Catch a Wave!
The Beach Buoys Redux

Bringing The Beach Buoys to Web Audio seemed straightforward enough. I’d already made enough progress on the other works to have the foundations in place: retrieving live data, navigating the problems of getting that data into a webpage through proxying and cross-origin restrictions, and reproducing the Pure Data sound tools I had used in my PhD portfolio as Web Audio API scripts. The basic architecture was the same as the original — retrieve wave period and spectral data and apply these to a sample of The Beach Boys singing. The wave period would determine the duration of the time-stretched loop, while the spectral data would shape the sound through filtering.

The audio engine was transformed. In keeping with the other works, the processor-intensive phase-vocoder time-stretching of the original was replaced with a more lightweight granulator, and I started to run tests. Then something strange happened: the piece would intermittently produce errors at the data retrieval stage. It turned out that the CDIP data, the foundation of the original piece, had been firewalled and severely rate-limited. Disaster.

This was not the first time a piece had been threatened by the changing nature of the web. My piece based on gold prices, for example, went through several different data sources during the writing of my PhD alone. But CDIP data was integral to The Beach Buoys in two particular ways. Firstly, geographically: the original piece was specifically concerned with wave data from the California coast. Secondly, and perhaps more importantly, CDIP provided a standardised form in which the spectral data was reported.

This was the first piece in the webification of my PhD portfolio where I would have to go back to the drawing board. Salvation came in the form of the National Data Buoy Center (NDBC), part of the U.S. National Weather Service. Its site provided access to the CDIP buoy data, re-broadcast in a slightly different format, as well as observations from other buoy stations around the world. Sticking with the geographical theme of the original, and the musical pun on The Beach Boys, I decided to use only buoys moored off the California coast. The thought did cross my mind that a ‘global’ version of the piece would be possible and I may look into that one day.

Turning this setback into an opportunity to rethink the piece led to two substantial changes — substantial enough to justify giving it a new name, Catch a Wave!, with a subtitle pointing to its predecessor.

The first change came from listening to the original. As part of the process of reconstructing it, I realised that I didn’t really like the sample I had taken from The Beach Boys. It is a heavy, unstable, jazzy harmony that I had originally thought had a floating, unresolved character. On repeated listening, though, it just struck me as rather dull and stale.

I went back to my original notes and to the first iteration of the piece, which had used an excerpt from Surfer Girl. The useful thing about this snippet was that it had a much broader range, with Brian Wilson’s soaring falsetto in particular. My original notes said that this experiment had not worked because the slight pitch bend in the sample sounded terrible when time-stretched. I solved this in the new version by taking a much shorter sample of the harmony, pre-stretching it in PaulStretch before submitting it to the granulation process that stretches it according to the wave period. Perhaps more luck than judgement, but the resulting sound had a sort of sparkle that I much preferred to the original.

There was initially a pang of guilt that I had somehow vandalised my own work, followed by the realisation that this is my work, and if I want to improve it, I can. The point of this PhD resurrection project is partly to faithfully restore the original works, but as long as the underlying idea remains — a Beach Boys sample transformed by wave data — I have no qualms about issuing a 2026 version and benefiting from hindsight.

The second change was to rethink how the filtering of the sample worked. In the original version, the uniform nature of the CDIP data — where the FFT information on wave energy was presented in 64 bins — was linked directly to a 64-bin FFT filter. Again, listening to the original version, I was struck by how ‘squashed’ and notched the sound had become. This was simply the by-product of such a fine-grained process, but it also seemed like an example of overfitting a potential musical model to the data based on an obvious mapping, rather than whether it sounded good: something I might well have criticised in the theoretical part of my PhD.

As with the original sample, I realised that this was an opportunity to improve the piece. This also avoided taking the FFT filtering approach into the browser, where it would have been considerably more computationally expensive.

I settled instead on a model in which a gentler set of bandpass filters — up to three, depending on the distribution of the data — would be applied. Their peaks coincide with the peaks identified in the spectral wave data, while their Q factor gently controls how narrowly those frequencies are emphasised rather than aggressively notching them out as the original architecture did.

The theoretical debates from my thesis re-emerged here, particularly the question of how closely a data-driven artwork should attempt to remain ‘true’ to its source data. In the end, I was reminded that these are data-driven artworks rather than scientific sonifications: the data provides the material and the structure, but my ear still has to guide the result.

There were some other, smaller changes. The new version has six layers rather than the original eight, and the live buoy observations are refreshed every thirty minutes rather than forming a fixed dataset. The result is therefore not quite a reconstruction of The Beach Buoys, but a new version of the same idea: a piece in which the changing behaviour of the ocean becomes a means of transforming a small fragment of The Beach Boys.

The original version of The Beach Buoys is documented in my PhD thesis.