Zitat:How does it work?
______________

As distinct from an FFT approach, this audio analyser uses a complex matched filter
for each harmonic of interest. This enables the magnitude, and if required, the phase
of each to be calculated. The THD, or total harmonic distortion, is the incoherent
sum of the harmonic components, expressed as a percentage of the fundamental.

Some other (hardware) THD measurements actually calculate this on the basis of:

...................Distortion
THD = ------------------------------------- x 100%
.............Fundamental + Distortion

This will give a somewhat lower figure at high distortion levels, as the figure
cannot exceed 100%, even when the fundamental is completely lost.

Initially, just the first 9 harmonics have been considered, mainly because this is also the
default for the SPICE .four analysis (although this can be changed explicitly) against
which it has been compared.

The disadvantage of this technique over the spectrum analysis approach is that only
the harmonics are measured; non-harmonic products and noise are ignored. So this
technique is only good for basically linear high level circuits. It would not be a good
solution when working with class D amplifiers or SMPSs.

There are several advantages over the .four method for calculating THD.
1. The level and frequency can be swept in one nested .tran analysis. The frequency
specified in the .four statement cannot be a variable, making a swept frequency
analyses impossible.
2. The measurement floor appears to be significantly lower than the .four method, at
least for small data sets, making the .tran analysis faster.
3. In LTspice, at least, the distortion figures from the .four routine in the errorlog cannot
be plotted in the waveform viewer. It requires that the errorlog be imported into a
separate application like Excel, and then manually generating the chart, although
the process could be accelerated by macros.

The accuracy of the method has been validated against .four with two circuits:
1. A calibrated non-linear behavioural amplifier, containing a sixth order
term in the transfer function, and
2. A (slightly) modified version of the simple class B audio amplifier that is in the
Educational directory of the LTspice standard installation.

The results, below, show excellent agreement. Both circuits are in the download.

Calibration Non-linear Test circuit at 1kHz
I/P THD, %
Vrms Analyser .four
0.1 0.0009 0.0025
0.15 0.0022 0.0032
0.22 0.0126 0.0127
0.32 0.084 0.083
0.46 0.572 0.565
0.68 3.89 3.85
1 26.52 26.27

Modified Class B amplifier shipped with Ltspice distribution at 1kHz
I/P THD, %
Vrms Analyser .four
0.010 3.32 3.27
0.012 5.51 5.49
0.015 4.87 4.87
0.018 4.02 4.02
0.022 3.22 3.23
0.026 2.64 2.65
0.032 2.16 2.17
0.038 1.76 1.77
0.046 1.43 1.44
0.056 1.17 1.18
0.068 0.97 0.97
0.083 0.8 0.81
0.100 0.67 0.67
0.121 0.56 0.56
0.147 0.46 0.46
0.178 0.38 0.38
0.215 0.31 0.32
0.261 0.26 0.26
0.316 0.21 0.21
0.383 0.17 0.17
0.464 0.21 0.2
0.562 1.07 1.03
0.681 7.79 7.73
0.825 14.55 14.51
1.000 20.2 20.17

Audio Distortion Analyser, Tony Casey, June 2010, tony@ritecom.com
____________________________________________________________________