The Relationship Between Chatter and Lobing

When the profile error of your cylindrical part exhibits a repeating, wavy pattern, it is said to have chatter or lobing. The two terms overlap and are often used interchangeably, although neither is clearly defined by standards or practices. In general, however, lower frequency undulations are called lobing, while higher frequency errors are called chatter. There are two problems with using this type of classification. First, neither indicates the calculation used to derive the result from the measured data. Second, neither indicates the dividing line in frequency where lobing becomes chatter. Therefore, a specification that defines “lobing” or “chatter” without methods and frequencies is inadequate. Let’s look at the methods used to calculate these parameters on your Adcole meter.

Chatter is calculated by subjecting the data set to a Fast Fourier Transform (FFT) and giving the results as an amplitude versus the number of times per revolution (called UPR or Undulations per Revolution). However, lobing dates to before the availability of modern computers, when it was not practical to perform Fourier transforms in full (if anyone even thought of doing so). Lobing was defined as peak-to-peak roundness error in an angular sector of the radial measurement set (the data recorded from the part profile while the part is rotated in the gauge). It is essentially a pie chart of the roundness data. In practice, a 5-degree “window” with lobes is the narrowest selected for lobe measurement. The widest is usually 45 degrees. If we assume a sinusoidal pattern around the part, with no error other than that of the waveform, a full wavelength occurring within a 45-degree window would represent 8 ripples. For 5 degrees, we would see 72 complete ripples.

Based on these method definitions, FFT can theoretically be used to detect ripple patterns at any frequency, including low frequencies, and the method for calculating lobing can detect the peak-to-peak values of high-frequency “chatter” (assuming that the lobe measurement includes a full wavelength of the chatter pattern and the pattern is adjacent around the log). So, it seems that the real question of what chatter is and what lobing is comes down to the demarcation line in frequency of the event, at least by generally accepted concepts. There are other proprietary, customer-specific methods, but these are the two general methods used.

In the real world, however, errors in profile or rounding are very rarely (if ever) a perfect sine wave. So, every measurement, whether it’s lobing or chatter, has a special use and advantage. But since the purpose of each overlap is the other, the two terms are often misused, interchangeably. A more serious problem arises when one replaces another problem. Instead of looking at lobing vs. chatter as a comparison of relative frequency, let’s look at this problem from a different angle: that of function. In other words, what are we trying to achieve with the lobing or chatter measurement?

As discussed earlier, the lobing measurement on your Adcole meter shows the maximum peak-to-peak amplitude within a limited sector of roundness data. It does not matter whether the difference is part of a recurring wave or a single-instance event. The Adcole Chatter measurement, on the other hand, is used to find the specific UPR and amplitude results for a recurring signal. The UPR component can then be used to identify the cause and can even identify the specific machine that caused it. Each measurement has advantages and disadvantages. FFT Chatter analysis attempts to identify cyclic patterns in the measurement data and can separate the different overlapping signals into their UPR and amplitude components. However, it is limited in its ability to determine the amplitude of non-contiguous patterns and is not the right tool for locating a single-instance event, such as a scratch or a flat spot.

Lobe measurement, on the other hand, can report the maximum peak and trough within the measurement window and is therefore better suited to finding something like a flat spot, a scratch, or a small non-clearing sector. With lobes, however, there is no practical way to separate the different frequencies without calculating for 360 or more windows. Even then, you would assume that the full peak-to-peak amplitude was the result of a specific infringing frequency. In reality, there are likely to be a number of frequency components that overlap and that can add or subtract from the total amplitude. Lobing is best used to detect rapid changes in data over a limited angular range by applying a tolerance to the window that is tighter than the window applied to the entire data set. For example, your specification may allow a roundness error of 6 microns but limit that error to no more than 2 microns in a 30-degree window.

The lobing specification is effective for finding isolated deviations, such as a step or scratch in the journal. Chatter is more useful for finding repetitive patterns that indicate vibration signals in the grinding equipment, or a situation where the rough operating pattern is still present (no runout). Both types of defects have functional implications for the crankshaft, and the right combination of chatter and lobing measurement can help determine and diagnose the cause. However, the two specifications are complementary, and one cannot completely replace the other.

By definition, Lobing looks at roundness data over a limited angle window; our standard FFT Chatter Analysis analyzes the full 360 degrees of data. However, if there is a non-contiguous chatter pattern, it may be useful to focus the analysis on a smaller portion of the dataset. A signal that occurs over a small angular range may appear to have a lower amplitude when the analysis is performed over 100% of the data. By limiting the analysis to a specific angular range, you can obtain a more accurate representation of the chatter amplitude. Adcole’s software packages for FFT chatter analysis include options to look at limited angle ranges for both journals and cam lobes. For a good analysis, make sure to use a sector that is wide enough. A sector that is too small can result in distorted results and/or a poor UPR determination.