Troubleshooting power steering pump noise and vibration issues Part 2: Source Decomposition by Jerry Nessler

When making acoustic or vibration measurements on or near the power steering pump, knowledge of the other sources of noise and vibration in the area of the measurement is required. The most dominant source will be the engine, but there are other accessory drives to consider. Since the engine speed is not the same as the power steering pump speed, the harmonics will be at different frequencies. The easiest way to distinguish between the engine harmonics and power steering pump harmonics is to have pulse tachometers on the engine and power steering pump.

The frequency spectra of the two tachometers then define the harmonics of the engine and of the power steering pump. The acoustic or vibration spectrum can be overlaid on the spectra of the tachometers and the contributions from each can be easily distinguishable as shown in the figure.



One potential problem with the frequency domain approach is if the speed is varying. If the engine is running rough or the speed cannot be maintained, such as when testing on the road (as opposed to the chassis dynamometer), then the frequency spectrum of any data (noise or vibration) which vary with speed will show smearing. Smearing occurs when a single frequency may change, for example by 10 percent, during the acquisition. The resulting narrow band frequency spectrum will show wide peaks smeared over several frequency bins. The smearing becomes much worse for higher harmonics to the extent that the data is not very useful. In order to avoid frequency smearing, the time domain data can be resampled to the revolution domain.

Data acquired in the revolution domain are sampled with a fixed number of samples per each revolution of the shaft, as compared to data acquired in time domain, which are sampled with a fixed number of samples per second. The speed varying data can be re-sampled using the pulse tachometer channel to the revolution domain, and this eliminates any smearing due to speed variation because the number of samples is independent of the speed of the shaft. The re-sampled data are then averaged in the revolution domain and the resulting spectrum is called an order spectrum because on the abscissa are orders (i.e. fundamental and harmonics) of the shaft rotation.

When looking at an order spectrum, the contributions of the component being tracked with the pulse tachometer will be at the integer orders. For example, the noise/vibration order spectrum of a 10-vane power steering pump will show peaks at the 10th, 20th, 30th, .. orders. The analyst doesn't have to know the speed of the engine to find the pump orders in the spectrum. All components not related to the power steering pump rotation will show up as non integer orders. For example, if the power steering pump is running at 1.1 times the engine speed, the engine will show up in the power steering pumps order spectrum at the 0.9 order. So, the re-sampled order domain data will eliminate the speed variation and identify the dominant contributions from the component being tracked.

Using this technique, components due to different rotating sources may be extracted from the original recording and listened to. As an example, listen to the following sounds:

- total noise recorded at driver position
- engine orders
- power steering pump orders