Originally published August 1996
Within the next 10 years, the electrical systems in some luxury automobiles will be so changed as to be almost unrecognizable. Although they will doubtless employ the old reliable 12-V lead-acid battery, their loads will be driven by a variety of voltages, both ac and dc, perhaps derived from a single ac distribution network. Designers will be able to match voltages to individual loads for best efficiency and performance--lights perhaps at 6 V ac, electronics at 5 V dc, active suspension at 350 V dc, and motors and actuators at 42 V dc. The digital signals controlling those loads will be carried by a separate communications network [Fig. 1].
That, at least, is how a working group of engineers from makers of autos and automotive components envision the electrical systems of luxury vehicles in the 2005-2015 time frame. The group, which gathered under the auspices of the Massachusetts Institute of Technology at the request of Mercedes-Benz AG, expects the technology to spread to less opulent models as costs drop [see , "Planning for 2005"].
Semiconductors, of course, are the enabling technology that will make it all possible (as well as making it all necessary). The costs of solid-state power converters, switches, and logic devices have been dropping steadily. The cost per watt today is less than half what it was in 1990 and is fast arriving at a level that is practical for certain automotive applications.
Two main forces are driving cars to multivoltage systems--the quest for ever-greater fuel economy and the emergence of new power-hungry automotive functions. Novel electrical equipment, like electromechanical valve actuators and active suspensions, will triple the aggregate electrical power demand in some cars--from 800 W today to an average of 2500 W and a peak value above 12 kW by perhaps as early as 2005. That power can be more effectively distributed and utilized at voltages much higher than today's 12 V dc.
All the same, a large, complex infrastructure now supports the 12-V system with components and services. Surmounting this obstacle will require agreement within the industry on many new system parameters, and reaching that agreement will take time.
The case for electrical efficiency
For better fuel economy, many devices now driven directly by the engine will be driven electrically. That way, the speed of, say, the water pump and the cooling fan can be varied to match the load or even turned off when not needed. To get the most out of switching to electric drive, the requisite electric motors should be operated at voltages substantially above 12 V to boost their efficiency.
Fuel economy also dictates that electricity be distributed at higher voltages, to reduce ohmic losses without resorting to a heavy and expensive harness of large-gauge wire. Improving the electrical system's efficiency so it lops 100 W off the average electrical load has the same effect on fuel economy as reducing the car's weight by 50 kg, as measured by the FTP (Federal Test Procedure) 75 standard profile of starts, runs, and stops [Fig. 2].
Even more to the point, the U.S. Corporate Average Fleet Efficiency (CAFE) standard prescribes a maximum fuel consumption rate for cars sold in the United States. The Federal government assesses a penalty of US $5 for every 0.1 mile per gallon (0.04 kilometer per liter) below 27.5 mi/gal (8.55 L/100 km) on every car the manufacturer sells. A 200-W electrical load accounts for about 0.4 km/L in the FTP 75 cycle test; so, if a manufacturer is delivering 25-mi/gal (9.41 L/100 km) cars, for example, it can justify spending more per vehicle on components to improve electrical efficiency.
At present, when U.S. and European fuel economy tests are conducted, only those electric loads essential to the operation of the vehicle are active--that is, the ignition and engine electronics. Lights, air-circulating blowers for the passenger compartment, entertainment electronics, power windows, and so forth are all turned off for the tests. But this could soon change. If a typical average electrical load is required during tests, the electrical performance of the car will become much more visible.
Nowhere will this visibility be greater than in the 80-mi/gal "green" car. A midsized car resembling a Ford Taurus or Chevrolet Lumina, it is the goal of the U.S. Partnership for a New Generation of Vehicles (PNGV), which comprises 11 Government agencies and the Big Three's U.S. Council for Automotive Research. (European manufacturers are pursuing the same goal, a car that consumes 3 L/100 km.) But unless the electrical system in today's car models is improved, over 25 percent of the green car's fuel will go to electrical loads. And if some mechanical functions such as air conditioning and power steering are electrified, that fraction could rise to 50 percent. If the partnership's green car is not to be a bare-bones model, then it will have to incorporate a truly superior electrical system.
Another strong motivator for electrical efficiency is the high cost of automotive electricity--a lot more than homeowners pay for theirs. The cost can be calculated in a straightforward way. Gasoline has a heat energy content of 43.5 megajoules per kilogram and a density of 0.73 kg/L, which gives it a volumetric energy content of about 32 MJ/L, or 8.8 kWh/L. Thus if the engine efficiency is 40 percent and the alternator/belt efficiency is 45 percent, a liter of gasoline furnishes approximately 1.6 kWh to a car's electrical system. Assuming a gasoline price of US 34 cents per liter, the cost of generating electricity in a car works out to about 21 cents per kilowatt-hour, substantially more than the 8 cents per kilowatt-hour average price of residential electricity in the United States.
Overcoming load-dump tyranny
Power electronics in future automobiles will perform two functions: simple on/off switching (now done by relays and manual switches) and controlling loads with logic, inverters, and dc-dc converters. Together, the two classes of power electronics involved will not only accommodate higher voltages but also overcome basic defects in the conventional 12-V system, such as widely varying system voltage and destructively high voltage transients.
The voltage in a 12-V system actually ranges from about 9 V to 16 V, depending on the alternator output current, battery age and state of charge, and other factors. Loads are sized to function properly at the lowest system voltage; thus, when the voltage is higher, they draw more current than necessary. Load components therefore need to be rated for continuous operation at the highest current.
Then there is the notorious load-dump transient, a voltage spike that appears on the system when a fully loaded alternator suddenly loses its load--for example, when a charging battery is inadvertently disconnected. The voltage behind the alternator's armature reactance then suddenly shows up on the system, a 40-V, 100-ms transient if the alternator is protected by avalanche diodes, and 80 V or more if it is not. The switches and load components therefore have to be rated for temporary overvoltages at least four times the nominal system voltage.
This requirement is an expensive one, especially for semiconductor switches. A load dump may never occur during the life of a car, yet components have to be ready to handle it. The result is that load devices have to be grossly overrated both for continuous current and transient voltage.
In future cars, however, power electronic converters will pro vide an interface between the alternator and the distribution system, making it unnecessary to overrate components. The alternator in such a car can be allowed to generate an unregulated output that varies with engine speed; power-conditioning circuitry will take that output and turn it into a constant, transient-free system voltage for distribution. It will then be unnecessary to overrate either the load components or the semiconductor devices that control them.
Semiconductor manufacturers are doing their part in developing advanced power devices for cars. For example, Siemens AG, in Munich, now offers smart power switches expressly for automotive service. These MOSFET devices not only do high-side switching, disconnecting the load from the supply voltage bus instead of from ground, but they also shut down if their temperature rises excessively. They also can protect themselves against overcurrents, and act as resettable fuses with the aid of associated logic, either monolithically integrated with the MOSFET or packaged with it as a hybrid integrated circuit.
Until recently, a big impediment to acceptance of power MOSFETs in automobiles has been their high on-state drain-to-source resistance, known as R ds(on). Typically, R ds(on) has been 100 m(omega) for a device switching 10 A, giving a 1-V forward drop at a junction temperature of 125 degrees centigrade and dissipating 10 W--not particularly efficient. Now, however, Siliconix Inc., Santa Clara, Calif., offers power MOSFETs, made by trench technology, that have an R ds(on) of only 16 m(omega) at a substantially higher junction temperature, 175 degrees centigrade. Other manufacturers are beginning to offer similar devices.
Further, as noted, the cost of power electronic converters is rapidly approaching the low levels that make them practical in cars. From 50 cents per watt in 1990, the cost has dropped to 15-20 cents per watt for converters with 100-3000-W ratings. When costs fall to 5-10 cents per watt, wide penetration of the automotive market can be expected.
Laying the groundwork
The MIT working group came up with a list of features it anticipated would be offered as options in luxury car models for the year 2005, and calculated their electric power demands for both summer and winter.
Some of these loads, such as the water pump, replace existing mechanical loads; some, like the electrically heated catalytic converter, are a response to regulatory mandates; and some are suggested by safety or driver convenience, the heated windshield, for example. Most of the new loads would benefit from a voltage higher than 12 V dc, and several, such as the electromechanical valves and active suspension, would require much higher voltages.
Incidentally, the group took an electrically driven air-conditioning compressor into consideration as a future electrical load. Its chief advantage would be packaging flexibility, since a drive belt would no longer be needed and the compressor could be placed anywhere in the vehicle. Also, with electric drive, the compressor speed could be varied to match the cooling load efficiently. But the load would be very large, and since it is now supplied effectively by a mechanical drive, the merits of conversion to an all-electric compressor drive in high-end cars are controversial.
There is little doubt, though, that the green car envisioned by the Partnership for a New Generation of Vehicles will be electrically air conditioned. That midsized car will have to be designed to cut the cooling load to 1.5 kW, however--about half of what today's equivalent cars consume.
The loads that are included on the list seem amply justified. A case in point is the active suspension system, whose introduction may very likely parallel that of air conditioners in automobiles. Thirty years ago, few cars had air conditioning; today it is difficult to buy a car in the United States without it. When first introduced, air conditioning was a very expensive option--up to 10 percent of the purchase price of the car. It also put a big dent in fuel economy at a time when the price of gasoline was higher in real dollars than it is today. But the comfort it afforded was so valued by consumers that within a decade, air conditioning became de facto standard equipment.
The MIT group foresees a similar future for active suspension systems, which keep the passenger compartment on a flat trajectory as the car wheels swerve and bounce over potholes, ruts, and rough roads. The systems sense a vehicle's vertical accelerations and energize electromechanical actuators to counter them [Fig. 3]. Expensive, heavy, and power-intensive (although not energy-intensive), an active suspension system will need voltages of about 350 V to operate efficiently. But it can improve a rider's comfort to the same extent that air conditioning did, and car buyers will probably embrace it just as enthusiastically.
No timing chain
Another case in point is electromechanical valve control. In today's engines, a camshaft acts on the valve stems to open and close the valves. As the crankshaft drives the camshafts through gears or a chain or belt, the timing of the valves' openings and closings is controlled by the cam design, and is fixed relative to piston position. This means that engine performance (in terms of emissions or fuel economy) is optimal over only a narrow range of engine speed.
If the valves were electromechanically actuated, however, they could be opened and closed without regard to crankshaft position. They could operate optimally at all engine speeds, torque levels, temperatures, and any other variables the designer includes. In fact, valve timing could be made part of a closed-loop emission-control system.
Moreover, an electromechanical system would eliminate the heavy and complicated camshafts and timing chains or gears. The valves would be actuated by sending current pulses through spring-loaded solenoids with the valve stems as their cores.
Electromechanical valves offer other interesting possibilities. For example, the valves can all be opened at the beginning of engine start-up, relieving compression and greatly reducing the cranking torque needed, so that smaller batteries and starter motors could be used. In fact, the peak power of the starter motor might be close to that of the alternator, so that a combined starter/alternator might become feasible. The starting torque might even be so low that the engine could be turned over through the fan belt at start-up, and the combined starter/alternator could then simply be mounted in the place now occupied by the alternator.
Start without moving
Even more intriguing, when combined with direct fuel injection, electromechanical valves may be able to start an engine statically, with no initial rotation whatsoever. Valves to the appropriate cylinders would be closed, and fuel would be injected into them and ignited, turning over the engine. If static starting should prove feasible, the battery could be designed for energy storage only, not for cranking power, and its size could be much reduced.
Another new kind of load, the electrically heated catalytic converter, is a direct response to environmental concerns and mandates. The electric heater will get the catalytic converter up to temperature quickly, which is important because a converter can reduce nitrogen emissions only when it is hot; the electric heater ensures that it gets hot within a few seconds after engine start-up instead of the many minutes required for heating by exhaust gases. In fact, one proposal calls for preheating the catalytic converter from a dedicated battery before start-up. Such heaters could eliminate a large source of pollution--emissions from cold engines.
Learning from MAESTrO
In examining the cost/benefit attributes of active suspensions, electromechanical valves, and other electric loads, the MIT working group used a computer program called MAESTrO (for Multiattribute Automotive Electrical System Tradeoff), which MIT's Laboratory for Electromagnetic and Electronic Systems originally developed for Mercedes-Benz. The program takes as its input the network topology and load types, voltages, powers, and duty ratios (load factors). It then determines the parameters of the loads (such as cost, weight, and efficiency), wires (such as gauge and loss), converters (weight and loss), transformers, and other components from its built-in models, and produces as output the system attribute values for specific designs showing how they compare [Fig. 4].
For example, design A may be lighter than design B but may cost substantially more. Design B may cost half as much as C, but C may be considerably more energy-efficient. With these MAESTrO-generated plots, it is easy to identify a "Pareto-optimal frontier"--that is, a line along which all designs are equally "good" but have different degrees of acceptability.
With this program, unacceptable designs can be eliminated, narrowing the field of acceptable possibilities and making final selection a tractable and structured task. With the help of this program, the MIT working group was able to probe the key issues and reach a consensus with deliberate speed.
And what is the consensus? Not surprisingly, the group quickly decided that the present 12-V dc system cannot be upgraded to handle future electrical loads; it would cost too much, weigh too much, and be too inefficient. Instead, the 12-V alternator will be replaced by a more efficient, higher-voltage design. Instead of being internally regulated at a constant voltage; the its voltage (and power) will increase with speed, so that its full capacity is exploited. An external power electronic interface will take that unregulated output and produce well-regulated ac or dc--or both.
The battery will still be a 12-V unit, though, because it will be the most reliable and economical electricity storage medium available. New battery technologies may eventually be practical for automobiles, but not by 2005. Doubling the battery voltage to 24 V would have a disproportionate effect not only on cost per ampere-hour, but also on reliability because of the greater number of cells and thinner plates.
An essential part of the new electrical distribution system will be load management. It will ensure that advanced loads, such as the active suspension, get the high power they need for the brief time they need it, coordinating power demands so that the alternator and power converters will not have to be sized unrealistically. It will also ensure that safety-critical loads such as power brakes and steering take precedence, and that "key-off" loads--the clock and security system--can draw power when the engine is off, but not discharge the battery to the point where the car cannot be started.
Loads will be switched on and off by semiconductor switches controlled through a data bus of optical fiber or copper. The power MOSFET is the switching device of choice because it is efficient and rugged and its manufacturing technology is well understood and widely practiced. Moreover, MOSFET power switches can be driven directly by a car's electronic control units.
Whether the distribution is ac, dc, or a mixture of both will depend on manufacturers' individual choices, and their assumptions about cost, performance, manufacturability, controllability, repairability, adaptability, and reliability. Whatever the electrical distribution architecture, it should be compatible with existing 12-V dc loads. This design will allow the present 12-V dc infrastructure to be used while new, more efficient, or functionally improved loads at other voltages are introduced.
As happens now, lamps and motors will account for most of the electrical load. Incandescent lamps will continue to be powered at 12 V, regardless of whether the power supply is ac or dc, because tungsten filaments for low voltages are shorter and less fragile than those that operate at higher voltages. In fact, short-filament 6-V incandescent lamps are so rugged and so much easier to focus than 12-V units that some manufacturers may return to them, especially for cars with ac distribution, where transformers make downconverting from the high-voltage bus especially easy.
For headlights, high-intensity gas or metal-vapor discharge technology will probably replace tungsten. They are already available as an option in the new Mercedes-Benz E-class car, providing twice the light at half the power.
Motors and actuators will definitely benefit from a higher, regulated voltage; at 12 V, motors lose 15 percent of their energy in the brushes alone. The dc motor, brushed or brushless, will be the choice for most applications such as fans and pumps because of its efficiency at high voltage, as well as its easy control, low cost, and high quality and reliability.
Where high torque at low speed is needed--in window lifts and windshield wipers, for instance--the ultrasonic motor with its low-profile pancake-like shape is promising. The Lexus already uses ultrasonic motors to adjust its headrests.
The upper limit on motor voltage will be set by safety concerns and semiconductor device voltage limitations. Many auto makers favor 42 V dc, which can readily be handled by power MOSFETs made by the standard 60-V process. It corresponds to the average full-wave rectified output of a single-phase 48-V rms ac distribution source. Also, it is not so high as to present a serious shock hazard to humans who accidentally touch it.
The typical model
If these predictions prove correct, what will a "typical" top-of-the-line automobile in 2005 look like? Although, certainly, variations among manufacturers can be expected, the model car has an alternator, directly driven by the engine, that produces a variable-frequency, variable-voltage output [Fig. 5]. From the alternator, a front-end converter creates two outputs: 25 kHz ac at 48 V rms for the main power bus and 12 V dc for charging the battery. The dc-ac portion of the converter is bidirectional so the battery can supply essential loads when the key is off. The starter is still 12 V and gets its power from the battery; the power-hungry electrically heated catalytic converter takes its power directly from the alternator.
To minimize wiring complexity, the ac main power bus feeds distribution boxes located throughout the vehicle. Containing MOSFET switches and fuses, the boxes are controlled by a separate data communication network that allows each box to convert the ac bus voltage into the voltages needed by the loads it serves. For example, a distribution box near the trunk might provide appropriate voltages to the tail lights, fuel pump, rear window defroster, retractable antenna, and an audio system. Another distribution box in the passenger compartment might control door locks, windows, side-view mirrors, compartment lights, and seat heaters.
Figure 5 can also be used to visualize an exclusively dc high-voltage architecture. In this case, the front-end converter will be an ac-dc-dc converter and dc-dc converters will replace the transformer-rectifier combinations to provide voltages different from the bus voltage, which will probably be 42 V dc.
The details of the models of the future will of course vary, and some manufacturers may choose to eschew ac distribution altogether. But clearly the new models will be more power hungry than today's already complex high-end automobiles, with their 1500 wires, innumerable branch points, as many as three dozen microprocessors, and more than two dozen sensors.
The 1918 edition of Putnam's Automobile Handbook--The Care and Management of the Modern Motor Car (by H. Clifford Brokaw and C. A. Starr: G. P. Putnam's Sons, New York) told readers: "It takes good 'juice' and lots of it to run a modern auto; not the kind Uncle Sam has put a ban upon, [but] the electric 'juice.' " With the possible exception of the nostalgic reference to the Prohibition Era, that observation is pertinent today and will be even more true 90 years after it was written.
About the Authors
John G. Kassakian (F) is professor of electrical engineering and director of the Laboratory for Electromagnetic and Electronic Systems at the Massachusetts Institute of Technology, Cambridge, where he works in power electronics and automotive electrical systems. He is founding president of the IEEE Power Electronics Society and coauthor of the textbook Principles of Power Electronics (Addison-Wesley, Boston, 1991).
Hans-Christoph Wolf is in charge of developing future power train management platforms at Mercedes-Benz AG, Stuttgart, Germany. He was previously responsible for developing advanced electric power distribution systems for the Advanced Engineering Group there.
John M. Miller (SM) is staff technical specialist, Vehicle Electrical Systems Department, at Ford Motor Co., Dearborn, Mich. His principal interests are control of electric machine drives and actuators, and power distribution system architecture. He is active in the PNGV.
Charles J. Hurton is manager, electrical subsystems and planning, at General Motors Corp.'s North American Operations Engineering Center Division, Warren, Mich. In earlier assignments at General Motors, he was manager, electrical component applications, and manager, medium-duty truck vehicle electrical systems.
Spectrum editor: Michael J. Riezenma
For a description of multiplexed digital communication buses in motor vehicles, see "The Thick and Thin of Car Cabling," by Mark Thompson, IEEE Spectrum, February 1996, pp. 42-45.
The authors will present their findings in detail in "The Future of Automotive Electrical Systems," at the IEEE Workshop on Power Electronics in Transportation, to be held in Dearborn, Mich., in October.
Two documents constitute the starting point of the discussions of the working group at the Massachusetts Institute of Technology. One is a seminal report by the Society of Automotive Engineers' (SAE) Dual/High Voltage Study Group, "Dual/High Voltage Vehicle Electrical System," by J. Vincent Hellman and R. J. Sandel, SAE paper 911652. Another is a comprehensive paper on high-voltage automotive systems, "Design Consideration for Higher Voltage Automotive Electrical Systems," by M. Matouka, SAE paper 911654. Both appear in the Proceedings of the SAE Future Transportation Technology Conference and Exposition, Portland, Ore., August 1991.
For information about emerging electrical functions in automobiles, see "The Future of Vehicle Electrical Power Systems and Their Impact on System Design" by G. A. Williams and M. J. Holt, Proceedings of the SAE Future Transportation Technology Conference and Exposition, Portland, Ore., August 1991, and "Control of Engine Load via Electromagnetic Valve Actuators" by Mark A. Theobald, B. Lesquene, and R. Henry, SAE paper 940816, February 1994.
For examples of MAESTrO analyses, see "Alternative Electrical Distribution System Architectures for Automobiles," a paper by K. A. Afridi, R. D. Tabors, and J. G. Kassakian in the Proceedings of the IEEE Workshop on Power Electronics in Transportation, Dearborn, Mich., October 1994.