(See index at

[This is the un-edited version of a story written by Paul VanValkenburgh for Road & Track.]

Would you buy new car just because you like the way it sounded? Probably not. However, all else being equal, such as performance and price, a person's preference between a Porsche, a Corvette, a Viper, and a Ferrari just might be based on the subliminal sound. And it wouldn't be surprising to find a correlation between a person's preference for engine sounds and music.

Recent scientific studies have begun proving what we all know intuitively: sounds affect our emotions. But the latest theories of music appreciation, and more generally sound appreciation, and even more generally, perception and emotion, explain how this happens. There is a natural human craving for pattern recognition, whether it's sound, sight, behaviors, or even in our earliest ancestors, weather and plant growing patterns. Steady tones are usually of little interest, becoming either boring or annoying. However, certain types of tone sequences, especially harmonics, are pleasantly anticipated, especially rising and falling patterns like musical scales -- especially when they include surprising but logical variations. Recall some of the most popular musical phrases of all-time, and see if they don't fit this pattern.

So it seems that sound appreciation is a matter of the mind setting up expectations, and having a pleasing satisfaction of those expectations, even at a subliminal level. In fact, it's been shown that sounds affect our emotions even before we're aware of them. It happens so rapidly in our pre-conscious minds, that we often react before we can explain why, and follow this with a logical explanation based on experience -- just as with other emotions like fear, love, and anger.

In the past, when we've tried to report on the quality of sound, we've been handicapped in not being able to present the actual sound itself. We've had to rely on subjective opinions presented on paper; at best, presenting eloquent poetry to represent the musical aesthetics. But, in fact, all sound, and especially music, can be measured and represented mathematically. As music is often related to colors ("blues" for example), all sounds can be broken down into their spectral components just the way all colors can. [There are rare instances of people whose brains are cross wired in such a way that they actually report "seeing" shades of colors that correspond to sounds they hear -- called "synesthesia."] But for the rest of us, sound is not represented by a rainbow of colors you might get from a prism, but by a graph showing the loudness of every individual sound frequency, over the typical human hearing range of about 20 to 20,000 cycles per second, or Hz.

The automotive industry has had good reason to quickly adopt the latest technologies in sound evaluation. for many decades, even we have used the most basic tool -- the dB averaging sound level meter. However, the dB meter is to music as the light meter is to photography: an absurd reduction of art to amplitude. More recently, dB meters have been designed with built-in filters so that they can measure the components of a given sound, the finest resolution being one-third of an octave, out of approximately a 10 octave hearing range. But even more incredibly sophisticated are computerized "real-time" mathematical analysers of sound, called FFT (for fast fourier transform) analyzers, that can break the sound down into its separate individual frequencies and peaks.

The earliest goals of sound engineers in the auto industry were simply to reduce the average sound level. That wasn't enough for consumers, however, who often complained about specific problem sounds even when the average sound was low. The one-third octave analyzers allowed engineers to focus on problems such as "rumble," "booming," and high frequency wind noise. Then, given FFT analyzers, they could look at one specific problem frequency peak, at one specific speed, and be able to match it to its mechanical cause in the vehicle, such as one specific gear or fan. Finally, they are learning to tailor automotive sounds, using the art of " psycho-acoustics." This may involve not only eliminating certain frequency peaks, but perhaps adding peaks to get the desired sound quality.

In our efforts to record the difference in sound quality between different high-performance cars, we used both a one-third octave analyzer for steady speed readings, and an FFT analyzer for dynamic readings under acceleration. The vehicles selected covered a wide spectrum themselves, from a six-cylinder Porsche, through a V-8 Corvette and a V-10 Viper, to a V-12 Ferrari.

For the dynamic readings, we took a series of spectrum readings from 1,000 to 6000 rpm in low gear (to reduce the effect of wind noise and tire noise), and the analyzer plotted them in three dimensions -- the third dimension, loudness, being represented by color. Surprisingly, as can be seen in the screen images, each of the cars was just as visually distinct as it was acoustically distinct. The loudest sound components, represented by the brightest reds, usually correspond to the cylinder firing frequency and its multiples, or harmonics. These can be seen as diagonal lines on the graphs, increasing in frequency as the cars accelerate. They begin at the little yellow "tails" at the lower left corner, which corresponds to engine firing frequency at idle.

The first notable exception is the 911 Porsche. Although the firing frequency can be barely made out at the lower end, most of the sound energy seems to be almost a random or "white" noise at much higher frequencies. This makes sense when you think about an air cooled Porsche, which has a distinct whine, or ring, generated primarily by the fan and cooling fins.

The other three cars have a very powerful and obvious exhaust note. The primary firing frequency is rpm divided by 60, multiplied by half the number of cylinders, or from about 50 to 600 Hz in our case. Harmonics show up as proportional diagonal lines of diminishing strength at higher and higher frequencies. The width of these peaks is also distinctive, from narrow "sharp" peaks on the Ferrari, to fat "broad" peaks on the Viper, as anyone who has heard both of these cars can relate to.

To continue our analogy with music, a "ringer" was thrown in. The last screen image shows the steady "acceleration" of a trombone, from f to b flat, and then held there. This corresponds roughly to our engine acceleration firing frequencies. But what is distinctive about the musical instrument, is that it is intentionally designed to accentuate an unlimited number of sharp harmonics, instead of trying to dampen or muffle them. The obvious implication here, is that we could probably intentionally design unmuffled exhaust pipes with "bells," as musical brass instruments -- and then "play" them with the throttle, at least over a limited range.

Interview with Gabriella Cerrato, at SDRC, sound engineering consultant on cars from Ferrari to the "big three".

[Can you identify what qualities of car sounds people like?]

This is an extremely difficult problem, but there are a few things to look at. For example, the primary engine firing frequency, and its multiples or harmonic orders, and how many half-orders you have. Also, the linearity of each order. Another thing is the relationship of these orders with the background noise. Some of your data shows a lot of broadband noise, which is generally bad, especially for a sporty sound.

In a sporty car you want to hear the firing orders. The cleanest signatures are from the Daytona and Viper. The even orders give you much more of the impression of power, especially if they have at least a 10 or 20 dB higher level than other orders. If you have more half-orders it gives you a rougher sound -- but which some people may like. It's more of a gutsy sound, dirty, muddy, impure, and not a pure sound like the Daytona.

The acoustic signature should be linear, meaning that peaks should increase linearly with increasing speed or rpm. Even in a sporty car, if you have that sort of sudden increase in resonance -- that can be extremely annoying.

[Yes, we noticed one of our test cars went through a sort of "wow" in the middle of its rpm range. And many cars will have a resonance at a certain rpm, which can be avoided by changing speed or shifting.]

This is probably not exhaust resonance, but acoustic cabin resonance. It's a big problem, especially when it happens at freeway speeds.

[It might also be desirable to owners of sporty cars who want to hear that sound when they're accelerating, but not all the time.]

One sporty car was designed with a muffler which was much too quiet. The buyers said, "I paid over 100,000 dollars for it, and I don't like its sound" But when you drive on long cruises and want to talk to a passenger, you don't want that constant booming.

[Yet some people prefer these sounds sometimes.]

It involves the psychology of customer expectations, which is why we run a lot of jury studies. If you present the same sound to a wide range of owners of different vehicles, they will usually want the most familiar sound. And there is a difference when you play the sound for a jury of expert listeners or technical people. [Do you have any training in psycho-acoustics?]

We deal with it a lot now, and other factors that may affect perception. We come up with sound quality descriptors, like fluctuation strength, loudness, roughness, roaring, powerful, gutsy, and we make a matrix of descriptors of each sound, and play it for a jury of 30 people. But it varies with the product -- whether it's an engine or a seat adjuster.

[Do you see variations in cultures and languages, in describing sounds?]

I came here from Fiat five years ago, where perception by the customers is totally different. And there are differences between Italian and English terms, like rumble, muddy, growl, that you don't find in a dictionary. I had to learn that, and now it's happening to Japanese engineers also.

[Do you have much experience in engineering exhaust systems, to match sounds you come up with in your simulations and jury evaluations?]

No, I mostly provide information to manufacturers of cars and components. Until recently, exhaust manufacturers would make lots of prototypes, run them on a chassis dyno, and take a lot of sound measurements, which is very expensive. Nowadays, they rely more on simulation software, using digital filters which can track with changing rpm.

[Do you know how to accomplish this with real mufflers?]

First, it is very important that a few engine orders are always present. Otherwise, people complain that they aren't getting enough auditory feedback to know how fast the engine is running.

[Do either of you know what to do to the muffler to get the sound you want?]

I don't. I just give the muffler designer the target to work at.

[How much do the engine mechanicals add to sound quality?]

A lot, especially engine vibration effects on resonance in the body structure, which may range between 80 and 120 Hz, and noises radiated directly out of the engine.

[What differences could there be in sound quality at the exhaust port between two engines, say a Ferrari and a Ford, besides compression ratio and valve timing, and maybe number of exhaust valves?]v I agree that if you have all the same engine parameters, you would have the same noise at the source.

[How would you create a Ferrari sound?]

If I looked at all the Ferrari sounds, I think I could come up with a "rule of thumb" combination of orders which could approximate the real sound. Give me enough sounds to look at, and time on my work station, and theoretically I could do it.

[Is active electronics the future of exhaust tuning?]

I spent four years working on active muffler control, and I don't see much future for it. On the other hand, canceling noise in the interior, with speakers, is less difficult.

[Wouldn't an active electronic muffler require a lot of power -- almost another engine to create strong enough canceling pulses.]

We built extra cavities in the exhaust system, and fit them with speakers to cancel the pulses at the tail pipe. But due to loudspeaker inefficiencies, we needed very powerful amplifiers. And conventional speakers can't take the temperature.

[I suppose that people like what they've grown accustomed to. For example, they may have never driven a Porsche and don't like the sound, but after they drive it for a while, they do like it - - as a reaction to the machine. In other words, maybe they find they like the car in spite of the sound.]

Yes, we can't make an absolute statement as to what is better. It's a matter of expectations.

(For more insight on the future of automotive technology, see
ng in psycho-acoustics?]

We deal with 5 w 5 B   b d   $   D F   f  f  *   M    _  #    H [    | 7 _ a  { <   P v x  4   H  H   h    ]     }  B   f  ) +  +   =    \  !     Z   |  2 4  4     O   t    ` ! Z! \! ! " E" G"  G" " # # w# # # $ 3$ $ $ <% >% s% % 7& & & & & U'  U' ' ( `( b( ( ) q) ) ) 6* j* * #+ + + + , c, , ,  , , - c- e- - . O. Q. . / / z/ / / / 0 y0 0 0 0  0 Q1 1 1 2 -2 2 2 '3 )3 3 3 Q4 ]4 4 4 r5 t5 v5 5 5  5 5 5 1 2 -2 2 2 '3 )3 3 3 Q4 ]4 4 4 r5 t5 v5 5 5    Arial d you want?]

I don't. I just give the muffler designer the target to work at.

[How much do the engine mec