How to set and monitor correct recording levels

Three simple concepts

Why start with levels? Setting correct recording levels may significantly improve recording quality. Unfortunately, there is no "set it and forget it" button anywhere in a field recording circuit. The right technique requires the understanding of three concepts:

• signal level,
• signal overload and clipping
• headroom

Signal level can be thought of either as a voltage level or a power (in decibels, when a reference power is given). In simple terms, signal level determines whether the recording is loud or soft.

This notion is closely linked to that of signal overload, which indicates a point at which a recording device (e.g., preamplifier) overloads and causes distortion. We typically think of signal overload as caused by an analog device's electrical and/or acoustic performance, but signal distortion can also occur as a result of clipping, which is thought of here as an error in digital processing. The distinction allows us the following generalization:

• a signal with analog overload may be captured without actual sample clipping
• a signal without analog overload may become digitally clipped due to errors in processing

Figure 1 shows illustrates the point. The recording in the left panel shows analog signal overload due to incorrect setting of recording levels on the microphone pre-amplifier. Note that no digital samples are actually clipped (RMS= 86 dB). The recording in the right panel shows a sample that was captured correctly, but underwent erroneous digital processing, which resulted in sample clipping (RMS = 96 dB). Obviously, the signal is bad in both cases and would undoubtedly result in inaccurate speech analysis.

Figure 1. Examples of analog overload and digital clipping

This brings us to the notion of headroom. The headroom of a recording device is the difference in power between the highest signal level present in a recording and the highest level available on this device before clipping occurs. Therefore, the notion of headroom is similar in both analog and digital domains. Simply put, we must make sure that our recording devices have allowed an adequate amount of analog headroom, which should facilitate obtaining an audio file with a sufficient amount of digital headroom to prevent clipping.

The general rule of thumb for setting recording levels is to set the preamplifier gain at a level that will allow one to capture a relatively level, with ample headroom and without causing overload and/or clipping. This heuristic works particularly well in close-talking configurations (e.g., with a headset microphone) and with high-quality microphone pre-amplifiers.

I must also mention that you may come across a philosophy for setting recording levels that is common in the broadcast and motion picture industries where the voice (or talent) for narration, voiceover, etc. should not be set too high at the recording stage in order to provide the sound engineer with adequate headroom to change levels (e.g., using a compressor) in post production. This is particularly true in 24-bit recording systems, which have more than 120 dB of dynamic range available (at least in terms of sampling). You, therefore, do not need to set recording levels too high. In fact, levels -12 dBFS should still get you around 100 dB of dynamic range. This philosophy applies particularly well to recording in sound-proof booths or quiet rooms, where extreneous noise is not a problem. One can then increase volume in post production; adding gain by RMS scaling does not increase noise.

In speech field research, we do not do much post production and we rarely record in quiet environments. Therefore, in a close-talking configuration, setting signal levels at around -12 dBFS might be necessary to achieve favorable SNR and good spectral detail, unless your specific microphone/pre-amplifier combination causes too much amplifier noise at this level. In this case, lower your gain by 6 dB or so.

VU, PPM, and dBFS meters

As it turns out, metering is probably the most complex aspect of field recording. In addition to controlling headroom, the recordist is often faced with the by the fact that, quite possibly, each device in a field recording chain may use a different level metering scale (e.g, dBu, dBV, dBFS, or VU). All they do is express a three-way variation is signal strength - the signal can go up, down, or remain constant. There are probably very good historical and engineering reasons for the variety but, with a little practice, you will be able to move across the different scales effortlessly.

VU meters are commonly found on analog microphone preamplifiers, tape recorders, and mixers. A VU meter (Figure 2) is, typically, a voltmeter calibrated in volume units (VU). Zero VU corresponds to the power delivered to a 600 Ω load when the voltage is 0.775 V RMS. Note that VU meters are calibrated with pure tones, so their response to speech may be a little different. VU meters are known to have a slow response to fast transients.

Figure 2. A typical VU meter; not the dBm and dbV values corresponding to 0 VU

Peak program meters (PPM) respond to the peak (or maximum) level of the signal, rather than its root-mean-square (RMS) value. Figure 3 compares peak and RMS values of a 100 Hz sinewave (left) and the vowel /ae/ synthesized with the F0 of 100 Hz (right).

Figure 3. Comparison of equivalent PPM and RMS values

PPM meters have better transient response than VU meters and are typically made as a series of light emitting diodes (LED). PPMs used on analog devices are commonly calibrated on the dBu scale, while those used on digital recorders are based on the dBFS (dB Full Scale) scale. Figure 4 shows a comparison of VU, PPM, and dBFS metering scales that are commonly found on field recording devices. 0 VU corresponds to +4 dBu on balanced and -10dBV on unbalanced equipment. 0 dB on a PPM EBU meter indicates the signal level of 0 dBu, while 0 dBFS refers to the maximum voltage level available in a 16-bit system. Any voltage over that level results in signal clipping.

Figure 4. Comparison of VU, PPM, and dBFS scales

Meters based on the dBFS scale are ubiquitous on digital systems, including audio recording and processing software. The value of 0 dBFS represents the highest possible value in a digital audio file, while - 96 dBFS (in a 16-bit system) represents the lowest possible value, thus defining the theoretical dynamic range of a 16-bit PCM audio file. Figure 5 shows recording controls of Sony Sound Forge Pro 10.0 audio editing software using he dBFS scale.

Figure 5. Sony Sound Forge Pro 10.0 live recording levels on a dBFS scale

Tone calibration

If several devices are connected together in a recording circuit (e.g., a microphone preamplifier, a mixer, and a digital recorder (or software), see Figure 6 below), signal levels across these devices should be calibrated in order to set and manage headroom properly. Calibration is typically done with a tone of known frequency and level (e.g., 1 kHz tone at +4 dBu). Figure X shows the calibration of an audio digitization (or digital recording) system. Some preamplifiers (e.g., the Sound Devices MixPre) and tape decks (e.g., Tascam 122 MK III) have built-in tone generators, which makes the calibration process easier. If a built-in tone generator is not available, an external generator should be used (e.g., Horita PT3 , Behringer CT100, etc.). Once a +4dBu tone is applied, the VU meter should be set to 0 VU, the PPM meter should be set to +4 dBu (in balanced connections), and the digital recorder's meter should be set to -12 dBFS. If the analog input comes from a microphone (e.g., in field recording situations), a 1 kHz tone at +4dBu should be generated by the preamplifier (or an external tone generator) and the digital recorder's meter should also be set to -12 dBFS, as well. The value of -12 dBFS is based on signals with average dynamic range. If rapid level peaks (such as those caused by crying or coughing) are expected and no limiter is used on the preamplifier, more headroom may be required, and so the digital recorder's levels should be set to -18 dBFS, instead. In a properly calibrated system, only the preamplifier's gain control should be used to further control signal level and headroom.

Figure 6. Signal flow in a tone calibration process

Practical benefits of tone calibration

So far I think I have managed to complicate things quite a bit with by recommending tone calibration. We certainly do not need any unnecessary complications in the field. Therefore, if we recommend a piece of technology, we had better be able to fully justify its use! Beyond the methodological benefits of obtaining recordings with adequate levels and headroom, tone calibration has an important practical benefit. Namely, once the recording circuit has been properly calibrated, you only need one set of signal controls to worry about. In other words, a 2 dB change in signal level on the microphone pre-amplifier, will result in the same change in the recorded sample. From now on, you can simply use the signal level controls on the microphone pre-amplifier to control overall levels and that is precisely what we want.

Finally, I just want to point out that you do not necessarily need to use a high-end signal generator, as they tend to be rather expensive and nearly impossible to use in the field. Instead, you just need a small, battery-powered unit, such as Behringer CT-100 cable tester. In addition to testing cable (a useful utility in its own right) the unit will generate a 440 Hz or 1 kHz test signal (pure tone) or at +4 dBu, -10 dBV, or -50 dBV. I have used the Behringer CT-100 tester for a number of years and I eventually gave it to a film maker friend of mine, who absolutely swears by it.