HarmonicTrak Distortion Analysis – What Are the Available Weighting Window Types and Which One Should I Use?

SoundCheck’s HarmonicTrak analysis algorithm performs step FFT analysis.  At each sine step of the response (as defined by the stimulus) a weighting window is applied and an FFT is performed. Spectral leakage occurs when the acquired data does not exactly correspond to one of the spectrum frequency lines. This leakage leads to amplitude accuracy errors as well as obscuring adjacent frequency peaks. For these reasons it is important to apply a Weighting function to obtain more useful information from a measurement. Windowing types affect the results of the measurement in different ways.

The available Window types are shown in the table below and can be selected on the Time tab of the HarmonicTrak analysis editor (Figure 1).  Not all window types are appropriate for distortion analysis but for those that are, each has a minimum requirement for the number of cycles per step in the response and when this requirement is satisfied, a minimum 2^nd harmonic level and lowest measurable distortion threshold can be achieved.

 

Window Type	Minimum Cycles per Step	Minimum H2	Lowest Measurable Distortion	Comment
None	5	n/a	n/a	A rectangular window (no weighting) is applied to the measurement. Not recommended for distortion analysis.
Hanning	5	-40 dB	1%	General purpose, high-speed but side lobe attenuation is low compared to other windows.
4 Term Blackman-Harris	10	-90 dB	0.003%	More precise but require more cycles per step. Good for most electroacoustic measurements.
7 Term Blackman-Harris	15	-120 dB	0.0001%	Most precise due to greatest attenuation of side lobes. Best window to use when measuring very low distortion devices, such as electronic circuits and electronic devices.
Flat Top	10	n/a	n/a	Excellent amplitude accuracy. Not recommended for distortion analysis.

Table 1 – Window Types and Settings for HarmonicTrak Distortion Analysis

 

Weighting Window Types – A Brief Description

None (Uniform) - Also referred to as Rectangular. No weighting is applied to the measurement. This works well with transients that are shorter in length than the measurement time. Due to the flat characteristic in the time domain, all parts of the signal are equally weighted.

Hanning - This is a smooth window function, which is one period of a cosine² function and tapers to zero at the beginning and end of the measurement. Hanning is recommended for the analysis of noisy signals. Its main advantage is that it has excellent frequency selectivity.

Blackman-Harris - This window has a low ripple (<0.87dB) in the pass-band and a low skirt (<-80 dB) in the stop-band.  Blackman-Harris is recommended for harmonic and order analysis. Its main advantage is that it has an excellent dynamic range combined with good frequency selectivity.

Flat-top - This window has very little ripple (<0.01dB) in the pass-band (in the frequency domain). The window’s main use is for level measurement of sinusoid (calibration), due to its negligible amplitude errors. Its main advantage is that it has excellent amplitude accuracy.

Figure 1 – Weighting window types are selected using the Weighting dropdown list found on the HarmonicTrak editor’s Time tab.

 

Figure 2 – Weighting Windows

 

 

Signal Type	Window Type
Sine wave or combination of sine waves	Hanning, Blackman-Harris
Sine wave (amplitude accuracy is important)	Flat Top
Narrow-band random signal (vibration detail)	Hanning
Broad-band random (white noise)	Hanning
Closely spaced sine waves	Blackman-Harris
Signals with harmonics	Blackman-Harris
Unknown content	Blackman-Harris

Table 2 – Signal Types and Recommended Weighting Windows