Automotive Packaging Is a Powerful, Problem-solving Tool
Protecting Silicon From Extreme Environments
BY ALEXANDER CRAIG AND STEPHEN MARTIN
Automotive devices run at high temperatures for long periods of time and may experience junction temperatures above 200°C for a short time during load dumps. Modern package designs minimize package contributions to RDS(ON), reducing the normal operating temperatures. Packaging also improves thermal resistance, further reducing junction temperatures.
As with all electronics, automotive electronics are being driven to miniaturization. Integrating control circuitry within the power device package minimizes size, but adds complexity to an already challenging goal for package design. Not only does the package have to provide good thermal dissipation for the power die, it also must electrically isolate the control die from the high voltages and currents associated with the power die. The drive to make things smaller also makes the thermal management task more difficult as the area for thermal dissipation decreases, even though the amount of power remains the same - or even increases.
Not only is the thermal density increasing, electronics are being used in higher temperature locations within the car. The environment can range from 200°C inside the transmission to an ambient temperature of 165°C when on a spark plug, to 150°C when mounted in the engine compartment, to the relatively benign environment of the passenger compartment at a maximum of 80°C. An automobile is estimated to have 6,000 cold starts over its lifetime, where the temperature could cycle from 40° to 150°C within the engine compartment. Protecting the silicon from the extremes of the environments and the associated stresses is part of the package function.
With the drive for circuit miniaturization and higher temperature environments comes the need for understanding the thermal limits and management of power semiconductors. This will ensure designs that will continue to provide the reliability required by the automotive market. Packaging has moved well beyond merely a chip carrying and chip-board interface element to become an even more powerful tool in solving problems.
Effect of Temperature on Silicon
Increasing temperature adversely affects the performance of power devices. For MOSFETS: the RDS(ON) goes up as the temperature increases, causing power loss; the brake down voltage (BVdss) goes up as temperature increases; leakage increases exponentially as temperature increases; and the threshold voltage goes down as temperature increases, making it difficult to turn off the gate at high temperatures.
For PIN diode, the forward voltage drop goes down as temperature increases; and the reverse recovery charge and time go up as temperature increases. For punch-through IGBTs like those used in ignition systems, the VCE (SAT) decreases as temperature increases; the threshold voltage goes decreases as temperature increases; switching time with inductive loads increase as temperature increases; leakage increases exponentially as temperature increases; and the BVdss increases as temperature increases.
From a power device perspective, Tj is the most critical factor. Most failures result from forcing Tj too high. This equation summarizes the point:
The ΔT represents the degrees above some content temperature at some distant infinite heat sink. For the vehicle, that infinite heat sink is the in-coming air and the content temperature used for vehicle designs is the classic temperature of 50°C. But that air is used cool the engine via the radiator. In general, the electronics modules experience a far hotter infinite heatsink temperature. For the power device in most modern power train designs, the infinite heatsink is the 105°C air moving across the module’s heatsink. Tables 1 and 2 show typical automotive conditions. The board temperature is often as high as 135°C.
Table 1. Examples of temperatures in typical automotive conditions.
Table 2. Required operating temperature for automotive electronic systems (courtesy of Toyota Motor Corp.).
In the past few years, power MOSFET devices have dramatically improved their specific on-resistance. For the silicon, die size is a large factor in the (VON × ION) and Rth portions of the above-mentioned equation. The improved on-resistance enables a smaller die to have the same resistance of a larger, older MOSFET. This smaller die, however, will have a higher thermal resistance. The onset of trench structures and resulting improvements to this technology has resulted in significant advancements, specifically to on-resistance (Figure 1). This means the power density has gone up by almost an order of magnitude in the past 10 years. Unfortunately, the thermal performance of the interface for the power device in the engine control unit (ECU) has not kept pace. In fact, the desire for surface-mounted devices has grown and modern ECUs no longer have the power devices connected directly to the heatsink. Where a power MOSFET was once in a TO-220 connected to a heatsink, today that same function is most likely performed with a DPAK soldered to a PCB with vias to an isolation pad to a heatsink.
Figure 1. Improved on-resistance with trench structure.
Smart power devices need to process both power and data. It is often more cost effective to use a silicon process optimized for signal processing for the smart functions of the device and use and entirely different silicon process optimized for power devices. The separation of processes leads to the need to reintegrate these different die into a package that provides interconnects between the power die and the signal-processing die and to the external circuitry. The package provides the power handling, die-interconnect power and signal connections, and possibly die substrate isolation along with the physical support and environmental protection. Low thermal impedance from the power junction to the case of the package is required to allow thermal cooling of the power devices. The thermal resistance impact is represented in Equation 1 by ‘Rth.’ Low thermal resistance is achieved by having the metal leadframe that the power die is attached to extend to the surface of the package. Solder die attach is required to offer the lowest thermal and electrical impedance to the back surface of the power device. Use of a non-conducting epoxy or a polyimide tape provides electrical isolation of the control die from the electrical potential of the back of the power switch die.
Figure 2. Inside a smart power device.
Figure 2 shows the inside of a 3-paddle, 5-die assembly. This multi-die packaging provides isolation between die substrates, low thermal resistance for the power die, and the ability to interconnect two separate smart power devices. In this device, the two control die have 12 interconnects, each using small gold bond wire to minimize control die size. The control IC is isolated from the power die by using non-conductive adhesive die attach. Power devices use thick aluminum bond wires for current handling and are solder die attached for high power dissipation. The solder connection of the power die to the DAPs and the DAPs to the circuit board provide minimal thermal resistance from the power die to heat-dissipating surfaces.
Figure 3. Scale comparison of PQFN package (left) and two control ICs.
The backside of a 12 × 8-mm power quad flat pack no-lead (PQFN) package, shown in Figure 3, combines the function of the two larger controllers into the one smaller package. This is accomplished by using the latest power MOSFETs for processing power and the latest IC technology for processing data.
Figure 4. Outline of a PFQN (left) and TO-252.
Using the PQFN-style packages minimizes size by eliminating the leads extending beyond the package body. Figure 4, which is shown roughly to scale, outlines two ignition IGBTs that use the same size die. But the PQFN package shown on the left is significantly smaller than the TO-252 (DPAK) on the right. This is of significant, because new automotive ignition designs are moving to switch-on-coil. This board space is a premium for such a circuit that has to sit in the pencil coil, which sits on the spark plug.
Automotive products typically are qualified to requirements found in AEC specifications Q100 for ICs or Q101 for discrete devices. Tests include operating life, temperature/humidity/bias testing such as HAST or H3TRB, power cycling, temperature cycling, high-temperature reverse bias (HTRB). In addition to reliability stress, characterization of the package material is necessary to understand performance trade-offs. Characterizations such as mold compound ionics, glass transition temperature, moisture absorption, and modulus at room temperature and elevated temperatures are some of the characteristics of interest.
Early in the evolution of semiconductors, packaging technology was primarily approached as a die-circuit board interface for signal processing. This is still true. With the increasing prevalence of power devices, however, the need to dissipate heat generated by the device to minimize inefficiencies in the semiconductor is driving the changes in packaging technology. The cost of silicon also is a factor, mostly affected by die size and process complexity.For example, since data silicon processing devices generally process low-voltage, low-current signals, the design and manufacturing processes for these devices is biased to make many low-power circuits - such as NAND and NOR gates and OPAMP - as small as possible. This process allows more die to be processed on the same wafer, lowering parts count costs. Complexity of design affects costs in the manufacturing process. Power devices, for instance, need to fend off high voltages and control the flow of large amounts of current.
The design and manufacturing process for power devices is biased to make a switch as efficient and small as possible. These two design goals traditionally have been quite different from the cost-saving goals of the manufacturing process. Today, new packaging technologies focusing on a single design and manufacturing process to produce a smart power device that is not optimized only for power or data control. Rather, modern package technology, such as PQFN, provides a silicon process to achieve a smart power device that optimizes silicon for both power and data control. This modern packaging process provides the reduced size, electrical, thermal, and environmental performance required for smart power devices in automotive electronics.
ALEXANDER CRAIG, staff applications engineer, and STEPHEN MARTIN, principal process engineer, may be contacted at Fairchild Semiconductor Corp., 82 Running Hill Road, South Portland, ME 04106; (207) 775-8100; e-mail: firstname.lastname@example.org, email@example.com.