Sounds Sweet!


Go to any ‘natter’ and you are likely to find a group of enthusiasts standing around a car listening to the engine and making comments such as “that engine sounds really sweet”, or “sounds as though it’s running a bit rough”. Considering the cacophony of noises made by an engine, especially earlier ones such as the XPAG, it is quite amazing people are able to distinguish these subtle differences in sounds to make comments such as this.

This came to me very forcibly when I took my TC to a rolling road to test the effect of using a petrol / kerosene mixture. I went prepared with a separate supply tank to bolt onto the fuel pump so we could change the mixture very quickly and pre-mixed bottles of petrol and kerosene. The first test was using “pure” petrol. The car was run up to 3,000 rpm and the mixture and timing checked. We then swapped to 5 parts petrol to 1 part kerosene, ran the engine up to 3,000 rpm again and that is when the three people present looked at each other in amazement. The engine sounded very much smoother.

Situations like this do not occur very often, specifically, where everything about the engine is the same, yet in the space of the few minutes it took to change the fuel, it sounded as though it was a different engine. Mixture and timing were rechecked and found to be the same as the previous run. The only measureable difference was the reduction in unburned hydrocarbons and NOX confirming that the engine really was running better.

What I still find amazing is that three people were clearly able to hear this improvement. The only possible explanation I can offer is that our ears were able to separate the sound made by the combustion of the fuel from the rest of the rattles, knocks, clicks and bangs.

This made me wonder if it were possible to actually measure this effect.

The human ear is a remarkable organ; it is not only able to hear very subtle differences in sounds but also to pick out those sounds from a very noisy background – something that presents a serious challenge to any measuring equipment. However, modern signal processing techniques may provide the means.

This article describes my investigations which give an indication of the possible value of this technique. Unfortunately, I have very little experience of signal processing and now we have the funding for more tests at Manchester University kindly offered by the MG Car Club, NTG and Totally T Type 2, I would like to include these measurements in the planned tests. I therefore request that anyone who might be able to help me with this contacts me.

Fourier analysis

Before presenting my findings, I apologise that I need to introduce some physics.

Display a picture on your computer and look at it under a powerful magnifying glass. You will see that a pink face is actually made up of thousands of small red, blue and green dots. It is only when you look at it from a distance that these primary colours combine to give the colour pink.

Sounds are very similar. All sounds are made up of a combination of sine waves of different frequencies. Close your lips and make a “hummmm” sound. This will be close to a pure sine wave. Press your tongue against our teeth and make a “zzzzzzzzzzzzz” sound of the same pitch. This will consists of the sine wave like the “hummmmm” sound along with many high frequency components (called harmonics). This is why it sounds harsher.

In practice, like the dots on a photograph, any sound can be broken down into a fundamental (normally referred to as the first harmonic) and a number of higher frequencies. The second harmonic is twice the frequency of the first, the third, three times the frequency, and so on. When added together with different amplitudes and phases, the harmonics change the shape of the sound envelope. For example:

You can see from the graph how the second harmonic changes the shape of the pure sine wave to be more what we would expect the combustion in the engine to look like. A rapid rise as the fuel burns followed by a slow fall as the piston goes through its power stroke. Adding the third harmonic smears the sine wave, something that could indicate an engine’s timing wandering or the fuel not burning consistently.

It is possible to analyse a sound recording to determine the amplitude of the different frequency components that make up that sound. This is called Fourier analysis which can produce a plot showing the amplitude of each of the harmonics that make up that sound. For example:

The graph on the left is probably a close approximation to the combustion pressure profile in the cylinder and has been produced by summing the first five harmonics. The chart on the right shows the relative amplitude of each of the harmonics that were summed to make the graph on the left. If you analyse a sound with the profile on the left into a Fourier analysis system, you will get the chart on the right.


In theory, it should be possible to reconstruct the pressure profile in a running engine just by recording the sound it makes. While this may sound impossible, the rolling road experience showed the human ear was capable of doing this, so I thought it worth a try. While the preliminary results show such an analysis is possible, more care needs to be taken both with the recording and analysis to produce anything meaningful. Hopefully, this can be achieved during the Manchester tests.

The first problem with making any such recording is the frequency of the first harmonic. For an engine running at 3000rpm this is 50Hz which is a very low frequency for sound and secondly is the same frequency as the mains which means any measurement could be affected by mains hum. In the tests I ran the engine at around 2750 rpm to avoid this problem.

The second problem with the XPAG is that it is a very noisy engine and with mine still using the mechanical distributor, it has poor timing consistency. However, one advantage is that, like the rolling road, I was looking for differences between fuels rather than any absolute measurement so anything that does not change should cancel out.

Armed with a low frequency microphone positioned in the engine bay to the rear of the engine, I started with my modern 2 litre Audi. The main problem here was I could not change fuels without running the tank dry and it was very difficult to maintain a consistent engine revs for the recordings, despite a block of wood under the accelerator.

Fortunately, by performing the Fourier analysis on only 10 seconds of the recording where the revs were consistent, I obtained the following Fourier spectra for two different fuels, A and B. In practice the car ran considerably better and smoother on fuel B:

Encouraged by the results with the Audi, I ran similar tests on the TC:

Some things were a little simpler. Firstly, I blocked up the back wheels and ran the engine in top gear to provide a load; secondly, I was able to fix the revs with the slow running control and finally, I used the small container fixed to fuel pump so I could change fuels more easily. This allowed me to complete the recordings within ½ hour rather over a couple of weeks as with the Audi.


The results for the Audi clearly show differences between the two fuels but also demonstrate the difficulties of making these measurements.

The recording on fuel B was taken at 2800 rpm and that of fuel A at 2600 rpm hence the difference in the frequencies of the harmonics between the two fuels. What is also interesting is that the 3rd harmonics from one cylinder at around 70 Hz indicates a problem. With a four cylinder engine running at 2600 rpm, there are around 44 “bangs / second”. However, each cylinder is only “banging” 22 times per second. With the microphone placed at the rear of the engine (to avoid the noise of the timing belt and alternator), it clearly “heard” one cylinder louder than the others which shows as its 3rd harmonic at ~70 Hz.

Despite these difficulties, differences can clearly be seen between the two fuels. Fuel A has broader peaks showing the timing between each “bang” is changing more than fuel B. While driving the car on fuel A, this appears as rougher running than with fuel B. In addition, the ratio of the 1st and 2nd harmonic peaks is different between the two fuels suggesting the combustion profile is different.

The results for the XPAG are equally interesting.

For each fuel the throttle setting was the same, controlled by the slow running control as is the load resulting from the transmission losses. The fact that the engine was running slower on fuel A, relative to fuel B suggest that perhaps it is producing less power. Certainly the car appears to drive a lot better on fuel B.

Compared to the Audi, the peaks are wider, something that would be expected considering the mechanical timing of the TC vs the electronic timing of the Audi. What is most obvious, however, is the dramatic differences in the relative amplitude of the harmonics between the two fuels which must imply a different combustion profile.
Where do we go from here?

For me, these results appear to offer a tantalising suggestion that it may be possible to electronically measure how “sweet” an engine sounds and this could provide a quantitative measure of what fuel should be used. The next stage is to improve on these results during the repeat engine tests at Manchester.

Perhaps, at last, the possibility of a little box that will tell you how well your engine is running is around the corner. However, I do not think it would ever replace the enthusiast’s comments such as “that sounds really sweet”, or “sounds as though it’s running a bit rough”.

© Paul Ireland
January 2015

5 thoughts on “Sounds Sweet!

  1. Graeme Hogg says:

    Hello Paul,
    Read you article with interest. I got the impression you were using a “free to air” microphone as your pickup. Just a suggestion, but perhaps you could utilise a simple detonation detector device based upon a simple 1W amplifier arrangement which utilises a piezo sounder as the microphone, the latter being clamped to the engine as close to the combustion chamber as practicable. This gives an audible analogue output which, with an earphone, allows you to listen the internal “symphony” but which is intended to highlight any pre-ignition noise. I am not an electronics expert however you might find such a device provides an analogue signal of sufficient range for your purposes and would help to exclude noise from external sources. I have a copy of a circuit and construction details should that be of interest.
    Graeme Hogg

  2. Brian tyler says:

    What’s in a name Doreen. ( apologies to CJ Dennis, an Australian poet)

    I had read of kerosene being added to petrol on anothe MG forum a number of years ago, and when I ran out of my normal fuel additive, Shelite, early one Sunday morning on the way to a vintage motorcycle rally, I rememberd this little gem of information and purchased some Kerosene along with my low octane fuel. The “Kero” was added at the rate of 10% by volume, the same as I would have done with the Shelite, and my expectation was a cooler running mix, as evidenced on previous use with the Shelite by the lack of blueing on the exhaust pipe. The day wasn’t particually hot as far as Australian conditions are concerned, but I found that I was moving my legs away from the general area of the engine as I was motoring, as the heat being given of was getting uncomfortable, and the smell was pretty bad- as anybody who has had to live with a poorly adjusted parrafin/kerosene heater will confirm.

    So we come to my question, are Kerosene, and Parrafine the same thing in the UK? Because here in Australia Kerosene is Parrafin, by another name, or is the U.K.’ s kerosene closer to Avgas (aviation fuel) or Shelite . It must be something different, because I can’t see any one using the Australian version to obtain better running results. I look forward to a fuel guru, or any one that’s knows commenting. Brian

  3. Paul Ireland says:

    Thanks for the comments:

    Brian – yes, kerosene, paraffin and 28 second heating oil are all basically the same. Since I first ran my tests using “standard supermarket” 95 octane fuel, I have found that my 7.25:1 compression ratio TC runs much better using 97 octane Super grade fuel (indeed as does my 2 litre Audi). In practice, I have found that different brands of fuel have a greater effect on engine heat than just adding kerosene. In practice, I still add 10% kerosene to my Super fuel both to save money and I feel the engine runs a little smoother. The reason for the Manchester engine tests is to, hopefully, get to the bottom of this.

    Graeme – thanks for your suggestion. I originally chose a low frequency “free to air” microphone as I was worried by the low frequency range I need to detect. 50 Hz for an engine running at 3000rpm or 25 Hz for a single cylinder. This is a very low frequency for sound and I was worried that a knock sensor would not detect these low frequencies. As you can see from the graphs, the Fourier analysis clearly “extracts” the key frequency components from the rest of the noise made by the engine. A further advantage is that I was only looking for differences. Fortunately, I have now made contact with people from the Electronic engineering department at Manchester, who have said they would be willing to help with these measurements.

    • Laurent CASTEL says:

      Hello Paul,
      Very good idea and congratulations for these preliminary results.
      As Graeme, I would rather use a contact sensor such as a piezzo to cancel any external noise mainly from the muffler or even acoustic properties of the garage. This kind of sensor will be much more sensitive to any mechanical noise like valves, bearings or cam followers. A drawback for combustion study or another advanced study for wearout detection ?

      Second point is the funding of the manchester university partneship. Why not launching a crowdfunding campaign on one of the dedicated websites ? Every classic car owner is concerned.
      Any one to manage such a campaign ? Unfortunately, I’m not the right person. Could it be a project for students in economy ?

  4. Paul Ireland says:


    Thanks for these comments. For the future tests the plan will be to install contact microphones and to improve the low frequency response we will also need to use high impedance amplifiers.

    In practice there is one other “trick” we are considering. By using the spark to trigger the start of recording and let it run for the next ~10 milliseconds or so, it will be possible to “listen” to each individual cylinder. With four contact microphones placed a similar positions, it will be possible to track the sounds made by each cylinder.

    I believe this has very exciting possibilities.

    I also think your suggestion of using sound to monitor the overall condition of the engine is also very exciting. I understand this technique is used in helicopter gearboxes to “listen” for wear.

    Also, thank you for your suggesting about funding. At the moment, I think we have sufficient funds but this is a possibility worth considering.

Comments are closed.