Why You Should Know About True Peak and Sample Peak Metering
Today we’re going to talk about a common problem in the audio mixing world: the need to normalize the peaks in your signal so they don’t go outside the correct operating system range. Has it happened to you? Of course. Typically, the active peak indicator will warn you when you need to lower the level if you want to introduce severe distortion within the peaks, hence its importance. The digital peak meter must show the limit of the digital audio signal, thus converting the digital signal to analog in order to be able to reproduce it in the speakers is often associated with certain problems that you should be aware of and minimize during the mastering process. And, if you spent some time in the mixing world, maybe you know how tough it is sometimes.
What is most important to understand is that the challenge for the peak meter has always been to provide us with a reliable reference so that our signal peaks do not exceed the right operating range. When we see an active peak indicator, we normally know that we need to lower the level if we don’t want to seriously distort the peaks. Now, in analog systems, there is usually a natural saturation that smoothes the peaks in a certain way, and can even take you to dangerous territory and ruin your work. However, in digital format, exceeding the full-scale level of 0dBFS implies some terrible clipping. Nothing like flattened beaks, only their straight scissor cut, which is too obvious to the ear in the form of a bright and annoying click. Have you heard it before?
Therefore, it is not surprising that while it was only a minor benchmark in analog, the peak meter is becoming pivotal in the digital environment. The desire to use all available headroom and achieve maximum sound levels makes the possibility of normalization appear in digital systems. Bringing the peak present in the record to full scale allows you to get a little more level and margin in the final product.
Since it is mixed many times at twenty-four bits and the final formats will be sixteen, normalization allows full use of the ninety-six decibels of the theoretical headroom allowed by sixteen bits. However, if the normalization is done with bit-computing thinking rather than using signal-driven sound thinking, some problems can arise. For example, boosting the signal so that the highest sample corresponds to the full scale of the digital representation, causing the resulting analog signal to demand higher levels in the converter.
Historically, this has not always been the case. There was a day when RMS and peak meters were all we needed to watch for the specificity of our audio signals, and it didn’t seem like we were living badly. But background recording methods are relatively recent and even more sophisticated. And it is with their maturity that the concern for quality assurance and interest in measuring systems and standards comes. New techniques and new qualities in sound recording and reproduction entail new metrics such as the one we deal with today.
For decades in the history of sound recording (before digitization became a form of recording,) the loudness meter and peak meter were our companions, with whom we lived to see how our signals were on par. The level meter offers a reference for the average rate, usually through a Root Mean Square strategy, which reports the average level. The peak meter gives us much more real-time information related to the instant behavior of the signal and, in this way, is able to warn us about the presence of potentially dangerous short peaks, which, when measured by RMS, simply disappear under the action of inertia and averaging, typical for measurements in the medium term.
As important as the digital peak is, it is striking that only recently there has been a general interest in characterizing “true peak”. What is that? Well, its name is due to the fact that we used something that was not a real peak.
Digitizing sound in an analog-to-digital converter means using a representation in which we store only instantaneous values at certain moments — sampling moments. The development of the signal during the time between two readings is not recorded, and it may well be that the true peak falls in the middle between two readings.
So, when you use normalization for digital recording, don’t force it at all. Any audio editor allows normalization, but we take unnecessary risks when we ask you to set the peak at full scale, leaving the new peak at 0dBFS so that the highest sample is at the end of its path. The simplest recommendation is to stay slightly lower when normalizing, even if it’s as little as one decibel. And now we understand one of the reasons. When the normalization software prompts you to set the new peak value to 0dBFS, tempting as it sounds, you should always include a small gap.
Nevertheless, it is more complicated than it sounds, and it is always best to trust professionals like Enhanced Media. If you need to master your project and don’t know how to keep it within the correct peaks, we can do that for you.