Currency Wars

Alongside weather data, financial markets were one of the first data sources to spark my imagination when it came to sonification, mainly because of their availability and constantly flowing nature. I was not the only one, and I’d estimate that sonifications of financial data are among the most common. Fertile but crowded ground. As I argued in my PhD, I sought to satisfy my definition of generative work: that each iteration is unique. To support that, I only worked with live data, providing a continual stream to drive the generative scheme. I was also not interested in the scientific auditory display of data (a rising or falling tone, say, which would allow a skilled listener to discern changes in the market) but rather an artistic response that somehow represented the data.

I pursued this with foreign exchange currencies by riffing on the fact that currencies are paired. The US dollar and pound sterling (USD/GBP), for example, have a relationship in which a change in their exchange rate can be described in opposite directions depending on which currency is used as the reference. This is linked to national anthems, and the basic unit of Currency Wars is a small snippet of one of these anthems — Ode to Joy in the case of the euro — triggered by a change in the relative value of a currency. Across eight major currencies — US Dollar, Euro, Japanese Yen, British Pound, Australian Dollar, Canadian Dollar, Swiss Franc and New Zealand Dollar — 28 unique currency pair relationships are constructed, each with the potential to trigger an anthem. The use of anthems as source material is itself a riff on Stockhausen’s Hymnen.

Sonically there are two channels. The first is a short ‘excitation’ burst of the anthem (less than 150ms), which feeds a long reverb to create a sort of spectral ghost or fingerprint of the anthem. In line with how the currency has moved, a window opens on this spectral sound. At any one time many of these events are happening at once, producing a continually changing spectral texture of all the anthems combined. A more ‘descriptive’ and longer snippet, between one and three seconds, allows fleetingly recognisable extracts of the anthems to be heard: “Australians all let us rejoice...” ... “O Canada!...” ... an orchestral flourish from God Save the King or The Star-Spangled Banner. Four versions of each anthem are available as a pool for the snippets to draw from.

One of the challenges in bringing this version to the web was the original work’s spatialisation, which tempered the deliberately chaotic soundworld by assigning each currency to its own speaker. Snippets of the anthems piled up into a texture, but their distribution across eight channels gave them room to breathe. Reduced to stereo, even with panning, this original scheme became quite busy. As with many other works in this webifying project, I took the opportunity to refine the piece for web presentation without losing the original idea. It also gave me the chance to fix a historical mistake that had hampered the work.

When I presented the original work live, a visiting researcher commented that it was a shame it was so metronomic. In retrospect, this observation pointed directly to a flaw in the way I had implemented the data feed. In testing, I had assumed that the irregular movements of the market would create a sort of texture by themselves. But when exposed to live data, I had inadvertently introduced a rhythm that I didn’t want because the data was polled every 10 seconds. The work became a landscape of regular bursts of activity followed by silence.

In this version, the data, from Finnhub’s WebSocket connection, comes thick and fast, with irregular timing based on whether the market data has been updated. Seemingly this solved the problem on its own, but on extended listening I realised two things. The texture was becoming far too dense: a stream of USD trades could fire off dozens of samples at once. I introduced a queuing system in which incoming trades are delayed and dispatched, creating some breathing room in the audio. I also introduced a rule that ‘dry’ extracts from the same anthem cannot overlap. This sets up some interesting questions about being ‘true to the data’ and manipulating it to achieve an artistic result. The queuing and sampling approach preserves a direct relationship between sonic events and changes in the data, even if that data has to be handled in such a way as to produce a successful artwork.

The original version of Currency Wars is documented in my PhD thesis.

Sheffield, September 2026