**5.3.2 Leakage and noise**

When a frequency signal (for example, a sinusoidal signal corresponding to the 10th order harmonics) is tested in the single tone experiment, other components (from the DC to the 60th order) aside from the input signal itself appear in the DFT computational results. These components should be zero but are non-zero in practice possibly because the output of the source is not a pure single tone but with side bands; this result may also be attributed to both leakage due to faulty compensation and noise.

To overcome the side bands of the source, 3 sets of data (7th, 8th, 9th) ahead of the input signal (10th) and 3 sets of data behind (11th, 12th, 13th) that same signal in the DFT computational results are excluded in the analysis below.

The rest of the data reflect mainly leakage and noise. An "enlargement test" is designed and implemented to distinguish between both parts. The effect of the non-integer-period is enlarged 10 times (Δ is enlarged from 0.04 to 0.4). The difference between the two groups of data is regarded as the influence from leakage only, and the remainder is regarded as noise.

The error of the *k*-th harmonic includes all leakages from the DC to the 60th order aside from itself. Therefore, all the contributions from the other harmonics are accumulated into the uncertainty of the *k*-th harmonic.

The contribution of the leakage from the experimental results is 2.5 μV/V, and the noise dependent on frequency is 6.0 μV/(*k* = 1).
