Ally Winning, European Editor, PSD
If it ain’t broke, don’t fix it is a commonly enough used phrase. The transistor is one thing that definitely ain’t broke. Since the first working bipolar junction transistor was produced in 1948 and the first CMOS one in 1963, those two designs have been the foundation of the technology that has completely changed the world and led to two industrial revolutions. Surely if there was a better way of designing transistors, it would have been discovered over the last 70 years?
Not according to David Summerland, CEO of Search For The Next (SFN) who is leading a consortium that has developed a new type of transistor that has been showing promising results during trials. Named Bizen, the new transistors are based on a traditionalbipolar design. The gate of the transistor is replaced with an isolated quantum tunnel connection. Inputs are based on Zener quantum tunnel current mechanics, resulting in switchingvoltages at the millivolt level. That small switching voltage lowers the dynamic power the device consumes.
The resulting device is completely reversible, with two anodes instead of the usual collector and emitter. The input structure allows more than one tunnel at a time, meaning that logic devices can be built from a single transistor. For example, a three input NOR gate can be formed from a single transistor using four process mask layers, around half the layers that would be required to implement the same block in CMOS. The whole process is confined to 8 process masks to reduce costs and for less complex manufacturing with shorter lead times.
In theory, and simulation, the results are interesting. The new process triples the gate density, reducing the required die size by two thirds. This increase in density doesn’t affect the performance of the new transistor, which retains or beats the performance and low power capabilities of CMOS.
Bizen transistors can be manufactured in standard fabs with no specialised processes. For the last two years, SFN has being working with Semfab at the company’s base in Fife in Scotland to refine the manufacturing process and see how the practical results compare with theoretical ones. According to Semfab’s CEO, Allan James the new devices have been manufactured on a 1 micron process to tri 4/5 standard, just below full production quality, and the initial results are close to the theoretical. James said the next stage was to use a 0.35micron process to check how theoretical results scale with the process node. SFN is also working on a Cadence PDK, which will allow CMOS designs to be translated to the new process.
The development is still at quite an early stage, but SFN is confident that things will move quickly and the company intends to develop a programmable junction transistor with an internal processing unit as a proof-of-concept in the near future. The power market is one that Summerland thinks would benefit from the technology, and initial prototypes will likely be aimed at the area. The technique should also be suitable for large digital processors and compound and wide-bandgap materials.