Demodulation Spectrum

A useful technique for measuring and analyzing data is a process called Demodulation. The demodulation process is effective for detection of high frequency low amplitude repetitive patterns that lie embedded within the time waveform. These are characteristic of certain types of mechanical faults, particularly rolling element bearing faults such as inner or outer race cracks and spalls that make a clicking or ringing tone as the rolling elements pass over the fault. Demodulation is useful as an early warning device, as it detects bearing tones before they are visible in a normal spectrum. As the fault progresses towards failure, the frequencies will spread out and appear more as an increase in the “noise floor” of the FFT spectrum as the amplitude increases.

The process works by extracting the low amplitude, high frequency impact signals and then tracing an 'envelope' around these signals to identify them as repetitions of the same fault. The resulting spectrum, with the low frequency data removed, will now clearly show the high frequency impact signals and harmonics.

The high frequency signals that demodulation aims to extract are do not travel well through large structures, therefore extra care must be taken to ensure the accelerometer is setup correctly. Ensure that:

1.        The accelerometer is mounted close to the fault source with the shortest direct path through the structure to the accelerometer.

2.        The accelerometer is well coupled, using either stud mounting or a very strong magnet on bare metal. A handheld probe or stinger is not recommended.

3.         The accelerometer mounting is consistent between visits. If not, a trend plot of overall RMS values will be meaningless.

The demodulation process can be graphically described in the following flow chart:

Demodulation Process Flow Chart

Below is a depiction of an acceleration time waveform with a repetitive high frequency component. Because of the large difference in amplitude and frequency, a very low amplitude high frequency signal could be overlooked during routine vibration analysis.


Acceleration Time Waveform with Fault

The high-pass filter removes the low frequency component of the signal, below:


Acceleration Time Waveform after High Pass Filter

The next step in the process is enveloping which lowers the frequency of the signal to that of the repetitive element.


Signal after Enveloping

The final step is to process the resulting time waveform signal into a frequency spectrum. Since the signal has been altered by removal of low frequencies and enveloping, it is referred to as the Demodulated Spectrum.


Demodulated Spectrum

A Bearing Detection Example of Demodulation

The following examples show CoCo screens in VDC mode being used to analyze a rolling element bearing with a slight defect.

CH1 Time Waveform and FFT with slight bearing defect

The following is the same signal with the demodulation spectrum on the lower trace:

CH1 Time Waveform and Demodulation Spectrum with slight bearing defect

As the bearing deteriorates, the defect typically becomes larger and generates a wider range of frequencies as the rolling elements pass over it. The following is the demodulation spectrum with slightly deteriorated bearing:

CH1 Time Waveform and Demodulation Spectrum with slightly deteriorated bearing

As can be seen in the screen below, the standard FFT spectrum shows the relatively high first order amplitude but only shows an elevated noise floor in the higher frequencies.

CH1 Time Waveform and FFT Spectrum with deteriorated bearing