Article - Issue 31, June 2007
Vehicle Sound Engineering
Dr Paul Jennings, Dr Garry Dunne and Roger Williams
Business and academe have got together to produce a simulator that is used to analyse, replicate and perfect the sounds made by motor vehicles. Dr Paul Jennings of WMG at the University of Warwick, Dr Garry Dunne of Jaguar and Land Rover and Roger Williams of Sound Evaluations, tell of their work creating car sounds appropriate to their marques.
Figure 1. Interactive evaluation of a new vehicle’s sound using a desktop simulator
Every sound from a door shutting and a motor whirring, to the engine idling and accelerating, has an influence on the choice of vehicle by a prospective car owner. A purr, a growl or a roar – sound can be the decider when it comes to choosing a car.
Over recent years cars have, overall, become much quieter. Automotive engineers and designers have spent decades taking the sound out of cars. Quieter engines, better manufactured vehicles, more aerodynamic shapes, and improved tyre technology have all contributed to the process of making quieter cars. However when most drivers accelerate they do not want silence, they want to get a feeling of power or sportiness, of luxury or refinement.
For the sound engineer, the challenge is to be able to capture these subjective feelings of customers and turn them into deliverable sounds for a new vehicle, enabling the vehicle noise to have character whilst still maintaining the impression of overall quality and solidity.
The aural mix
Potential target sounds for vehicle manufacturers can be created by manipulating or mixing existing sounds, or by using engineering simulation tools. There will often be a variety of constraints on developing vehicle sound, such as cost, weight, ride, handling, styling and manufacturability. During vehicle development there will be continual monitoring of these elements and revision to achieve the target sounds.
The intended sounds for new vehicles need to be assessed and compared with existing and competitors’ vehicle sounds. Direct inputs from both customers and from Noise, Vibration and Harshness (NVH) experts are also needed, so that all relevant options can be properly considered and evaluated.
A new approach
In the past, traditional on-road testing has allowed vehicle sounds to be assessed by both NVH experts and by customers. However, it is difficult to ensure consistency and repeatability with this method, and it is not possible to test and experience vehicles that don’t yet exist!
For the last few years a collaborative team have been working on this multidisciplinary problem, drawing on their combined expertise in sound and vibration, computer-based modelling, artificial intelligence, product development, vehicle manufacturing, psychology and marketing.
The first research was undertaken to develop appropriate jury evaluation techniques within the controlled environment of a listening room (Figure 2). Here, back-to-back listening comparisons were made possible and consistency and repeatability were improved. These techniques were based on methods used widely in psychology and consumer industries. They include paired comparison and semantic differential approaches. As well as successfully adapting these approaches to the needs of automotive sound quality, they led to a new target-setting process being implemented within Jaguar Cars.
A lot of new knowledge was gained about the subjective appraisal of vehicle sounds. For example, it was established that during the appraisal of the wide-open throttle acceleration sound in second gear, the assessors’ subjective impressions could be broken down into measures of two independent qualities – how powerful they perceived the vehicle to be and how refined. This enabled different vehicles to be positioned in a two-dimensional target space, as shown in Figure 3. Intriguingly, it was found that vehicles of the same brand clustered together. Hence it is possible for manufacturers to set targets for future vehicle sounds relative to their current vehicle and competitor brands.
These techniques have proved successful in linking the engineering data behind the target sounds to subjective appraisal. There are, however, limitations. The assessments are not able to be made in the right context – sounds should be assessed whilst driving, and in the presence of other stimuli which significantly influence the perception of sound. Furthermore, listening to a fixed, pre-prepared sound of a repeatable vehicle operating
condition, such as second gear wide-open throttle, is not representative of a typical on-road experience.
So we have developed a noise and vibration simulator to address these limitations. It now forms the basis for an improved NVH process by providing:
A unified NVH data model, which contains the characteristics of current vehicles, competitor vehicles, targets and design alternatives in a single database which is simple to update.
Interactivity and context.
A variety of interactive evaluation tools which all use the same data model to ensure a consistent and robust approach to target setting and decision making.
Simply put, the NVH Simulator we have produced can now accurately recreate the noise and vibration of a vehicle in a realistic driver-in-loop interactive environment. As components are altered, the simulator reflects those changes in the vehicle’s NVH. It gives carmakers and customers the opportunity to drive, feel and hear a potential new car without actually committing to a concept vehicle. It also allows them to directly compare the NVH of this new car with existing and competitor vehicles which are also replicated within the Simulator. There are two versions, one for use at a desktop and the other a full vehicle simulator.
This Simulator (Figure 1) is a fully interactive system for use in an office or listening room environment, enabling the user to experience the sounds of the virtual scenario. This is primarily a tool for the NVH engineer to develop targets, understand problems or evaluate design alternatives. It can also be used to capture subjective opinions by performing interactive jury evaluations.
The full vehicle
This Simulator (Figures 4a and 4b) is where the assessor sits in a real car stationed in a room and ’drives’ it using the normal controls whilst experiencing the synthesised sound and vibration at interface points (for example, seat, pedals, steering wheel). The complete vehicle is situated in front of a large projection screen on which a virtual visual scenario is displayed. The main purpose of such a simulator is to provide a more realistic environment for non-experts to make assessments and also to investigate the combined effects of sound and vibration.
Testing the limits
In both cases, a car is represented in the NVH Simulator by a hierarchical data model (Figure 5), which can include any or all of the contributions of the physical sources and components that can be heard inside the vehicle when it is operated over its full driving envelope. Each NVH component in the model is represented by one or more sound objects containing the data and metadata, together with the rules that determine how the sounds will be synthesised in real time based on driver inputs (see box).
As well as the NVH data, the car simulation requires a performance model which is an empirical representation of each car’s driving dynamics. This defines the relationship between the actual position of the controls (throttle, brake, gear) and the vehicle’s road speed and engine revolutions per minute (rpm). The performance model is usually derived from on-road measurements but may also be calculated with a simple parametric model
In summary, the NVH Simulator allows any NVH data which can be represented in the time, frequency or order domains (see Driver Inputs box) to be evaluated interactively as a virtual car whose sound, vibration and driving performance are indistinguishable from the real car. A parallel would be the ability to play a symphony, with a symphony orchestra, and to be able to change the sound of any individual instrument whilst everyone else continues to play unchanged.
The evaluation capabilities of the NVH Simulator can be used together in a logical step-by-step process to define and deliver the required sound quality in a new vehicle. The simulator can also provide improved methods for understanding and solving noise and vibration problems in existing vehicles. This process is shown in Figure 6. The same data model is used for all the different types of evaluation, but the data is presented to each category of assessor in the optimum way for their individual tasks.
This interactive, stimulus based approach has been successfully applied in several recent new Jaguar vehicle programmes. One of the most significant benefits to this approach is that it introduces input into the NVH process which has not previously existed. Firstly, it enables engineers, decision makers and real customers to share a common experience of the sound of a new car before and during its development. Secondly, it provides a tool which directly connects subjective opinion about NVH to the relevant engineering data. People have always had a very personal and close relationship with their cars, often talking about their ‘personality’. Perhaps we are now nearing a time when as well as choosing the shape, colour and texture of the finished product, the car you buy will also be the one that sounds best for you too.
The total engine component needs two sound objects – a set of phase-aligned engine orders to represent the harmonics, and a set of masking spectra to represent the broad band contributions. The orders are continuously mapped from the revolutions per minute (rpm) domain to the time domain, based on the instantaneous rpm value.
Masking sounds are a blend of the two random waveforms or spectra which are closest to the current instantaneous vehicle speed or engine rpm.
Secondary sounds can include:
Gear whines which are also defined by a set of phase aligned orders for each significant gear and load combination. In this case the orders will have their own phasors which can be a non-integer multiple of engine rpm or vehicle speed depending where in the drive-train the gear sets are located.
Transients such as key-on during starting, horn and indicator, rattles, discrete impact events from the road surface or traffic. These are either triggered manually or by the visual scenario.
We have established that the sound of cars is important for automotive companies. The soundscapes of urban environments are also important in providing a unique and appropriate identity. Soundscapes in hospitals, for example, can help to promote the ambiance in healing environments and disguise or mask distressing sounds.
Simply reducing the levels of sound is not necessarily the best approach, as there are many other factors that influence the positive perception of sound.
The authors are therefore going to apply the learning and techniques developed for automotive applications to new challenges. An initial aim is to design a Simulator to recreate and evaluate the hospital sound environment based on the approach used to successfully develop the automotive NVH Simulator.
BIOGRAPHIES – Dr Paul Jennings, Dr Garry Dunne and Roger Williams
Paul is an Associate Professor at WMG, the global innovation specialist based at the University of Warwick. He leads two research teams, one working on hybrid vehicles and the other on linking the subjective perception of products with engineering targets. The product perception interest is founded on the described research on Automotive Sound Quality Engineering.
Garry is the Technical Leader for Vehicle NVH at Jaguar and Land Rover. He joined Jaguar Cars in 1990 and in his current role as Technical Leader at Jaguar and Land Rover he is responsible for managing the technical delivery of the NVH & Sound Quality attribute across all of Jaguar and Land Rover’s Vehicle programmes. He has particular responsibility for defining the strategy and delivering the Powertrain Sound Quality attribute.
Roger is a Physics graduate with over 30 years of consulting experience in the field of product noise and vibration. His main areas of expertise include Sound Quality Engineering and the development and application of new Sound and Vibration Synthesis Techniques, Tools and Processes. He is a founder and Technical Director of Sound Evaluations Ltd and is also a Director of Novisim Ltd, a collaborative venture focused specifically on the development of Vehicle NVH Simulators.
The authors would like to acknowledge the support received from Warwick’s Innovative Manufacturing Research Centre who funded some of the research and provided the photographs for this article.