For the PDF version of this article, click here.
Electro-hydraulic and electric power steering (EHPS and EPS) systems are demanding applications for automotive electronics. As vehicle curb weight increases, system torque and power requirements increase, giving rise to trade-offs between cost and thermal management. Once power levels of 800 W to 1 kW are reached, designers need to carefully consider these trade-offs.
Figure 1 shows a typical configuration for an EHPS system. In this system, a brushless dc (BLDC) motor is driven by an electronic control unit (ECU), which contains a power (inverter) stage necessary to convert dc power from the battery into 3-phase ac required to drive the BLDC motor. The motor is used to drive a hydraulic fluid pump, the fluid providing steering assist via a rotary valve and rack mechanism similar to that used in traditional hydraulic power steering systems. A sensor on the steering column measures the driver's inputs to the system as angular position.
A simple schematic of the ECU's power inverter stage is shown in Figure 2. The figure shows a die-on-leadframe (DOL) approach pioneered by International Rectifier.
DISCRETE MOSFET PACKAGING: D2PAK
As we probe how recent technology advancements have enabled efficiency improvements and cost reduction in a typical 1 kW EHPS system, the following sections will cover discrete MOSFET packaging improvements, power module thermo-mechanical design, and algorithms for reducing torque ripple.
As silicon technology has evolved to lower RDS(on) capability, packaging technology has become the limiting factor in the current carrying capability of discrete plastic packages — the limiting factor in a 3-lead package usually being the source lead for larger die. For example, a standard 3-lead D2Pak is limited to between 75 A and 100 A, depending on the lead cross-section and supplier's specification method. For 1 kW EHPS systems, this becomes a limiting factor, forcing designers to use multiple discrete devices in parallel or a module approach.
System cost reduction in 1 kW EHPS (and EPS) systems can be achieved by increasing the current handling capability of the discrete package (usually D2Pak). Some suppliers have addressed this by increasing the cross-sectional area of the source lead, while others have increased the lead count in a standard D2Pak footprint, such that the current is carried by five leads in parallel rather than one. The latter has the advantage of reducing the lead temperature at the board level, resulting in greater reliability of solder joints and doubling the current handling capability of a standard 3-lead package. Furthermore, the extra leads allow a larger lead-frame T-post inside the package, resulting in additional wirebond area capability and reduction of the die free package resistance to less than half the standard package. This leads to an overall reduction of RDS(on) in the discrete part.
DIE ON LEAD-FRAME TECHNOLOGY
Figure 3 shows a photo of the package (a) and the internal wirebond configuration (b).
For systems in the 1 kW range, a cost-effective module technology for the power inverter can have advantages over a discrete solution, especially for higher-end systems requiring lower loop resistance and inductance. Also, package improvements, specifically in the area of lower thermal and electrical resistance, mean smaller MOSFETs (higher RDS(on)) can be used in the design. This can reduce the overall cost of the module, as the largest single component of cost in an EPS or EHPS power inverter is typically the silicon power MOSFET devices. Conversely, for the same-size silicon, lower losses and greater efficiency can be achieved in a given module.
Most automotive modules today use IMS (insulated metal substrate), DBC (direct bonded copper, to a ceramic substrate), thick film substrates or PCB-based “modules” in lower current applications, which primarily use discrete power devices. A new concept in power inverter packaging in this medium to low power range is DOL. In this case, there is not a separate substrate, but the MOSFETs are mounted directly onto the same insert molded lead-frame used to make the terminals for external connection, which are continued into the housing of the module. Figure 4 compares a DOL approach (a) vs. a more conventional approach using die on DBC (b).
The advantage of DOL is that it eliminates the substrate without adding an equivalent cost to the housing/leadframe. Typically, a PCB containing the primary system control and gate drive circuitry is located in close proximity, resulting in a cost-effective approach to system design.
Table 1 compares typical values of junction to heatsink thermal resistance and loop resistance and inductance (from battery plus terminal to motor and back to battery minus terminal) for DOL and DBC approaches in a 1 kW EHPS system. In the DOL case, measurements were performed on the module shown in Figure 2. As can be seen, the DOL solution provided lower values in all cases, leading to lower dissipation, improvements in efficiency and an overall lower bill of materials cost than the DBC solution.
CONTROL ALGORITHMS: TORQUE RIPPLE REDUCTION
The DOL power module demonstrates advantages in many package performance metrics. However, one disadvantage is that it is not an electrically insulated module. While other module package technologies using substrates have the desired electrical insulation, the inherent insulation has the disadvantage of leading to higher thermal and electrical resistance. To gain the maximum performance from DOL modules, a highly filled silicone adhesive was used to secure the bottom surface of the module to the heatsink. This material provided good thermal transfer and electrical insulation. The thickness of this material must be carefully controlled, as thermal performance is a function of the adhesive thickness shown in Figure 5. Nubs in the plastic housing on the bottom surface of the module were used to control the adhesive thickness.
Further improvements were made by using a proprietary control algorithm to reduce the torque ripple effect in the system. This is a key figure of merit because it affects the driver's steering feel and can influence the driver's impression of the entire vehicle. Note that it is more of an issue with EPS systems due to the absence of hydraulics between the ECU and the steering gear.
The ECU is both a source of torque ripple and a means to compensate for some of the parts of the system. There are two sources of torque ripple: 1) reluctance or cogging torque, and 2) unmatched flux linkages. Cogging torque can be explained by Figure 6. Flux naturally seeks a path of least reluctance. Therefore, if the motor exhibits magnetic saliency, the flux has a natural tendency to align itself with the stator teeth, as shown on the right-hand side of the figure. This represents the position of minimum reluctance. This variation in reluctance creates a torque pulsation that is a function of the rotor position around a stator tooth and is inherent in the motor design.
The second source of torque ripple is unmatched flux linkages. If the shape of the current waveform deviates from perfectly matching the shape of the flux waveform, then the interaction between the current and flux (the flux linkage), which creates torque, will not be constant. Torque pulsations will result, as illustrated in Figure 6b.
Ideally, the motor would have sinusoidally distributed flux to avoid saliency and ultimately reluctance torque. To avoid unmatched flux linkages, a perfectly sinusoidal flux, and therefore perfectly matched sinusoidal currents, is required. In practice, however, the flux often deviates from sinusoidal (motor-inherent) as does the current (ECU-inherent), resulting in torque ripple.
The currents may deviate from perfect sinusoids for several reasons:
- Inverter dead-time (time between turning off one switch in a totem pole and turning on the other).
- Discretization of position feedback.
- Discretization of current feedback.
- In general, discretization, inaccuracies and nonlinearities inherent to the ECU.
Advanced algorithms are used in EPS systems to compensate for inverter dead-time, usually by intentionally modifying the switching patterns to produce the desired resulting sinusoidal currents. In other words, dead-time results in the actual voltage produced by the ECU being different from the commanded voltage. Dead-time compensation algorithms modify the commanded voltage (PWM switching pattern) so that the resulting voltage, after the application of dead-time, is as desired.
Figure 7 shows a 59% reduction in torque ripple achieved by adding a dead-time compensation algorithm to a traditional vector control approach. In the figures, angular position of the rotor is shown in yellow while torque is shown in green.
The ECU was used to compensate for the cogging torque inherent in the motor. Because ECUs need significant computing power to implement their functionality, these computations were achieved with little or no additional cost. Since cogging torque can be thought of as higher-frequency flux harmonics, adaptive filtering techniques were used to inject curent harmonics to essentially cancel the motor's flux harmonics.
Figure 8 shows the algorithm block diagram to implement this approach. The torque reference was modified by a position-dependant function that attempted to reduce the torque ripple. The coefficients of this function were chosen adaptively by an “adaptation law.” Convergence of the error to zero was achieved by using Lyapunov's stability theory. As time increases, the algorithm adapts such that the harmonics existing in the flux become present into the current reference and cause the motor current to better match the nonsinusoidal flux.
Figure 9 illustrates a 35% reduction in torque ripple with use of adaptive torque ripple compensation.
Three technologies: higher current discrete packaging, die on leadframe technology and advanced control algorithms, can be used to improve the efficiency and cost of EPS or EHPS systems as they transition to higher curb weight vehicles.
DOL can improve efficiency and cost by eliminating the module substrate interface and improving thermal and electrical interfaces. Further improvements can be made by using an advanced control algorithm to improve steering feel and reduce torque ripple.
In the case of discrete packaging, additional current handling capability coupled with the latest silicon technology, namely trench MOSFETs, can result in a cost-effective solution for low-power EPS or EHPS systems by eliminating the need for parallel components or transition to a module approach.
ABOUT THE AUTHORS
Jim Tompkins is director of engineering for International Rectifier Automotive Systems, Halifax, Canada. He has 16 years of embedded systems development experience and is focused on steering systems.
Jean P. Quirion is a development engineer at International Rectifier Automotive Systems. He has three years of experience in embedded system development, focusing on control software design.
Bill Grant is director of engineering, advanced packaging for International Rectifier's automotive systems design group in El Segundo, Calif.
Anthony F.J. Murray is responsible for technical marketing of automotive components for North America for International Rectifier, El Segundo, Calif.