-frequency boosts. Cutting around 3kHz to 7kHz tends to reduce harshness. Try a 4kHz lowpass filter on distorted guitar amps to take the edge off, and try multiband compression on voices that get raspy. Set the compressor so it kicks in above 3kHz or so during loud passages. Also, many condenser mics have high-frequency peaks that can sound brittle or sibilant. Apply a high-frequency roll-off, or better yet, use ribbon mics or flat-response condensers. Try moving the mic to a spot that sounds mellower, as well. Mic a vocal at nose height if it sounds too edgy when miked at mouth height. Mic a trumpet off-axis, and position the mic near the edge of the speaker cone on a guitar amp. Finally, use a tube mic preamp that can produce euphonic even-order distortion that’s toasty warm.
  —Bruce Bartlett