Normally-On devices and circuits, SiC and high temperature: using SiC JFETs in power converters

Normally-On devices and circuits, SiC and high temperature: using SiC JFETs in power converters
of 7
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
  Normally-On devices and circuits, SiC and high temperature : using SiC JFETs in power converters Dominique Bergogne, Hervé Morel, Dominique Tournier, Bruno Allard,Dominique Planson, Christophe Raynaud, Mihai Lazar dominique.bergogne@insa-lyon.frAMPERE, INSA-Lyon, F-69621, France Abstract SiC is compared on a physical basis to Si. SiC devices offer lower On-resistance and operating temperatures over 300°C. It is shown that inverters using normally-On power switches do not differ from inverters based on normally-Off devices when the gate driver is taken into account. Basic building blocs of a Gate driver are proposed using solely SiC Normally-On JFETs. Principal issues with SiC JFET power converters are listed. 1SiC and high temperature   1.1SiC and other semiconductors In this paper we will consider SiC-3C and Si only, as they are commonly used for producing commercial devices or engineering samples. From now, SiC-3C will be noted SiC. Other semiconductors such as SiC-4H, SiC-6H and GaN appears on the graphics for information purposes, see fig. 1. As for all semiconductors used as switches, one can define two different states : conduction (On-state) and isolation (Off-state) implying to two different physical mechanisms. Such physics are not within the scope of this paper. The plots used here, fig. 1 have been computed by H. Morel using standard, physical or not, known models and material data from the literature [1]. SiC power switches offers much higher power capabilities than Si for two main reasons that can be deduced form two standard graphics, one for each state. 1.1.2 Specific resistance vs breakdown voltage Semiconductors are used to control the current flow in circuits, and therefore, must exhibit the lowest resistance as possible while conducting : the On-state. Devices designers have to accommodate for contradictory variations of performance versus design parameters such as doping, geometrical dimensions... The physics of semiconductors implies an increase of resistance to achieve a higher voltage breakdown. Conduction losses are directly impacted : high voltage devices will have high forward voltage drop due to high resistance. This is why designers have produced bipolar devices where conduction is not directly a matter of resistance, but, as always, this property is counter-balanced by an almost fixed voltage drop. This leads to a practical limit of interest between unipolar (resistance like behavior) and bipolar (diode like characteristic). It is well known that “for low voltage MOSFET (unipolar) are better than IGBT (bipolar)” because it is possible to reduce the resistance to such levels that the voltage drop is lower than the voltage threshold of bipolar devices.  Fig. 1: Specific resistance versus optimal breakdown voltage of common semiconductors.  Unipolar devices present lower losses due to conduction (forward voltage drop) at a given nominal current compared to Bipolar in the left part of the diagram of fig. 2. For high breakdown voltages, Bipolar takes oven in performance as due to an almost constant voltage drop. The performance curves cross at a point defining the practical limit for the choice between Bipolar and Unipolar with breakdown voltage as a parameter.It is commonly accepted that the practical limit in Si is in the 600V range and for SiC, it is around 4kV. This is why SiC-JFET are good candidates for household and industrial applications.The specific resistance versus breakdown voltage graphic, fig. 1, shows a significant lower resistance of SiC compared to Si, a ratio of 20. Hence, for a fixed breakdown voltage and a given current density, a theoretical reduction of conduction losses by a factor of 20. 1.1.3 Thermal runaway vs breakdown voltage When in the Off-state, semiconductors are used to disconnect voltage sources within a circuit. The performance here is the ability to sustain a high voltage while not conducting. Practically, thermal runaway occurs when the semiconductor's junction is not able to sustain the voltage. Temperature is set by the ambient and by losses within the device. Losses depend on electrical operating conditions and are also affected by temperature. Avalanche during switching, over current or short-circuit will raise the temperature within the device to eventually trigger thermal runaway.In other words, thermal runaway occurs when a power semiconductor device is no more able to block the voltage. Physically this is related to the property of the blocking junction of the device, i.e. a PIN diode. It is well known that the optimal breakdown voltage of a PIN diode corresponds to optimal values of the low doping layer width, W B , and doping level, N D . The intrinsic concentration increases with the temperature, so when the doping level N D  is reached, the junction looses its blocking capabilities. Based on physical models, H. Morel has plotted this limit for several semiconductors and placed commercially available devices on the graphic, see fig. 3. Practically again, it means that the device as turned into a fuse with no possible control. A short-circuit at full nominal voltage on a Si power IGBT will destroy the device in little more than 10 us. For a SiC power JFET it takes up to 10 ms. The difference, a ratio of 10 +3 , can mainly be explained by fig. 3. For a blocking voltage of 1200V Si devices show a temperature limit around 190°C while SiC exhibit a 950°C limit. SiC has a much higher headroom allowing an important temperature rise within the device before runaway occurs. As a consequence, SiC is a good candidate for high temperature operation. Several laboratories worldwide have demonstrated power converters using SiC power devices at temperatures over 300°C [2], [3] and [4]. 1.2 SiC power devices today Citing 'semiconductor TODAY' [5] : the only commercially available product is the Schottky barrier diode, which is now reaching the 1200V and 20A range. It is mostly used in high end power factor correction. At a recent seminar [6], Transic has presented the first commercially available SiC bipolar transistor rating 6A/1200V. Other companies provide engineering samples at a performance and quality level close to industrial standards, for example SiCED [7]. Nevertheless, twenty key player companies worldwide are massively cited in papers on SiC MOSFETs, JFETs, BJTs and IGBT. From data in [5] it appears that the most studied device is the MOSFET, see table below, but papers relate majors reliability difficulties related to the oxide layer.  Research activity  MOS- FET  JFETOn IGBTBJTJFETOff  Popularity75%40%30%15%15%  Fig. 2: Forward voltage drop versus optimal break-down voltage at fixed current density. In other words, conduction losses as a function of voltage rating. Fig. 3: Critical temperature versus optimal breakdown voltage. The voltage rating of planar junction deter-mines an absolute temperature limit.  Normally-On JFETs come second and are, from [8], in the process of being mass produced. Engineering JFET samples are rated 1200V with On-resistance as low as 50m Ω . Our feeling is that oxide-less devices, JFET and BJT, could arrive on the mass market soon. Packages for such devices are under study or development as the end users and electrical power designers are pushing for higher operating temperatures. 1.3Issues with high temperature Thermal stress, mechanical stress, packaging, solder, assembly, associated passive components, very high power cooling densities are amongst the usual addressed issues. High temperature means three things : • high temperature materials • availability of components • increased thermal cycling amplitudeStandard Si device packages and die attachment techniques do not meet requirements. New techniques for dies attachment at low or high temperature are developed along with packages. At now, the high temperature in the industrial high end applications means 175°C. For example a running industrial programs aims at 200°C [16] and academic study, CoTHT [9] addresses temperatures up to 300°C. The situation today, concerning high temperature systems at a pre-production state could be summarized in a chart, see fig. 4. The 200°C frontier is visible, above 250°C the lack of capacitors and control, driver circuits leaves an interesting space for research work. 2Normally-On devices and circuits Until recently, power semiconductor switches where normally-On devices, that is, in the absence of energy on the control electrode, the semiconducting structure was in the Off-state. Fewer and fewer electrical engineers and designers recall the time of thermionic valves, normally-On devices. Today the power JFET brings back memories, with a difference, Power is now mainly switching circuits. At power-up or during idle operation the switches are in the Off-state. It is 'natural' to accept that with no further thinking...The first paragraph deals with normally-On switches and converter structures. In the second paragraph we will look at both normally-On and normally-Off devices and the consequences on circuit behaviour and design issues for MOSFETs and JFETs. Then, JFET based circuits are presented to be used in high temperature all SiC converters. 2.0Normally-On and structures 2.0.1Voltage and current fed inverters Voltage fed inverters are commonly used using normally-Off devices, note that normally-Off devices are almost the only devices available. It seems natural to connect an inverter leg made off series connected open switches in parallel to a voltage source as in fig. 5. Anti-parallel diodes are connected to the bipolar transistors (BJT) to provide a path for the reverse current imposed by the load during operation. Some rule says : it is not possible to short a voltage source, which indeed is true! The drivers associated or not with the signal controller must implement via software and preferably hardware, and exclusion rule between a=1 and b=1 to present a short circuit on the voltage source. At power-up or during idling, a=b=0. In the event of an over load, the control circuit or the driver will set a safe state where load current is nulled by applying a=b=0, the load is disconnected from the voltage source.  Fig. 4: Tentative map of components for high tempera-ture electronics. Fig. 5: Voltage fed inverter using normallyOff BJTs  Voltage sources can be left open just as current sources can be connected to a short circuit. This fact makes us think that a current fed inverter should use normally-On switches to prevent any open circuit, as the rule says. Fig. 6 presents a current fed inverter leg implemented using normally-On JFET, series diodes are compulsory to ensure the reverse voltage blocking capability needed in such a structure.In such a structure the voltage across open switches is determined by the load. The worst case is an open load or an 'under load' leading to over-voltages on the switching devices. So a=b=0 must be avoided by using adequate logic functions implemented in the drivers or the control system. During power-up or idling, the inverter must present a short circuit impedance to the current source, thus a=b=1 is the safe state. The voltage across the load is forced to zero by switches being in the On-state.Current source inverter are commonly used in power induction heating systems. For moderate supply voltages, IGBTs are used in series with a diode. This shows that normally-Off devices can be used in series with current sources despite the feeling that normally-Off transistors could be in the Off-state with out the will of the designer. 2.1The differences between : Normally-On and Normally-Off 2.1.1 On the 'power side' The most common converter structure is the inverter leg, two power switches are connected to a voltage source and are controlled to be alternatively conducting, producing a square wave voltage on the output terminal, see below, fig. 7.Using normally-On or normally-Off devices has no impact on the 'power side', but on can note an inverting driver to keep logic identical. Input control signals are 'a' and 'b'. If 'a' is active, the high side switch is conducting.From the circuits of fig. 7 it is obvious that setting both 'a' and 'b' to the active level will produce a short-circuit on the voltage source. In both cases, normally-On and normally-Off, a=b=1 produces a short circuit on the voltage source. But the circuit, with switches drawn open, looks safer than the circuit with closed devices.In the real world, a=b=1 is never allowed by hardware, it might however happen due to electromagnetic interference (EMI) or severe malfunction of the driver, the usual consequence is the destruction of power devices. SiC JFET superior robustness to short circuit could provide extended fault tolerance, particularly to misfiring due to unpredictable EMI. 2.1.2 On the 'control side' Normally-On power switches are controlled using a  gate driver. This subsystem receives a control signal   to set the switches's state. The control signal is provided by the converter's controller. Fig. 8 represents a gate driver including some kind of  protection circuit   sensing the Drain voltage. A logic  core processes control and protection information. The Gate of a power JFET is biased by the output stage . It can be noted that a negative voltage is needed to set the JFET in the Off-state.  Insulation  is needed as power switches are connected to different high voltage sources. The Source reference is connected to the highest potential of output stage.Normally-Off devices such as MOSFETs are also controlled with a gate driver. The difference is the connection of the reference voltage of the output stage, see fig. 9. For a JFET, ground and Source are at the highest voltage level while a MOSFET Source and ground reference are at the lowest. A MOSFET requires a positive gate voltage to be 'On' while a negative bias will turn Off the JFET. For both devices, high gate level means On-state and low gate level is Off-state. Thus the logic circuit within the gate driver is the same for both JFET and MOSFET.  Fig. 6: Current fed inverter using normally-On JFETs. Fig. 7: Voltage fed inverter using (left) normally-Off devices and (right) normally-On devices. Fig. 8: Normally-On JFET gate driver.  Protection circuit senses Drain voltage with respect to Source. For MOSFET the reference is at the lowest level so the protection circuit can use the same power supply as the output stage. For the JFET, the gate driver 'lies below' the reference voltage (negative supply) so to sense a positive voltage with an op-amp, for example, a positive supply must be added.Conclusion, on a functional point of view, the same driver structure can drive both normally-On JFETs and normally-Off MOSFETs : a change of connection for the Source reference will suffice. Even the resistors used to adjust the turn-On and the turn-Off transients are placed in the same current loop. The difference will appear in the supply voltage of the output stage as gate threshold voltages differ between JFET and MOSFET. A positive voltage supply could be needed for sensing Drain voltage in the JFET driver. 2.2Examples of circuits using JFETs The recent availability of SiC BJT make possible to translate quite easily old fashioned Si schematics into high temperature circuits to control high temperature SiC power converters. This possibility is not addressed in this paper. The authors have taken a look at the feasibility of a gate driver using Normally-On JFETs only in the project to obtain a high temperature driver [9]. 2.2.1Examples of Analog functions This example present a possible protection circuit building bloc : the Drain voltage detector. On fig. 10 Q1 is the power JFET to be monitored. Note that a positive voltage is needed as mentioned early about JFET gate driver requirements. The comparator is transistor Q2, a P-type low power normally-On JFET, it's output remain at Ground level during normal conduction of Q1, D is conducting maintaining Q2 On. In the event of over current, the voltage across Q1 rises, turning off diode D, the gate voltage of Q2 is allowed to reach +VDD, turning Off Q2. The voltage reference of the comparator function is the gate threshold voltage of Q2. The output voltage is now fixed by a resistor to -VDD, this level indicates a over current in Q1. 2.2.2Examples of Logic functions At first glance fig. 11 presents an AND gate along with a NOR gate. But using JFET requires negative gate bias, so inputs Va and Vb are negative in respect to ground. On the other side, the output is strictly positive, leaving us with three logic levels : +VDD, ground and -VDD. Which one is One? Do the lowest voltage level correspond to Low logic level?By choosing the following convention it is possible to get some logic out of these circuits : for positive voltages, the high logic level (One) is +VDD, low logic level (Zero) is Ground. For negative voltages, high logic level is Ground and low logic level is -VDD. This can be illustrated by circuit of fig. 12 , a circuit to translate negative voltage logic levels to positive voltage logic levels.  Logic levelVoltage level  Vin01Gnd+VDDVout01Gnd-VDD  Fig. 10: Analog Drain voltage detector with logic out- put. Power transistor is Q1, detector is Q2. Fig. 9: NormallyOff MOSFET gate driver. Fig. 11: Basic logic gates using JFETs. Fig. 12: Logic voltage level translator and truth table.
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!