If you are of a certain age, when the term ‘supercascode‘ is mentioned you might think of a TV – the ‘Admiral’ 24-inch series 30. It boasted a ‘Super cascode chassis with extra tubes for unparalleled power and interference-free reception, with a built-in Omni-scope antenna’. It even had a remote control – on the end of a length of wire. Today, supercascode means something else; you can follow the origins of the term forward from tube voltage stabilizers in 1939 through early audio amplifiers to stacks of bipolar transistors for high voltage applications. It’s not clear if the ‘super’ in the TV chassis description just meant ‘fantastic’, but the claim of extra tubes is a hint that the arrangement might be an early version of what we think of as a supercascode today – a stack of SiC semiconductor switches controlled by a silicon MOSFET.
So how did the idea re-surface? Switching power supplies and inverters are the norm in modern power conversion and the semiconductor types used vary, depending on the power and voltage level. IGBTs are a low-cost and proven solution, but they have to be switched at low frequency to keep losses down, necessitating large and costly associated magnetic components. Si-MOSFETs can be used at higher frequencies but are limited to around 1000V without resorting to specialized, expensive parts. They also become inefficient at high power and voltage – on-resistance is high giving significant conduction losses and their body diode is of little use with high energy recovery levels. A solution here is to use an external parallel diode for ‘third-quadrant’ operation, along with an additional low-voltage blocking Schottky to steer current away from the MOSFET’s body diode, but that adds further cost and conduction losses. Paralleled MOSFETs are a solution to the conduction loss problem but this only makes dynamic losses higher, current monitoring complex and voltage rating is still limited.
SiC semiconductors are a better solution, with inherent high voltage capability, but implemented as SiC MOSFETs, they still have a relatively poor body diode and need careful gate drive for efficient operation. At this point the ‘cascode’ or ‘SiC FET’ comes in – a combination of an Si-MOSFET and a normally-on SiC JFET which forms a fast, normally-off hybrid switch with low conduction losses, a low-loss body diode effect and an easy, non-critical gate drive.
The SiC FET is a major advance towards the perfect switch and is available from UnitedSiC with ratings up to 1700V, but above that, IGBTs might still seem to be the only practical solution. However, the cascode or SiC FET can look to its heritage and be configured as a ‘supercascode’ with a stack of SiC JFETS instead of a single device, for higher voltage ratings. See Figure.
The passive components in the circuit are all small types for biasing and balancing of the voltages across the series JFETs J1-J5 and the Si-MOSFET M1 is a low voltage type with a standard gate drive. More SiC JFET devices or complete supercascode modules can be stacked, for even high voltage ratings. For example, UnitedSiC has demonstrated a module which switches 40kV/1A with a total of 30 stacked SiC JFET die, each rated at 1700V and showing a combined on-resistance of just 30 ohms.
A nice benefit of the supercascode approach is that current is now easily monitored in the single Si-MOSFET with isolated cells, for a typical 1:1000 sense ratio. Desaturation detection also becomes easier as the Si-MOSFET drain can be monitored, which never exceeds a few volts in conduction or blocking states.
Perhaps the biggest advantage of the technique is the opportunity to use standard off-the-shelf parts in the stack. These can be field-proven and low-cost giving an overall saving compared with parallel MOSFETs or even IGBTs, when the system benefits of high frequency switching are factored-in. End-product development time is shortened and de-risked.
Supercascodes are the future for high-power, high-frequency switching with lowest losses and will find applications in fast EV chargers, traction inverters, renewable energy and more. They will be available in standard module packages but not, I suspect, in maple, walnut or rosewood tonings, like the Admiral TV.
Sign up for our quarterly newsletter and receive important technical information on all new products, app notes, white papers, and blogs.