Loudness, RMS, and Peak, Oh My!

Right. Loudness. Often spoken about, rarely understood…what do you need to understand about it in order to get some nice tasty levels on your audio? Let’s start with four key definitions, and later, I’ll tell you what I recommend you master to, for a few different projects.

So: the definitions. First up: volume. This is what most people think of when they’re trying to describe how loud something is. “What’s the volume?” “Turn down the volume!” “Pump up the volume! Dance! Dance!” In reality, “volume” isn’t a real thing. When we talk about volume, it’s simply the output setting on whatever device you’re using. It varies from person to person and device to device. See also: Spinal Tap.

Next definition: peak amplitude. This is the loudest part of a given sound. If you’re looking at a waveform, it’s the highest point reached on the graph. It can’t go higher than 0dB, since that’s equivalent to the full signal output of whatever software you’re using. I talked about that more in my August 30th episode, creatively titled “It’s All About the Decibels, Baby!” I was proud of that one. But I digress. Peak amplitude is a bad indication of how “loud” something is because it may have only been reached for a split second. To relate amplitude back to the perceived loudness of a sound, you really need to start looking at an average over the whole sound. That’s where RMS comes in.

RMS stands for “Root Mean Square.” It’s a mathematical way to calculate the average of a function that trends both above and below the x-axis when you graph it out. Sound waves fall into this category, and loosely speaking, to calculate the root mean square of a signal, you square the values of all the samples, average them, and take the square root of that number. The clue’s in the name, really. Theoretically, that would give you a pretty good idea of the average amplitude (and therefore the perceived loudness) of a sound. And in many cases, you’ll see “loudness” and “RMS” used interchangeably. In reality, it’s a little more complicated than that, and it’s all the fault of the International Telecommunications Union (or ITU).

In 2006, the ITU published an extensive set of guidelines recommending audio measurement algorithms for determining what they call “subjective programme loudness.” In the guidelines, they take into account that perceived loudness may change based on source and programme type, or other factors, and they attempt to correct for that. As a result, we have what’s commonly referred to as “LUFS,” or “loudness units relative to full-scale.” You might see it written as “LKFS” – that’s “loudness, k-weighted, relative to full-scale” – and you might also see the code “ITU-R BS.1770-4” – that’s the name of the ITU recommendation it’s referring to. Colloquially, all of these terms are equivalent to what we call “loudness.” Loudness is nowadays considered the standard measurement for the perceived volume of a given sound, and it’s what you should be looking at most of the time.

Now, loudness is actually somewhat regulated, at least in television. In 2010, the US passed an act of Congress that forbids the broadcast of TV commercials that are louder than the program that they accompany. This Act is called the Commercial Advertisement Loudness Mitigation (or CALM) Act. It uses a similar standard as the ITU one, but for television, from the Advanced Television Systems Committee (or ATSC). There’s currently no such regulation for radio, but most broadcasters work to a loose set of industry standards nonetheless.

But how do you measure loudness? If you’re using Adobe Audition, like I do, then it’s simple. In the waveform view, run Amplitude Statistics and it will show you all the measures. If you want to conform your audio to a desired loudness, you can do that in a single step, using Match Loudness. For other software, it could be more complicated. Not everything ships with a loudness meter. Audacity, for instance, will require a third-party extension – I used Wave Stats for about a year. But whatever you use, make sure that you’re using something that’s measuring loudness, not just amplitude.

So what should you be mastering to? The wrong loudness is incredibly distracting and annoying for listeners, who have to compensate by changing volume all the time. There are some best practices that you can start with, which I’ll explain, and why they are the way they are.

First let’s look at radio. If you’re producing for broadcast on the radio, aim for -23dB, and a peak of -0.1dB. This gets you about as good a dynamic range as you can get, and stays consistent with the almost universal TV target of -23. Often, radio and streaming services will use automated compression and normalization to bring up the loudness at point of broadcast, but then at least you can be sure that any loss of quality is the broadcaster’s issue, not yours.

Next up: Audiobooks. If you work through ACX, the audiobook creation exchange, their regulations call for levels between -23 and -18dB, and a peak of -3. I usually go for -20dB loudness and -3dB peak, right in the middle of the range. Again, that’s so you get a nice dynamic range and the -3dB peak is relatively arbitrary; most people can’t hear the difference between -3 and -0.1.

Finally, for podcasts, I’d recommend -16dB loudness and -3dB peak. While you’d think it would be similar to radio, the current “best practice” for podcasts is actually somewhat louder, possibly because of the trend of listening on mobile devices.

All three of these recommendations are very much NOT set in stone, and it’s important to keep an eye on what the industry (or industries) are doing. The goal here is the best possible listener experience, and you’ll get that by having all programming being at similar loudness.

It makes a huge difference, just that little adjustment, and it’s one of the tricks that I use, to make sure my audio always sounds great. So I hope it helps you, too.

Further reading: