A Review of Enveloping and Demodulation Techniques

Introduction
Enveloping/Demodulation is an advanced tool for vibration analysis; and it is used extensively for fault detection in bearings and gearboxes. The technique focuses on the high-frequency zone of the spectrum. Using a high-pass filter (allowing high frequencies, but blocking lower ones), the analyzer focuses on the low-level high-frequency data content. The analyzer essentially tries to pick up some peaks that would otherwise be lost in the noise floor (or at times called the carpet, consisting of extremely small amplitudes across the whole frequency range of the spectrum) of a narrow band spectrum. To understand the concept of enveloping and demodulation, let us consider the example of a defective bearing which has a single spalling (where a piece of metal chipped off) on the track of the outer race. Every time a ball passes through this spall, it would generate an impact; and let us just 'call it a click for simplicity. See figure below.

High frequency vibrations

Outer race defect

Acceleration time waveform

As a specific example, let us say in this bearing 8.6 balls pass over this spall defect during every revolution. Then 8.6 clicks per revolution are generated. If the shaft were rotating at 1800 rpm we would see a peak at 15,480 cpm on the FFT display. This peak may be visible in a regular FFT, but generally the low-frequency area in a spectrum is crowded, and it may be difficult to notice the small peak and relate it to the bearing or gearbox defect. For this reason, the enveloping technique is used. Returning to the ball bearing travelling over the spall, we should note that along with 15,480 cpm low-frequency peak, there is something else that is happening. Every click of the ball bearing passing over the defect is analogous to someone striking a bell. When someone rings the bell, we hear two kinds of sounds. One is of the striking frequency and the other is the ringing sound of the bell. The ringing sound is the resonance of the bell, which is basically a high-frequency vibration. Now, let us carry this analogy over to the defective ball bearing. Every click of the ball passing over the defect is similar to the bell being struck. Just as the bell rings, the bearing resonates in a similar manner, generating high frequency vibrations. The ringing high frequency of the bearing is dependent on the mass and the stiffness of the bearing. Thus, the frequency of 15,480 cpm in the FFT is the same as the striking frequency, and the resonance of bearing components relates to the ringing of the bell. To understand this technique, we need to look at the high-frequency ringing vibrations of the bearing. As the balls pass over the spall defect, the bearing continues to generate the ringing frequency. At the instant of the ball impact on the defect, the vibration amplitude reaches maximum value, after which it begins to die out until the next ball impacts and the amplitude rises again.

This process continues endlessly, giving rise to a waveform that looks like the one shown below.

Amplitude Modulation This phenomenon of amplitude rising and falling with time is called amplitude modulation in wave terminology. The next step is demodulation. In this process, the ringing of the bearing is removed and the process leaves us with a trace that looks like the wave shown.

High-frequency Information Removal If this enveloped waveform was analyzed, it would appear on the FFT chart as an amplitude peak, with a frequency of 15,480 cpm, which happens to be what we called the click frequency. Thus, by looking at a small bandwidth in the high-frequency region, we can trace a low frequency wave. It is important that this lowfrequency vibration be impacting in nature to cause amplitude modulation of a high-frequency wave. This is exactly the kind of vibration generated from defective bearings and gears. Thus, the method of enveloping and demodulation is used for analyzing bearings and gears. Commercially, there are several techniques of enveloping/demodulation; and these are available from a number of vendors, including Commtest, CSI, Entek, SKF, and SPM instruments. These are used for analyzing defects in the high-frequency range. These techniques are referred to as Demod, PeakVue, Spike Energy (gSE), Spectral Emission Energy (SEE), and the Shock Pulse Method (SPM).

Let us first examine the Commtest technique, as it is implemented in the vb instrument family. Demodulation As above, we will pick the example of a bearing that is developing a crack in its outer race. Each time a ball passes over the crack it creates a high-energy burst of vibration, with each burst lasting for a very short time.

When the vibration signal is passed through the FFT analyzer to create a frequency-based spectrum, we would normally expect to see a big spike or peak occurring at the same frequency (or rate of occurrence) that each ball passes over that crack. Unfortunately we do not; instead we will commonly see what is referred to as a highfrequency 'haystack', as displayed below.

Why does the FFT analyzer display a haystack instead of distinct peaks? The reason is that the FFT process looks for repeating sinusoidal (wave-shaped) components at individual frequencies, and due to the short duration and low amplitude nature of these impact events, the FFT correctly tells us that there is not enough occurring at the BPFO frequency to stand out on its own. This can be further illustrated by picturing yourself trying to listen to someone tapping on a drinking glass filled with water while at the same time you are standing next to a loud and low-toned gong or church bell. What happens is that the longer and louder vibration event of the gong drowns out the softer and shorter sound coming from the tapping event; this is true even though both events are occurring at the same time. Our ears hear the combination of both events, but we are unable to separate and distinguish the two. Reviewing the time waveform at the top of the page, we note that repetitive noise bursts are occurring. So, how can we use the vb instrument to 'hear' them? Enter the mode Envelope Demodulation also known as High Frequency Enveloping. The Demodulation process looks for the repetitive patterns created by impact events that lie embedded within the time waveform. The vb instrument will then display a demodulated frequency spectrum that will greatly emphasise these fault frequencies. But how does it work?

The first step is to apply a high-pass filter that removes the large low-frequency component(s). The illustration below gives an idea of what the FFT analyzer then 'hears' only the tapping on the water-filled glass:

Next we trace an 'envelope' around the bursts in the waveform to identify the impact events as repetitions of the same fault.

And now this enveloped signal is passed through the FFT process to obtain a frequency spectrum. The signal now clearly presents the BPFO peaks (and harmonics) as we first hoped for.

Other similar technologies are discussed briefly below. PeakVue The PeakVue method offered by CSI can be briefly described as: A high-pass filter removes low-frequency content in the acceleration signal. It passes through the analog-to-digital converter. Each sample is scrutinized and only the amplitude level that exceeds a specified trigger level is assigned a digital value.

That is, if the analyzer is set up to take 1024 samples from the analog signal, there will be 1024 digital peak values if there are strong impacts. The digital time waveform will only have the highest positive values, one per sample. If there are no defects, there will be no high pulses, and only low amplitude noise (from the signal or instrument) may show. A pulse from an impact due to a defect bearing or gear has very short time duration and can be periodic; thus, there will only be high peak values in the cells or samples where the pulse occurred. The FFT algorithm then processes this digital information and the resulting spectrum will only show a fundamental peak and harmonics that have a frequency equal to that of the pulse.

With PeakVue, the signal is extracted from high-frequency data, but there is no rectifier or enveloping done by a low-pass filter.

gSE (Spike Energy) Spectrum The gSE spectrum is performed as follows: A high-pass filter removes the low-frequency content (long time period data) in the acceleration signal. The resulting signal then passes through a rectifier, which has an RC time decay to emphasize the impact events.

The filtered signal (step 1) passes through a peak-to-peak detector, which not only holds the peak-to-peak amplitude, but also applies a carefully selected decay time constant. The decay time constant is directly related to the spectrum maximum frequency (Fmax). The output signal from a Spike Energy peak-to-peak detector is a saw-tooth-shaped signal. Even though the rectified signal has only positive values, it is made up of many periods of highfrequency values. It is then 'enveloped' or smoothened electronically and passed through the analog-todigital converter. After the analog-to-digital converter digitizes the acceleration signal, the digital values represent the total energy in each time sample, not just the peaks. If there were no defects, there will be no outstanding pulses, perhaps only signal/instrument noise of low amplitude (energy) in each sample. To remove the high-frequency content, the digitized information is sent through a low-pass filter. The resulting waveform is the shell or the envelope of the modulation. The data is then subjected to a peak-to-peak detector that determines how often the pulse is occurring (the fundamental period), and also determines the duration of the pulse. Thus, when digital information is processed using the FFT algorithm, the resulting spectrum will only show a fundamental peak and harmonics that have the pulse frequency.

The differences here can be summarized as follows: Spike Energy uses a rectifier. In Spike Energy, the signal is enveloped. The amplitudes are not true peak acceleration values. They are gSE (Spike Energy) values. Spectral emission energy - SEE A spectral emission energy (SEE) device used by SKF takes high-frequency acoustic measurements and combines them with enveloping techniques to diagnose bearing condition. During enveloping, the vibration signal is filtered to leave only the high frequencies, which removes most of the vibrations caused by structural vibration, misalignment and other factors. The remaining defect signals are reduced in energy content, but are still occurring at the same time intervals. High-frequency vibrations are generated each time a defect bearing is rolled over. The SEE provides detection of metal-to-metal contact when the lubrication film is broken due to a bearing defect or other similar faults. This contact generates high frequencies in the region of 250-350 kHz. The SEE method monitors bearings by using high-frequency acoustic emission detection using a wideband piezoelectric accelerometer. The SEE method differs from normal spectral analysis that only go as high as 20 kHz and of enveloping techniques at 5-60 kHz. In this case, an acoustic emission transducer is used. The transducer's contact with the machine should be free of grease or any air gaps that could cause loss of signal strength. A signal of 250-350 kHz is filtered and enveloped. A low-pass filter then ensures that only low-frequency components remain. This enveloped signal is analyzed digitally using normal analysis equipment.

Shock Pulse Method - SPM The Shock Pulse Method is a technique patented by SPM Instruments. It uses vibration signals measured from rotating rolling bearings as the basis for condition monitoring of machines. This technique was invented in the late 1960s. On the impact of a roller with a race defect, the shock pulse transducer reacts with a large amplitude oscillation to the weak shock pulses, because it is excited at its natural frequency at 32 kHz. Machine vibration at a much lower frequency is filtered. The analysis comprises of the following stages: The vibration signal from the machine, with superimposed transients at the resonance frequency caused by shock pulses, is passed to an electronic filter. The filter passes a train of transients at 32 kHz. Their amplitudes depend on the energy of the shock pulses. The transients are converted into analog electronic pulses. The converted shock pulse signal from the bearing is finally converted to a rapid sequence of stronger and weaker electrical pulses.

The filtered transducer signal reflects the pressure variation in the rolling interface of the bearing. When the oil film in the bearing is thick, the shock pulse level low and without distinctive peaks. The level increases when the oil film is reduced, but there are still no distinctive peaks. Damage causes strong pulses at irregular intervals. Shock pulse meters measure the shock signal on a decibel scale, at two levels. Concluding Remarks While there are a variety of methods having different names and implementation, they all strip off the normally occurring low-frequency content in an acceleration signal and focus on the high-frequency (ultrasonic) data. Independent studies have shown that these methods all provide the desired results in detecting high frequency impacts imparted by early signs of failure in rolling element bearings and gearboxes.