From an eco-design context, the total value chain of a product is a long one. It starts with the raw materials, and continues with the design of the product, its production, the transportation to the end customer, the usage, the recycling, and the consequential consumption. At each stage of this life cycle, energy consumption and the production of hazardous waste are considered and the impact quantified. The "unwanted" output consists of emissions such as heat, waste water, greenhouse gases, process chemicals, even more greenhouse gases during transport and usage, and upon recycling even more chemicals and landfill. Quantifying all the impacts in dollars and cents, and summing it all up results in what is called the "lifecycle cost".
It is obvious that in this case less is more. What can be done? Since the biggest impact - at least with energy-using products - is typically to be gained with improved efficiency during operation and standby, a lot of the attention must turn to power supplies and electric motors.
In our daily life, most applications contain either power conversion or motion control subsystems, as the picture below implies. Structuring the application landscape into this dichotomy makes the challenge much easier to address.
The power conversion subsystem basically addresses all AC-DC and DC-DC conversion. Taking a closer look at the topologies used, in the overwhelming majority, a DC voltage is converted to another DC voltage using a switch-mode conversion circuit. Even in the offline power supply, one of the first things to find at the input is a rectifier, so - strictly speaking - an AC-DC power supply is in most cases a DC-DC converter, anyway.
Click on image to enlarge.
The motion control subsystem basically works the other way around. Here, a DC input voltage is used to generate alternating waveforms, appropriately shaped to make a motor turn. Sometimes the term DC-AC is used for these systems and other times, it is called frequency inverters. Most motors have three phases, and these inverters have three outputs, again using a switch-mode circuit to create the phase-shifted waveforms for the motor. Three phases really is the minimum to determine a sense of rotation and make the motor start in the desired direction (with the exception of DC brush and switched-reluctance motors).
The key drivers for improvement for all three are improved performance of the switches and control circuits used; the trend to integrate more and more into the power electronics system; and a change in the value chain of manufacturers who produce of this equipment. For instance, many manufacturers consider a power supply as a necessary evil because it is difficult to design, costs space and money and generates heat, but does not add marketable advantages and features to their end product. These manufacturers tend to focus on other aspects of the product more important to them. In these cases, the power supply design must come from somewhere else, and the semiconductor supplier can make a big difference by providing good solution support.
Switch-mode topologies have been around for a long time, but with most of the energy still being lost in the power switches, most of the potential in improving these designs is through improved power switches. It is interesting to note that over time, with characteristics of power switches improving over the decades, the choice of conversion circuits has evolved and changed as well. Today, flyback converters are being used up to 150W and beyond, with the power range up to 400W - that used to be covered with full-bridge converters - now is addressed with half-bridge converters. Not to forget the improvements in control circuits and passive components, where new control schemes and tighter tolerances have allowed a widespread use of resonant topologies, improving efficiency and reducing electromagnetic emission further.
Click on image to enlarge.
The device landscape has seen huge developments in the last twenty years. Bipolar transistors have been replaced by MOSFETs, where big gains in RDSON and robustness have been made, not to forget the combination of a bipolar transistor with a smaller MOS to drive it monolithically, the so-called IGBT.
Click on image to enlarge.
The picture above shows a cross section of a vertical planar MOSFET, as a schematic representation and a cross-section.
A insulated gate bipolar transistor, although containing two elements instead of one, is not much more complex (as shown below):
Click on image to enlarge.
The symbol on the right indicates how a bipolar transistor is being driven by a MOSFET, and the left picture shows the vertical structure of the device as it is implemented in silicon.
Since MOSFETs are capable of higher switching frequencies (hence, smaller inductors), the applications that require relatively low current and fast switching operation or linear characteristics of I-V are usually built with MOSFET, whereas higher-power applications that require higher gain as well as high current and moderate switching operation are usually built with IGBTs. These are also easier to scale up in breakdown voltage, with the most common values of 1200V, 1700V and 3500V for higher-power systems - values that are next to impossible for MOSFETs, let alone commercially.
So where are the big differences between these and the ideal switch? First of all, to drive the real switch some power is required. This power has to be provided by the gate driver. As both the MOSFET and the IGBT represent a capacitive load, the power required can be calculated with the gate capacitance, the required drive voltage, and the switching frequency. A bipolar transistor requires a base current that, in the case of an IGBT, is elegantly taken from the high power rail and sunk into the load. Since this power can be quite high, bipolar transistors are being used in switch-mode power supplies in only very few cases these days.
As the gate represents a capacitive load, high peak currents can occur when the gate driver switches. These peak currents are in direct correlation to the switching speed of the main switch, and that can be a good or a bad thing. Usually, fast switching is desired since the device will spend less and less time in the "linear" region (between fully on and fully off), but faster current change dI/dt in the circuit can lead to unwanted side effects, like high peak voltages that can destroy the switches or other components. And, fast switching inevitably creates electromagnetic emission that needs to be filtered away to comply with the regulations.
Another difference between the ideal switch and the MOSFET or IGBT is that the on-resistance of both devices is non-zero, leading to conduction losses. In the case of the IGBT, it is even worse - a more or less constant voltage drop across the device will lead to on-state losses that can be quite high especially at low loads.
The third difference is that parasitic capacitances in the devices store energy and release it exactly when the device is changing state from on to off or vice versa. These losses can be significant, and they cause power dissipation even when there is no load attached.
As our original quest is to improve the efficiency of our power subsystems, it is the power switch that has most impact, and here mostly these three effects (gate drive, switching and conduction losses) that offer a path to improvement. So what's new? The picture below shows the path across various device structures from planar to vertical and super junction MOSFETs.
Click on image to enlarge.
The picture above shows the classical planar MOSFET like above, in comparison the picture on the right shows a super junction MOSFET produced with multiple epitaxial layers.
Moore's law is unfortunately not applicable here. As lithographic equipment improves, device structures can be reduced and more active transistors can be squeezed in the same area. But two effects limit this: First, the electric field must remain below a certain strength, else the device structures will break down internally. Second, should an overvoltage be applied to the device, it is desirable to absorb the energy in a controlled way, requiring not only specific structures but also enough silicon volume to not destroy the device.
But not everything is bad. With the reduction in size of the individual transistor cells (usually arranged in stripes), the on-resistance could be vastly improved. So, for a given RDSON value, the chip became much smaller and more cost-efficient, but also required less gate drive power (this was also driven by improved device structures that reduce the internal capacitances at the gate). Today, a power MOSFET with a breakdown voltage of 600V at an on-resistance of less than 85mOhm (in a TO220 package) is possible, and that will reduce power losses in that particular spot more than 2x over previous generations.
A significant further improvement came with reducing the resistance of the bulk semiconductor material in the vertical transistor. In a power MOSFET, most of the action happens in just a few micrometers from the surface, and the thickness is just required for mechanical handling and to allow the depletion zone to stretch into the device depth, to not exceed the maximum electrical field strength. It is not surprising that significant development has taken place in making the wafers thinner, involving improved handling, to get rid of that unwanted resistance in the current path. In the newer device structures called "super-junction" MOSFET, the n-doping of the device is increased to reduce this resistance even further, and that is offset with p-doping brought into the bulk of the device so that the overall charge balance is preserved.
In the case of IGBTs, applying trench technology to reduce the size of the on-chip lateral isolation structures helps to reduce the chip size while maintaining performance. But these trenches need to support significant isolation voltages, so this technological step was not easy to achieve. The result is 25% lower conduction losses, and 8% lower switching losses, compared to previous generations. For heating applications, the widespread use of IGBT-based induction heaters then allowed to improve the efficiency from approx. 40% (with gas) to over 90%.
How will this new devices shape the application landscape? These days, energy savings and new regulations around the topic are more important than ever. These regulations can be addressed with existing circuits and devices. But it cannot be done while maintaining existing cost levels. Power switches from Fairchild Semiconductor provide a new cost-performance ratio that makes life easier for many R&D engineers that are asked by their company and their customers to make the power subsystem more efficient.
It is important to note that these new, smaller devices, really enable efficient integration of power subsystems in multi-chip packages too, for covering reasonable power levels. With previous technologies, power was always too limited, since with the thermal limitations of the packages and the excessive heat these solutions made no sense. This is changing now, with a multitude of IGBT modules mainly for motion control, induction heating and welding applications, and first power supply modules beyond the FPS power range already available from Fairchild Semiconductor, and more to come.
It was (again) possible to improve the state of the art where it matters most in power subsystems. Developing such a new technology and making sure it works under all imaginable circumstances can take several years, and is a multi-million dollar undertaking - not something for the faint at heart. But it is worth it, as more and more products are being scrutinized for saving energy, leaving more money in the end users' pockets and saving the planet in the process. So, it's just a switch - or is it?
