Simulated dataset that we can reconstruct perfectly

Lucy and I are trying to come up with a test dataset where we simulate two electrodes, and then reconstruct one location equidistant between them. The idea would be to have a test where we know exactly what the reconstruction should be, and we can make sure our code recovers it exactly. we tried using 2 sinusoids where of frequency f and 2f, but the problem is that they are orthogonal so the model breaks down. Any ideas for 2 timeseries that we can use to create this test?

@jeremymanning - not a rush on this, just wanted to jot down the idea