Second Distribution Transformer Analysis

We have analyzed 26 distribution transformers in DC, Maryland, Virginia, Iowa, Helsinki, and New Hampshire. This analysis builds on the foundation set in the initial distribution transformer analysis; of most relevance is the section "What are we looking for?", and the separate overview that discusses the physical phenomena behind a transformer's acoustics.

The observations below of distribution transformers show that.

Selecting the Right Fourier

Fourier transforms combine many different parameters, and selecting them for the situation is a form of art. Fourier transforms are primarily categorized by the frame size (how many samples we're looking at). While padding zeros in the data allows for greater precision when determining the underlying frequencies (since it increases the frame size), it does not offer greater accuracy — the same result could be determined through interpolation.

Frame Size (Fourier Frequency)

The Fourier transform is subject to a basic tradeoff that is similar to the Heisenberg uncertainty principle (in German, "the unsharpness relation") which says that you can know when something happened or what something was, but not both. In this context, as we increase the accuracy of the frequency, we decrease the accuracy of the time.

Knowing this, we can look at a pothead transformer hoisted 20 feet in the air.