Illustration: Dmitry Panich
Every good engineer knows that proper level management is critical to making good recordings, and that the ultimate weapon against setting improper levels is a pair of well-trained ears. Meters, whether in the form of old-school bouncing needles or colorful ladders of LEDs, are important visual aids that help quantify what the ears are hearing.
Meters show up on almost every type of audio device, from preamps to consoles and from recorders to software. Generally speaking, they show us the current level of input or output signal relative to some reference point to help us find that sweet spot between the lower (noise floor) and upper (overload) limitations of a device.
To make the best use of your meters, it's important to understand exactly what they're telling you. To that end, I'll discuss the different types of meters you may encounter, what they really show, and how things change between analog and digital environments.
IN MY VU
FIG. 1: The classic VU meter, shown here on an Avalon VT-737sp compressor, is a specialized voltmeter with a needle indicating audio levels relative to 0 VU, a reference indicating the optimum level.
The classic needle-based VU meter was introduced more than 60 years ago, in part as a visual reference for radio announcers to check that their speaking volume hovered at the optimum level for broadcast (see Fig. 1). VU stands for volume unit, and a level of 0 VU is calibrated to correspond to a specific input or output voltage.
The physical response of the needle, or its ballistics, is deliberately slow, yielding a relatively stable display of the current overall level rather than bouncing rapidly with changing volumes. As a result, a VU meter is a decent indicator of a sound's average loudness. In other words, the needle doesn't react quickly enough to capture the exact level of an instrument's hard attacks or the short gaps between notes. Instead, it settles on the level of the sustained parts of a sound, which is (roughly speaking) the part that our ears perceive as the sound's actual loudness.
That's useful for balancing the levels of different sounds, but it's of little use for ensuring that signals don't distort at their loudest peaks. As a result, VU meters always have a certain amount of headroom built in. Headroom is defined as the difference between the optimum level and maximum level (the point above which distortion occurs).
As an example of how this all fits together, let's look at how to use the VU meter on an analog recorder. Quite simply, you raise or lower the output level of the device feeding the recorder (a console or a preamp, perhaps) so that the recorder's input VU meter hovers at 0 VU during the loud portions of the music. If the meter occasionally bounces 2 to 3 dB higher during the loudest portions, not to worry — that's part of what the headroom is there for. The rest of the headroom is there because if the average level of the music reaches +3 dB, the peaks (the parts the VU meter is too slow to measure) are reaching several decibels higher than that, and we don't want them to distort.
The amount of headroom that you have to work with varies with the device and the circumstance. For example, in broadcast situations it's desirable to compress a signal so that average level is closer to maximum level, but in classical recording the difference between the average and maximum levels can be quite large.
MY INTEREST IS PEAKED
FIG. 2: Peak meters are found in analog and digital devices and generally use ladders of green, yellow, and red LEDs to indicate normal, high, and maximum levels, respectively.
To keep closer track of peak levels, the peak meter was introduced (see Fig. 2). A peak meter reacts much more quickly than a VU meter, allowing it to track a signal's brief excursions above the average level. That gives an engineer the ability to know precisely how much headroom is being used.
For most signals, though, peak meters give little or no indication of the perceived loudness a VU meter shows so well. For organ sounds, in which the tone starts at a certain amplitude and maintains that amplitude until the tone ends, peak and average levels are the same. For percussive tones such as a piano or Clavinet, however, the brief but powerful attack is much louder than the sustain, producing a wide discrepancy between average level and peak level. (That's why normalization, which automatically raises the level of tracks so that their highest peaks are at maximum, often results in one track sounding much louder than another.)
Peak meters are typically rows of green, yellow, and red LEDs. The colors mean slightly different things on digital equipment than on analog equipment, but the red indicator is always the maximum level and the yellow LEDs — like an amber traffic light — warn of the approaching red.
Peak meters are particularly important in digital audio, because the onset of distortion is so immediate and harsh. For example, when digital clipping occurs meters actually run out of numbers to describe a peak that is too high. The result is an ugly-sounding “flat-top” waveform. Using a peak meter to keep a close watch on peak levels in digital gear is standard practice. (Analog distortion, by contrast, can be relatively minor, so that if a peak exceeds maximum level only slightly the distortion will probably be inoffensive.)
As a constant reminder of the absolute nature of this ceiling, we ordinarily measure digital levels downward from maximum. The highest level, or full scale, is designated as 0, and all other levels are measured as so many decibels below full scale (dBFS).
I'M SEEING RED
Although it's good practice to avoid tripping the red 0 dBFS indicator on your peak meter, seeing red doesn't necessarily mean a take is ruined. Remember, full scale refers to the highest number that our digital system can assign to a voltage, and hitting that number for one sample just means we've made maximum use of our system's dynamic range. Two or three samples at full scale, however, means that the voltage must have exceeded 0 dBFS for a short time and that the signal has been clipped.
That's why digital peak meters often have an overs indicator, which lights to show that a certain number of samples in a row have reached 0 dBFS. Sometimes the overs indicator is actually a counter, while other times the number of samples that constitute an over is arbitrarily set by the manufacturer. There is no real industry standard, so check your documentation or contact the manufacturer if you want to know how your meters are calibrated.
Ultimately, of course, your ears are the best judge of whether a recording has been ruined by clipping. Depending on the program material, you may find that several full-scale samples in a row exhibit no audible distortion. That doesn't mean it's a good idea, though — it just means you got away with it!
I WANT IT ALL!
So VU-style average-level metering is useful for making musical judgments about the loudness of a signal, and peak metering is a necessity for ensuring distortion-free digital recording and processing. At the dawn of the 21st century, do we really need to sacrifice one for the other?
Heck, no! Meters are now increasingly being designed (especially in software) to allow the engineer to choose which kind of metering is appropriate. Often, both types of metering can be displayed simultaneously.
FIG. 3: This meter display from Cakewalk Sonar 1.3 combines average (RMS) metering with peak metering and includes a peak-hold function.
Fig. 3 shows how this is implemented in Cakewalk Sonar 1.3. The drop-down menu allows the user to choose Peak metering, RMS metering, or both. RMS stands for root mean square, a method of averaging values that yields a truer representation of perceived loudness than does the good old VU meter.
In Fig. 3, the solid green bars at the lower end of the scale indicate RMS levels hovering at around -16 dBFS. The short yellow and red bars above indicate the current peak levels of -7 dBFS in the left channel and -5 dBFS in the right channel. Note that the peaks are 10 to 11 dB higher than the average (RMS) levels.
Near the top of the scale are two thin white bars indicating the highest recent peaks. This peak hold function is common in peak meters and usually retains its value for from one to three seconds. Sometimes “infinite” peak hold is available to retain peak levels until they are reset manually. Because the values in Fig. 3 are at -0.5 dBFS and -3 dBFS, we had better hope that represents the loudest point in the recording session!
If you're accustomed to peak meters, you may be surprised to find that the perceived loudness (as represented by the RMS level) is more than 15 dB lower than the peak value at the loudest part of the recording. This particular example is of a collegiate symphonic band playing a particularly dynamic and percussive piece and is representative of the sort of peak-to-average differences that one often finds in large classical ensembles.
Compare that with what would happen if we recorded a pipe organ and set levels to peak near 0 dBFS. Because the peak-to-average differences are virtually nil with an organ, the perceived loudness of the organ recording would be more than 15 dB louder than the symphonic band recording — a result that would be extremely unnatural sounding. Therefore, having both peak and VU- or RMS-style meters and knowing how to use them is clearly essential to making great recordings.
THE CASE FOR K
A great example of peak and average metering living together in harmony can be seen in mastering engineer Bob Katz's proposed K System. The K System, as detailed on Katz's Digital Domain Web site (www.digido.com), integrates metering, monitor calibration, and standardized level practices, but here we'll look at just the meters.
FIG. 4: Here, the K System K-14 meter is shown as implemented in Metric Halo Labs Spectrafoo. Note that full scale is labeled +14 dB, indicating 14 dB of headroom.
K System meters feature solid bars to indicate RMS average levels, with lines or dots above to indicate instantaneous peak levels (see Fig. 4). Peak-hold indicators can be set to 10-second hold or infinite hold, and an overs counter is recommended.
So far that sounds similar to the Cakewalk Sonar meters discussed earlier, but the K System meters don't count down from 0 dBFS. Instead, the 0 dB point is set 20, 14, or 12 dB below full scale, reflecting the VU meter-style practice of setting levels by perceived loudness instead of by peak. Full scale is designated +20, +14, or +12 dB. These three variations, called K-20, K-14, and K-12, respectively, reflect different amounts of headroom and are intended for different applications.
K-20 metering accommodates the demands of very dynamic music, such as symphony orchestras, audiophile recordings, and film sound. More compressed styles such as pop, rock, and R&B use the K-14 scale, and the K-12 scale is reserved for broadcast production.
Consider the ramifications of the K System for recording the symphonic band and pipe organ mentioned previously. For the band, setting levels to hover around 0 dB on the K-20 meters when the band is playing forte (full, but not maximum, volume) would leave enough headroom for the 15 to 16 dB difference between average and peak levels with a small cushion before clipping. Setting the organ's forte levels to 0 dB would appear at first glance to “waste” some dynamic range, but it would result in a level that compares naturally to the band recording.
I'D LIKE TO METER
Meter design and implementation is remarkably inconsistent. Even though the characteristics of VU and peak meters are well defined, it's hard to know whether your meters conform to those specifications precisely. The only defense is to learn how your meters respond and to compare what you see with what you hear. Find out whether your meters offer more than one type of display and how many overs set off an alarm.
If you have access to peak and average metering, use them both. If you have only one, use your ears and your good sense to extrapolate the information that your meter isn't providing. Armed with the knowledge of what your meters are telling you, you'll be able to manage your levels sensibly.
Brian Smithersis a musician, engineer, and educator in Orlando, Florida. He teaches Audio Workstations at Full Sail Real World Education.