A Simple Automotive Noise Test

In a recent article we described how the Prosig P8000 hardware and DATS software had been used to help Dalmeny Racing diagnose a problem with an exhaust bracket on their Formula Ford racing car. Whilst the car was instrumented for structural tests on the exhaust the opportunity was taken to carry out a simple automotive noise test. It was felt that these would provide some useful “real world” data as well as maybe providing some extra information regarding the exhaust bracket failure. After analysing and animating the hammer data it became clear that the engine runup data wouldn’t be needed. However, it was decided that some analysis should be carried out to see if the noise and vibration data backed up the conclusions of the other tests.

This measurement consisted of an accelerometer on the exhaust and two microphones positioned roughly at the position of the drivers right and left ear, as shown in the picture at the top of the page. A run up and run down test were performed.

Whilst also recording a fourth channel containing a twice per engine rev tacho signal, the engine speed was increased to maximum and then back down to idle again. This gives a good view across the engine speed range and is useful in identifying any unusual patterns of noise or vibration that could lead to problems.

With all the measurements done and reviewed quickly on site, the Prosig engineers were happy that they had collected a good set of readings. Now it was back to base for some analysis. A mobile system, such as the DATS/P8000 combination, can be used to take readings and analyse results “on site”. However, once the data had been safely captured the car could be freed up for other test and development work.

You can listen tot he runup (acceleration) and rundown (deceleration) below.

First, the engine run-ups and run-downs were analysed to look for any unusual patterns in the frequency v. engine speed domain. The data had been captured as a single continuous signal containing both the engine acceleration and deceleration. The signal from one of the microphones is shown below in Figure 2.

Figure 2: Original raw signal
Figure 3a & 3b: Signal split into two

The DATS software was used to split this into two separate datasets, a runup and a rundown, ready for further processing. (NOTE: Actually, the DATS software will analyse the runup and rundown from the original combined data. Physically splitting it has only been done to help make the description in this article clearer).

The first issue, as is so often the case, was the tacho signal. This was taken from the ignition voltage at the coil. Unfortunately, whilst this gave a two pulse-per-rev signal, it did not provide a very clean signal.

In order to carryout a waterfall analysis it is essential to have a satisfactory tacho signal. At first glance the tacho in Figure 4 seems fine.

However, if we look at the speed v. time curve shown in Figure 5 that was generated from this tacho, it is very rough.

Not only is it rough but we have speeds in excess of 15000 rpm. The engine is not capable of such speeds! If we look in more detail at the tacho then it is clear that on occasion there is a double pulse. The red line on the graph was the tacho threshold level chosen by the software. Increasing this level a little would improve matters. But we would do better if we could ignore the secondary part of the tacho.

Fortunately, the latest tacho processing in the DATS software has additional features that will overcome the “multiple“ triggering.

One of these is a percentage “hold off” parameter. By setting this to 40% then the Software will skip over the secondary part of the tacho.

The speed curve in Figure 8 shows the much improved result.

Figure 4: Tacho SignalFigure 5: Initial speed v time curve
Figure 6: Analyzing the tacho pulsesFigure 7: DATS Tacho processing parameters
Figure 8: Improved speed v time curve

There are however still defects at around 6.5 seconds and 11 seconds. Looking at the original tacho in more detail as shown in Figure 9 reveals the problems.

Figure 9: Anomalies in tacho signal

At the 6.5 second region there is a missing pulse and at around 11.16 to 11.17 seconds there is an additional pulse. Both of these events may be removed by using the DATS “Tacho Repair” function. The result is shown in Figure 10.


Figure 10: Final speed v time curve

We now have a very reasonable looking speed curve that matches the known performance of the engine. Now we can begin to get reliable waterfall analyses.

Waterfalls were then calculated for the exhaust vibration channel and left and right noise channels.


Figure 11 : Waterfall plot of exhaust vibration

Figure 11 shows a colour map of the waterfall data from the exhaust vibration channel. The graph is shown with the speed range restricted to 5000rpm to 6500rpm. This roughly approximates to the working range of the engine at racing speeds. As expected from a four cylinder engine, even orders, mainly second, dominate the data. However, there is clearly some unusual vibration in the region of 300 to 400Hz. This can be seen most clearly at 5500rpm where there is a large vibration at around 370Hz.

The following two tabs change content below.

James Wren

Solutions Engineer and Sales & Marketing Manager at Prosig
James Wren is a Solutions Engineer and the Sales & Marketing Manager for Prosig Ltd. James graduated from Portsmouth University in 2001, with a Masters degree in Electronic Engineering. He is a Chartered Engineer and a registered Eur Ing. He has been involved with motorsport from a very early age with special interest in data acquisition. James is a founder member of the Dalmeny Racing team.

2 thoughts on “A Simple Automotive Noise Test

  1. b. l.

    What caused the tachometer signal problem in the initial setup? That is, why was part of the data set ok but some of it not? Explain how hold off works and how it corrected this. Thanks!

  2. James Wren Post author

    Hello B.L.

    Thanks for asking a question on our blog.

    As the article states the initial problems with the tachometer are caused by extra erroneous pulses and missing pulses, which are probably caused by mechanical problems from the rotor inside the distributor on this particular automotive engine.

    Also as you can see from the article the tachometer pulses themselves are very noisy and each pulse actually has a spike at the leading edge and the trailing edge of differing magnitudes over time.

    This means any fixed amplitude peak detection is going to involve errors, if you simply count the peaks, as the article states.

    The DATS tachometer repair function simply analyses the signal, this gives the software the ability to predict when the next pulse should occur, for example the speed is not going to change by half in less than one pulse, therefore any pulses detected earlier then 50% of expected time can be considered erroneous. The process is actually much more complex, but this analogy gives a brief understanding of the concept of the hold off time, as a percentage of how much of the pulse is going to be ignored, or held off. The DATS tachometer processing software suite has further advanced features not referenced in this article, for example variable tachometer amplitude tracking.

    If you have any further questions, please feel free to ask.

Leave a Reply

  1. We welcome any feedback, questions or comments
CLOSE
CLOSE
Optimization WordPress Plugins & Solutions by W3 EDGE