One area that has long been of interest to cylinder head tuners is the analysis of flow velocities within a port and around the valves. For decades, pitot tubes have been used to take spot pressure readings from inside the ports, and most flow benches come supplied with a port to accept a pitot.

A pitot tube can be used in various ways, either by being placed in a static location to measure flow velocity at that point or moved around like a wand, to try to identify areas of high or stagnant flow. While this makes them a useful tool for obtaining a general idea of flow velocity within the port, however, they have their limitations, the most notable of which is the fact that inserting a pitot tube into a port introduces an obstruction that can interfere with the airflow and skew the accuracy of readings.

This problem has led some head tuners to develop methods of taking pressure measurements from within the ports in such a way that there are no obstructions to the flow. For example, the author has seen one example of a test head where the tuner drilled a series of holes (known as pressure taps) around the valve seating area, with each attached to its own manometer. This allowed pressure readings to be taken without any impact on the flow conditions around the valve as there was no obstruction present.

The data it generated was also interesting, giving a snapshot of how pressures differed around the valve and changed with valve lift. It also allowed for comparative testing between different port designs.

There is one big disadvantage with this approach though, in that the heads that were drilled in this way could not be used for anything other than testing.

This brings us to a second novel approach to assessing differences in flow across the valve. It involves using a specially made valve that incorporates a pressure tap in the valve head. The tube from the pressure tap then runs up the valve stem to a tube that can be connected to the pitot port present on many flow benches.

As the head is run on the flow bench, the pressure at the point on the valve where the pressure tap is located can be recorded, either manually or via a flow bench software package. Once the pressures at that point of the valve have been recorded through the range of valve lifts, the valve can be rotated to record at a different point; adding an indexed collar on the valve stem means this can be done in a repeatable fashion.

The end result is a full pressure map of flow at the valve head throughout the valve’s entire lift range, providing an invaluable insight into a particular design’s flow characteristics.

These two systems just go to show that, although CFD and complex flow visualisation systems are the favoured approach these days for large racing operations, there is still plenty that can be learnt from a good old-fashioned flow bench.

Written by Lawrence Butcher

In ultrasonic testing, high-frequency sound waves are reflected from flaws in predictable ways, producing distinctive echo patterns that can be detected, displayed and recorded. Its ease of use and non-destructive nature of makes it a very useful tool for the quality control of engine and transmission components, particularly cast or fabricated components. In this month’s article we will take a closer look at the details behind it.

Ultrasonic testing relies on the fact that sound waves travelling through a medium will be reflected or transmitted in different ways if the composition of the medium varies. They can therefore be used to identify flaws or voids in materials. Using an ultrasound transducer to generate an ultrasonic sound wave, a wide range of materials can be tested for integrity.

There are two methods of receiving the reflected ultrasound waveform: reflection and attenuation. Using the reflection (or pulse-echo) method, a single transducer performs both the sending and receiving of the pulsed waves, as the process relies on the sound waves that are reflected back from the material being tested. The reflected ultrasound that is received by the transducer from an impervious obstacle such as the back wall of the object or an imperfection within it. These results are then displayed in the form of a signal with an amplitude representing the intensity of the reflection and the distance from the sensor, which is represented by the time difference between sending and receiving. As the transducer passes over a flaw, the signal displayed will change from that produced by homogenous material.

With the attenuation (or through-transmission) mode, a transmitter sends ultrasound through one surface and a separate receiver detects the amount that has reached it on another surface after travelling through the test sample. Imperfections or other conditions in the space between the transmitter and receiver reduce the amount of ultrasound transmitted, thus revealing their presence. This type of testing is used to check large-volume components.

There are several features of ultrasound testing that make it particularly attractive for use in motorsport. As mentioned, it is non-destructive, meaning valuable parts do not need to be chopped up to check their integrity. Ultrasonic testers are also portable, so they can be easily used trackside for checking components, and this portability means they can also be used to examine large assemblies. They can also be used for tasks ranging from checking welds to assessing castings for porosity. Overall, ultrasonic testing is an exceptionally useful technique that can provide an invaluable capability in any race ’shop.

Written by Lawrence Butcher

For example, during the development of its Formula One V8 engines, the Honda team wanted to improve the driveability of its engine through closer control of the air-to-fuel ratio (AFR) entering the combustion chamber. The demands placed on an engine such as those found in the previous generation of Formula One car are considerable, not least the need to be able to move from fully closed throttle to wide-open throttle almost instantly. This requires exact control of the fuel injection events, and if this is not achieved then the AFR – particularly in the immediate vicinity of the spark plug – may end up rich or lean, hindering flame front propagation and impacting driveability.

By measuring in-cylinder pressures, Honda had found that misfires were occurring under certain transient throttle conditions, which it thought was down to an incorrect AFR. Naturally it wanted to find out why these misfires were occurring and take steps to improve the situation. To do so, the team used a micro-Cassegrain sensor incorporated into a spark plug, which was able to measure the chemi-luminescence of different elements in the flame front inside the combustion chamber. By studying the intensity of this luminescence, the rate of flame propagation and the AFR could be established.

To ensure that the sensor system was not interfering with the normal operation of the spark plug, Honda ran sensors in only four of the eight cylinders and compared the variation in combustion pressures between instrumented and non-instrumented cylinders. Also, the micro-Cassegrain sensing elements were placed behind sapphire glass shields to protect them from the extreme heat and pressure in the combustion chamber. By coupling the sensors to a high-speed data acquisition system, the team was able to record data at high resolution, with measurements accurate to within 1º of crank angle at 18,000 rpm.

The results Honda obtained from the measurement system allowed the team to gain considerable insight into the combustion behaviour of individual cylinders. For example, it was found that variations in the design of the air inlet scoop had a considerable impact on the AFR around the plugs in particular cylinders, something that had not been apparent before.

However, it was the results obtained during transient testing that were most revealing. As mentioned, the team had already found that certain cylinders were misfiring, but could not confirm why. The sensor provided the insight needed for this confirmation, and it was found that the AFR around the plugs from cylinder to cylinder varied considerably under transient throttle conditions, from rich to lean, causing the misfires that had been detected.

As with any form of testing, it is often the case that being able to record something that was previously un-recordable leads to more questions than answers. In Honda’s case, it discovered the reason behind its engines’ driveability issues, but finding a solution that would allow sufficiently precise mixture control to combat it was a different matter. However, without the ability to study exactly what was occurring in the cylinders, any solutions the team may have devised would have been mere guesses, as the exact problem it was trying to address would not have been clear to see.

Written by Lawrence Butcher

In simple terms, a CAN bus is a network of individual electronic controllers that communicate using a protocol which automatically gives important signals priority over less urgent ones. As vehicle electrical systems grew in complexity, CAN systems were developed to reduce the amount of wiring needed. This reduction is achieved owing to the fact that each controller needs only two wires to transmit signal data.

For example, if a car’s stability control system featured wheel speed, brake pressure and suspension displacement sensors on each wheel then, without CAN, each of them would need to be hardwired to the stability control ECU. With a CAN system, however, the sensors can input into individual control boxes located either at each corner of the car or, say, one front and one back, which then process the sensor signals and communicate with the main ECU via simple twin wires. In the old system, anything up to 20 individual wires would need to run to the ECU, while with CAN this can be reduced to just four if only two processing boxes are used.

From a data engineering perspective, these systems make it very easy to add or remove functionality from a particular data recording set-up. If properly implemented, CAN will also provide much better system redundancy in the event of components getting damaged.

There are two key ways of going about this with a production car already fitted with a CAN system. The first and easiest option is simply to connect to the factory ECU (if used) through the diagnostics port. With the correct interface and data logger, the information the ECU is receiving from its various sensors can be easily recorded. The alternative is to run a system piggybacked onto the existing wiring, sharing the CAN data signal being sent to the ECU.

“If you are doing a test and you want to add some functionality to the car, the CAN is very useful,” explains Steve Dunlop, an engineering and racecar constructor with the JRM team, which has extensive experience of running both rally cars and track machinery. “One particular area that that comes to mind is when you want to add a wideband lambda sensor instead of the narrowband type usually found in production cars; it is very easy to add it to the CAN stream and integrate it with your existing system. It also means you do not need to cut into the existing wiring looms, and the new additions simply run in parallel with the sensors already present.” 

The benefits of CAN-based vehicle electronics have also seen their growing adoption over the past five years in motorsport-specific engine control systems. For example, the Fiesta R5 rally car features a near-comprehensive CAN-based wiring system, with individual control modules designed specifically for the demands of motorsport sited throughout the car. The result is less cabling in the car, meaning simpler and thus cheaper and lighter wiring, combined with easier maintenance thanks to the improved fault diagnostics capability the CAN system has over a regular wiring loom.

Written by Lawrence Butcher

The first, and by far the cheapest, is simply to use a powerful desktop computer. The latest generation of high-end machines have an excellent cost-to-performance ratio. They feature multiple core processors, which – if configured correctly – can be harnessed to effectively handle the requirements of basic CFD computations. For more complex calculations though, the very large number of computations required becomes a limiting factor: while they can be undertaken, the time taken to do so is considerable. When it comes to simulating engine flows, anything much more complex than basic steady-state flow analysis of simple geometries is beyond the capabilities of a regular PC.

The next step takes us into the world of high-performance computing (HPC), a term often bandied about in relation to CFD simulation and one that covers a wide range of computer types. There’s no fixed definition of how powerful a computer needs to be for it to be considered as ‘high performance’. This is because the performance of microprocessors has increased in an exponential way for many years, so any such definition is soon out of date. It’s more usual to consider a computer to be high performance if it uses multiple processors (tens, hundreds or even thousands) connected together by some kind of network to achieve well above the performance of a single processor. Using multiple processors in this way is also often referred to as parallel computing.

When it comes to CFD calculations, the benefit of such systems are considerable: with many processors and processor cores handling calculations, more calculations can be undertaken in a shorter time. Even with a properly configured single-processor machine, with say, a quad-core processor, only four calculations of a particular solution can be undertaken at any one time. Use, say, 20 quad-core processors on the same solution though and clearly the processing time will fall considerably.

At one end of the HPC spectrum you have the multi-million pound supercomputers used by the likes of Formula One teams, featuring thousands of processing cores and capable of very rapid and complex CFD solutions. With the best of these, the sky is (almost) the limit, allowing for example high-resolution simulation of entire engines. At the other end, and the one applicable to most who need a decent level of CFD functionality, are interlinked clusters of individual desktop machines. While each machine on its own is nothing special, considerable performance can be achieved when linked together correctly. The benefit of this type of system is that regular workstations can be used, provided they are connected by a sufficiently high-speed local area network. That means they can be used for other tasks when not needed for CFD work.

For most small-scale powertrain CFD users, this approach is the most useful, negating the need to have a dedicated and expensive HPC cluster. At this point, and in order to avoid confusion, it should be noted that cluster computing can also refer to a cluster of processors housed in a single machine. The mode of operation is the same, but instead of consisting of a cluster of workstations, the processors are all centrally located.

The final option, which has only become a viable one in recent years, is the use of ‘cloud’ computing resources. This essentially means that computing tasks are outsourced virtually to HPC facilities around the world. It is a nascent industry at the moment, but its obvious benefit is that it removes the need for in-house computing resources. For companies who may only have an infrequent need for simulation, the availability of such resources opens up the possibility of undertaking far more complex simulations than would have been feasible a few years ago. 

Written by Lawrence Butcher

This is a particular issue when it comes to sensors such as strain gauges and thermocouples, which have very low output signals. Whereas sensors like pressure transducers can have an output voltage of 1-10 V, the output of strain gauges and thermocouples will be only a few milli- or microvolts. That puts them in the same range as much of the electronic interference found in a racecar, particularly the new generation of hybrid Formula One and sports cars with high-voltage electrical circuits.

As a result the background ‘noise’ can have a considerable impact on the recorded output of a sensor, as the ‘signal-to-noise’ ratio is very low. This ratio is the level of a particular signal’s strength compared to the level of background noise. So if, for example, a sensor had a 5 V output signal from a sensor and a background level of signal noise of a few microvolts, the signal-to-noise ratio would be very high; however, if the sensor output is only in the millivolt range, the signal-to-noise ratio would be much lower, so it would be much harder to distinguish the signal from the noise.

While a lot can be achieved using complex signal processing algorithms to ‘clean up’ noisy signals, effective circuit design and shielding will go a long way towards improving sensor performance, reducing the need for such measures. The most obvious solution is to keep the cable length between a sensor and its signal amplifier as short as possible; long cables act as antennae, picking up electrical and magnetic interference. Twisted cable pairs also help to prevent signal noise by reducing induction between the wires.

Adding shielding around cables is also imperative. Shielding a cable simply involves enclosing the insulated signal cables in a conductive layer, normally braided metal wire or copper tape, that acts as a Faraday cage around the signal cables, reducing the impact of electrical or magnetic interference.   

Installation considerations must also be taken into account in order to reduce signal noise. For example, strain gauges are often mechanically and electrically attached to the component being measured, which can result in ground-loop currents forming that will contribute to signal noise. These occur where a difference in potential develops between the sensor ground and the signal amplifier. However, careful attention to detail when it comes to earthing the sensor, shielding and the signal amplifier can prevent them.

Things get a little trickier though when dealing with very high voltage systems, such as those in hybrids with powerful electric drive components and energy storage systems. The electromagnetic fields these produce can create problems not only for sensors used to measure their operation but any other sensors that may be close by. Fortunately, the frequencies of these fields can be identified and accounted for in the signal processing, but where very strong fields are present, the only solution is to increase the level of shielding on the sensor circuits, otherwise the EMI can actually be sufficient to damage the circuitry itself.

One interesting solution to the problem of measuring in very high EMI environments is the use of optical rather than electrical sensors. Where conventional electrical sensors use transducers to covert physical phenomena into electrical signals, optical sensors use light, and fibre optic cables instead of wires. Such sensors are immune to EMI, although they are still nowhere near as common as electrical sensors, and we will take a more in-depth look at their potential applications in a future RET-Monitor article.

Written by Lawrence Butcher

However, even this cannot ensure complete control over the quality of finished parts. For example, the make-up of parts produced using SLS (selective laser sintering) processes can vary from one component to the next, so a means of non-destructively examining components internally can greatly reduce wastage. One such method that is seeing increasing use is CT (computed tomography) scanning, which allows for accurate and detailed analysis of both internal geometries and material density.

Most people will be aware of the term CT scanning thanks to its widespread use in medicine, but its use in an industrial context is less well known. Put simply, a CT scanner uses an X-ray source to produce tomographic images, or ‘slices’, of a component, which are then processed to create a 3D image of the component’s internals. The first generation of CT scanners took nine days to produce a single slice image; now though, even highly complex objects can be scanned in less than an hour thanks to advances in computer processing power.

A CT scanner for industrial use will feature an X-ray source (an open or closed X-ray tube) that produces a conical beam of electrons that penetrates the object to be analysed. The electrons are detected and interpreted by a 2D sensor as a ‘digital radiograph’ image. The object being examined is positioned on a precision rotary stage, and an image is acquired each time the stage is rotated. Depending on the resolution required, each step is usually 0.1-1º, producing between 360 and 3600 individual images. The scan usually covers a rotation of 360º, but for specific applications the scan can be a limited angle.

Once a component has been scanned, the resulting image can be used to study different parameters. For example, if a scan was taken of a cast component then variations in material density throughout the part can be examined. With software, problems such as inclusions in the material, cracks or voids left during the casting process can be highlighted. Whereas previously the only way to check for these would have been through destructive testing, CT scanning allows for any component to be checked and then used. Leading on from this application, CT scanning also opens up a route to effective reverse engineering of components. The most common form of reverse engineering is laser scanning, however, this has limitations in that it is hard to measure internal geometries. With CT scanning, this is not an issue and with the correct analysis software, highly accurate 3D models can be created from scans, which can then be transferred into regular CAD packages.

One particularly interesting application for CT scanning in the motorsport environment is in checking composite structures. The strength and stiffness of structural carbon fibre parts is determined by both the orientation of the carbon fibres and the fibre-to-resin ratio. When designing such parts, engineers will use FEA (finite element analysis) to determine the optimum weave arrangement; however, it is impossible to check how closely the final part conforms to the design parameters without using destructive test methods. Using CT scanning, the exact internal structure of the part can be checked and thus its real-world mechanical performance more accurately predicted.

These are only a few of the potential uses for this interesting technology, but it has the potential to provide engineers with a highly accurate method of checking a plethora of previously hard-to-examine parts. The only downside is cost – CT scanners are very expensive, so having such a facility in-house is not feasible for most operations. (which is why most CT inspection is performed at 3^rd party CT inspection labs as opposed to in house inspection).  However, with an ever-increasing number of parts made using 3D printing, and a subsequent need to closely examine internal forms is likely to be an increase in third-party suppliers providing CT services, making it easier to outsource parts for checking. 

Written by Lawrence Butcher

The sensors work by measuring changes in SAW properties across a substrate as the stress within the substrate varies. In its simplest form, a SAW transducer consists of two arrays of thin metal electrodes deposited on a piezoelectric substrate such as quartz. The polarities of the electrodes alternate, and when an RF signal of the right frequency is applied across them it causes the surface of the crystal to expand and contract, generating the surface wave.

The surface velocity of the acoustic waves is one ten-thousandth the speed of light, so for example, a signal at 100 MHz with a free-space wavelength of 3 m would have a corresponding acoustic wavelength of about 30 microns. The upshot is that SAW transducers can incorporate a very large signal bandwidth in a small volume, allowing for high-resolution measurements.  

In a SAW transducer designed for strain measurement, the piezoelectric substrate is attached to the material to be stressed – for example, the output shaft of a dynamometer. What is known as an input interdigitated transducer (IDT) is mounted on one side of the surface of the substrate, and a second, output IDT on the other side of the substrate. The acoustic wave created at the input IDT travels across the substrate to the output transducer, which converts the wave back into an electrical signal. Any changes in the frequency of the mechanical wave between the input and the output can then be recorded.

The transmitted waves’ frequency depends on the distance between the electrodes on each electrode array, and the waves travel at right angles to the electrodes. Therefore, any change in shaft length alters the electrode spacing and thus operating frequency. Tension increases the frequency while compression reduces it.

Using a suitable signal processing device, these changes in the wave’s frequency can be used to determine torque as a function of strain in the surface substrate. In effect, the transducer acts as a ‘frequency dependent’ strain gauge. One of the biggest benefits is that SAW transducers are not affected by strong magnetic fields, as the wave is ‘mechanical’ rather than electromagnetic, making them particularly useful for measuring electromagnetically noisy devices such as electric motors.

The signals generated by SAW transducers can be transmitted using a non-contact capacitive coupling, which also supplies sufficient power for the transducer to operate. The coupler consists of a stator and a rotor that may be a pair of discs or coaxial cylinders. The rotor is mounted directly to the sensing component – in the case of a dynamometer, the output shaft – and rotates with it. The rotor carries a 360° microstrip wired to the SAW sensor, while the stator is mounted statically and also carries a microstrip wired into the signal processing unit.

Although still a relatively new technology, the use of SAW transducers on dynamometers has a number of benefits. The most notable is the ability to conduct contactless measurement, removing the need for devices such as slip rings, which are necessary for signal transmission when using traditional strain gauges. SAW sensors can also be used on any material, unlike magneto-elastic sensors which require the test components to be made from specific metals with suitable magnetic properties. Other benefits include the compact nature of the sensors and the fact that they are inherently rugged, having few parts that can be damaged or displaced.

Overall, this emerging technology is likely to have a number of useful applications both in both the powertrain testing and wider automotive worlds.

Written by Lawrence Butcher

Aspects such as these can be investigated using modern CFD techniques, although this is still a very specialised field and an avenue not available to many engine builders. This has led some tuners and cylinder head manufacturers to investigate methods of physically visualising port flow conditions through a process known as ‘wet flow’ testing; illustrated in the video link below.

The goal with any intake port development, in addition to maximising flow, is to design the port so the fuel is vaporised or mixed with the inlet air stream as uniformly as possible. This is difficult to accomplish because even the tiniest fuel droplet has significantly more mass than air, which means it is much easier for air to turn a corner or move around a valve seat than fuel. Thus it is not uncommon for fuel to fall out of suspension and collect on the port walls, ultimately entering the combustion chamber in a stream.

This can be a particular problem where the geometry of the inlet tract is far from optimal, for example in the case of V8 engines with a centrally mounted four-barrel carburettor, such as those in many drag and stock car series. Obviously this is not conducive to efficient combustion, so any improvements in port design that help keep fuel in suspension will aid performance. This is where wet-flow testing can provide significant benefits, allowing for fuel distribution in the inlet to be assessed and the impact of changes on behaviour analysed.

The mechanics of wet-flow testing are straightforward. The intake port is pressurised and liquid is introduced into the air stream using an atomiser. A clear plastic cylinder sleeve allows the operator to observe and record the behaviour of the fuel droplets in the valve bowl and combustion chamber. The liquid is then separated from the air and captured in a recovery canister.

The liquid that is introduced needs to be of the same specific gravity as the fuel that will ultimately flow through the ports, in order to produce behaviour as close to the real operating conditions as possible. In addition, the liquid contains a fluorescent dye, visible under ultraviolet light, to allow for its movement to be easily tracked and recorded using still or video imaging.

The resulting information can provide an invaluable insight into areas of potential improvement in a cylinder head’s design. For example, one head manufacturer found considerable real-world performance gains between one of its heads optimised using traditional flow bench techniques and one using wet flow. While both heads produced very similar ‘dry’ flow figures, the one developed using wet-flow techniques produced better numbers on the dyno, thanks to improvements in fuel atomisation and distribution.

The testing method is still in its infancy, but beyond basic port development it has also proved useful in the area of carburettor design, particularly with the four-barrel types that are common in US racing. No doubt as more engine builders and tuners are able to adopt wet-flow testing practices, they will find even more applications and make useful gains with components that were already thought to be ‘optimised’.

Example of a two-valve head being tested using the ‘wet flow’ technique 

Written by Lawrence Butcher

Material stress is normally calculated by measuring the specimen’s cross-sectional area and relating this to the measured load obtained from a calibrated load cell while loading the specimen via grips or adapters. The problem with this type of testing though is that it can be difficult to test some samples, particularly those that fail with a lot of force, making it hard to asses dimensional changes accurately or those that need to be tested over a large range, where travel limits on strain gauges can be an issue. It is here that the non-contact nature of a video extensometer can prove effective, as the method can provide measurements up to and including the point of failure, as well as accommodate large or oddly shaped samples. 

Method of operation

A video extensometer captures a continuous image of the specimen under test, using a rig fitted with a high-frame-rate digital camera placed in a fixed location to the specimen. The specimen is marked with two indicator points that contrast with the base material, allowing them to be recognised by post-processing software. It is essential that the markers are as sharp and as different in contrast as possible, to ensure correct automatic target recognition and tracking. The target position is detected at the edge of a contrast transition and is hence not affected by changes in target width.

As the material is elongated, the software measures the pixel distance between the two markers and compares it to a calibration value to provide a strain measurement. The calibration value is obtained by first testing a calibration specimen that has known properties and often accurately etched positional markers. A correctly calibrated video extensometer should be capable of recording at a resolution of 1 µm.

One factor that can affect test results from this system is changes in ambient lighting conditions. If the lighting changes then the contrast between the markers and the test specimen can also change, impacting accuracy. To combat this, filters are used on both the lights used to illuminate the subject and the video capture lens, removing any possible variations.

Beyond simple strain measurements, some video extensometers can also be configured to provide strain distribution analysis. This is achieved by applying multiple markings to the surface of the test specimen and tracking the positional change of these marks using processing software. This information can then be used to ascertain the level of deformation as load is applied, and the nature of the deformation – a useful feature where the dimensional properties of a part under load are important.

Video extensometers are a proven technology and can be very useful where standard testing methods are impractical. They can also be a very cost-effective solution, as systems are available for retrofitting to more traditional extensometers, providing the option of both contact and non-contact measurement in the same machine.

Written by Lawrence Butcher

When it comes to testing and analysing factors such as injector spray patterns and the combustion process itself,  much reliance is placed on CFD simulation. However, there is still no substitute for real-world data, which is where combustion imaging systems come into play. As has been covered in previous RET-Monitor articles, in-cylinder pressure sensors can provide useful information of the combustion process, although pressure data does not deliver any insight into the spatial character of combustion, which can only be obtained by optical analysis, which allows the actual injection and combustion process to be captured visually, enabling the fuel distribution and flame front travel within the combustion chamber to be assessed.

Imaging the combustion process is by no means a new development, and for many years high-speed photography, combined with bespoke test engines featuring viewing windows, have been used for such tasks. While effective, the cost of such development is very high, a factor that is now being addressed with the introduction of non-intrusive imaging systems that can be retrofitted to existing engines.

One such system uses an endoscope to provide images from within the combustion chamber. This can be used to gather both visible and ultraviolet light from the combustion process. Mixture distribution within the combustion chamber can be observed using the visible light spectrum, while the ultraviolet spectrum is used to view the combustion event itself.

Different types of processing equipment are required to record images of these two processes. To capture images within the visual spectrum a CCD (charge-coupled device) very similar to those found in a regular digital camera is used. For in-depth analysis of combustion though, a CCD cannot capture all of the UV spectrum, so instead the images are directed at a series of photomultipliers that can record the entire UV spectrum. The endoscope includes about 10,000 optical fibres in an inner tube, while an outer, concentric tube provides additional mechanical protection; the space between them is filled with compressed air for cooling.

The compact dimensions (4 mm in outer diameter) mean the system can easily be adapted to series engines without any major modifications in engine design. The installation of the endoscope needs only a small access hole in to the combustion chamber, similar to the ones used for pressure transducers. The endoscope is mounted into a probe, which is protected by a fused silica window. The endoscope allows signals to be captured within an observation cone of about 80º. When only images within the visual light spectrum are being captured, the combustion chamber needs to illuminated; this is provided by a secondary fibre optic illumination probe.

The ability to image the combustion chamber on any engine – provided of course that access can be created for the probe – is highly beneficial. Not only can the spray pattern of injectors be assessed but conditions such as knock can also be identified and isolated, which is of particular importance given the ‘lean’ nature of DI engines.

Written by Lawrence Butcher

Sample rate is the measure of the number of readings that can be taken in a set period of time, and is measured in Hertz (Hz), the SI unit for cycles per second. Each input on a logging system will be able to receive and process signals at a particular sample rate, and it is important because it dictates the resolution of the data available for analysis; measuring different aspects of an engine’s operation requires differing levels of data resolution. Ultimately you will only be able to study certain events in detail if you have a sufficiently high sampling rate.

Every conceivable engine parameter can now recorded by a dynamometer’s data logger, covering everything from oil pressure to, if suitably instrumented, cylinder pressures. For some areas, such as oil pressure, the sample rate need not be very high – 50 Hz is more than enough, and even that is probably excessive. However, for detailed engine performance and combustion analysis, much higher rates are needed.

Taking the most basic measurement undertaken on an engine dyno, that of power output, sampling at 100 Hz or preferably 200 Hz is the norm. Between cylinder firings there is an instantaneous drop in torque and crankshaft rpm, which is indiscernible to a racecar’s driver but is important from a performance analysis perspective. Taking an engine rotating at 6000 rpm and a data sampling rate of 50 Hz, you will only be recording torque every other rotation of the crank.

The biggest impact of a low sample rate such as this is inaccuracy in the averaged power curve of the engine produced using the gathered data. Some data samples will be in sync with the firing of the plugs, others won’t. By doubling or even quadrupling the sampling rate, sufficient data is acquired to even out the sample range and provide a more reliable average of the power curve – not to mention providing more data points for in-depth analysis of transient engine behaviour.

Moving on to more complex engine testing, sample rates become even more important. For example, if you want to take four samples per crankshaft rotation, of combustion pressure for example, at the same 6000 rpm then you need a minimum sampling rate of 400 Hz. Now, very few race engines operate at such low revs, and logic dictates that to study the same data at 12,000 rpm, you are going to need a sample rate of 800 Hz.

So if, for example, one was looking at combustion pressures in an engine running at 12,000 rpm, a sample rate of 800 Hz would provide sufficient data over a sample of 500 combustion cycles to give a good average of combustion pressure throughout a combustion cycle. Obviously, the validity of such averaging also depends on the accuracy of the crank position measurements.

Two other factors must be considered when choosing a data logging system – the number of channels and their sampling rates. For example, if you want to log cylinder pressures on a V8, you are going to need at least eight channels capable of recording at a sufficiently high rate. Consequently, you also need the ability to process all these channels at their maximum rate simultaneously. The reason high-end equipment carries a hefty price tag is functionality such as this, but the insight it provides into engine performance is exceptional.

Written by Lawrence Butcher

In the past, comprehensive simulation – for example of gas flow within an engine – was the preserve of high-end commercial software packages. However, this situation is now changing thanks to the appearance and subsequent development of software packages aimed at users who may not have the resources of OEMs or works racing teams. These programs fall broadly into two groups: ‘basic’ simulation packages, which can provide a close approximation of engine performance, and more complex ‘open source’ packages.

The basic packages can range from free-to-use trial version of mainstream software to dedicated commercial packages that do not carry the hefty licence fees of more powerful programs. The open source systems tend to be free to use and are potentially very powerful; however, they usually lack the user-friendly interfaces of commercial systems, so it takes longer to gain useful results from such software. The one major advantage though of this type of software is its open source nature, enabling users who understand it to expand the functionality using their own code. There are even open source CFD solvers suitable for simulating complex in-cylinder gas flows in three dimensions – something that is definitely not included in any ‘basic’ simulation software.

Despite their limitations, even basic engine simulators can provide considerable benefits to engine builders, both from a component optimisation and a time/cost-saving perspective. These types of simulation package tend to focus on a particular area of an engine, helping to keep complexity (and thus the aforementioned cost) low. So which aspects of an engine can be modelled with some of the more commonly available systems?

One of the most frequently investigated areas is camshaft and valvetrain design. Most simulation software will allow users to model lobe profiles and assess their effect on valvetrain motion. Also, depending on a particular program’s capability, the results obtained for such simulation can show factors such as contact stress and oil film thickness at the lobe-lifter interface, as well as more in-depth insights into the valvetrain’s behaviour. For example, the impact of differing profiles on valve spring specification can be determined or, if looking at a pushrod-actuated valve, the impact of pushrod/rocker geometry on efficiency can be examined. The results gained from dedicated valve simulation packages can also often be used as inputs for overall engine models.

Similar to valvetrain simulation, there are also a number of dedicated cranktrain analysis packages available commercially. Beyond the basic requirements for assessing crank, rod and piston movement, some packages allow for other useful factors to be assessed. Wrist pin and rod bolt sizes can be determined, for example, with some software even providing preset piston loadings to work from. In more advanced packages, the con rod design can also be scrutinised for fatigue resistance under particular operating conditions.

Provided that the input data is reliable, even the most basic simulation system can prove very useful for engine development. Much as is the case with the use of CFD for car aerodynamic development, being able to gain an estimate of which combinations of components will produce the best gains – without having to physically build the components – greatly reduces the resources needed for engine development.

While the packages that can be classed as ‘affordable’ will not produce absolutely perfect results, they are still accurate enough to be useful development tools. The free-to-use, open source packages also have the potential to be very powerful, provided that the user has the correct expertise. However, both approaches give engineers tools that were previously only available to high-end manufacturers with deep pockets.

Written by Lawrence Butcher

The mounting of cylinder pressure sensors can present engineers with a number of choices and challenges. The easiest course of action, and an especially useful one if an engine is not a purely experimental unit, is to use a sensor integrated into a spark plug or glow plug. The recent appearance on the market of portable combustion analysis systems has opened up the possibility for using plug-based sensors in trackside applications. This is an exciting development and brings a level of analysis to the trackside that was previously available only in the dyno room.

Conversely, if it’s possible to modify a cylinder head to accept sensors then dedicated pressure transducers can be used. While there are only a few different types of plug-based sensor on the market, the range of non-plug based sensors is much wider, presenting a greater choice in terms of sensor specification. These sensors will invariably be mounted in threaded holes into the combustion chamber, and are generally reserved for use on dedicated test engines.

Recent developments have also seen the appearance of a third mounting method – the incorporation of pressure transducers into a multi-layered head gasket. Here, probe-type pressure sensors, complete with signal amplifiers, are built into the structure of a cylinder head gasket, with one sensor per cylinder. This provides a permanent pressure-sensing capability without the need for the structure of the cylinder head to be altered.

Sensors designed for measuring in-cylinder pressures generally use a piezoelectric sensing element. When pressure, force or acceleration is applied to a quartz crystal, a charge is developed across the crystal that is proportional to the force applied. The fundamental difference between these crystal sensors and static-force devices such as strain gauges is that the electric signal generated by the crystal decays rapidly, making such sensors unsuitable for measuring static forces or pressures but useful for dynamic measurements.

The use of piezoelectric crystals in many high-end pressure sensors means that vibration or shocks near the sensor can disrupt the output signal. Obviously this is undesirable, and sensor manufacturers take steps to mitigate the effects.

Different sensors have varying degrees of sensitivity to acceleration (caused by the aforementioned vibrations). For example, certain transducers use water cooling to increase their sensitivity and durability; however, this also increases their sensitivity to acceleration. This means that for pressure-measuring applications in close proximity to components that create shockwaves – notably the intake and exhaust valves of high-rpm race engines – non-cooled sensors with low acceleration sensitivity are used.

There are, however, sensors that can provide the signal resolution of a water-cooled sensor yet still be used in high-acceleration areas. These sensors are designed to compensate for shock- or vibration-induced signal variations. One method of achieving this compensation is by adding a seismic mass and a separate ‘compensation crystal’ of reverse polarity to the sensor. These components are scaled to cancel out exactly the inertial effect of the masses (the end piece and diaphragm that make up the sensor body) which act on the pressure-sensing crystal stack when accelerated.

Pressure sensors are a key tool in engine development, and manufacturers of such devices recognise that motorsport applications put demands on their equipment beyond those presented by most OEM engine manufacturers. To this end, most sensor suppliers will produce sensors specifically for the motorsport industry, and can provide advice on the correct specification to choose.

Written by Lawrence Butcher

The most common form of strain gauge found in motorsport is an electrical resistor constructed from a foil section bonded to a dielectric backing. When the gauge is stretched or compressed, its electrical resistance varies, and this change can be related to the amount of force on the material. However, the variation is very small, and to measure the strain accurately requires precise measurement of very small changes in resistance. To measure such small changes in resistance, strain gages are almost always used in a bridge configuration with a voltage excitation source. The general configuration is a Wheatstone bridge [Fig. 1] which consists of four resistive arms (the ‘bridges’) with an excitation voltage, V_EX, applied across the bridges.

Fig. 1 - schematic layout of a Wheatstone bridge

From this schematic, it is apparent that when R₁/R₂ = R₄/R₃, the voltage output V₀ is zero. Under these conditions, the bridge is said to be balanced, and any change in resistance in any arm of the bridge results in a non-zero output voltage. Therefore, if you replace R₄ in Fig. 1 with an active strain gauge, any changes in the gauge’s resistance will unbalance the bridge and produce a non-zero output voltage. It is then possible to create an algorithm in the data acquisition software to convert this output into a unit of force. The basic principle is expressed in the following equation:

This is a very simplified example of the operation of a strain gauge, and careful consideration has to be paid to external factors that can affect the readings obtained. The most significant of these is the effect of temperature on the electrical resistance of the gauge material. Most conductive materials will change electrical resistance as ambient temperature changes, which can lead to false readings.

The solution is to use two strain gauges in each bridge, one of which is a ‘dummy’ gauge placed transversely to the applied strain direction, meaning the strain has little effect on it. The reading from this gauge can then act as a reference for the variation in resistance with temperature, which can then be subtracted from the reading from the gauge fixed along the axis of strain. Both will be affected by temperature in the same way, with the difference between the readings being a measure of the strain. It should be noted though that while this minimises temperature effects it does not completely eliminate them, and other factors also need to be taken into consideration, such as the resistance in any cables running to a signal amplifier and electronic noise from nearby sources.

The advantage of strain gauges is that they can be incorporated into a wide variety of components made from a range of materials, for example integrated into the end of a suspension pushrod as shown in Fig. 2. Often these components are designed to incorporate a strain gauge in their structure, reducing any possible compromise on the part’s intended function. Overall, such components provide engineers with an invaluable insight into the forces components are subjected to when in use.

Fig. 2 - A strain gauge incorporated into the end of a Formula One suspension component (Photo: Lawrence Butcher)

Written by Lawrence Butcher

The vast majority of conventional systems for measuring torque operate by measuring the torsional deflection induced by the applied torque, by one of two methods: measurement of twist angle and measurement of surface strain changes.

The twist angle method of torque measurement generally requires a portion of the dynamometer’s output shaft to be reduced in diameter, to allow it to twist under load; a pair of toothed discs are then attached at opposite ends of the slim portion. The twist angle can be determined from the phase difference between magnetically or optically detected tooth/space patterns on each of the discs. Because the discs are rotating at the same rate as the output shaft, this allows for the torque to be measured at each revolution of the engine; however, the addition of extra toothed wheels at different angles could refine this measurement.

For the second method, measurement of surface strain can be achieved through the use of piezoresistive strain gauges attached to the shaft. These strains are generally too small to be accurately measured directly, so common practice is to use four gauges arranged in a Wheatstone bridge circuit. With rotating shafts, a coupling method such as rotary transformers, slip rings or local telemetry is required to feed the excitation current to the gauges and to acquire the signal from the bridge circuit in a non-contacting manner.

These sensor types are more than adequate for most dyno operations; however, their resolution is not sufficient for measuring very small torque differences, with their frequency response rarely exceeding 1 kHz. For example, if an engine is being powered by an input torque of 100 lb-ft, and the frictional losses from the bearings accounts for only 1% of the overall losses, then an accuracy of less than this is required from the instrumentation to measure the effect. This is where the latest generation of magneto-elastic torque sensors come in to play.

Magneto-elastic torque sensors produce signals that are a function of torsional stress, not strain. As a result they are generally much stiffer mechanically than the conventional elastic torque sensors. They also offer a far higher frequency response, typically of the order of 2-4 kHz. Measuring surface stress by magneto-elastic methods also provides a non-contacting system for measuring torque in a more compact construction than those required for either the twist angle or surface strain methods.

These sensors can be broken down into two separate groups, which measure magnetic quantities related to the surface shear stress in different ways. The first method measures magnetic permeability changes in the shaft surface caused by the stress-induced magnetic anisotropy affecting the permeance of a magnetic flux path, and uses a magnetising source and a pick-up (sensing) coil. These are referred to as PB Type 1 sensors. The second method is referred to as PB Type 2, where the stress-induced magnetic anisotropy causes a permanently magnetised magneto-elastically active member to generate a measurable magnetic flux.

A key advantage of a Type 1 sensor over traditional sensing methods is wireless transduction - removing the need for physical contacts with the rotating member - combined with mechanically robust construction. However, despite their various benefits, Type I permeability-based magneto-elastic torque sensors suffer from a number of disadvantages that limit their use in a testing environment. These problems derive from the fact that the variable being measured, permeability, does not depend solely on the applied torque. In any one material composition, even in a controlled environment, permeability can vary with both temperature and magnetisation. The result is that in many real-world environments, the changes due to these factors can exceed the changes in permeability that are a function of torque, making the measurements useless.

Type 2 sensors are much more suitable for use in real-world applications. They have many of the benefits of Type 1 sensors but also overcome most of the problems. The sensors can be constructed either with a thin ring of magneto-elastically active material rigidly attached to a shaft, or by using a portion of the shaft itself as the magneto-elastically active element. In response to the magneto-elastic energy associated with the bi-axial principal stresses by which torque is transmitted along the shaft, each moment will rotate towards the nearest positive principal stress direction and away from the nearest negative principal direction.

This re-orientation of the originally circular magnetisation results in a net axial magnetisation component. The divergence of this component at the edges of the polarised bands is the source of a magnetic field in the space around the shaft, which can be readily measured with one or more magnetic field sensors. This means that the problems associated with the measurement of permeability are not encountered, allowing for very accurate and repeatable readings to be taken. Additionally, the polarised bands can be easily incorporated into the output shaft of a dyno, or even into the driveshafts of a whole vehicle.

Advances such as these allow far smaller differences in performance to be accurately measured on dynos, an especially important factor in race series where power outputs can vary by factions of a percentage point.

Fig. 1 - The basic layout of a shaft for use with a magneto-elastic torque sensor

Written by Lawrence Butcher

The dictionary definition of accuracy is: "The degree to which the result of a measurement, calculation, or specification conforms to the correct value or a standard", whereas for repeatability it is: "The ability of a measuring instrument to repeat the same results during the act of measurement".

From these definitions it is quite possible to surmise that if a machine's results are described as repeatable, it does not necessarily mean they are accurate, because no reference to accuracy is made. It is perfectly feasible to have a machine that reads 1 mm as 5 mm every single time; it is repeatable, but definitely not accurate.

Accuracy in engineering terms is gauged in terms of +/- tolerance. For example, a micrometer can be accurate to within +/- 0.005 mm, so measurements taken will never be more than 0.005 mm from the absolute measurement. The smaller the tolerance, the higher the accuracy, and the same principle can be applied to weights, power outputs and so. So from this, it is safe to presume that an accurate machine is also a repeatable one, otherwise it could not be classed as accurate.

As a general rule, accuracy is achieved by measuring something as directly as possible. For simple distance or clearance measurements, this is easily achieved by using a micrometer or verniers. For something like the power output of an engine though it is a little more complex. The most accurate method is to measure from the crankshaft - provided of course that the measuring equipment used operates to tight tolerances - as this is the most direct connection possible. However, if you want to measure the power output at the wheels, taking into account the parasitic losses present in the drivetrain, the most direct method of attachment would be to the hubs. If the drivetrain is already installed in a vehicle, removing the wheels from the equation increases the accuracy, as factors such as tyre carcass distortion and grip between the tyres and dyno rollers will not come into play. For every link in the measurement chain there will be an added tolerance, and accuracy will fall.

Both accuracy and repeatability can be affected by factors in the measurement chain - or those that are external to the chain but have an impact on the end result - and which vary depending on testing conditions. These can include, but are not limited to:

Environmental - inlet air temperature, measurement point, atmospheric pressure, weather/altitude, relative humidity and geographic location/season

Other - fuel type and quality, octane number/energy contents, fuel temperature, drivetrain temperature, engine cooling fluids, gearbox oil and rear axle oil

While it is nigh-on impossible to control all these factors precisely, even given a state-of-the-art test cell, variations must be accounted for in order to achieve both accurate and repeatable results. This is where correction factors come into play, and most dynamometer systems will have these calculations built into their processing systems. However, this introduces yet another link in the measurement chain, with the accuracy of components such a barometric pressure sensors influencing final power output figures.

Written by Lawrence Butcher

These days, FEA simulation removes the need for a large amount of physical testing, but simulation packages still need data, and some people cannot afford software capable of modelling composite structures. Properties for metallic components can be readily found in the literature from suppliers, with variations being present depending on factors such as heat treatment. However, for composites the variation in mechanical performance relies on a far greater number of factors such as the number of plies used, resin density, resin type, orientation of fibres and so on. It is therefore often necessary to test specific combinations of fibres to ascertain the best construction method for a specific task, which is achieved in a number of ways.

The mechanical testing of composite structures to obtain parameters such as strength and elastic limits is a time-consuming and often difficult process, and can be somewhat simplified by the testing of simple structures such as flat coupons - small sections of the material in question. The data obtained from these tests can then be directly related with varying degrees of simplicity and accuracy to any structural shape.

The make-up of a test coupon usually consists of a central section, referred to as the 'gauge length', which is the area where failure is expected to take place, and two end sections where the coupon is clamped to the testing apparatus. For composites, the end pieces will often include aluminium tabs to prevent the composite being crushed by the test machine's grips. The surface of the test coupon also needs to be polished to remove surface flaws that could lead to stress risers and premature failure.

A coupon of this type can then be used for a basic tensile test, using standard equipment, the most common being the universal testing machine. This has two crossheads - one is adjusted for the length of the specimen and the other is driven to apply tension to the test specimen, and either hydraulic or electromagnetically powered machines are available. The test process involves placing the test specimen in the machine and applying tension to it until it fractures. During the process, the elongation of the gauge section is recorded against the applied force.

Basic tensile testing then is easy enough, but often a component will be subject to both tensile and compression forces and be asked to operate in shear. It is therefore often necessary to undertake testing to ascertain a specific lay-up's performance in shear, specifically the point at which the separate plies begin to delaminate. This is the point where the resin bonding the layers of fibres together fails, leading to the catastrophic failure of the part. It is this bond that is improved by processes such as autoclave treatment, which produces a far more homogenous material structure than would be possible otherwise.

There are many different types of shear test methodologies, but one of the most common for measuring shear delamination is the short-beam shear test [Fig. 1], where a small specimen (less than 30 mm long) is loaded in three-point bending until a delamination forms in the centre plane at one end of the specimen.

These are just a few examples of the many tests that can be carried out, but the results they can generate are invaluable in ascertaining the integrity of different composite construction techniques.

Fig.1 - A three-point short-beam shear test

Fig. 2 - A universal testing machine for tensile testing operations

Written by Lawrence Butcher

Engine testing is essentially a scientific experiment and, as such, as many factors as possible need to be tightly controlled to ensure reliable and repeatable results. The biggest factor is the environment within the test cell.

This may seem initially like a simple matter of installing a basic HVAC system to ensure the temperature and humidity remains stable, but a common mistake made when building a test cell is the level of heat rejection from even a small-output I/C engine. For example, an engine delivering 100 flywheel horsepower to the absorber (74 kW per hour) radiates up to 32 kW per hour into the cell from its exhaust, cooling radiator and block surfaces.

Even at idle, when zero crankshaft horsepower is available, significant heat energy must be dealt with, as the engine is still burning fuel in overcoming its internal friction. As a conservative estimate, a test cell with the exhaust routed outside the cell will need about 2000 cfm of air-moving capability per 100 bhp of engine output.

In a well-designed cell the oil and water coolers will also be located outside the cell, which will reduce the level of radiated heat, but the volume of cooling air required will still be substantial. Other equipment such as double-walled and insulated exhaust systems can also be used to reduce the level of radiated heat. Many manufacturers of test systems will also be able to supply suitable HVAC equipment, with the specification based on the specific requirements of a particular dynamometer.

The extraction of exhaust gases also needs to be considered. Some basic cells will use a hood that's very similar in concept to those found on a domestic oven (but obviously much larger) and that draws air and exhaust fumes out of the test cell. While simple, this system is not ideal and cannot ensure that all fumes are extracted; a far better solution is to use a sealed extraction system. The use of a hood also prevents the use of jibs or cranes to move engines and equipment into the test cell.

Carbon monoxide inside the test cell is another important consideration. Inadequate airflow, improperly sealed construction or positive pressure inside the test cell will result in CO leaks. Since it is odourless, colourless and deadly at low PPM levels there should be a small amount of negative pressure (0.25 in of water) in the cell at all times during testing, to prevent CO from being pushed out into other areas of the facility.

CO present in the combustion air will also displace oxygen and degrade the power output of the engine under test. Exhaust extraction systems that connect directly to the header pipes can be used to extract exhaust gases; however, the system introduces negative pressure into the exhaust side of the engine and ultimately affects performance, air-fuel ratio tuning and timing aspects of the engine while it's on the dynamometer.

One final factor that needs to be considered is the control of these ancillary systems. Having four or five different control panels dotted in and around the test cell is not an ideal solution. It makes more sense for all the HVAC and other services to be handled from the dyno control room where they can be controlled and monitored by the dyno operator without distracting from the primary task of running the engine test.

Fig. 1 - Careful planning of the layout for a test cell will make testing operations easier to control for the dyno operator."

Written by Lawrence Butcher

Generally speaking, most dynamometer manufacturers will supply sensors and a data management interface with their test systems, often featuring their own proprietary data analysis software. The benefit of such a system is that set-up and integration of the components is simple. Often the data acquisition system and dyno controls will be integrated into a single console, with a Windows-based PC to deal with software operation.

However, depending on the manufacturer, the capability and versatility of these systems can be somewhat limited. The most basic instrumentation systems necessary to record general engine operating parameters will include the following sensors:
" Brake torque
" Engine speed
" Engine power
" Exhaust gas temperature
" Intake air pressure and temperature
" Coolant temperature
" Oil pressure and temperature
" Fuel and air mass flow
" Air-to-fuel ratio

Although most of these sensors are already fitted to many engines, for testing purposes their roles will be duplicated by those linked to the dyno's data system. This allows factors such as the ECU's signal conditioning performance to be analysed, without the risk of the results being compromised by any potential interference from the data acquisition system.

When testing becomes more involved, for example once parameters such as cylinder pressures need to be measured, the number of data inputs required increases considerably. A fully instrumented test bed can have anything up to 150 sensors, all recording in real time and, depending on their role, sometimes needing to be logged at a very high rate. There will also be a mixture of digital and analogue signals that need to be handled, with each analogue channel requiring a signal conditioner to process its output before it is received by the data analysis system.

Again, most dynamometer manufacturers can supply expansion boards for their systems, which can often be daisy-chained together to increase capacity further. At the same time, the data analysis programs are also scalable, allowing users to create new channels and determine the values of the data shown. A typical expansion board will feature a mix of different input and outputs, such as dedicated pressure, load cell and frequency inputs as well as outputs for controlling components such as throttle servos.

Once you have all of the required inputs and outputs, the challenge then becomes one of organising the mountains of data that can be gathered. It is here that the usability of a given software analysis package comes into its own. The most highly developed systems allow for many different channels to be arranged logically, and information from different test sessions easily overlaid for comparison. Generally, there are also facilities for creating custom maths channels to calculate specific parameters, which may be unique to a particular test.

Overall, whatever you may want to investigate - be it the simple optimisation of an ECU or a comprehensive engine endurance test - there are sensors and data systems out there that allow you to do so.

Fig. 1 - A typical display screen from a dynamometer control and logging console, showing key engine operating parameters and channels of logged data

Written by Lawrence Butcher

It is often thought that this sort of analysis is the preserve of manufacturers building clean-sheet engine designs, but the benefits of combustion analysis make it relevant even to small-scale race engine builders and tuners. The number of parameters that can be investigated given the equipment available on the open market is legion, although for the purposes of this article we will look at just one of the most useful, the pressure-volume (PV) loop. The reason is twofold - it provides a very clear graphical representation of a cylinder's behaviour, and any changes or corrections made to an engine are immediately visible.

Equipment
The production of a PV loop for an engine requires at least one pressure sensor per cylinder, a means of accurately recording crank position, and computing resources to gather and present the recorded data. For purely experimental engines, pressure sensors can be built into the cylinder head architecture, but a far more practical approach is the use of spark plug sensors. These are specially designed plugs that incorporate a high-speed piezoelectric pressure sensor that can record the pressures generated in the cylinder.

There are several companies producing such sensors, which come with features such as 'flame guards' to reduce errors produced by thermal shock experienced during the combustion process. Crank position can be measured using an engine's built-in sensors, but these generally lack resolution so specialised sensors are preferable. Most companies specialising in such test equipment will normally be able to supply an integrated data acquisition system complete with plug-and-play interfaces for sensors and dedicated software to process the large volume of data generated.

The PV loop
A PV loop is a line graph that displays in-cylinder pressure at all points in a particular cylinder's cycle. This simple graphical representation provides an invaluable insight into the combustion behaviour of an engine as the cylinder volume changes with the crank stroke. This representation also allows for very easy comparison of different engine configurations.

Fig. 1 - The PV loop representing the ideal Otto Cycle

Fig. 2 - A typical PV loop for a spark-ignition engine

Fig. 1 shows the ideal PV diagram for an Otto cycle engine, while Fig. 2 shows the PV diagram for a typical combustion engine, with the key events annotated. It can be seen from the comparison of the two diagrams that the ideal cycle has no pumping loop (grey shading). This indicates that the gas exchanges from the intake manifold into the cylinder, and from the cylinder into the exhaust manifold after combustion, should ideally occur without any associated losses. In practice, this can never be realised, and work is always expended drawing air into the cylinder and expelling exhaust gas from the cylinder.

The work output of an internal combustion engine is indicated by the difference in area contained within the power loop and the pumping loop. This means that improvements in an engine's efficiency can be identified by an increase in the size of the power loop relative to the pumping loop. In simple terms, if the combustion efficiency of the engine is increased, this will be represented as a greater area within the power loop, with no change in the pumping loop.

However, useable power can also be gained by reducing pumping losses, which can be identified by a reduction in the area of the pumping loop while the power loop remains constant. With an engine instrumented to record this information, the effects of changes to factors such as valve events, ignition timing and inlet/exhaust design can be compared very easily.

Producing PV graphs for individual cylinders also allows each cylinder to be compared with the others, and treated as an individual engine, the goal being to make all cylinders as efficient as possible. This is the key benefit of combustion analysis: where a dyno allows you to look at only the overall output of an engine, individual cylinder analysis gives a far more complete picture of an engine's performance.

In future instalments of RET-Monitor we will look at the individual areas of the PV graph in more detail and relate them to potential performance gains, as well as address some of the other aspects of combustion analysis.

Fig. 3 - The pumping section of PV graph for a six-cylinder engine. Each line represents the pressure trace for a separate cylinder. Note the variation between the values for each cylinder

Written by Lawrence Butcher

Water brakes use water flow proportional to the applied load to create resistance to the test engine's output. A controlled flow of water through the inlet manifold is directed at the centre of the rotor in each absorption section; the water is then expelled towards the outside of the dynamometer body by centrifugal force. As it is directed outwards, the water is accelerated into pockets on the stationary stator plates, where it is decelerated. This continuous acceleration/deceleration of the water creates the applied load to the motor. Through this transfer of energy, the water is heated and discharged.

However, water brakes are generally used only for steady-state measurements, as opposed to dynamic testing, because of the non-linear torque characteristics of the water brake system and the problems this creates in terms of regulating load control. Water brakes can also be prone to instability as a result of high-frequency pressure pulses created by the movement of the water, which causes problems such as excessive vibration and results in torque and load oscillations.

One method of controlling or limiting the effect of these pulsations is through improved rotor design. Advances in CFD (computational fluid dynamics) over the past decade have allowed for far more accurate modelling of the fluid behaviour inside the dynamometer. This in turn allows manufacturers to refine the design of the rotor and stator components to reduce the occurrence of the high-frequency pulses, or ensure that they occur outside the dyno's normal operating window. The main benefit of these refinements is that they allow water brakes to run at higher revs while still retaining accuracy and repeatability.

Beyond simply refining the design of the internal components, other approaches have also been used to improve the stability of water brakes. One notable method was pioneered by General Electric, to address stability problems encountered when testing very high output motors. The solution uses the injection of high-pressure air into the brake housing, which has the effect of raising the partial pressure within the brake, preventing or dampening any high-frequency pulses. The volume of air injected varies depending on the load applied to the dyno, with the volume being determined based on a mathematical model specific to the type and size of water brake.

While this approach no doubt improves the performance of the water brake system, it also brings a considerable increase in complexity, not only in the construction of the brake itself but also the control systems, with a separate microcontroller needed to service the air injection system.

The same developments in CFD that have aided improvements in stability are also increasing the potential uses of water brakes in a dynamic role. By being able to accurately model the torque characteristics of the brake, which are dictated by the positions of the inlet and outlet water valves, the non-linearity of the torque curve can be accounted for. This data can be used as a reference map that is then used by a feedback controller to accurately govern the torque of the water brake as an engine is run dynamically. Note though that other factors such as inlet and outlet water temperatures also need to be accounted for, as these have a considerable effect on the torque characteristics.

These developments are in their early stages, but they go to show that although the water brake may be relatively old technology, it still has potential as a versatile engine test bed.

Fig. 1 - Map showing water brake torque in relation to inlet and outlet water valve position (From "Inverse Torque Control of Hydrodynamic Dynamometers for Combustion Engine Test Benches", by TE Passenbrunner, M Trogmann and L del Re, American Control Conference 2011)

Written by Lawrence Butcher

To recap, there are two main types of chassis dynamometers (dynos) in widespread use these days - inertia and eddy current.

Previous issues of RET-Monitor have covered the operation of both types of dyno; the important factor here is that even if all other conditions are equal, the power numbers produced by one over the other will invariably differ. This is very important to consider, especially if circumstances dictate that a test programme takes place in multiple locations, on different dynamometers. Engineers have long recognised these discrepancies and thus 'correction factors' are applied to raw dyno data, the most common being SAE J1349 and SAE J607, which take into account factors such as humidity, barometric pressure and air temperature.

When comparing dyno results, always ensure that the numbers are corrected to the same standard. However, even with proper application of these correction factors, atmospheric conditions can play an additional role, having an effect on ignition timing in modern ECU-controlled engines.

The SAE correction factors account only for the change in the density of the air due to atmospheric conditions, and cannot account for things like engine borderline spark sensitivity. As inlet air temperature increases, an engine's ECU will generally retard spark to prevent detonation using the particular octane of fuel for which it was calibrated. Correction factors cannot account for this because different engine designs can have different spark sensitivity and different sensitivity of torque relative to ignition timing. Beyond the factors that can have an effect on an engine's power output, there are also a number of factors related to the nature of chassis dynos and specific vehicle drivetrains that can skew figures.

Generally, chassis dyno tests are performed using the 'roll-on' method, where the vehicle's drive wheels are accelerated in a particular gear from low to high speed (generally to the rev limit of the engine) in one continuous sweep. Because of this constant acceleration, engine and transmission inertia, drive wheel inertia, tyre characteristics, gear ratio and axle ratio can all affect the final measured horsepower. For example, a heavier wheel will take more torque to accelerate at the same rate as a lighter wheel, so heavier wheels will tend to reduce the measured wheel horsepower.

This is countered by allowing the vehicle to 'coast down', which involves allowing the wheels to slow down with the clutch depressed; this isolates the transmission drag from the engine drag. The dynamometer control systems can then subtract the known inertia of the rollers from the overall inertia during coast down, thus ascertaining the power loss through the transmission. However, this figure can also be subject to error; for example, if a vehicle is strapped down to the dyno under a different tension from run to run, the resistance of the wheels on the rollers will vary.

While there is no doubt that chassis dynos are an invaluable tool for the engine tuner, there are a lot of factors to consider when trying to obtain comparative data. For meaningful results, it is imperative to ensure that as many factors as possible are kept consistent.

Fig. 1 - A NASCAR Nationwide car on a chassis dyno. NASCAR teams will often test a car before and after a race to check for performance drop-off

Written by Lawrence Butcher

NASCAR Cup racing is some of the most closely fought motorsports in the world, and any performance improvement, no matter how small, can prove decisive. This is especially the case on the superspeedways, where the mandated restrictor plates equalise engine performance. With gains ever harder to find, teams have been looking to the transmission for an improvement in performance. Being able to measure these improvements accurately is a key facet of any development project. While transmission dynamometers are fairly commonplace, those powerful enough to replicate a 750 hp racecar - or sensitive enough to track tiny efficiency gains reliably - are not.

This has led one company to develop its own system that is tailored to the needs of NASCAR. Unlike most dynamometers, which rely on an electric motor to operate, this system uses a series of hydraulic pumps; these have excellent power density and can be operated much like a DC electric motor.

The use of an oil-based hydraulic medium also allows the dyno to simulate most of the loadings a transmission is likely to experience. There is a series of input and output pumps, with the power generated by the output pumps being recycled to drive the input pumps, greatly reducing the power consumption of the whole system. The system can also be instantaneously reversed in order to simulate overrun scenarios, while built-in accumulators allow for transient power scenarios.

The final and potentially most important feature of the rig is its ability to tilt, simulating cornering g-forces - vital given the quantity of time a NASCAR spends turning. Engineers can then asses oil distribution under load when the car is on steep banking and use this data to improve the internal casing design. Measurement of input and output torques is achieved with a series of torque transducers that provide a level of sensitivity far beyond that available on commercial dyno systems of the same power output.

Since the facility opened there has been a steady stream of teams using the services on offer to gain a greater insight into transmission performance. Most of the work undertaken is in assessing new components in an attempt to gain greater efficiency, with improvements of even 0.5% being seen as a performance advantage. However, it does not stop there, and many teams use the facility to run endurance tests on components, at a far lower cost than track testing or running an entire vehicle on a chassis dyno. Factors such as gear vibration have also been investigated, using accelerometers installed on the transmission casings.

Fig. 1 - Ring and pinion gears are just some of the components that have been improved using data from a transmission dyno

Written by Lawrence Butcher

The use of lasers to measure the movement of the valvetrain and other components is fairly well documented, however a brief summary is beneficial. In essence, a large electric motor is used to spin an engine through its operating rpm range, while a laser beam is used to measure factors such as valve lift and rocker arm deflection.

The main system available on the market uses a single laser to achieve these measurements, but one particular NASCAR engine operation found this set-up to be insufficient for its needs. The specific areas the company wanted to study were valve movement, in order to validate its simulation models, and piston rock, to identify areas for improvement in piston skirt design.

In particular the engineers found that the resolution provided by the single laser, which was accurate to 20 microns, was too low. The solution was to replace the single laser with a twin array, with each of the new lasers accurate to 10 microns. For the valvetrain testing, the twin laser array was directed through a cutaway section of cylinder block at the inlet and exhaust valves, which were covered with a thin film of retro reflective tape to improve the strength of the return signal.

A baseline reading for valve movement was taken at 3000 rpm, which allowed an accurate trace of the valve movement to be created and where factors such as bounce on closing would not be present. With a baseline established, reading could then be taken at high rpm, up to the 9000 rpm redline, to obtain an accurate record of the influence of valve spring harmonics, as well as identifying any undesired traits such as valve 'loft' or 'bounce' on closing. The end result was that potential problems, which would not usually become apparent until component damage occurred, could be addressed, and the process is now used to validate new valvetrain designs.

The second area to be investigated was piston rock, a particularly relevant issue in NASCAR Cup engines, where generous piston clearances are the norm (in the region of 0.010 in). The same twin laser array was used, although here the cylinder head was removed to allow access to the piston crown. Retro reflective tape again helped to improve the strength of the return signal. The lasers were pointed at opposite sides of the piston crown, and the engine cycled through its operating rpm range.

While undertaking this testing it was found that oil mist from the crankcase degraded signal strength; the solution was to use a fume extractor nozzle placed close to the bore. As a result of this testing cycle, the degree of piston rock could be accurately assessed at each stage of the crank's rotation, providing data that allowed the piston skirt design to be modified in order to reduce frictional losses.

Fig. 1 - The twin laser array used for measuring valve and piston movement
Fig. 2 - Retro reflective tape was used to aid return signal strength for the lasers

Written by Lawrence Butcher

The advantage of an electric active dyno is clearly its ability to motor, allowing it to drive the engine and simulate conditions such as transmission drag, gearbox

downshifts and so on. However, this capability can also be used when the engine is not running to narrow down the source of losses.

Speaking to one engineer from a well known testing and development company, it became clear that the average testing process for a new engine went well beyond basic drive cycle simulation. The engineers would regularly test the engine with components such as the cylinder heads or pistons and rods removed, to ascertain the losses related directly to these components. Not only does this avoid the need for specific test rigs to be built, but the engine can be motored with all other parameters - coolant temperature, oil temperature, oil flow and so on - at the correct levels, thanks to the ancillaries associated with the test cell.

In recent years, this type of testing has been aided by the development of ever more accurate torque sensors, which are vital to obtaining correct data. After all, if an engine is being powered by an input torque of 100lb-ft, and the frictional losses from the bearings accounts for only 1% of the overall losses, then an accuracy of less than this is required from the instrumentation to measure the effect.

The latest generation of magneto-elastic sensors are capable of providing the level of accuracy required for these tasks. The non-invasive nature of their operation, which does not interrupt the torque flow from the dyno in any way, means they can be installed much closer to the prime driver input than usual. This allows detection of transient torque peaks of significantly higher amplitude than those normally recorded with conventional dyno torque sensors such as telemetric strain gauges.

Moving off at a slight tangent, if accurate testing of this type is to be conducted, the overall installation of the dynamometer is of utmost importance. The often very small nature of the losses being measured means factors such as drive shaft misalignment are more relevant than ever, with any offset translating into additional drag in the system, which will consequently be picked up by the torque sensor.

During my conversation with the development engineer another, slightly unexpected benefit of active dynos came to light - fault finding. Imagine you have hooked everything up in the test cell, yet the engine simply refuses to fire. Enter the ability of the dyno to motor the engine.

Most modern race engines are heavily instrumented from the factory with an array of position sensors, pressure sensors and anything else the ECU needs to run the engine. Provided that the ECU is working, an active dyno allows for the engine to be 'run' without firing and all the sensors interrogated for faults using the test bay instrumentation. It's a far quicker approach than having to measure or replace components individually, giving more time for constructive testing.

Next month I will look at dynamometers designed specifically for assessing and measuring reciprocating components, above and beyond those possible with an active dyno designed for use with running engines.

Fig. 1 - A top of the line active engine dyno cell provides engineers with a vast array of information regarding an engine's operational parameters. (Courtesy of Toyota Motorsport)

Written by Lawrence Butcher

A passive dyno usually consists of a power absorber - a water brake, hydraulic brake or eddy current brake - and a control system that varies the resistance of the absorber to match the engine's torque output, or in some instances allows for a pre-programmed acceleration rate. An active dyno system, on the other hand, allows for a far wider range of drive situations to be simulated, incorporating data collected from circuits under race conditions.

The absorber on an active dyno normally uses a permanent magnet motor-generator, operated by a fully programmable computer control unit. This control allows for three different types of operating mode:

• Basic power measurement - where the dyno acts like a passive unit to measure engine torque.

• Transient loss testing - achieved by driving the engine and measuring the input torque required.

• Drive cycle testing - a combination of power measurement and loss testing to simulate 'real world' running conditions.

This third mode of operation is the most useful to development engineers, allowing for accurate simulation of engine loads under race conditions. Data collected during the running of the car can be transferred to the dyno controller, and specific scenarios can be run through without the logistics and expense of actually putting a car on track. This technology has been widely adapted by many top-level motorsport outfits, from NASCAR to Formula One.

Taking NASCAR as an example, each circuit places a unique set of demands on the engine. At one extreme is Talladega Super Speedway, a 2.66-mile oval where running is primarily at full throttle, with the engine operating in a narrow band of around 300 rpm, with engine speed being dictated by the deceleration caused by cornering forces. In contrast to this, the 0.55-mile oval at Martinsville sees the engine running at anything from 5800 rpm to 9500 rpm, with heavy braking events for corners. Road courses place an even wider range of demands on the engine, from short sharp rises in rpm to prolonged pulls from low revs on exiting slow corners.

An active dyno can simulate all of these conditions, so long as data on throttle position, rates of engine acceleration and deceleration and shifting/braking points is inputted. From this data, the controller can apply the correct throttle openings at the right time, while providing sufficient resistance to represent the mass of the vehicle and aerodynamic forces acting on it. The load can also be rapidly changed to represent full-throttle shifting and load spikes encountered upon clutch release, or accelerate the engine with the throttle closed, as would happen during a downshift.

In recent years several key advances have allowed for more accurate drive cycle testing using active dynamometers, some of which have also resulted in unforeseen cost benefits for teams and manufacturers. First, improvements in computing power have vastly increased the capabilities of closed-loop control systems, with data systems capable of logging more channels of information at ever increasing rates. The parameters for closed-loop control can consist of - but are not limited to - throttle position, engine speed, engine torque, manifold absolute pressure (MAP), manifold vacuum.

Additionally, the latest systems have user-definable parameters that allow the engineer to add almost any control input they can think of. This means that more data collected from track testing and wind tunnel programmes, which have also seen an increase in density of data gathered, can be incorporated into engine simulation scenarios.

The second key advance has been the development of high-speed drive motors, capable of exceptionally fine and rapid speed adjustment. Similar to the benefits of higher data sampling rates, this new generation of motors allow for a more accurate replication of engine load scenarios and hence greater correlation between on-track and test cell results. Unrelated to track performance, motor development has also resulted in potential cost savings for operators, with some manufacturers now offering regenerative motor systems that use the power absorbed by the dynamometer to generate electricity that can be returned to the grid. Given that many systems can have electric motors rated up to 1200 hp, which consume prodigious amounts of power, any potential offsetting of energy costs is a welcome benefit.

The overall result for engineers is a far deeper understanding of an engine's traits, combining the benefits of on-track testing and the demands this places on a motor, with the accuracy of operating in a controlled environment. The one thing that is not present is the g-loads seen by a car on track; however, test beds are available to simulate such conditions, but that is another story altogether.

Fig. 1 - This overview of a dyno test cell shows the volume of equipment and sensors needed to assess the performance of a modern engine accurately (Courtesy of Delphi)

Written by Lawrence Butcher

At this point it is perhaps worth pointing out that while there are any number of methods of determining fluid flow, only those that don't appreciably change the pressure in the fluid are suitable for measuring gasoline. Any method (such as a simple orifice plate) that can introduce a vapour state by reducing the pressure in the fluid - even if only instantaneously - surely has to be avoided.

One of the easiest ways to measure volume flow, and that used successfully by myself for many years, is what I refer to as the 'pipette' system. Consisting of a number of different-size vessels or bulbs shaped like a pipette - 25 ml, 50 ml and sometimes even 100 ml depending on the fuel flow rate and accuracy required - the time was measured for the level of the fuel to flow down from a mark at the top of the pipette through the glass vessel to a mark in the lower pipe. With the time measured using a handheld stopwatch, in the days when engineers would routinely work from within the cell, this was a cheap and surprisingly accurate method of measuring volumetric fuel flow.

A little more sophisticated but relying on more or less the same principle is the electronically controlled fuel balance. Triggered by pressure or position sensors in a balance system, the time taken for a defined amount of fuel used is measured and the output of average fuel consumption then fed directly into a data acquisition system. Increasingly however, as engineers need to know more about the actual rate of fuel usage, a totally different concept is used - that of the Coriolis meter.

To anyone not entirely familiar with it the Coriolis principle states, "A mass revolving in a circle, which is then also subject to rotation in a plane at right angles to the first plane of rotation, will experience a force in the plane at right angles to the other two planes". It sounds a bit convoluted but the effect can be demonstrated by holding a spinning bicycle wheel between two hands. If you subsequently try to turn it parallel to the ground, it will twist in an unexpected direction.

Applying this to a quantity of fuel flowing through a pipe, if the fuel flowing around a loop is subject to a vibration in a plane at right angles to it, a force will be generated much like the force on our bicycle wheel. And as the fuel flow rate increases, this Coriolis force will also increase in proportion. The effect can also be produced in a straight tube excited by some form of electromagnet in the centre. Furthermore, while single-tube devices can measure mass flow, if the flow is split, twin-tube devices excited in an equal and opposite direction can not only measure mass fuel flow but density as well, and will have the additional advantage of being less susceptible to general vibration within the test cell.

Best of all, however, with typical turndown ratios of 100:1 and claimed accuracies of ± 0.1%, Coriolis fuel flow meters are a welcome enhancement to any test cell.

Fig. 1 - The fuel balance
Fig. 2 - The Coriolis meter

Written by John Coxon

Within a modern engine build and development business the most expensive part of the operation is usually the engine test cells. By the time the building has been designed and built, and individual cells kitted out with dynos, control systems, data acquisition and all the services - water, fuel, air and exhaust handling equipment, not to mention combustion analysis or any other exotic equipment - you may well be capitalised to the thick end of £1 million ($1.5m) per cell.

It is little wonder therefore that in all but the most ivory of towers a facility of this nature has to earn its keep. And to keep the bean-counters happy - who as we know would have appeared to have to pay for it all out of their own money - this means shaft-turning hours and keeping the facility working sometimes 24 hours a day and up to 7 days a week! In such an environment, taking a cell out of operation for even a few days for such luxuries as 'preventative maintenance' can be the business equivalent of a suicide note. But whereas our cheap and cheerful water brake may soldier on for many thousands of hours the eddy-current dyno, while still rugged and a fine product, needs a little more loving care.

The eddy-current dynamometer relies on the principle that an electrically conductive shaft or disc moving through a magnetic field will create a resistance to that movement. In practice this means a highly permeable magnetic steel disc spinning within the variable magnetic field of an energising coil. The resistance to the torque generated by the engine is then converted into heat and conducted away into the cooling water system by the loss plates, which are mounted in the dyno casing and adjacent to the spinning rotor.

A precision instrument, these loss plates are critical to the performance of the machine and must allow for controlled radial expansion, while at the same time distribute the heat as evenly as possible around them to avoid any unnecessary distortion. Even so, the temperature of any cooling water should be kept below 60 C to avoid any of the thermal protection systems cutting in. Overloading, loss of cooling water pressure (even momentarily) can result in excessive thermal distortion, causing catastrophic contact between plate and disc.

Insufficient corrosion inhibitor or impurities in the cooling water - nitrates, sulphates, chlorides and so on - can be another issue to trap the unwary, for although the surfaces of the waterways are most often coating in electroless nickel plate, machines of this type will still need regular checking if they are to have a long and productive life. In general, manufacturers will recommend de-scaling every 50 hours if customers still insist on using hard water and no corrosion inhibitor.

With its low rotating inertia and rapid response to variations in load, the eddy-current dyno is the professional's favourite for general test work. But although manufacturers strive to make them as rugged as possible, a bit of preventative maintenance can go a long way.

Fig. 1 - Cross-section of an eddy-current dyno

Written by John Coxon

Essentially the device consists of a system of rotors and stators that transfer the working fluid (water) back and forth between them; the reaction thus created loads the engine. In controlling the flow of water on either the inlet side or outlet side - and sometimes both - using either manual or servo control, the amount of fluid passing through is adjusted to balance the forces and maintain the engine at the required speed.

Mounted on trunnion bearings at the front and rear, the resistance torque is measured by a simple linear load cell at the end of a beam attached to the casing of the machine. Generally designed to run in only one direction of rotation, simple machines of this type may have output flanges at each end to accommodate both clockwise- and anticlockwise-rotating engines.

Requiring only a steady supply of water (but greater than that normally found from a domestic supply) at a constant pressure of about 45-60 psi, dynamometers of this type are simple to install but will use large amounts of water if installed in a total loss system. For this reason, local water authorities may need to be contacted to ensure that there will be a sufficient supply of water, to ensure adequate control of the engine and also not denude any of the surrounding properties of their water pressure while you are testing.

In converting all the engine output power into heat energy, and assuming an upper limit on the outlet temperature of the dyno of about 60 C, the minimum supply of water should be of the order of 45 litres or 10 imperial gallons per minute per 100 bhp. Since water is a limiting resource in many parts of the world and possibly metered, the use of water brakes using total-loss systems may become progressively more expensive. For permanent installations therefore, the provision of a cooling tower to both cool and reduce the amount of water used should be considered.

Inexpensive to buy and install, the water brake does however have many disadvantages when compared to other dynamometer machines. While machines of this type may have low inertia and good speed control capability (often quoted as +/-5 rpm), adequate control of the load is generally considered to be poor. For this reason water brakes are not often used when dynamic test cycles need to be undertaken. Useful for engine break-in and general development work, water brakes may be ancient technology but still have a place in the world of modern motorsport.

Fig. 1 - Typical modern water brake

Written by John Coxon

Yet while the limitations to the power absorbed by an engine dyno will clearly be determined by examining the dyno specification, the limits to the power absorbed by the chassis dyno is a little less certain. It's all down to the tractive force generated between the tyre and the chassis dyno rollers.

When used to simulate a drive-by cycle, the tractive effort on the chassis dyno is normally small and is expressed as the sum of the frictional forces on the vehicle. These forces are: rolling resistance - caused by bearings, tyres and so on, aerodynamic effects, and those required to accelerate the vehicle, and all have to be programmed into the rollers against vehicle speed, to simulate the loads experienced by the engine.

Now to accurately simulate the forces between the tyre and the ground the roller on which the driven wheels sit has to be as large as possible. This is because when travelling along the road the tyre effectively sees a 'roller' of infinite diameter. However, in trying to absorb the full engine power, as in the case of wide-open throttle performance testing, the tractive effort produced will almost certainly exceed that on a single roller with just the weight of the vehicle to hold it down.

For full-throttle testing, therefore, it would seem much better to use a system of twin rollers when the driven wheel centres itself in the valley between them. With such an arrangement the vehicle can be tied down, usually with flexible straps to give an extra level of security during the test and regain some of the traction area lost.

When twin rollers are used, simple geometry limitations dictate that their diameters need to be much smaller, so the contact patch between the roller and tyre will also be much smaller. However, given that there are now two contact patches - one for each roller - the frictional force may or may not be equivalent to travelling on level ground.

What is certain though is that in trying to increase the maximum tractive force available (by strapping the car down on the rollers even tighter against its springs), energy that would otherwise have been used to power the rollers will have been expended in the increased deflection of the tyres, and therefore the power figures seen on the dyno will not necessarily reflect that produced by the engine. Furthermore, if testing for any length of time then the tyres will invariably overheat, shredding great chunks of rubber around the dyno cell and almost certainly becoming unusable thereafter.

One way around this is to use wheel-driven dynos that bolt directly onto wheel drive flanges. This surmounts the tractive effort issue but, in not deflecting the tyres, reduces the driveline frictional losses and is therefore unrepresentative in a different way.

Chassis dynos, where the drive is taken through the vehicle tyres, certainly have their place but in my opinion for accurate full-throttle work you can't beat an engine dyno.

Fig. 1 - Chassis dyno testing

Written by John Coxon

In the design of any heat engine where fuel is burned, a thermal gradient inevitably exists across the unit. This gradient brings with it thermal expansion of the metals used, creating distortion and internal stresses over and above those caused by the pressures of combustion and component dynamics.

To dissipate this heat, the flow rate of the coolant can be increased, at which point the temperature rise across the engine will fall. But in doing so the power required by the coolant pump will increase and may eventually affect the effectiveness of the heat transfer between metal and water as a result of rapid changes in localised velocities, cavitation and so on. To avoid this, engine designers prefer to design the pump such that the temperature rise of the coolant across the engine is somewhere near 4-5 C when the engine is running at full load. This balances the power required to run the pump against the thermal stress introduced in the engine.

But this is where the problem starts. Trying to measure the exact temperature of both the coolant inlet (at about 78-80 C) and outlet (consequently at 82-85 C) to give an exact temperature difference to within 0.1 C is not easy. A mercury thermometer, carefully designed and calibrated, will probably do the job reasonably well, but they are not easily integrated into data-logging systems, and gone are the days when engine testers will even enter the test cell when an engine is at full chat.

The most obvious choice by some would be to use a simple thermocouple. A type 'K' Chrome / Alumel would give a useful level of sensitivity, at about 40 microvolts per degree C, but it is when we look at the measurement error that the alarm bells ring. As installed against an electrically generated reference temperature, the typical measuring error will be as much as ±1.5-2 C across randomly selected units, falling to ±1 C across the same batch.

With an error such as this on each temperature measurement, the error in computing the heat rejected to the coolant could approach 40-50%. Careful calibration will of course improve on that, as well as wiring the thermocouples 'back-to-back' to measure the temperature difference, but for a general-purpose thermocouple that needs to be robust enough to withstand the rigours of the engine test bed, the type 'K' is hardly suitable for such accurate work.

For more accurate readings, resistance temperature devices - the most common of which is the Platinum Resistance Thermometer - are much more accurate, but since they use sophisticated electronic circuitry they are also more expensive. In three- or four-wire form, and when carefully calibrated, accuracy can be as good as ± 0.03 C, but when wired to give a differential output between the inlet and outlet temperatures of the engine coolant, the overall heat-to-coolant error is much more acceptable.

Fig. 1 - Three-wire Platinum Resistance Thermometer

Written by John Coxon

As a development engineer I put this down to the water vapour in the air suppressing power-sapping 'knock' because at no other times during the winter was the carburettored car anywhere near as responsive and rewarding to drive. And that was most probably the case. But the effect of water vapour in the atmosphere can have a significant effect on the performance of an engine in other ways.

An engine on the dynamometer will generally be tested under controlled conditions. Even if the facility cannot manipulate the barometric pressure, temperature and humidity of the incoming air it will generally be within a narrow range.

When an engine is calibrated, therefore, it has been done so under a restricted range of conditions. When it goes out into the big wide world, however, it can experience extremes of temperature and humidity that can have a significant effect on its performance.

In a fuel-injected unit, much of this variation can be compensated for in the target air:fuel ratio, and changes in the air density will automatically be offset by the engine controller. But for an engine equipped with carburettors the situation is very different.

As well as the changes to the ambient air temperature and day-to-day variations in atmospheric pressure, the amount of water vapour in the air displacing the air itself also has to be considered. In the cold (4 C) dry air of a desert region the vapour pressure displacing this air can be very little, something like 5 mm of Hg at 20% humidity. Along the road at the coast and later in the day, when the temperature has risen to 34 C and an energy-sapping 90% relative humidity, that partial pressure will now be nearer 36 mm of Hg.

The difference in the air consumed could therefore change by as much as 4% as a result of changes in humidity alone. Coupled with the drop in density due to the temperature, this will have a major effect on the jetting of a carburettor tuned to lesser extremes.

Unfortunately there is no easy solution to this dilemma. But armed with the test conditions of pressure, temperature and humidity on the dyno - which I'm sure you've all kept somewhere - the relative change in air density can be calculated.

Because of the way a carburettor works, a 10% change in relative density may not require a 10% reduction in main jet size - something like 3-4% might be nearer the truth. But what it will tell you is which way to go (either up or down) in terms of jet size to return to maximum performance. The rest is down to you, and the copious notes you will endeavour to keep thereafter.

At the end of it all, however, you will look rather enviously at those with fuel injection.

Fig. 1 - Hygrometric chart

Written by John Coxon

When the car slows down, the energy otherwise wasted as heat during braking is stored in the flywheel and released back into the driveline at the next opportunity. In an engine, however, the main purpose of the flywheel is to even out the speed fluctuations in the crankshaft. But when it comes to dyno testing, this inertia - indeed, the whole inertia of the engine - can have a significant effect on the power measured depending on the test method used.

The power produced by an engine at any particular speed is described as the rate of doing work at that speed. When dyno testing, an engine is run at a series of fixed engine speeds and the torque produced is measured at each one. Running at a constant speed and wide-open throttle, these figures should be accurate and repeatable, giving a good representation of the potential of the engine at each speed.

At a fixed, constant speed the engine can be fully optimised for best fuelling or ignition setting, and the data easily inputted into the appropriate matrices in the engine control system. Using this method, the engine can be run at a series of constant speeds and the performance curve assembled from the data of a series of constant speed points.

Competition engines, however, rarely if ever run at a constant speed. For the vehicle to accelerate, the engine power has to be greater than the load necessary (wind resistance, friction and so on) just to maintain progress and so, as the vehicle accelerates, the engine speed increases as well. This continues in each of the gears until either the engine reaches its rev limit or the road load and engine power balance each other out. Either way, the engine speed is constantly varying, and only part of the power produced is being used to accelerate the vehicle; the other part is being used to accelerate the engine itself.

An engine with a light flywheel or one with lightweight internal parts will therefore expend less of its power on accelerating the engine and more of it on accelerating the vehicle. Fitting lightweight flywheels to race or rally engines consequently not only speeds up the gear changes but can also produce a much livelier engine, which in turn can also increase (however slightly) the rate of acceleration of the vehicle.

Bearing all this in mind, performance testing 'on the fly' more accurately reflects the conditions in the car. On the plus side, engine performance curves take less time, and although the torque generated will appear to be slightly less than the constant-speed version, with competition engines having a finite life - measured in hours, not tens of thousands, as in the case of a roadcar engine - much more of the engine's life will remain for the track.

But having decided to test this way, what sort of acceleration rate should be used? Well, NASCAR use 300 rpm per second for the superspeedways and 600 rpm per second for the short tracks. The faster an engine accelerates, however, the less power is measured at the dyno. Also, the higher this rate of engine speed increase, the more difficult is the control and the greater the variability of the data produced.

And unlike constant-speed testing, reducing the mass of the flywheel should be reflected in the performance curve.

Fig. 1 - Testing on the fly

Written by John Coxon

Many of you will no doubt be familiar with the large and unmistakable throttle lever alongside the console of an older engine test bed. Linked directly to the engine throttle via an equally unambiguous control cable or metal rod, throttle control was left entirely to the tester, and should anything become amiss or the engine detonate for so much as a split-second, the throttle would be wrenched away from fully open or rammed shut as an instinctive reaction.

With the dyno in 'speed' mode - as opposed to constant torque or power law mode - the engine revs would be restricted and the engine therefore safe from overspeeding. Here, speed control was purely down to the choice of dynamometer. In the case of hydraulic versions, the water pump impeller, situated as it was on the dyno shaft, would pump more water should the engine speed choose to rise, increasing the load. As this load increased, the engine speed would fall and therefore some level of control was inherent in the system. But I seem to remember it wasn't always foolproof.

For the eddy current dyno, things were a little more complicated. Here the magnetic resistance of the rotors, acting as it did on the stator assembly, was governed by the excitation current passing through the field coil.

In its simplest form, control would therefore be obtained from a potentiometer feeding the coil from a DC supply. For a given excitation current the torque absorbed would rise quickly with engine speed and then less rapidly until saturation. At this time the torque absorbed would stay constant.

Inherently some level of slow speed control was obtained but unless the excitation current was increased at higher speeds, control would be lost. When speed stabilisation was necessary at higher speeds, this excitation current would be used to modify the torque/speed characteristic and an electronic method used.

With modern dynamometer systems, almost regardless of design, control is relegated to the software, in particular the presence of the Proportional, Integral and Derivative (PID) loop. Here, speed control (or indeed any form of parameter control, be it temperature, load or whatever) is obtained by taking the error signal - the difference between the Set Point and the Measured Point - and adjusting the output of the controller to reduce it. Closed-loop PID routines can therefore be found everywhere in modern digital control systems.

In the case of an engine speed control system, the Proportional part adjusts the signal instantaneously in proportion to the change in engine rpm. For instance, if the error is reduced by half, then this element of control will act to reduce the next error computed by the same factor. This component controls the bulk of the response but, as readers will probably realise, the set value can never properly be achieved since the error is only halved after each loop.

Sometimes referred to as the 'gain' in the system for the Measured Point and the Set Point to coincide there are two other correction factors.

One of these is the Integral component. This serves to reduce the error over the longer term. So long as the error is never zero this element will continue to increase. When the instantaneous error is zero it has little effect, but if the set point is overshot, the error value changes its polarity and a negative term added to the integral. The term 'drift correction' might be a better description here.

The third component, the Derivative, has a damping effect on the system but also improves its response time. Linked to the rate of change of the error, this component is therefore larger when sudden changes occur. It is highest during the initial transient change and has the effect of 'flattening' the response curve, reducing the tendency to overshoot.

So the next time you are called a control freak, take it as a sign of genius and consider it a compliment! But remember the terms - Proportional, Integral and Derivative.

Fig. 1 - Variations in control in response to a digital input

Written by John Coxon

I know I'm being slightly pedantic, but engines - at least the ones we are more familiar with - are designed to produce torque at a particular

rotational speed, the product of which is then power. Torque is described as the actual turning effort produced by the crankshaft, and power the rate at which that torque is produced.

Likewise, in response to the question, "How much power does it give?" I can be equally pedantic. Depending on the mood, in answering such a question, the rejoinder might well be, "It depends on what you want", and in all seriousness the point will have been made.

You see, the power of any engine burning a fossil fuel is proportional to the amount of air - or, more correctly, oxygen - consumed. So when the atmospheric air pressure is high there is more oxygen in it and so more power produced.

The same goes for the temperature of that oxygen but in reverse. In this case, the higher the intake charge temperature the less dense the air, and the power produced will consequently fall.

Taking this a step further, if the air has water vapour in it, this will contribute towards the overall atmospheric pressure and must therefore be compensated for in any calculation. So when stating the power of an engine, the conditions under which it was tested need to be carefully stated.

To help, or possibly even confuse the situation, various authoritative bodies around the world have come up with engine test codes designed to alleviate the problem. Thus we have SAE, ISO, ECE, JIS and even DIN standards which in one way or another attempt to set certain atmospheric references to which all engine output characteristics are to be corrected.

The SAE, ECE and JIS standards all now seem to be converging on the atmospheric references of 990 mb of pressure and 25 C of dry air, while the German DIN standard refers to 993 mb and 20 C With a higher pressure and lower temperature, the DIN standard generates slightly higher corrected values so one needs to be a bit cynical when hearing this reference.

The latest version of the SAE standard - J1340 - specifies 990 mb and 25 C but the older reference (J607), still favoured by some, uses 1000 mb and 15.5 C For the really brazen, however, we need to go to the old STP standard where power is corrected to 1013 mb and 15.5 C. Engine powers and torques when corrected using this approach will be as much as 4% higher. No wonder it is still used by some parts of the performance industry.

Using the correction factor C.F. = x for wide-open throttle performance of naturally aspirated spark-ignition units (where subscript 's' refers to the reference temperature and pressure, and 'o' denotes observed values), engine performance values can, in theory, be readily compared from day to day, week to week and even location to location. This of course only takes into account the indicated power and is adequate for most forms of engine development activities when the precise figure is not necessarily needed. Since friction power is not affected in the same way prefer as pressure and temperature, these standards - particularly J1349 - have been modified to assume a constant mechanical efficiency.

So if you are ever asked the question, "How much power does it produce?" just pause for a moment and reflect before retorting, "How much do you want?"

Fig. 1 - The engine dyno

Written by John Coxon

The word suggests making the best use of what we have. We can therefore optimise the flow of air running through a radiator, making sure the temperature drop is at a maximum for a given amount of air flow, or we can optimise the amount of material in a gearbox casing. In each case we are making maximum use of a limited resource, and I would have hoped that anyone reading this would generally understand the processes involved.

With engine dyno testing, however, it's a different ball game. To optimise an engine can mean a number of things - do we want best power, best fuel economy or even best power to weight, say? While the dyno engineer doesn't ordinarily get a look-in at this last example, he tends to focus his skills very much on the other two, and because of this his vocabulary changes.

So instead of using 'optimise' our dyno engineer might use MBT, WMMP or even LBT. In some cases he may use RMMP or RBT, but before we consign him to the sanatorium or that place where all dyno engineers eventually go - home perhaps (?) - let me explain.

MBT stands for Minimum (ignition) Advance for Best Torque whereas WMMP stands for Weakest Mixture for Maximum Power, sometimes referred to as LBT (Leanest mixture for Best Torque). MBT is obtained by placing the dyno in 'speed' control mode and the throttle in 'position' mode, and with a wide open throttle, altering the ignition advance at an air:fuel ratio to give maximum performance. This ratio is typically about 12:1 for standard gasoline.

Exact practices will vary from tester to tester, but with his eyes glued to the torque output gauge and a well-calibrated ear to listen out for 'knock', the ignition will be advanced in 2 (crank) degree steps. Continually correcting any fuelling drift, once maximum torque has been reached and the power doesn't rise any further, then backing off 1º will give an approximation to MBT.

Another way, and one that is slightly less subjective is to increase the ignition advance until power drops off by 1%. Retarding again through peak torque until the power drops off on the other side by 1%, the MBT point is roughly half way. Should the engine start 'knocking' at any point, the ignition may need to be retarded quickly by several degrees if the engine is not to be damaged. Sometimes, however, occasional' knock' can be tolerated.

WMMP or LBT are simply variations on the same theme but instead of outright power, fuel flow will be targeted. WMMP/LBT are most often sought at part throttle since running gasoline engines lean at wide open throttle can risk detonation damage.

So 'optimise' is a term rarely spoken by dyno engineers but if they talk continually about MBT or WMMP, just cut them some slack.

Fig. 1 - The dyno tester's office

Written by John Coxon

Last year (2009), the organisers of the British Touring Car Championship made great play over the emissions testing of all their competing cars. Monitoring only carbon dioxide emissions (and dutifully simply ignoring all of the real toxics coming out of the exhaust) and using one of the best emission test facilities in the UK, the programme concluded that while racing cars could be made to be as 'efficient' as road cars, running a test cycle that both could achieve sensibly, the EU CO2 emissions target for the immediate future of 120 gm/km, simply does not equate with a racing car (or indeed a road car) doing 10 - 15 miles per gallon. Sadly, the real numbers of the work will never be published but in doing it, the BTCC have somehow inadvertently highlighted the real difficulties of measuring vehicle emissions, especially those from competition cars.

The idea behind any kind of emissions testing is to test the vehicle under the conditions in which it is most likely to be used. Thus for road going vehicles we have a mixture of urban and highway duty cycles covering speeds around and up to 50 km/h and 120 km/h respectively. Following a similar argument for race vehicles typical speeds would need to be anywhere between 110 km/h and say, 240 km/h. I don't know about you, but sitting in 150 mph racer is scary enough but strapping that onto a dynamometer and repeating it within the confines of a vehicle test cell is perhaps more than the nerves may be able to take. And don't forget, front wheel drive vehicles are fundamentally unstable on chassis rollers and so any testing of this type would require the vehicle to be well and truly tied down!

Competition engines also tend to be highly dynamic. Rapid changing of engine speeds and loads would need a dyno not just capable of absorbing the 300 bhp of a modern BTCC car, but also the capability of controlling those loads to the required accuracy. The only vehicle dynamometer with very low inertia, fast response and the ability to absorb power as well as motor the vehicle, (to simulate the incidence of heavy braking when the inertia of the vehicle reacts against the engine) is an AC unit. Each vehicle would still have to undertake 'coast-down' testing to replicate frictional losses and a suitable drag coefficient determined to simulate the aerodynamic forces. All in all quite a lot of what I would call unnecessary effort to achieve what would appear to be very little. Of course, we could transfer all of this to a transient engine test bed, a là that found with Formula One engine manufacturers, but that's getting a little out of hand, isn't it?

Undertaking a test cycle simulating a race vehicle running between 110 km/h and 130 km/h swapping between 5th and 6th gears was therefore hardly representative but the BTCC, I suppose, did produce the opportunity to compare results with their OE counterparts. No, the only sensible way of measuring any emissions on a race car would be to do it on track with some kind of mobile sensor or analyser. While I hear that one emissions analyser company may be interested in picking up the gauntlet again to develop an on-board emissions measuring system, wouldn't it be easier simply to measure the fuel consumption and, since the amount of carbon in any fuel can be accurately established, calculate back?

Fig. 1 - On the Rollers

Written by John Coxon

The dynamometer consists of a shaft carrying a rotor, which revolves inside a water-tight casing. In each face of the rotor there are a number of radial vanes set at an oblique angle to the axis of the rotor, the spaces between forming cups of semi-elliptical section. On each face of the casing and facing the others, is a similar arrangement forming another set of cups and while the rotor vanes face forward in the direction of rotation, these casing vanes are presented in the opposite direction.
With the rotor attached to the drive or output shaft of the engine, water at a constant input pressure (around 25 lbs/in2) is flung outwards by centrifugal force and thence into the cups in the casing. These cups serve as a guide to return the water to the inner part of the rotor where it is flung out again to repeat the cycle once more. The resistance caused by the action of the water between the rotor and its casing reacts against the casing which tries to turn on the trunnion bearings on which it is supported. Counteracted by a damped load arm, weights and a spring balance are used to measure the engine output torque. When taking any reading it was essential therefore that the load arm was precisely horizontal and aligned with a datum point on the machine for that purpose. This was achieved by operating a system of sluice gates using a hand wheel. These sluice gates when moved in or out, exposed more of the vanes to the water which either increased or decreased the resistance torque offered. The energy comprising the engine power output was converted into heat, which was dissipated by passing it through a cooling tower before being returned to the dyno.

Dyno testing back then was a question of manually balancing the output torque of the engine from within the engine test cell by adding or taking off balance weights and or using the dyno hand wheels to bring the load arm back to the datum. Often a spare nut or washer could be placed on the load arm to increase or decrease the 'power' the engine was giving to suit your purpose but that is altogether a different story!

Working from within the test cell things may have been noisy, it may have been very hot and certainly you were always aware of any strange noises coming from the engine, but by jolly it kept you well fit!

Fig. 1 - Cross-section of a Froude DPX dynamometer.

Written by John Coxon

Engine dynamometers are generally the province of serious tuners and the OE engine business. Attaching the engine to a fixed means of absorbing the load, be that hydraulic or electric, is at first sight a simple exercise. The independent control over engine speed and load enables many tasks to be undertaken especially those conditions where engine stability is of paramount importance. Within the confines of the test cell, coolant and oil temperatures can be accurately controlled and with a little more care, so can the air inlet temperature. Any one of these can independently have a major influence on engine performance and so for any serious engine development an engine dyno in a test cell is an absolute must.

But there are times when for convenience or simply on the grounds of cost, testing on a chassis dynamometer may be the better solution should the engine installation need to be representative of that in the vehicle. Installing full vehicle exhaust systems in the test cell has always been a problem. Even if the exhaust manifold clears the engine mounting frame, at some time the remaining pipe work will invariably need to pass right through some other fixed part of the cell, normally the dyno itself. Getting the correct under bonnet engine airflow just right can also take much effort and time. When, say, mapping an engine for fuelling and ignition as well the exhaust system, the correct air intake arrangement is also imperative. Especially important for road going vehicles, the only way that this can be done relatively easily, is on the chassis dyno.

Chassis dynos allow the full vehicle to be installed quickly and easily by placing the driven wheels on a set of rollers and measuring the engine output indirectly via the gearbox and differential. Far less repeatable than the engine dyno, it is at least a way of evaluating the wide open throttle performance of an engine under some level of control, albeit basic. Due account will need to be made of the losses through the transmission system and tyres and much care will have to be taken in ensuring the dyno will absorb the full torque of the engine. Many dynos are only suited for emission test work where high engine powers are never experienced. Such emission test dynos may only absorb up to a maximum of around 60kW and so are totally unsuited to most types of performance work.

The other concern is that of the tractive force required between tyre and roller at wide open throttle powers. With some rollers, even with the vehicle tied down heavily against the vehicle springing, the power of the engine can overcome the maximum tractive force available. In the short term this can overheat and burn the tyre leaving a rather acrid smell. If continued it will almost certainly cause the tyre to fail, possibly even burst with all manner of horrendous consequences.

Since the vehicle is stationary while the driven wheels move, the airflow over and around it is also hardly representative of the real situation out on the road or track although in some installations the fan speed can be synchronised to match the road wheel speed thus to some extent simulating this airflow. At best however, this is only an approximation.

But while the engine dyno is necessary for accuracy and the chassis dyno for the convenience, in the end there is no substitute for the real thing out on the track.

Fig. 1 - Chassis Dyno test set-up.

Written by John Coxon

But in any form of reliability testing it is essential to get the right balance between realism and repeatability.

In a Grand Prix car the engine duty cycle varies from lap to lap and depends on many factors; tyre temperatures and wear, are but only two. As the tyre heats up it produces more grip and with more grip comes more speed, and as a result the engine duty cycle changes. Towards the end of a driving stint the tyres wear, lose grip and the engine then may spend more time on the speed limiter. At this point the driver will then come off the throttle, become less aggressive and the duty cycle changes yet again. Over the course of a weekend we mustn’t forget either that the track will “rubber-in” progressively changing the duty cycle, and then perhaps an overnight shower of rain could wash everything away and so we will be back again where we started. A simple change in wind direction can also change the test parameters. A tail wind on the main straight may cause the team to run a taller gear to avoid the engine running on the rev limiter well before the end. With just a change in the wind, we would already be 200 rpm down on where we were earlier in the day.

If all that wasn’t enough, traffic is another variable, which has to be taken into account. When one car is following another, not only does the drag change, but the engine may spend much more time on the limiter. Cooling to the engine water and oil systems will also change significantly, producing spikes in the running temperature. And of course we can’t lose sight of the driver. He is probably the greatest variable of all! At one time he may be driving quite sensibly but when the ‘red mist’ comes down, his driving can become either totally irresponsible or very inspired. In any form of realistic reliability test cycle therefore, all of these variables have to be covered.

However, realism is one thing, repeatability is another. For reliability testing you need the test to be as repeatable as possible but if you want more of one then you almost certainly need to compromise on the other.

Written by John Coxon.

A perfect example of this is the engine dynamometer.

Coming in a variety of types, shapes and sizes, and surrounded by a certain amount of mystique by the outside worldwide, a dynamometer is simply a device to apply a load to an engine. A simple dynamometer can be nothing more than a crude mechanical brake applying a load to the rim of a flywheel. Attaching the load arm to some form of spring-balance or in engineering speak, load cell, and the maximum torque generated at a constant engine speed, can be quickly calculated. Multiply this by the speed of the engine speed in revolutions per minute and divide by a constant appropriate to the units in use, and the power of the engine at that particular speed can be readily determined. For a system such as this to be practical the heat from the friction in the brake would need to be cooled by a continuous supply of cooling water but essentially all engine dynamometers are only a variation on this simple theme.

Dynamometers can be hydraulic, electric or eddy current.

While early dynamometers were hydraulic, using a system of ‘sluices’ or buckets to control water flow in something akin to a simple gearbox torque converter, most steady state dynamometers today are of the eddy current type. Similar to an electric motor in some ways, an eddy current machine consists of a stator supported on free to rotate trunnion bearings and a rotor which is driven off the back of the engine. The rotor consists of a central hub from which thick rectangular section spokes project radially. Either side of the spinning rotor, smooth faced stationary loss plates can be found and surrounding these plates, is a large coil. All contained within a substantial steel casing, an AC current is passed through the coil (or coils if more than one is used), which produces a rapidly fluctuating magnetic field. The action of the spinning radial spokes interrupting the magnetic field induces eddy currents into the loss plates and the heating effect thus generated has to be cooled using a re-circulating water-based liquid passing through the many passageways machined into the plates. Shaft power from the engine is therefore resisted by the action of the eddy currents and totally converted into heat, which is carried away by the cooling system. Although controlled by the AC current, the dynamometer does not generate any current or voltage as would an electric version.

Like any rotating machine the absorption power envelope is limited on one hand by the torque, generated by the eddy currents, and then ultimately by the maximum safe speed of the rotor. Sometimes rotors can be cropped to increase their maximum speed but this also reduces the amount of torque absorption. Engine control however, should only be performed inside this operating envelope. Operating outside it will mean a possible loss of engine control and can therefore seriously damage your wallet.

Written by John Coxon.

The engine in question was the first ever example of a Le Mans Prototype V8 and the author’s role was to do nothing other than warm the engine up on a base map which would be followed by a check over before the real mapping work commenced.

For around five minutes (no-one ever went through the data in much detail) everything went smoothly and nothing of any note happened; the engine hummed away and the hustle and bustle of the workshop continued apace in the background.

The world was a peaceful place and a lot of idling away was done, and as the author contemplated deeply exactly what was happening inside the engine, the engine built up temperature.

Then without any warning whatsoever the revs started to rise dramatically as if the engine had gone to full throttle. The engine continued to pick up speed and it was instantly obvious that is was in danger of over-revving and over-revving quickly; the author didn’t know exactly on the day (and still doesn’t actually!) how the rev cut strategy worked on that particular ECU.

A quick look at the throttle lever (this was well in the days of throttle cables) confirmed that it hadn’t moved and so there was only one thing for it.

As any dyno man will tell you there is only one reaction you can have in any unexpected circumstances; you find the big red button and you hit it!

The author did so and the engine died instantly. The soundproofing, being what it was in the day, meant that the regular dyno operator (along with most of the workshop) heard what had happened and came running.

Subsequent analysis of the data showed that the engine had over-revved but not to a point where anything had been damaged and after a borescope was deemed OK to continue. The data also confirmed that the throttle had not been opened.

Ten minutes later investigation of the dyno cell revealed the cause of the near disaster; an hydraulic oil return line from the dyno controller had come adrift, allowing the fluid to drain away until control was lost, which caused the dyno to lose all load during running. With no load on, a 4.0-litre V8 will happily spin up even on idle.

So the main lesson learned that day was the fact that anything can and at some point will go wrong.

Nowadays the world is a different place and a test cell can be rigged up such that an engine is automatically cut in the event of a problem but that is not always the case.

So if the author has one piece of advice for you; always know where your red button is!

Written by Tom Sharp.

The company recently came into possession of a brand new 4-cylinder Honda road car engine, donated to them by Honda as the car it had come from could not be sold and therefore needed to be disposed of.

They were delighted; the Honda would replace an ageing Ford engine which had seen better days and the new engine would give the technicians a bit of a change.

The company in question installed the Honda engine on the dynamometer, modified the fuel, oil and water systems to suit, built a new control board onto which they installed the car instrumentation, throttle pedal and ignition switch.

They were all ready to go and went for the fire up, only to find that there was no output from the dynamometer rev counter and the load figure seemed a bit random.

Subsequent investigation quickly identified the problem; the Honda engine ran in the opposite direction to every other engine the company had ever dealt with, anti-clockwise when viewed from the front (the non-output end). Within a short time, the dynamometer was (literally) turned around and rebuilt and no damage was done, but it left a clear lesson; never assume anything.

The give away from the cut away drawing is obviously the timing belt; the tight side is to the left and the slack side, with the tensioner in it, is to the right. Not that this company would be the first to make this mistake.

It is always worth checking that things are going the right way round; the author once encountered a gear type oil pump which pumped in the wrong direction because its designer assumed that oil passed between the two shafts rather than around the outside.

Alternators are another classic example. Although they will generate electrical power whilst rotating in either direction, the cooling fans are typically directional.

Bi-directionality is one of the advantages of an eddy current dynamometer, and that can be a surprisingly useful feature for a company who test a variety of engines.

Written by Tom Sharp.

The system incorporates the latest hardware and software technology in order to improve the dynamometer data acquisition capability to the benefit of the company internally and to its customers.

The company, which manufactures Judd racing engines, has three dynamometer cells, all of which contain a Froude G490 water brake dynamometer. The new system is portable and can be configured for, and run in, each cell independently.

Judd engines have dominated the Le Mans customer engine market in recent years, in terms of sales, performance and reliability, so clearly the system is working.

For the last 15 years the company have used a Magnetti Marelli data acquisition system which operated satisfactorily, but after Le Mans 2008 EDL decided that a more flexible and modern system was required, as John Judd Jr explains;

“The previous Marelli dynamometer data acquisition system was entirely independent of the engine and ECU; it would log all of the dynamometer parameters, but that was it. It had a limited number of channels and was not very configurable.

“Our new dynamometer data acquisition system uses a Motec ADL combined dashboard / data logger. This is an up to date system that offers us excellent flexibility; the software is easily configurable and it can log up to 26 channels; we actually use 20 from the dyno. The main benefit though is that we can export data from an engine ECU via a CAN link into our Motec system. This means that we have one system which is logging everything we need from the engine and the dynamometer.

“All our sports car engines run EFI Euro 12 ECUs, which have built in data acquisition systems. That is a really useful feature on an ECU from an engine manufacturer’s perspective as it allows us to monitor engine parameters independently of the team.

“We installed the system earlier this year and already we’re seeing the benefits. The new system makes it much easier to identify faults during running, and if an engine problem occurs during the dyno test or on PDI (pre-delivery inspection) we can very quickly analyse the engine and dynamometer data.

“The other big advantage is seen during the testing of engines which are run on ECUs which don’t have on board data logging; and we have plenty of those! A typical example would be a 1990s Formula One engine; depending on the individual case we can now more easily adapt the new Motec system to log data from the engine sensors.

“We have an obligation to supply a high quality engine to the customer and having this additional capability benefits us both. The data allows us to check many more parameters before the engine is delivered and should any problems subsequently arise in the car, it helps in the analysis.

EDL have, typically, adapted the system to their own requirements.

In each dynamometer cell, two LCD screens display output from the new system. The dashboard, which actually is the data logger, is mounted on an in-house control box which has switches incorporated in it allowing the operator to change between different pre-set screen configurations for warm up and power test. The operator can also reset alarms and set timing beacons in the data.

John Judd Jr explains the choice of Motec;

“The Motec system is a cost effective solution; dynamometer testing is an expensive business. We have to be able to offer customers a service they can be entirely confident with, but at the same time we have to sell dynamometer time in a competitive environment.

“In commissioning the new dynamometer data acquisition system we’ve achieved our aims; it’s a modern, flexible system which we can adapt to future requirements.”

Written by Tom Sharp.

When we talk of dynamometers we are often referring an engine dyno, but there are other types of dyno’s that we use on racing cars which have nothing to do with the engine, the Damper Dyno being a very common example. Here we ‘cycle’ the damper by applying a varying (usually sinusoidal) Force input, and then measure its resistance and the variation in velocity that results.
On the damper dyno we are working with linear units, Forces and linear velocities, and this is also the case when we put a car onto a chassis dynamometer, or ‘rolling road’, although we get our results in the form of bhp at the wheels. In fact what we are measuring is Tractive Effort at the wheels, which is directly related to the engine Torque through the gearing. Analogous to the relationship between Power, Force and Velocity is the following which we use for rotational motion and Torque. The relationship is usually given as Power = Torque x rpm.

Damper dyno’s are fairly common in most professional racing teams, whilst any serious engine builder will have access to an engine dynamometer, but a type of chassis dynamometer that is much less common, usually being available only to projects that have a major road car manufacturer behind them, is a combined engine and transmission dynamometer.

Here we mount the complete engine, transmission and driveline and then run the engine as if on an engine brake, whilst the software operates the clutch and transmission, the trend towards semi-automatic sequential paddle shift operated systems in recent years making this essentially a software exercise.

For many years it was often the case, particularly in endurance racing, and especially at the Le Mans 24 Hours, that the weak link in the reliability chain was not the engine, or the suspension, but the gearbox. The problem was not just that many transmissions were subject to prolonged high loading, but that much of the duty cycle consisted of shock, and reverse loads. This type of cyclic loading is often associated with metal fatigue, and this was often at the root of component failure.

A modern, sophisticated engine and transmission dyno can take, as its input, a typical duty cycle for the complete 24 hour race. This means that the engine will be throttled, on the brake, as if it were completing a lap of the circuit, the transmission will be shifted as per the requirements of the lap, and the complete driveline assembly, including driveshafts and wheel hubs will undergo this simulation over a period of perhaps 28-30 hours. This is necessary to ensure that an engine-driveline aggregate that has achieved 24 hour durability, but was about to fail after 24 hrs and 5 minutes is not signed off! There will of course be a statistical factor of safety, or in this case longevity that needs to be accounted for, in much the same way that when stressing for fatigue we look at number of cycles to failure, and then add a factor.

This type of ‘simulation’ is a subtly different use of a dyno from the typical engine test unit, in which performance is quantified, along with attempts to improve it. On the combined engine and driveline dyno the main focus is on reliability, although it can be used to assist with transient engine mapping and also is a handy tool in optimising gear change strategy.

In removing the variations in shift quality inherent in a manual gearshift, the adoption of the electro- hydraulic or pneumatic gearshift has been a big factor in improvements in gearbox reliability in recent years, but the use of the dyno, in the background, has also played a significant role.

Written by Peter Elleray.

The PCC consists of two rolling road dynamometers, Cell 1 and Cell 2. Hatton Systems Ltd is responsible for the dynamometer equipment and control systems and HORIBA for the gas analyzing technology.

Vehicles running on carbon-based fuels (gasoline, diesel, LPG, etc) fill their tanks by volume – gallon or litre. It’s generally easier to meter fuel by volume than to weigh it. Vehicle exhaust emissions are regulated and have to be measured. This is done by measuring volumes rather than trying to weigh the gases.

If a carbon fuel burns efficiently carbon dioxide and water are produced. Inefficient combustion is cleaned up by a vehicle’s catalyst; with everything working as expected 99.9% of the fuel is emitted as carbon dioxide. This has become a measure of a vehicles ‘green-ness’, the lower the amount of carbon dioxide emitted, the greener the vehicle is perceived to be.

For the BTCC test program, carbon dioxide emissions are measured over a drive cycle, which has been specified for the purpose. While the vehicle is being driven through the drive cycle, exhaust gases are measured using a special type of flow meter. This is referred to as a Constant Volume Sampling system (CVS). This overcomes the problem with pulsating flows by diluting the exhaust with as much clean air as necessary to stop condensation of the water vapour and to maintain a constant volume at the flow meter.

A sample of the diluting air is taken throughout the test period and analysed for carbon dioxide. The proportion of carbon dioxide found in the sample is applied to the volume of diluted exhaust gas measured by the CVS which gives the volume of carbon dioxide emitted during the test period. Multiplying the volume by the density of carbon dioxide and dividing by the distance travelled during the test period gives a result for carbon dioxide in g/km.

The proportion of carbon dioxide in the sample gas – the sample concentration – is obtained by the use of a fast response and highly accurate emission analyser such as the top of the range HORIBA MEXA-7000. The HORIBA MEXA systems make use of the NIDR (Non-Dispersive Infra-Red) technique. Carbon Dioxide has a strong absorption at a specific wavelength region of the infra-red spectrum. The analyser works by measuring the intensity of this absorption and comparing it that of a known CO2 calibration gas. A sample of the ambient air used to dilute the exhaust gas and a sample of the diluted exhaust itself are piped to the CVS system and stored in special bags made from Tedlar. From here they are sent into the analyser which measures the concentration or amount of carbon dioxide in each bag. The mass of carbon dioxide can then be obtained.

Obtaining the concentration of other emissions components follows the same systematic flow but uses different methods. There is a clear need for very accurate emissions analysis systems in automotive testing and this is one such example.

Written by Tom Sharp.

The original specification for the Cell 1 dynamometer was firmly focused on building a world class facility to meet the future needs of the vehicle testing market. This started with creating a 4WD facility for the emission testing of petrol and diesel vehicles to EPA standards. One key element of this is having a dynamometer with full road load simulation.Vehicle testing using a chassis dynamometer with 48” rollers and centre-mounted AC motors is the recognized ‘gold standard’ for emission testing to EPA requirements. That determined the basic format of the machine. From there the specification took on what would be needed to carry out emissions tests for cars of the future with increasingly low emissions.JLR were also keen to push the boundaries of the dynamometer’s specification to a position well ahead of contemporary visions. This extended the scope of the project to enable the new facility to carry out power testing and vehicle development on a wide range of products.The AC motor specification was increased from a typical 125 kW to a continuous rating of 255 kW per axle. Up to 1MW can be absorbed for short periods of time. The physical range of the machine was increased; wheelbases between 2,300 and 4,000 mm can be accommodated as can axle loads of up to 3 tons.

Several automated features were included in the specification. The wheels are automatically centred to consistently position the vehicle on the crown of the rollers. Automated roller covers allow the unused part of the roller to be covered for increased operator safety.A unique automated dead-weight calibration facility was also added. This runs through an EPA calibration sequence handling the 670 kg of calibration weights all by remote control from above ground. This avoids the health and safety issue of handling many large weights and a calibration beam. The individual calibration weights go up to 280 kg and are certified to with 250 mg to National Physics Laboratory traceable standards.Cell 1 also features a bespoke vehicle restraint system. Retaining a vehicle on a dynamometer to the required standards is complicated. The vehicle must remain within ± 25 mm of the crown of the roller, which is not a stable position. The restraint system cannot contact the wheels during the test and no vertical forces can be applied to the vehicle. A multitude of different attachment points and towing eyes must be accommodated. No paint damage is allowed during fitting and removing of the restraint system. The restraint system must withstand the considerable forces that can be applied during power testing – over 4 tons – and the restraint must not block the flow of vital cooling air during the test. Finally the restraint must not cause a health and safety issue by introducing tripping hazards into the test cell.Achieving this goal is made even more complicated as many current cars use air spring suspension which can change their ride height at fixed speeds. Simple restraining straps are simply not feasible in a world class facility like the PCC; particularly as they will stretch, allowing the vehicle to move.JLR designed a solid vehicle restraint specifically for Cell 1; a unique and much admired solution. Having such a stiff restraint gives the drivers enormous confidence and allows them to concentrate on the difficult job of driving the vehicle to the required standard.Next month: Future development plans for Cell 1 and a look at Horiba’s latest gas analysing technology.