This ain’t a scene it’s a God damn Arms Race!
No actually I’m not a Fall Out Boy fan, but it is a good chorus, and it might illustrate some of the things I will be banging on about to talk about arms race competition. There’ll even be models an’all later, but here’s some of the basics in a couple of little cases and a long explanation of my favourite engineering tool.
(I probably didn’t mention, I have a bit of a background in sitting at the console in a recording studio so if there are technicals that a) don’t seem right to fellow recordists then I’m probably oversimplifying for non recordists and b) if there are technicals that don’t make sense to non-recordists then I’m probably trying too hard to get across their purpose etc. Either way, I’m probably annoying both of you, and I’m very sorry)
You’re there, watching television, comfortably settled in front of the sofa, when a programme reaches it’s commercial break. Suddenly, the adverts begin, seemingly twice as loud as the programme. You fumble for the controller, perhaps wishing you’d just downloaded the show from the internet, where some kind soul would have pre-edited out the adverts.
The loudness of adverts is a minor annoyance in the grand scheme of things, deserving of the “first world problems” dismissal at best. However, they also form an interesting case in both the uses and abuses of technology and in the type of incentives that lead to loudness competition. The technology in question, the family of devices known as limiters or compressors, is one of the most useful tools in a sound artists kit. They are used for giving punch and presence to mixes, gluing disparate sounds together or sweetening vocals and their use is common throughout the recording or broadcasting industries. However, they can also be put to use in making the overall sound product ‘louder’. It is this use which is the main pre-occupation of the story.
Before we get to that stage it’s worth considering what a compressor or limiter actually does in principle. The device is actually an arrangement of two simpler components. The first, an envelope follower, takes the output of a signal and, as its name suggests, follows it. When the level of the signal is above a given, generally user specified, value, the thresh-hold, the envelope follower sends a signal to the second component, the levelling amplifier. This reduces the level of the parts of the signal by a specified amount, usually denoted by a dial marked ‘ratio’. The signals are then re-combined, often being passed through a second stage of amplification, generally called “make-up”. The result of this process is that sounds whose peak volume is greater than the thresh-hold are attenuated. The dynamic range of the whole signal is therefore ‘squashed’, showing a smaller difference between peak and troughs in its volume.
If it seems counter-intuitive that an amplifier that reduces the volume of loud signals would make the whole signal louder, you need to consider the difference between peak volume and average volume. The peak is, as it’s name suggests, the level of the loudest signal (‘transient’) in the whole. Reducing the gap between peak and floor and then re-amplifying the signal afterwards means that the average volume of the signal will be higher. Peak volume, in broadcasting terms, is given as an iron limit that may not be exceeded. Average volume, on the other hand, is not so simple to define, and as a consequence technical standards for loudness are far more recent. So when an advert wants to grab your attention, the avenue of shouting at a higher peak volume is closed. Instead, it gets attention by shouting every word.
The use of loudness as a competitive tool is not restricted to broadcast advertising. In recorded music, listener attention is also a desired reaction, and the cheapest way to get it is to simply make the average volume of your song higher. It will come as no surprise that the average volume of recordings has been rising consistently since the introduction of digital recording to the mass market in the 1980s. The process pretty much reaches it’s apotheosis with Metallica’s Death Magnetic album, which I find pretty much unlistenable on the grounds of its loudness. This process is generally held to have a couple of different components. The first, the technological change component is an expression of firstly the greater headroom afforded by digital recording and secondly the greater prevalence of devices tailored to using that headroom. The second is the demand component and is a little more complex in the sense that it has to account for 1) greater concentration among media outlets leading 2)greater homogeneity of outputs across media “strands” which means that 3) competition becomes less about diversity of niche and more about being the biggest thing in the centre of the market place.
This process is generally called “The Loudness War”, and the important things to remember, as I said, before I even think about giving any of the modelling stuff away, is that it is
- Dynamic. Arms races are expressed in terms of the change in something over a period of time. This becomes crucial when we finally do take the plunge to modelling.
- Bounded. This can either be in terms of carrying capacity (for ecological arms race models), diminishing marginal benefit (economics models), or technology (in fact in this case a limiting factor is that the ultimate of a continuously amplified signal is just a square wave.
- Impose External costs. The participants in an arms race competition act according to their own interests. Costs or benefits may fall on non-combatants. In this case the ultimate listeners of Metallica’s Death Magnetic album pay the ultimate external cost of the participants need to attract radio attention.
(There are few resonances you might have noted here. One is that the process of competition leading to external costs when individual participants are engaged in an arms race is very similar to the cost of increased safety in SUVs. Brilliant Paper by Michelle White (2004) here. The second is that this framework is a bit, well Marxish. Actually that’s true: Marx’s insight that capital and labour fight over profits is approximately equivalent to a Lotka Volterra framework with negative feedback. An arms race is the same type of framework with positive feedback: i.e. Bees and flowers rather than foxes and bunnies. Finally the interesting feature is the runaway use of something that is actually quite useful in another context, which is generating that nice bit of round, warm presence in your mixes.)
Arms race competition is not a bug, it’s a feature. It is absolutely central to any type of competition where one of the key dimensions is technology. (Technology is present everywhere, but it’s not always the most important part of the competitive environment). The next time you do the fumble for the remote at the ad-break, remember, it’s not particularly personal, you just got in the way of a gunfight.