Audio levels (and their cousin loudness) has been an ongoing area of confusion and annoyances from the 1930s on. Adding digital to the mix has done little to simplify the situation.
Here we discuss some of the background and the relationship between the VU Meter, the Peak Programme Meter, and digital meters reference to 0 dBFS (Full Scale).
Rane’s Web site succinctly summarizes the history and specifications of the VU Meter:
VU meter (volume unit) The term volume unit was adopted to refer to a special meter whose response closely related to the perceived loudness of the audio signal. It is a voltmeter with standardized dB calibration for measuring audio signal levels, and with attack and overshoot (needle ballistics) optimized for broadcast and sound recording. Jointly developed by Bell Labs, CBS and NBC, and put into use in May, 1939, VU meter characteristics are defined by ANSI specification “Volume Measurements of Electrical Speech and Program waves, ” C16.5-1942 (which is know incorporated into IEC 60268-17). 0 VU is defined to be a level of +4 dBu for an applied sine wave. The VU meter has relatively slow response. It is driven from a full-wave averaging circuit defined to reach 99% full-scale deflection in 300 ms and overshoot not less than 1% and not more than 1.5%. Since a VU meter is optimized for perceived loudness it is not a good indicator of peak performance.
In Europe, the Peak Programme Meter was widely used. Once again the Rane Web site provides an excellent definition:
PPM (peak program meter) An audio meter originally developed in Europe to accurately measure and display peak audio signals (as opposed to average audio signals; see VU meter). The PPM augments the VU meter and it is normal to find both in modern recording studios. The PPM is particularly valuable for digital audio recording or signal processing due to the critical monitoring required to prevent exceeding 0 dBFS and reducing overs. There are two standards: IEC 60268-10 for analog meters and IEC 60268-18 for digital meters. [These are available to buy on the IEC website.] An interesting aspect of PPM design is that rather than respond instantaneously to peaks, they require a finite 5 ms integration time, so that only peaks wide enough to be audible are displayed. IEC 60268-10 translates this into a response that is 1 dB down from steady-state for a 10 ms tone burst, 2 dB down for a 5 ms burst, and 4 dB down for a 3 ms tone burst — requirements satisfied by an attack time constant of 1.7 ms. The IEC specified decay rate of 1.5 seconds to a -20 dB level can be met with a 650 ms time constant.
It is crucial to remember that with the VU meter, one sets the lineup tone and mixes to the same point, 0 VU which is usually +4 dBu or +4 dBm.
Similarly, with a PPM, one sets the lineup tone to mid scale (typically -8) but mixes to 0 on the scale. 0 on the PPM scale corresponds to +12 dBu or +12 dBm, 8 dB above 0 VU on the VU meter. This is the 8 dB peak to average ratio demonstrated by the VU meter and Peak Programme Meters.
Note, that the Peak Programme Meter still has a 10 ms integration time so there are still peaks which are let past.
In analog recording, tape compression would start about 8-12 dB above the lineup tone (which would be set to the operating level in nWb/m). So, there was a soft limiter built into analog recordings.
With the onset of digital, it was easy and necessary to provide metering that indicated decibels below full scale. Once you run out of numbers, it just can’t get any louder and “overs” are generated. Overs are sequential full-scale values for the digital audio value. Since there is no soft clipping, standards have been developed to capture almost all of the peaks that are present in a wide variety of program material.
In the video world, lineup tone is set to –18 dBFS (decibels below digital full scale) in Europe and –20 dBFS in North America for television use.
CDs tend to be normalized so that the highest peaks are set to 0 dBFS, so there is no connection between analog lineup tone and 0 dBFS on the final recording.
For transferring material, one should probably use the highest level that does not provide any overs. To be on the safe side, if there is a lineup tone, setting that to –16 dBFS to –20 dBFS is defensible. I wouldn’t set lineup tone much above –16 dBFS.
One trick that I use is to set the record levels conservatively (approximately –18 dBFS) and ingest at 24 bits. This way, we still have “active” bits (even if they are just noise) when we normalize peaks to –0.3 dBFS.
Articles of interest: